Clinical Magnetic Resonance Imaging PDF

Clinical Magnetic Resonance Imaging: 3-Volume Set
by Robert R. Edelman, John Hesselink, and Michael Zlatkin

● Hardcover: 4200 pages
● Publisher: Saunders; 3 edition (October 21, 2005)
● Language: English
● ISBN-10: 0721603068
● ISBN-13: 978-0721603063
Description:
The MRI reference that the American Journal of Roentgenology called "hard to beat" is back
in a state- of-the-art New Edition! It comprehensively examines all of the newest technologies
and clinical applications relevant to MR imaging of the heart, brain, head and neck, spine,
body, and musculoskeletal system.
4,700 beautifully reproduced illustrations - including hundreds of new full-color images - help
readers accurately diagnosis a broad spectrum of conditions. This exhaustively revised 3rd Edition
delivers more than 70% new content and authors and a new full-color format that covers all
the important technologies as you see them.
Audience:
Diagnostic Radiologists, Radiology Residents, Health Science Libraries.
Table of Contents
VOLUME 1
I. PHYSICS, INSTRUMENTATION, AND ADVANCED TECHNIQUES
1. History
2. Basic Principles
3. Practical Considerations and Image Optimization
4. Instrumentation: Magnet, Gradients, and Coils
5. Pulse Sequence Design
6. Biochemical Basis of the MRI Appearance of Cerebral Hemorrhage
7. Advanced Imaging Techniques
8. Parallel Imaging Methods
9. Principles of Functional Imaging of the Brain
10. Diffusion-Weighted Imaging
11. Diffusion Tensor Imaging
12. Perfusion Imaging of the Brain
13. Contrast Agents: Basic Principles
14. Tissue-Specific Contrast Agents
15. Molecular Imaging
16. Functional Imaging of the Body
17. Magnetic Resonance Spectroscopy: Basic Principles
18. High-Field Imaging
19. MRI-Guided Interventions
20. MRI-Guided Intravascular Interventions
21. Screening MRI
22. Image Artifacts and Solutions
23. Image Processing: Principles, Techniques, and Applications
24. Bioeffects, Safety, and Patient Management
25. The MR Imaging Center
26. Measuring the Capacity, Productivity, and Costs of Service of an MRI Center: The Service Costing Approach
II. HEART
27. Magnetic Resonance Angiography: Basic Principles

28. Basic Principles and Clinical Applications of Flow Quantification
29. Principles and Optimization of Contrast-Enhanced Three-Dimensional Magnetic Resonance Angiography
30. Magnetic Resonance Angiography of the Body
31. Magnetic Resonance Venography of the Body
32. Cardiac Imaging Techniques
33. Coronary Arteries
34. Myocardial Perfusion
35. Myocardial Viability
36. Valvular Heart Disease
37. Adult Heart Disease
38. Pediatric Congenital Heart Disease
VOLUME 2
III. BRAIN
39. Brain: Indications, Technique, and Atlas

40. Adult Brain Tumors
41. Brainstem, Cranial Nerves and Cerebellum
42. Pituitary Gland and Parasellar Region
43. Perfusion and MRS for Brain Tumor Diagnosis
44. Infectious & Inflammatory Diseases
45. Intracranial Hemorrhage
46. Trauma
47. MR Imaging of Epilepsy
48. Practical Clinical Applications of Functional MRI
49. Aneurysms & Vascular Malformations
50. Stroke & Cerebral Ischemia
51. MR Angiography of the Head and Neck
52. Diffusion & Perfusion MRI
53. White Matter Disease
54. Diffusion Tensor Imaging
55. Neurodegenerative Disorders
56. Toxic and Metabolic Disorders
57. Developmental Disorders
58. Pediatric Brain Tumors
59. Pediatric Anoxic/Ischemic Injury
60. Functional MRI in Neuropsychiatric Disorders
61. MR Spectroscopy of the Brain
IV. HEAD AND NECK
62. Orbital and Intraocular Lesions
63. Skull Base & Temporal Bone
64. Paranasal Sinuses & Nasal Cavity
65. Nasopharynx & Deep Facial Compartments
66. Lower Face & Salivary Glands
67. Neck
V. SPINE
68. Spine Atlas
69. Spinal Cord and Intradural Disease
70. Degenerative Disease
71. Positional and Kinetic Spin Imaging
72. Post-operative Spine
73. Pediatric Spine: Congenital and Developmental Disorders
74. Vertebral & Paravertebral Abnormalities
75. MR Neurography
VOLUME 3
VI. BODY
76. Chest, Including Lung Function
77. Breast Cancer
78. Breast Implants
79. MR Cholangiopancreatography
80. Gallbladder
81. Focal Liver Disease
82. Diffuse Liver Disease
83. Liver Transplant Imaging
84. Pancreas
85. Bowel, Peritoneum, and Mesentery
86. Kidneys
87. Adrenal Glands
88. Bladder
89. Prostate
90. Scrotum and Testes
91. Malignant Disorders of the Female Pelvis
92. Female Pelvis: Benign Conditions
93. Pelvic Floor Imaging
94. Fetal MRI
95. Pediatric Body
VII. MUSCULOSKELETAL SYSTEM
96. Musculoskeletal MRI Techniques
97. MR Arthrography
98. Kinematic MRI
99. Shoulder
100. Elbow
101. Wrist and Hand
102. Hip
103. Knee
104. Ankle and Foot
105. TM Joint
106. The Musculotendinous Unit
107. Bone and Soft Tissue Tumors
108. Marrow Disorders
109. Cartilage
110. Pediatric Musculoskeletal Disorders
111. Synovial Disorders
112. Extremity Scanners
Printed from: Clinical Magnetic Resonance Imaging, 3rd editio... http://www.clinicalmri.com/content/printpage.cfm?ID=S1
HYSICS NSTRUMENTATION AND DVANCED

ECHNIQUES
page 1
page 2
page 2
page 3
ISTORY OF AGNETIC ESONANCE

Roy Irwan
Matthijs Oudkerk
Great advances have been made in recent decades in the development of (nuclear) magnetic
resonance (imaging). The imaging community often omits the word "nuclear" and attaches the word
"imaging" so that NMR becomes MRI. The former change is largely due to public relations concerns,
while the latter refers to the imaging. For convenience and consistency, MR is used for both NMR and
MRI throughout this chapter unless otherwise stated.
A knowledge of the history of MR allows us to better appreciate the remarkable progress in the field.
According to the Roman philosopher, Marcus Tullius Cicero (106-43 BC), those who have no
1
knowledge of the things that took place before their birth will remain a child.
2-5,14-36
The fundamentals of conventional MR have been expounded in a number of texts. It is our intent
in this chapter to chronologically review the most important milestones related to the development of
MR, including the first commercially available MR scanners. Such a review, especially in a relatively
short chapter, inevitably leads to difficult compromises. Those interested in more detailed discussions
are referred to the cited references and references therein.
We confine the scope of the discussion in this chapter to a description of the relevant contributors to
the physics both before and after the Nobel Prize in 1952, a year that is often regarded as the birth of
MR.
In addition, a guided tour of the Fourier transform will be given in a separate section, before we
discuss the development of MR imaging. Many manufacturers have developed pulse sequences and
have often used different names for the same technique. For this reason we strive to classify the main
classes of pulse sequences and list them according to the major manufacturers.
Finally, it is not our goal to cover clinical applications or contrast agents, which are dealt with in other
chapters later in this book.
Printed from: Clinical Magnetic Resonance Imaging, 3rd edition (on 11 April 2010)
© 2010 Elsevier
1 of 1 11-04-2010 23:22
Printed from: Clinical Magnetic Resonance Imaging, 3rd editio... http://www.clinicalmri.com/content/printpage.cfm?ID=HC001001
OVERVIEW OF THE HISTORY OF MAGNETIC RESONANCE

Early Scientific Contributions
Although the basic discovery of MR was often related to the Nobel Prize in 1952, the fundamental
phenomenon of MR is much older and may be traced back to the Fourier transform which is a real
watershed in the history of MR.
Fourier
Jean Baptiste Joseph Fourier (Fig. 1-1) was born on March 21, 1768 in Auxerre and died on May 16,
1830 in Paris. Fourier served three years as the secretary of the Institut d'Egypte at the beginning of
the 19th century and later became prefect of the Isère département in France.6 Furthermore, he was
one of the chief engineers on Napoleon's expedition to Egypt, where the torrid climate appealed to him.
The focus of his life, however, was mathematics and without his Fourier transform we would not be
able to create MR images. A brief overview of the Fourier transform will be given in a separate section
later in this chapter.
The 1920s were extremely fruitful scientifically, particularly due to the success of quantum theory and
quantum mechanics. The milestones in the field of MR are summarized below.
Pauli
In 1924, an Austrian physicist, Wolfgang Pauli (Fig. 1-2), proposed a quantum spin number for
electrons. He is best known for the Pauli exclusion principle, proposed in 1925, for which he received
12
the Nobel Prize in 1945.
page 3
page 4
Add to lightbox
Figure 1-1 J Fourier,1768-1830, the founder of the Fourier transform, which is the basis of most
(medical) imaging modalities today.
This principle says that two identical particles (fermions) cannot exist in the same quantum state.4
1 of 17 11-04-2010 23:28
Furthermore, prior to World War II, Pauli was the first to recognize the existence of the neutrino, an
uncharged and massless particle that carries off energy in radioactivity.12
Uhlenbeck
In the same year as Pauli proposed his exclusion principle, George Uhlenbeck (Fig. 1-3), introduced
the concept of a spinning electron, with resultant angular momentum and a magnetic dipole moment
arising from the spinning electrical charge. It was Pauli's exclusion principle that led Uhlenbeck to arrive
4
at this idea. He wrote :
"… it occurred to me that, since (I had learned) each quantum number corresponds to a
degree of freedom of the electron, Pauli's fourth quantum number must mean that the
electron had an additional degree of freedom - in other words the electron must be
rotating."
The concept immediately excited a number of great scientists at that time such as Bohr, Pauli,
Einstein, Heisenberg and others interested in quantum theory. Besides this work, Uhlenbeck also
contributed significantly to atomic structure and the kinetic theory of matter. He extended Boltzmann's
equation to dense gasses and wrote two papers on Brownian motion.11
Rabi
Add to lightbox
Figure 1-2 W Pauli, born on 25 April 1900 in Vienna, received the Nobel Prize for Physics in 1945 for
his exclusion principle. (Reproduced by permission of the Nobel Foundation.)
2 of 17 11-04-2010 23:28
Add to lightbox
Figure 1-3 G Uhlenbeck (left) and N van Kampen (right) during the Boltzman conference in Vienna in
1973. Uhlenbeck proposed the concept of electron spin in 1925. (Courtesy of N van Kampen.)
During the early 1930s Isaac Rabi (Fig. 1-4), born in Raymanov, Austria, on July 29, 1898, set up a
laboratory at Columbia University in New York which later became a major center for atomic and
13
molecular studies.
page 4
page 5
Add to lightbox
Figure 1-4 I Rabi was born in Austria in 1898 and awarded the Nobel Prize for Physics in 1944 for his
investigation on molecular beam magnetic resonance methods. (Reproduced by permission of the
Nobel Foundation.)
3 of 17 11-04-2010 23:28
Rabi's successful research was influenced by the visit of Cornelis Jacobus Gorter (see below), a Dutch
physicist, in September 1937. Gorter and his co-worker Broer reported unsuccessful attempts to
observe nuclear magnetic resonance in pure crystalline materials.7 This first publication with the name
"Nuclear Magnetic Resonance" in Gorter's paper provided important clues to Rabi, who accepted and
realized Gorter's suggestions concerning his experiments, modified them and was finally able to
2
observe resonance experimentally. This led to the publication of "A New Method of Measuring Nuclear
Magnetic Moment" in 1938 where the first MR signal from LiCL (lithium chloride) (Fig. 1-5) was
reported.8
Although this publication refers to Gorter's visit and unsuccessful experiment, it does not acknowledge
his suggestions. Gorter's reaction to Rabi's publication was rather furious9:
"I cannot deny that I felt some pride, mixed with the feeling that my contribution was
somewhat undervalued though my advice was acknowledged in the Letter."
Rabi was eventually awarded the Nobel Prize for physics in 1944 for his investigation on the molecular
beam magnetic resonance methods.
Gorter
Gorter himself (Fig. 1-6) was born in Utrecht on August 14, 1907. He went to school in The Hague and
studied physics in Leiden. He was the first to demonstrate the phenomenon of paramagnetic
relaxation10 and narrowly missed the discovery of nuclear spin resonance.
Add to lightbox
Figure 1-5 First reported MR signal from LiCl by Rabi. The beam intensity is measured as a function
of various values of the magnetic fields.2,11 One ampere corresponds to approximately 1.84 × 10-4
Tesla (T).
4 of 17 11-04-2010 23:28
Add to lightbox
Figure 1-6 C Gorter, a Dutch physicist, attempted to observe resonant heating of a substance in a
strong magnetic field, without success. His negative result, however, provided the important clues to
Rabi's successful experiments. (Courtesy of Leiden University.)
His approach was to use a resonance property of the nuclear spins when they are placed in a
magnetic field B0. At the Larmor frequency,
where γ is the gyromagnetic ratio, Gorter knew that a magnetic
dipole transition should occur if an alternating radiofrequency (RF) field B1 is applied perpendicular to
B0.
page 5
page 6
Add to lightbox
Figure 1-7 First measurement for electron paramagnetic resonance on copper at 4.76 mT carried out
5 of 17 11-04-2010 23:28
by Zavoisky at the Kazan State University.
Gorter expected to detect the heat produced by resonant absorption in the sample by using a sensitive
calorimetric method that he had successfully employed in studies of electron paramagnetic relaxation.
Unfortunately, he did not observe heat absorption from the resonant process and concluded that a long
spin-lattice relaxation time caused the spin system to be partially saturated when RF energy was
4
absorbed.
Later it appeared that he used excessively pure samples that produced longer relaxation times,
causing the partial saturation mentioned before and therefore no detectable resonance. He then
missed the Nobel Prize.
As a person, Gorter was both an experimentalist and skilled at developing theoretical ideas. He died in
Leiden on March 30, 1980.
Zavoisky
The first detection of electron paramagnetic resonance (Fig. 1-7) was carried out at Kazan State
University, Russia, by Evgeny Zavoisky (Fig. 1-8) in 1944-1945. Zavoisky had first attempted to detect
MR in 1941 but, like Gorter, he had failed.
He used a sample of water to which paramagnetic ions were added to shorten the completely
unknown relaxation time. He was inspired by Rabi's resonance studies in molecular beams but he was
also aware of Gorter's unsuccessful 1936 attempt to detect MR.
After the German invasion of Russia caused several laboratories to move from Moscow to remote
Kazan, he continued trying to detect electron resonance in both solids and solutions of paramagnetic
salts. Eventually, he obtained signals with equipment working in the range of 1 GHz and published a
paper13 which is generally accepted as the first reported observation of electron paramagnetic
resonance.
Add to lightbox
Figure 1-8 E Zavoisky, born on 28 September 1907, a major contributor to MR from Kazan, which was
part of the former Soviet Union.
Zavoisky's name is written in the history of science because of his detection of electronic paramagnetic
6 of 17 11-04-2010 23:28
resonance and his brilliant work on nuclear physics, controlled thermonuclear synthesis and physical
electronics.4 His work resulted in important progress in MR, although Russian contributions to MR
were hardly discussed in the West.
The practical application of the discoveries made so far came from a breakthrough for which two
American physicists received the Nobel Prize.
The Nobel Prize for Physics 1952

Although an incredible amount of fundamental work on MR had been done long before World War II,
1946 is commonly regarded as the year in which MR was discovered. During this year, Felix Bloch14
15
(Fig. 1-9) and Edward Purcell (Fig. 1-10) independently detected the MR phenomenon, for which
both shared the Nobel Prize for Physics in 1952.12
Their work was particularly accredited to a property of atomic nuclei having an odd number of nucleons
that precess at RF in a magnetic field, the frequency depending on the magnetic strength.
Purcell
Purcell was born in Taylorville, Illinois, USA, on August 30, 1912. 12 As a leader of a fundamental
research group at the MIT Radiation Laboratory, he proposed trying an experiment to detect the
transition between nuclear magnetic energy levels using RF methods. He was initially unaware of
Gorter's similar unsuccessful experiment.
page 6
page 7
Add to lightbox
Figure 1-9 F Bloch first observed the property of atoms in a magnetic field, which he referred to as
"nuclear induction." Together with E Purcell, he shared the Nobel Prize for Physics in 1952.
(Reproduced by permission of the Nobel Foundation.)
Together with his colleagues Torrey and Pound, he prepared a resonant cavity to study the absorption
of RF energy in paraffin. After some trial and error, they verified the signal and found the resonance.
15
Their report of this discovery was published in Physical Review in 1946.
7 of 17 11-04-2010 23:28
Bloch
Bloch was born in Zurich, Switzerland, on October 23, 1905 and taught at the University of Leipzig until
12
1933. He came to the US in 1933, joined Stanford University at Palo Alto in 1934 and became a US
citizen in 1939. He died in 1983 in Zurich.
Add to lightbox
Figure 1-10 Purcell independently discovered the MR phenomenon in 1946. (Reproduced by
permission of the Nobel Foundation.)
Add to lightbox
Figure 1-11 Bloch's schematic representation describing how the RF field B1 induces rotation of
magnetization towards the transverse plane (left) and the rotating frame behavior (right).
In contrast to Purcell's experiment, Bloch and his colleagues used what they called "nuclear induction."
8 of 17 11-04-2010 23:28
Bloch described the experiment as measuring an electromotive force resulting from the forced
precession of the nuclear magnetization in the applied RF field.14 This principle is shown in Figure 1-11,
which is a schematic representation most commonly used to describe the concept of MR today.
Furthermore, the behavior of the magnetization vector M shown in Figure 1-11 is described by the
so-called simplified Bloch equation, given by:
where B includes the various magnetic fields applied.

page 7
page 8
Although Bloch and Purcell were the first to see the possibility of using MR for medical imaging, it was
not until the 1960s that the first MR prototypes were built for medical purposes.
First Magnetic Resonance Images

Lauterbur
The year 1973 was very important for medical imaging technology. In that year X-ray based computed
tomography (CT) was introduced by Hounsfield27 and MR imaging (MRI) was first demonstrated on
two small tubes of water by Paul Lauterbur (Fig. 1-12), who used a backprojection technique similar to
that of CT.28
Lauterbur published his work in Nature, in an article entitled "Image formation by induced local
interaction; examples employing magnetic resonance."28 Despite the fact that the paper was nearly not
published, having been initially rejected by the editor as not of sufficiently wide significance for inclusion
in the journal, his work represented the foundation for a revolution in imaging.
In this paper Lauterbur described a new imaging technique which he termed zeugmatography (from
the Greek zeugmo meaning joining together), which was later replaced by MRI. This zeugmatography
referred to the joining together of a weak gradient magnetic field with the stronger main magnetic field,
allowing the spatial localization of two tubes of water (Fig. 1-13).
Moreover, he introduced the use of gradients in the magnetic field. By analysis of the characteristics of
the emitted radio waves, he could determine their origin. This made it possible to build up
two-dimensional pictures of structures that could not be visualized with other methods.
Add to lightbox
Figure 1-12 P Lauterbur, at that time a professor in chemistry at the University of New York at Stony
9 of 17 11-04-2010 23:28
Brook, first demonstrated MRI on small tube samples in 1973, the same year that CT was invented.
(Copyright Bruno Press.)
This imaging experiment, therefore, moved from the single dimension of MR spectroscopy to the
second dimension of spatial orientation and thus became the foundation of MR imaging. MR also owes
a debt to CT as it was developed initially on the back of CT but quickly outpaced that technique.
For this reason, Lauterbur shared the 2003 Nobel Prize in Physiology and Medicine with Peter
Mansfield (discussed below) for their discoveries concerning MRI.
Mansfield
Add to lightbox
page 8
page 9
Figure 1-13 The first MR image of two tubes of water demonstrated by Lauterbur using a
backprojection technique, upon which CT is based.
10 of 17 11-04-2010 23:28
Add to lightbox
Figure 1-14 Sir Peter Mansfield devised a new MR imaging technique called echo-planar imaging
(EPI) in the late 1970s. EPI is considered to be the first ultra high-speed imaging technique and was
used to demonstrate the first clinical MR images. (Copyright Bruno Press.)
The contributions of Sir Peter Mansfield (Fig. 1-14) and the Nottingham group are numerous and
fundamental. They include:
NMR diffraction in solids

slice selection
active magnetic shielding of gradient coils
echo volume imaging
active acoustic shielding methods that lower noise levels produced by gradient coils.
Of great relevance to the field of fast MR imaging, and in particular to diffusion, perfusion and
functional imaging of the brain, Mansfield further developed the utilization of gradient magnetic fields.
He showed how the signals can be mathematically analyzed which later gave rise to the echo-planar
imaging (EPI) technique in 1977.35 EPI was the first ultra high-speed imaging technique and many of its
28
variants are now in use.
Furthermore, Mansfield was the first to demonstrate clinical MR images using his technique. A color
version of a cross-sectional MR scan of a human finger in vivo is shown in Figure 1-15.
Thus, modern MR imaging of human internal organs with exact and noninvasive methods was born. For
this reason, the Nobel Assembly at the Karolinska Institute decided to award the Nobel Prize in
Physiology and Medicine for 2003 jointly to Mansfield and Lauterbur, as discussed earlier.
First Commercial Magnetic Resonance Scanners

Aberdeen Prototype
The first MR scanners appeared almost at the same time in the 1970s. However, the whole-body
magnet (Fig. 1-16) was first built by Oxford Instruments Ltd in cooperation with the University of
16
Aberdeen. Work first began in Aberdeen in 1972 by a group under the direction of John Mallard (Fig.
1-17) and led by James Hutchison. They originated the "spin-warp" method of spatial localization for
MRI, now universally used.
11 of 17 11-04-2010 23:28
Add to lightbox
Figure 1-15 First clinical MR image showing a cross-section of a human finger as demonstrated by
38
Mansfield in a color version.
Add to lightbox
Figure 1-16 The Aberdeen MR imaging prototype. Hutchison himself is in the position of the patient.
The group was very interested in the T1 relaxation time and extensive studies were performed on
normal and pathologic tissues.16 T1 values of normal and malignant animal and human tissues have
been presented and their implications for in vivo clinical MR imaging have been discussed. 17
FONAR
In 1971, Raymond Damadian (Fig. 1-18) reported a marked difference in relaxation times between
normal and abnormal tissues of the same type, as well as between different types of normal tissues. 18
12 of 17 11-04-2010 23:28
page 9
page 10
Add to lightbox
Figure 1-17 J Mallard at Aberdeen University led research on the first whole-body MR prototype
depicted in Fig. 1-16.
This discovery is the basis for the tissue contrast of every MR image today and hence the creation of
the MR industry. This milestone is also often regarded as the first successful MR method for medical
diagnosis in which MR could detect disease.20 Damadian filed a pioneer patent for the practical use of
his discovery in 1972 (Fig. 1-19).
Moreover, he noticed that centering the MR object in the magnet produced a good signal and moving
the object too far off center caused it to vanish. Based on this idea, he designed a scan using a
saddle-shaped magnetic field so that only the center was at the resonant frequency, later known as
Field fOcusing Nuclear mAgnetic Resonance (FONAR).18
Philips
13 of 17 11-04-2010 23:28
Add to lightbox
Figure 1-18 R Damadian, an American medical doctor at the State University of New York in Brooklyn,
demonstrated MR of the whole human body.
Add to lightbox
Figure 1-19 Damadian patented the relaxation differences and their use in MR scanning.22 (Courtesy
of FONAR Corporation).
The pioneering MR scanner, the result of a joint project between Philips Research Laboratories
Eindhoven and Philips Medical Systems Best, is shown in Figure 1-20. Luiten led the PROTON Project
which commenced in May 1978 and used a resistive magnet with coils of copper tubing which acted as
the conductor and also conveyed the coolant, producing a magnetic field strength of 0.15 T.21
The magnet had a field diameter of approximately 1 m and at that time, it was the biggest and the
14 of 17 11-04-2010 23:28
strongest scan magnet in the world with a power of 60 kW. A couple of years later superconducting
magnets were designed which allow for much higher magnetic field strengths.
page 10
page 11
Add to lightbox
Figure 1-20 The first Philips MR 0.15 T scanner built in 1978 at Philips Research Laboratories,
Eindhoven. Frame dimensions are 1.5 × 1.9 × 1.9 m (l × w × h). (Courtesy of Philips Medical
Systems, Best.)
Add to lightbox
Figure 1-21 The first transverse MR image (Luiten's head) acquired using a 0.15 T Philips resistive
magnet in 1978. Slice thickness was 10 mm. The method was selective excitation and 2D Fourier
imaging.24 (Courtesy of Philips Medical Systems, Best.)
15 of 17 11-04-2010 23:28
Add to lightbox
Figure 1-22 One of the early resistive 0.2 T MR scanners built in 1979. (Courtesy of Siemens Medical
Solutions, Erlangen.)
Add to lightbox
Figure 1-23 Siemens' first transverse MR image of a volunteer acquired using a 0.2 T in 1980.
(Courtesy of Siemens Medical Solutions, Erlangen.)
Unlike the backprojection technique proposed by Lauterbur, which used one-dimensional Fourier
transform (1D-FT), the first successful MR image produced by Philips was obtained with the
two-dimensional Fourier transform (2D-FT) imaging.21
Despite a relatively low signal-to-noise ratio (SNR), the image shows many anatomical details, such as
optic nerves, skull bone and internal carotid arteries (Fig. 1-21). Not until after the advent of
superconducting magnets in 1982 could better MR images be produced.
Siemens
Siemens' involvement with MR development began in the late 1970s. An early resistive 0.2 T prototype
magnet is shown in Figure 1-22. This prototype was the first Siemens MR to be installed in a clinical
environment.22
page 11
page 12
16 of 17 11-04-2010 23:28
Add to lightbox
Figure 1-24 The first MR image of Siemens, acquired in 1979 and recalculated in 1980. (Courtesy of
Siemens Medical Solutions, Erlangen.)
22
The first Siemens head image (Fig. 1-23) was produced using this prototype in March 1980. The first
images of a red pepper (Fig. 1-24) in 1980 were the signal to get Siemens off to a flying start on the
next phase of MR product development, which culminated in the Magnetom. The first Magnetom, with
a superconducting 0.35 T magnet (Fig. 1-25), was delivered to the Mallinckrodt Institute in St Louis,
USA.22
General Electric
Although General Electric (GE) Medical Systems had built one of the early electromagnets (Fig. 1-26),
they did not become really interested in the MR market until 1982.
The GE Research and Development center in Schenectady began to explore the fields of MR imaging
and spectroscopy following the arrival in 1980 of Paul Bottomley and Bill Edelstein, who had had
significant experience in Aberdeen.4
By November 1982, they had obtained MR images that were shown at the annual meeting of the
Radiological Society of North America. A year later, GE succeeded in producing images with 256 ×
23
256 resolution with 4 mm slices and 100-s scan times using a 1.5 T magnet. Bottomley et al reported
1 31 13
the first H MR imaging and localized P or C chemical shift spectroscopy as two new noninvasive
1
diagnostic tools in 1983. Previously, H MR imaging systems all operated at magnetic field strengths
below 0.65 T.
© 2010 Elsevier
17 of 17 11-04-2010 23:28
A GUIDED TOUR OF THE FOURIER TRANSFORM
Add to lightbox
Figure 1-25 The first Magnetom with a superconducting 0.35 T magnet. (Courtesy of Siemens Medical
Solutions, Erlangen.)
Add to lightbox
4
Figure 1-26 The early electromagnet built by General Electric in 1947 and designed for 0.15 T.
The Fourier transform (FT) is a major foundation of most modern imaging techniques. Introduced into the MR
field by Richard Ernst,34 it is now possible to go back and forth between time signal and its frequency
spectrum with enough speed to create a whole new range of applications for this mystical mathematical
device. An overview of the development of algorithms to perform FT faster and faster can be found in a paper
by Cooley et al.26
Although its role in MR is of considerable importance, many MR physicians simply regard the FT as a mystical
mathematical tool without any real understanding of it. For this reason we attempt here to explain briefly the
basic principles of the FT without going into great detail, before we move on to the next section.
page 12
page 13
1 of 4 11-04-2010 23:32
Add to lightbox
Figure 1-27 Three cosine waves with their corresponding Fourier transform. A, 2 Hz frequency cosine wave
with its FT being two spikes. B, 4 Hz cosine wave with half the amplitude of A. The spikes are twice as far
apart than in A because the frequency of the curve on the left is twice as large as that of the curve in B. C,
Their sum (A + B) and its frequency spectrum.
The discussion in this section is centered around the following questions.
What is the Fourier transform?

What can it do?
What is its relevance to MR?
A more in-depth coverage can be found in many texbooks.24,25,29
For some time the FT has served as a bridge between the time domain and the frequency domain. The FT
shows how any signal, one-, two- or three-dimensional, can be broken down into a sum of sine waves of
different frequencies, phases and amplitudes. These waves are called frequency components because they
are identified by frequency, as explained later. For convenience, let us discuss 1D-FT first.
Time-signal cosine waves (Fig. 1-27A) correspond to a pair of spikes in the frequency domain positioned
exactly at the frequencies of the cosine wave (Fig. 1-27B). This pair of spikes is symmetrical about zero
because the cosine wave is identical for positive and negative frequencies.
For example, a cosine wave that has a frequency of 2 Hz (there are two cycles per second) corresponds to a
pair of spikes at ±2 Hz at the frequency axis (Fig. 1-27A). The height of the spike is proportional to the
amplitude of the cosine wave. Similarly, Figure 1-27B depicts a cosine wave with a frequency twice as high
and an amplitude half that of Figure 1-27A. As expected, there are a pair of spikes at ±4 Hz at the frequency
axis.
Furthermore, Figure 1-27C demonstrates a very important property of the FT stated earlier, namely that the
FT of the sum of two or more signals is the sum of each signal's FT.
2 of 4 11-04-2010 23:32
The left-pointing arrow in Figure 1-27 symbolizes the inverse FT which returns a Fourier spectrum expressed
by its frequency components to its original temporal representation.
Mathematically, the FT results in complex numbers that consist of real and imaginary parts that are always
90° out of phase. In practice, complex numbers are just another way of representing a magnitude and a
phase of a signal.
To understand how the FT can be related to the MR, it is necessary to expand 1D-FT into 2D-FT which is
often referred to as a data matrix or an image, as will be discussed below.
Magnetic Resonance Imaging

page 13
page 14
Add to lightbox
Figure 1-28 k-space representation. Each row in the data matrix is associated with a phase-encoding
gradient pulse applied to a MR signal. Each column in the data matrix represents the frequency of each
sample of the signal.
A data matrix is an alternative representation for an MR image and is often called a frequency domain or
29,30
k-space representation of an MR image. The samples in a data matrix are identified by co-ordinates kx
and ky. The frequency domain is interesting in MR imaging because MR images begin as data samples in this
domain.
The concept of k-space was patented by Likes31 and later, independently, proposed by Tweig.32 In principle,
a data matrix can be performed in any order depending on the pulse sequence design, as explained later in
this section.
Furthermore, because a phase-encoding gradient is applied to each row in a data matrix, the vertical axis (ky)
is referred to as phase-encoding. Similarly, a frequency-encoding gradient is applied to each column in a data
matrix and therefore the horizontal axis (kx) is referred to as frequency encoding (Fig. 1-28). The phase and
frequency encoding will be explained later. Since the early 1980s a typical size for the data matrix has been
256 × 256 (phase × frequency encode).
After data acquisition is finished, an MR image can be reconstructed using 2D-FT which transforms any image
from the frequency domain to the image domain and vice versa. Since the data matrix consists of rows and
columns, 2D-FT normally performs each row of the data matrix, from top to bottom. Sometimes it is said that
inverse 2D-FT should be performed to obtain a MR image. Both statements are correct as the alternatives
differ mainly by a scale factor.
Figure 1-29 demonstrates the 2D-FT in which a row in the k-space is shown simply as a 1D-FT of the same
row in the image space. In other words, the frequency signal on the left shows a projection of anatomy in the
3 of 4 11-04-2010 23:32
image space.
Before a MR signal that comes from the coils, often referred to as the free induction decay (FID) signal, can
be meaningfully interpreted by the Fourier transform, it has to undergo several processing stages, two of
which are discussed below.
Frequency Encoding
Frequency encoding in MR imaging is a procedure to resolve spatial information along one direction (usually
horizontal) of a MR image. In the absence of any gradients, the excited protons are spinning at the same
frequency and the received signal is an exponentially decaying curve. This MR signal is often referred to as
T2* FID signal and is illustrated in Figure 1-30.
The frequency of a MR signal is equal to the Larmor frequency at the location from which the signal is
emitted.30 It is the function of the frequency-encoding gradient to spread out the Larmor frequency to
distinguish different locations along one direction. In this case, the protons will be spinning at frequencies that
depend upon their position along the gradient. The received signal is thus the sum of the signal at each
frequency (Fig. 1-31). The FT is then used to break up the sum signal into its separate frequency
components.
Phase-Encoding
Spatial information is encoded into the phase of MR signals by the process of phase-encoding. Phase-
encoding resolves structures along the direction that is perpendicular to the frequency encoding.
As phase is related to angle, phase-encoding can be most easily understood, for instance, as viewing the
object from a series of perspectives at different angles (Fig. 1-32). Hence, the steps involved in phase-
encoding along the vertical direction (see Fig. 1-28) can be roughly compared to the rotation steps in CT.
While the phase-encoding gradient is on, each position along the phase-encoding direction has its own unique
Larmor frequency, a point that is similar to the frequency encoding. In contrast, when the gradient is switched
off, each position along the phase-encoding direction has an identical Larmor frequency.
However, the phases of nuclei in different spatial positions are not identical once the phase-encoding gradient
has been turned off. The amount of the phase shift caused by the phase-encoding gradient is the key to
identifying the location of structures along the phase-encoding direction of a MR image.
The effect of the phase-encoding gradient causing a position-dependent phase shift is illustrated in Figure
1-33. The phase shift increases as long as the phase-encoding gradient is applied. The figure on the left
represents the phase dispersion after a brief period of t seconds, while the figure on the right shows the
accumulated phase difference after 2t seconds.
© 2010 Elsevier
4 of 4 11-04-2010 23:32
DEVELOPMENT OF MAGNETIC RESONANCE IMAGING

Since its introduction as a diagnostic tool in the 1970s, MRI has undergone dramatic improvements in
all the features that define image quality, such as resolution, signal-to-noise ratio (SNR) and speed.
page 14
page 15
Add to lightbox
Figure 1-29 Fourier transforming each row in the k-space projects the anatomy along the frequency-
encoding direction. The data samples of each row in the k-space are converted into a plot of intensity
in the image domain.
1 of 11 11-04-2010 23:32
Add to lightbox
Figure 1-30 No gradient is applied during data acquisition which produces a T2* FID signal.
Add to lightbox
page 15
page 16
Figure 1-31 A frequency-encoding gradient is applied during data acquisition to spread out the
Larmor frequency into distinguishable frequencies. The received signal is broken up into signals at
different frequencies.
2 of 11 11-04-2010 23:32
Add to lightbox
Figure 1-32 Phase-encoding can be thought of as viewing an object from different angles.
This improvement in resolution and speed has been made possible by the development of stronger
magnets and faster gradients producing higher gradient amplitudes (mT/m) and shorter rise times
(ms). Simultaneously, the pulse sequences have also grown rapidly over the years since the EPI
discovery by Mansfield.
For this reason, we have organized our discussion of the development of MRI into three parts, i.e., the
magnet, gradient and pulse sequences that make up the main components of a MRI scanner.
Magnets
As the most important and expensive component, the magnet predominantly determines SNR and
therefore the image quality. The SNR increases approximately linearly with field strength, provided the
37
same sampling bandwidths are used. However, the stronger the magnetic field, the greater the static
field inhomogeneity.
Resistive magnets were used in the beginning (see previous section). However, these are limited by
their electrical power requirements to around 0.3 T and hence are only suitable for low field strength.
Add to lightbox
3 of 11 11-04-2010 23:32
Add to lightbox
Figure 1-33 After a phase-encoding gradient pulse has been applied, phase shifts along the vertical
direction occur and keep increasing. Left: after t seconds, right: after 2t seconds.
Subsequently, permanent magnets were proposed, which consist of large blocks made from
ferromagnetic alloys, e.g., in the form of a C-shaped magnet.33 These types of magnets have a
permanent magnetic field with field strength below 0.5 T.
The last category and the most widely used are the superconductive magnets with field strengths
between 0.5 T and 3 T (research systems may reach 7 T or more). This type of magnet arose from
the pioneering contribution to the theory of superconductors and superfluids of Abrikosov, Ginzburg
22
and Leggett (Fig. 1-34), for which they were awarded the Nobel Prize for Physics in 2003.
A superconducting magnet has a strong magnetic field generated by the electric current flowing in
large coils. In contrast to resistive magnets, superconducting magnets have no resistivity at very low
temperatures close to absolute zero. Therefore, a constant, high current will flow for years without an
electrical voltage.
The early superconducting magnets were very long (2.55 m) and heavy (≈ 8 tons for 1.5 T). However,
magnet design has made considerable progress over the years since passive iron shielding was
introduced (Fig. 1-35). Therefore, the magnet length has also been reduced to an overall length of
around 1.6 m for 1 T and 1.5 T.22
In summary, in the last two decades a gradual increase in static field strength for MRI has occurred,
ranging from 0.15 T to 1 T and later 1.5 T. The typical field strength for routine clinical imaging is now
1.5 T, whereas 3 T and more is commonly used for research activities.
Gradients
The gradients used for spatial encoding are one of the essential components in MRI as they determine
the resolution of the image reconstruction. The spatial resolution is inversely proportional to the
gradient time integral over the readout period, TRO. The hatched areas shown in Figure 1-36 represent
this time integral.
page 16
page 17
4 of 11 11-04-2010 23:32
Add to lightbox
Add to lightbox
Add to lightbox
Figure 1-34 A Abrikosov (A), V Ginzburg (B) and A Leggett (C) shared the Nobel Prize for Physics in
2003 for their pioneering contribution to the theory of superconductors and superfluids which are used,
for example, in MRI. (Copyright Bruno Press.)
5 of 11 11-04-2010 23:32
Add to lightbox
Figure 1-35 Magnet with passive shielding allowing smaller and lighter magnets. (Courtesy of
Siemens Medical Solutions, Erlangen.)
Add to lightbox
Figure 1-36 Gradient time integral (hatched area). To attain the higher level of spatial resolution,
gradient amplitude GR (mT/m) must increase and simultaneously the rise time Trise (ms) must
decrease.
Table 1-1. Development of MR Gradient Amplitude and Rise Time

Period GR (mT/m) Trise (ms)
1983 - 85 3-6 1.5 - 1.6
1986 - 89 10 1
1991 - 93 25 0.6
1999 - now 40 0.1 - 0.2
Moreover, as the gradients cannot turn on and off in zero time, i.e., it takes time to ramp the gradients
up to the desired amplitude, the rise time, Trise, has become an issue. When the measurement time for
a particular MR image is to be shortened without sacrificing resolution, the gradient amplitude, GR,
must increase and the rise time must decrease as again demonstrated in Figure 1-36.
For this reason, much research has been carried out to develop and implement higher gradient
amplitudes and shorter rise times. One major problem is that large numbers of turns in the gradient
6 of 11 11-04-2010 23:32
coils tend to cause high inductance per unit gradient strength, resulting in difficulty in ramping quickly.
The development of gradient amplitudes and the rise times commercially available are summarized in
Table 1-1.22
As development continues, a gradient amplitude of 80 mT/m and a rise time of 0.1 ms have been
prototyped and tested, particularly for head and small animal studies. 22 However, the price of these
systems is still too high to allow them to become widely available commercially.
Sequences
As mentioned earlier, in order to form an image, a number of MR signals must be acquired. This signal
acquisition can be programmed using pulse sequences, which is often considered as the heart of a MR
measurement.
page 17
page 18
Add to lightbox
Add to lightbox
Figure 1-37 EPI k-space trajectory maps using the single-shot method (A) and the blipped phase-
encoding (B).
7 of 11 11-04-2010 23:32
Add to lightbox
Figure 1-38 EPI pulse sequence used to create the k-space trajectory of Fig. 1-37B. Excitation is
limited to a single slice by transmitting the RF pulse in the presence of GS. The brief pulses or "blips"
of GP cause the trajectory to move up one line at a time in the k-space.
Using the concept of k-space, Mansfield36 introduced the first rapid imaging technique (EPI), which
uses a series of gradient echoes to traverse the whole of k-space in a rectangular raster. This
technique is achieved by using single-shot methods that generate trains of echoes after a single RF
pulse. Each echo is then encoded with different phase-encoding information. A k-space trajectory map
of the single-shot method is depicted in Figure 1-37A.
Another variant of the single-shot sequence is blipped phase-encoding where small-amplitude steps
are applied between readout gradient reversals. A sketch of the k-space trajectory of this sequence is
shown in Figure 1-37B.
Figure 1-38 shows a straightforward version, using only a single excitation pulse to produce a blipped
k-space trajectory shown in Figure 1-37B. The RF pulse is made slice selective by simultaneously
turning on a gradient along the slice selection axis.
Note that the positive/negative alternation of the readout gradient gives the alternating positive and
negative velocities in the readout direction, while the brief pulses or "blips" of the phase-encoding
gradient move the data from line to line along that axis.
Table 1-2. Echo-Planar Imaging Terminology

Term Characteristics
Single shot one excitation pulse
Multishot more than one excitation pulse
Blipped phase- small amplitude steps supplied between readout gradient reversals
encoding
Spiral phase-encoding phase-encoding gradient has alternately positive and negative
polarity
Note: terminology is the same for all the manufacturers.
The multishot EPI sequence acquires more points at the same time in the k-space and is therefore
8 of 11 11-04-2010 23:32
faster. This sequence uses more than one excitation pulse to acquire echoes.
When EPI was first developed, it was thought that it would have its greatest impact in providing
real-time MR images. In practice, its greatest application appears to be in the area of functional MR
imaging of the brain. However, because of the gradient-echo nature of the detection technique, all EPI
techniques are sensitive to T2* effects, such as field inhomogeneity.
Table 1-2 summarizes the various EPI techniques that have been implemented on commercially
available systems by various manufacturers.
Spin-Echo
Nearly as early as Mansfield's description of EPI, the use of (multiple) 180° refocusing RF pulses
following an excitation pulse to generate (multiple) spin-echoes was proposed as shown in Figure
1-40.38
This concept was then implemented at the Delft University of Technology where multiple echo single
shot (MESS) and multiple echo multiple shot (MEMS) were developed.40
Hennig et al39 (Fig. 1-39), from the University of Freiburg, modified and improved the sequences and
designated them as RARE (rapid acquisition with relaxation enhancement). This technique is better
known under the commercial names of fast or turbo spin-echo (TSE), fast spin-echo (FSE) and Half
Fourier acquisition turbo spin-echo (HASTE) (Table 1-3).
The main difference from the multi-echo SE sequence is that the multiple echoes per excitation are
separately phase-encoded, so that data are collected faster (see Chapters 3, 5 and 7). This is
achieved by applying a number of consecutive 180° refocusing pulses per excitation, in order to create
shots or segments with a corresponding number of differently phase-encoded echoes.
Additionally, in terms of scan time, the TSE method reduces the scan time for T2-weighted imaging by
an order of magnitude, allowing measurements of eight slices with a 256 × 256 matrix in about one
minute.40
Gradient-Echo Sequences
page 18
page 19
Table 1-3. Spin-Echo Sequences

Single-echo Multi-echo Hybrid Gradient
Manufacturer spin-echo spin-echo Echo train spin-echo echo - Spin-echo
Siemens MS Single Spin-echo Turbo spin-echo Turbo Gradient
spin-echo (TSE) spin-echo (TGSE)
Half Fourier
acquisition turbo
spin-echo (HASTE)
Philips MS Spin-echo Multi spin-echo Turbo spin-echo Gradient
(MSE) (TSE) spin-echo
(GRASE)
Ultrafast spin-echo
(UFSE)
GE MS Spin-echo Multi-echo Fast spin-echo (FSE) GRASE
multiplanar
(MEMP)
9 of 11 11-04-2010 23:32
Single shot FSE

(SS-FSE)
Table 1-4. Gradient-Echo Sequences

Manufacturer Spoiled Post-excitation Pre-excitation Magnetization
Siemens MS Fast low angle Fast imaging with Reversed FISP TurboFLASH
shot (FLASH) steady-state (PSFIF)
precision (FISP)
Philips MS T1 contrast- Fast field echo T2 contrast- Turbo field echo
enhanced fast (FFE) enhanced (T2 (TFE)
field echo (T1 CE-FFE)
CE-FFE)
GE MS Spoiled GRASS Gradient acquisition Steady state free IR-prepared
(SPGR) in the steady state precession fast (GRASS)
Fast spoiled (GRASS) (SSFP)
GRASS
(FSPGR)
Multiplanar
spoiled GRASS
(MSPGR)
Fast multiplanar
spoiled GRASS
(FMSPGR)
Add to lightbox
Figure 1-39 J Hennig, who designated the improved spin-echo sequence as RARE. (Courtesy of J
Hennig.)
10 of 11 11-04-2010 23:32
At about the same time, FLASH (fast low angle shot) appeared, creating a new class of pulse
sequences, i.e., gradient-echo sequences. This sequence was developed at the Max Planck Institute
by Haase, Frahm (Fig. 1-41) and their co-workers.41
The gradient-echo sequences collect only one gradient echo per RF excitation. 42,43 Moreover, as
opposed to spin-echo sequences, gradient-echo sequences do not use a 180° refocusing pulse; the
echo signal is, however, formed by applying gradient pulses of opposite polarity in the readout
direction (Fig. 1-42).
Add to lightbox
Figure 1-40 Multi-echo spin-echo pulse sequence timing diagram, two echoes illustrated. The
repetition time (TR) and the echo time (TE) are the two variables of interest.
There are a number of both positive and negative effects caused by the absence of the 180° pulse.
For instance, it reduces the RF power deposited in the patient and therefore lessens tissue heating.
The absence of the 180° refocusing pulse, in contrast, makes the relative contributions from fat and
water in gradient-echo sequences dependent on TE.
Table 1-4 summarizes several common gradient-echo sequences, including the terminology currently
used.
© 2010 Elsevier
11 of 11 11-04-2010 23:32
CONCLUSION
page 19
page 20
Add to lightbox
Add to lightbox
Figure 1-41 A Haase (A, Courtesy of Bayerische Julius-Maximilians-Universität Würzburg, Fakultät
für Physik und Astronomie, Würzburg, Germany) and J Frahm (B, Courtesy of Biomedizinische
NMR Forschungs GmbH, Max-Planck-Institut für biophysikalische Chemie, Göttingen, Germany,
44
Copyright 2003) devised and developed gradient-echo sequences, FLASH.
1 of 5 11-04-2010 23:33
Add to lightbox
Figure 1-42 Gradient-echo pulse sequence timing diagram. This class of sequences is characterized
by the absence of a 180° refocusing pulse. Echo is formed, however, by applying gradient pulses of
opposite polarity in the readout direction.
In this chapter we have tried to put together the fundamental research, evolution and latest
developments of various aspects of MR. It is impossible to cover all aspects in detail, so there are
many aspects that we have not even mentioned.
The timeline of the most important milestones in MR as a growing science is given in Box 1-1, a list of
Nobel Prizes related to MR in Box 1-2, and of Nobel Prizes in other fields for individuals also active in
MR in Box 1-3.
The first commercial MR scanners appeared about 20 years or so after the work of Bloch and Purcell.
A research group at Aberdeen started to build a MR prototype in 1972 whereafter several other
manufacturers followed.
Box 1-1 Historical overview: the development of MR. As a

comparison, the invention of CT is included
1800s Fourier transform - Fourier
1920s Basics of MR established by various scientists
1946 MR phenomenon - Bloch and Purcell
1952 Nobel Prize - Bloch and Purcell
1973 CT - Hounsfield
1973 First MR images on samples - Lauterbur
1975 Ernst's introduction of FT methods to MR
1977 First clinical MR images - Mansfield
2003 Nobel Prize - Lauterbur and Mansfield
Furthermore, the FT revolution has certainly opened up the bridge between a k-space and the image
reconstruction, thus making MR a far more dynamic and innovative technique. For this reason, we have
also presented a basic understanding of the FT without any mathematical formulas.
The development of MR pulse sequence design continues at a fast pace and the clinical demand for
MR imaging continues to grow throughout the world. Scan times were very long in the early 1980s
2 of 5 11-04-2010 23:33
when a T2-weighted scan using a spin-echo took at least 15 minutes. Today, single-shot scan times
can be as fast as 50 ms for a single image using EPI.
In conclusion, MR continues to thrive on innovation and to broaden its scope, apparently without limit.
The foundation laid down by the contributions of the scientists mentioned above, however, still remains.
Acknowledgments
We are grateful to several people for supplying images and making this chapter more readable,
including N van Kampen from Utrecht University, H Diebels from Philips Medical Systems, P Kreisler
from Siemens Medical Systems and M Greuter from Groningen University Hospital.
page 20
page 21
Box 1-2 Nobel Prizes Directly Related to MR. (From ref. 44

with permission)
Name Year Category Description
Norman F 1989 Physics "For the invention of the separated oscillatory fields
Ramsey method and its use in the hydrogen maser and other
atomic clocks"
Hans G Dehmelt 1989 Physics "For the development of the ion trap technique"
K Alexander 1987 Physics "For their important breakthrough in the discovery of
Müller superconductivity in ceramic materials"
Nicolaas 1981 Physics "For their contribution to the development of laser
Bloembergen spectroscopy"
John H Van 1977 Physics "For their fundamental theoretical investigations of the
Vleck electronic structure of magnetic and disordered systems"
Alfred Kastler 1966 Physics "Optical methods for studying Hertzian resonances"
Box 1-3 Nobel Prizes in Other Fields, Awarded to Individuals

Who Also Contributed to the Development of MR. (From ref.
44 with permission)
Name Year Category Description
Paul C 2003 Medicine "For their discoveries concerning magnetic resonance
Lauterbur imaging" Glasgow 2001
Sir Peter 2003 Medicine
Mansfield
Kurt 2002 Chemistry "For his development of nuclear magnetic resonance
Wüthrich spectroscopy for determining the three-dimensional
structure of biological macro-molecules in solution"
Richard R 1991 Chemistry "For his contributions to the development of the
Ernst methodology of high resolution nuclear magnetic resonance
(NMR) spectroscopy"
Felix Bloch 1952 Physics "For their development of new methods for nuclear
magnetic precision measurements and discoveries in
connection therewith"
Edward Mills 1952 Physics
Purcell
3 of 5 11-04-2010 23:33
Isidor Isaac 1944 Physics "For his resonance method for recording the magnetic
Rabi properties of atomic nuclei"
REFERENCES
1. Clayton E: The Internet Encyclopedia of Philosophy. Available at http://www.utm.edu/research/iep/c/cicero.htm
2. Ramsey NF: Early history of magnetic resonance. Phys Perspect 1:123-135, 1999.
3. Partain CL, James AE, Rollo FD, Price RR (eds): Nuclear Magnetic Resonance (NMR) Imaging. Philadelphia: Saunders,
1983.
4. Grant DM, Harris RK (eds): Encyclopedia of Nuclear Magnetic Resonance: Historical Perspectives, Vol. 1. Chichester, UK:
Wiley, 1996.
5. Vaughan B: A short illustrated history of magnetic resonance. Australas Radiol 33(4):390-398, 1989.
6. Herivel J: Joseph Fourier. The Man and the Physicist. Oxford: Clarendon Press, 1975.
7. Gorter CJ, Broer LJF: Negative result of an attempt to observe nuclear magnetic resonance in solids. Physica 9:591-596,
1942.
8. Rabi II, Zacharias JR, Millman S, Kusch P: A new method of measuring nuclear magnetic moments. Phys Rev 53:318, 1938.
9. Gorter CJ: Bad luck in attempts to make scientific discoveries. Phys Today 20:76, 1967.
10. Gorter CJ: Paramagnetic Resonance. Amsterdam: Elsevier, 1947.
11. The MacTutor History of Mathematics archive. Available at http://www-groups.dcs.st-and.ac.uk/~history/
12. Nobel Prize website. Available at http://www.nobel.se/
13. Zavoisky EK: Discovery of electron spin resonance. J Phys USSR 9:211, 245, 1945.
14. Bloch F: Nuclear induction. Phys Rev 70:460-473, 1946.
15. Purcell EM, Torrey HC, Pound RV: Resonance absorption by nuclear magnetic moments in a solid. Phys Rev 69:37-38,
1946.
16. Mallard JR: NMR Imaging at Aberdeen: A Historical Perspective. Philadelphia: Saunders, 1983, pp. 15-25.
17. Mallard JR, Hutchison JMS, Edelstein WA, et al: In-vivo NMR imaging in medicine: the Aberdeen approach, both physical
and biological. Phil Trans R Soc Lond B 289:519, 1980.
18. Damadian R: Tumour detection by nuclear magnetic resonance. Science 171:1151, 1971. Medline Similar articles
19. Damadian R: Apparatus and method for detecting cancer in tissue. US Patent 3,789,832, 1972.
20. Damadian R: Nuclear magnetic resonance: a noninvasive approach to cancer. Hosp Pract 12:63-70, 1977. Medline
Similar articles
21. Luiten AL, Locher PR, van Uijen C, et al: NMR Clinical Results: Eindhoven. Philadelphia: Saunders, 1983, pp. 250-260.
22. Kreisler P: 20 Years of Development and a Constantly Improving Performance. Magnetom. Magnetom Flash, Siemens
Medical 1:124-127, 2003.
23. Bottomley PA, Hart HR, Edelstein WA, et al: NMR imaging/spectroscopy system to study both anatomy and metabolism.
Lancet 2(8344):273-274, 1983.
24. Papoulis A: Signal Analysis. New York: McGraw-Hill, 1977.
25. Bracewell RN: The Fourier Transform and Its Application. New York: McGraw-Hill, 1986.
26. Cooley JW, Lewis PAW, Welch PD: Historical notes on the fast Fourier Transform. IEEE Trans Audio Electroacoustics
15:76-79, 1967.
27. Hounsfield GN: Computerized transverse axial scanning (tomography). Part I, Description of system. Br J Radiol
46:1016-1022, 1973. Medline Similar articles
28. Lauterbur PC: Image formation by induced local interactions: examples employing nuclear magnetic resonance. Nature
242:190-191, 1973.
29. Nishimura DG: Principles of Magnetic Resonance Imaging. Course notes, Stanford University, 1995.
30. Liang Z-P, Lauterbur PC: Principle of Magnetic Resonance Imaging: A signal processing perspective. New York: IEEE
Press Series, 2000.
31. Likes RS: Moving gradient zeugmatography. US patent 4,307,343, August 1979.
32. Tweig DB: The k-trajectory formulation of the NMR imaging process with applications in analysis and synthesis of imaging
methods. Med Phys 10:610-621, 1983. Medline Similar articles
33. Redpath T: MRI developments in perspective. Br J Radiol 70:S70-S79, 1997.
34. Kumar A, Welti D, Ernst RR: NMR Fourier zeugmatography. J Magn Reson 18:69-83, 1975.
35. Mansfield P, Maudsley AA: Medical imaging by NMR. Br J Radiol 50:188-194, 1977. Medline Similar articles
4 of 5 11-04-2010 23:33
36. Mansfield P: Multi-planar image formation using NMR. J Phys C Solid State Phys 10:L55-L58, 1977.
page 21
page 22
37. Hoult DI, Chen D-N, Sank VJ: The field dependence of NMR imaging II. Arguments concerning optimal field strength. Magn
Reson Med 3:730-746, 1986. Medline Similar articles
38. Mansfield P, Picket IL: Biological and medical imaging by NMR. J Magn Reson 29:355-373, 1978.
39. Hennig J, Nauerth A, Friedburg H: RARE imaging: a fast imaging method for clinical MR. Magn Reson Med 3:823-833,
1986. Medline Similar articles
40. van Vaals JJ, Groen JP, van Yperen GH: Recent progress in fast MR imaging. Medica Mundi 2:152-167, 1991.
41. Haase J, Friedburg H, Stroebel B: Rapid non-tomographic approach to MR myelography without contrast agents. J CAT
10:375-378, 1986.
42. Elster AD: Gradient-echo MR imaging: techniques and acronyms. Radiology 186:1-8, 1993. Medline
Similar articles
43. Haacke EM, Frahm J: A guide to understanding key aspects of fast gradient-echo imaging. J Magn Reson Imag
1:621-624, 1991.
44. Boesch C. Nobel Prizes for nuclear magnetic resonance: 2003 and historical perspectives. J MRI 19:517-519, 2004.
© 2010 Elsevier
5 of 5 11-04-2010 23:33
Printed from: Clinical Magnetic Resonance Imaging, 3rd editio...http://www.clinicalmri.com/content/printpage.cfm?ID=C00203068
ASIC RINCIPLES
John P. Mugler III
Our goal in this chapter is to provide a basic understanding of how magnetic resonance (MR) images
are generated and to lay the foundation for the discussion of advanced magnetic resonance imaging
(MRI) techniques in subsequent chapters. The chapter is divided into four main sections that address
the following questions:
How is the nuclear magnetic resonance signal generated?

What are the important characteristics of the signal behavior?
How can we determine the origin of signals from within the body to form an image?
How can the signal be manipulated to create different types of image contrast?
In-depth discussions of the salient physical principles are presented through detailed descriptions of
the concepts and liberal use of diagrams, avoiding extensive mathematical formulations. In some
sections, additional technical or mathematical detail is provided for the interested reader. To facilitate
recognition of this advanced material, the text passages are clearly labeled and set in smaller type;
these passages can be skipped without loss of continuity.
In certain discussions, particularly early in the chapter, reference is made to classical versus quantum
physics. The classical viewpoint (that is, not involving the fundamental quantization of various physical
quantities), although not correct for rigorously describing the behavior of subatomic particles such as
protons and neutrons, is nonetheless useful to provide an intuitive grasp of many concepts. However,
the quantum viewpoint will be necessary in some instances.
© 2010 Elsevier
1 of 1 11-04-2010 23:35
ORIGIN OF THE NUCLEAR MAGNETIC RESONANCE SIGNAL

Nuclear "Magnets"
The starting point for our discussion is the nucleus of an atom. The nucleus is composed of some
number of protons and neutrons, each of which intrinsically possesses angular momentum. Recall that
angular momentum, commonly symbolized by a vector directed along the axis of rotation, is a quantity
that represents the intensity of rotational motion. In classical physics angular momentum is equal to the
product of the angular velocity of a rotating body, for example a toy spin top, and its moment of inertia
with respect to the axis of rotation. The intrinsic angular momentum of a proton or neutron is called
spin.
page 23
page 24
Add to lightbox
Figure 2-1 A, Conceptual representation of a proton "spin" and its magnetic moment vector. B, In the
absence of an applied magnetic field, nuclear spins move in response to thermal energy exchanges
and are randomly oriented. Thus, the sum of their magnetic moments, also known as the net
magnetization M, is zero.
Protons and neutrons also possess a magnetic dipole moment that is proportional to their intrinsic spin.
From a classical-physics perspective, a proton or neutron can be regarded as a particle that contains
charge and possesses angular momentum, therefore giving rise to a magnetic moment. (A neutron has
no net charge, but is composed of charged particles. Although the topic of our discussion is nuclear
magnetic properties, note that electrons also possess intrinsic spin and thus a magnetic moment.)
Qualitatively, the magnetic properties of a proton or neutron can be thought of as arising from a tiny
bar magnet. Because spin gives rise to the magnetic properties of nuclei, which, as we will see later in
the chapter, in turn yield the nuclear magnetic resonance signal, nuclei with magnetic moments are
often loosely referred to as simply "spins" in the MRI literature. Figure 2-1A shows a conceptual
representation of a proton inspired by the classical viewpoint; the arrow through the sphere denotes its
intrinsic magnetic moment and the curved arrow indicating rotation about the axis of the moment
denotes its intrinsic spin. An important physical quantity in MRI is the ratio of the magnetic moment to
the spin, which is called the gyromagnetic ratio, conventionally denoted by the symbol γ.
In a nucleus, the magnetic moments of proton-proton or neutron-neutron pairs tend to align in

opposition such that the net magnetic moment of the pair is zero. Hence, nuclei with an even number of
12 16
protons and neutrons such as C or O possess no net magnetic moment. On the other hand, when
a nucleus has an odd number of protons or an odd number of neutrons, there is an unpaired proton or
neutron and the nucleus has a net magnetic dipole moment. Note that the net magnetic moment may
come from a proton, as for 1H, or from a neutron, as for 3He. Table 2-1 lists the nuclei that are
1
primarily of interest for medical imaging along with their characteristic gyromagnetic ratios.
Table 2-1. Gyromagnetic Ratio and Spin for Nuclei of Interest in Medical
1 of 11 11-04-2010 23:37
Imaging
Isotope Gyromagnetic Ratio (Mhz/T) Spin
1 42.58 1/2
H
3 32.44 1/2
He
13 10.71 1/2
C
17 5.77 5/2
O
19 40.08 1/2
F
23 11.27 3/2
Na
31 17.25 1/2
P
129 11.86 1/2
Xe
So far we have discussed that the nuclei of certain atoms, such as those of hydrogen atoms in a water
molecule, possess a magnetic moment and angular momentum, or from the classical-physics viewpoint
might each be regarded as a tiny ball of electric charge that is spinning about its axis, which gives rise
to magnetic properties. Because of the high concentration of hydrogen nuclei in the body in water and
fat molecules, and their relatively high gyromagnetic ratio, these nuclei are the source of signal for
nearly all clinical MRI exams. Therefore, the remainder of our discussion focuses on the behavior of
hydrogen nuclei, although most of our comments apply equally to the other nuclei listed in Table 2-1.
Since hydrogen nuclei are protons, descriptions of MRI often use the term "proton" interchangeably
with the term "hydrogen nuclei." Next, we explore how proton spins react when placed in an externally
applied magnetic field, beginning with the classical-physics perspective.
Behavior of Nuclear Spins in a Magnetic Field

In the absence of an externally applied magnetic field, the proton (hydrogen nuclei) spins in a
substance, such as the tissues of your body, move in response to thermal energy exchanges and are
oriented in random directions (Fig. 2-1B). The vector sum of the magnetic moments in a volume of
interest, for example the sum of the magnetic moments shown diagrammatically in Figure 2-1B, is
called the net macroscopic magnetization, conventionally denoted by the symbol M. This quantity,
which is among the most important in the theory of MRI, is often referred to as the net magnetization,
or simply the magnetization. For the randomly oriented spins in Figure 2-1B, M equals zero.
Now let us consider what happens if a static (that is, constant in time) external magnetic field, denoted
by the symbol B0, is applied to one of our proton spins. The external magnetic field exerts a torque on
the magnetic moment, but instead of aligning it with the applied magnetic field, this torque causes the
spin (magnetic moment) to move at a right angle to the plane that, at any instant, contains the
magnetic-field and angular-momentum vectors, as shown in Figure 2-2. (Note in Fig. 2-2 that we
introduced a standard Cartesian [rectangular] x-y-z coordinate system. By convention, the applied
magnetic field B0 is directed along the positive z-axis. B0 is the symbol typically used to denote the
static magnetic field produced by the main magnet of an MRI scanner.) This motion of the spin about
the axis of an externally applied magnetic field is called precession, and its frequency ω0 is given by
the well-known Larmor equation:
The frequency of precession, also known as the Larmor frequency, is directly proportional to the
gyromagnetic ratio for the nucleus (Table 2-1) and the applied field strength.
page 24
page 25
2 of 11 11-04-2010 23:37
Add to lightbox
Figure 2-2 Precession of a nuclear magnetic moment μ about an externally applied magnetic field B0.
The angular frequency of precession is ω 0. The dashed oval shows the path that is traced out by the
tip of the magnetic-moment vector as it precesses about the z-axis.
Precession occurs due to the combination of the magnetic forces and the angular momentum of the
proton. This behavior may not be intuitively obvious, but it is identical to that which occurs for a toy spin
top or toy gyroscope under the influence of gravity. Anyone who has played with one of these will
remember that such a spinning toy neither remains standing straight up, nor does it fall over; it instead
wobbles around, or in other words precesses, as it spins. In this case, gravity is analogous to the
applied magnetic field and the top possesses angular momentum because it is spinning.
For the interested reader: The vector form of Equation 2-1, ω0 = -γB0, shows that the
angular velocity vector is antiparallel to the applied magnetic-field vector, which explains
the sense of the rotation as shown in Figure 2-2. The equation is written in terms of the
magnetic-field induction B instead of the magnetic-field strength H because we are
interested in the magnetic fields that exist (or, in other words, are induced) within
substances such as biological tissue. In the MRI literature, the magnetic-field induction is
often loosely referred to as the magnetic-field strength, although this is not strictly
correct. Nonetheless, we will follow this widespread nomenclature.
With regard to the classical-physics perspective portrayed in Figure 2-2, it is

inappropriate from the quantum viewpoint to describe the behavior of a single nuclear
spin in the manner shown. However, the time-dependent behavior of the expectation
value for the magnetic moment of a single spin, or of the expectation value for the
average moment resulting from an ensemble of noninteracting spins, can be described by
the classical law of motion for a nuclear magnetic dipole in an external magnetic field.2
Thus, in the interest of understanding the main concepts involved, it is reasonable to use
the classical-physics viewpoint.
3 of 11 11-04-2010 23:37
Add to lightbox
Figure 2-3 Measurable energy states of hydrogen nuclei (protons) in an externally applied magnetic
field B0. The high- and low-energy states are commonly referred to as "spin down" and "spin up,"
respectively. [hstrok] denotes Planck's constant divided by 2π.
The next step is to explore the behavior for a collection, or population, of nuclear spins. For this task
we must draw upon quantum physics, which tells us that a nuclear spin in a uniform static magnetic
field B0 has a discrete number of measurable energy states, equal to twice the nuclear-spin value plus
one. The spin of a nucleus with an odd number of protons or neutrons depends on the detailed
structure of the nucleus. Most nuclei of medical interest have a spin of ½ (see Table 2-1), and thus the
number of measurable energy states in a uniform magnetic field is two. Figure 2-3 illustrates this
concept. As shown in the figure, the difference in energy between the two states is directly
proportional to the gyromagnetic ratio for the nucleus and the applied field strength; this is the same
dependence on these two quantities that we saw above for the frequency of precession.
Drawing on the concepts from classical physics, several terms are commonly used to refer to a spin
measured in one of the two possible energy states. The high-energy state is called "spin down" and
the spin is said to be aligned away from or "antiparallel" to the external field. Analogously, the
low-energy state is called "spin up" and the spin is said to be aligned toward or "parallel" to the
external field. Connecting the energy-state and precession concepts, there are two energy states
because the angular momentum is quantized, that is, it can only have one of two values. This is what
the spin number tells us-it indicates the number of allowable angular momentum values. This is in
contrast to a toy top, which for practical purposes has a continuously variable angular momentum.
At typical applied field strengths, for example 1.5 tesla, and body temperature, the energy separation
∆E between spin states is relatively small compared to the level of thermal energy in the system. If we
were to compare a collection of spins at two points in time, we would find that a number of the spins
that were initially in the high-energy state had made transitions to the low-energy state by exchanging
energy with their surroundings, and vice versa. Nonetheless, at any instant in time, the mean number of
spins in each energy state remains constant, with slightly more spins in the low-energy state. Under
these conditions, the spin system is said to be in thermal equilibrium.
page 25
page 26
4 of 11 11-04-2010 23:37
Add to lightbox
Figure 2-4 When your body is placed in the magnetic field B0 created by the main magnet of an MRI
scanner, a net thermal-equilibrium magnetization M0 is induced within the tissues. A, B0 is parallel to
the long axis of the body as occurs in a standard cylinder-shaped high-field MRI magnet. B, B0 is
perpendicular to the long axis of the body as occurs in a C-shaped low-field MRI magnet. In either
case, the magnetization is aligned with the applied field and is directly proportional to its magnitude.
For example, the smaller B0 in B compared to A results in a smaller thermal-equilibrium
magnetization. (The actual magnitude of M0 compared to B0 is much smaller than depicted in this
diagram.)
For a large population of spins, the net excess in the low-energy state can be calculated by using the
Boltzmann distribution. We find that this net excess is only a few spins out of each million for the field
strengths typically used in MRI. Recalling that the net magnetization is the vector sum of the individual
magnetic moments, the excess of spins in the parallel state results in a net magnetization, M0, aligned
with the external field B0 (Fig. 2-4). M0 is called the thermal-equilibrium magnetization; it is directly
proportional to B0 and inversely proportional to temperature. As discussed later in the chapter, the
signal that is measured in MRI is directly proportional to M0. The fact that M0 increases in direct
proportion to B0 is one of the primary motivations for developing MRI systems with increasingly higher
field strengths.
For the interested reader: At thermal equilibrium, the ratio of the mean number of spins
in the spin-up state (Nup) to that in the spin-down state (Ndown) is given by the Boltzmann
distribution as:
where [hstrok] is Planck's constant divided by 2π, k is Boltzmann's constant, and T is the
absolute temperature. Using this result along with the quantum mechanical expectation
values for the magnetic moment of a proton spin in the spin-up or spin-down state in a
static magnetic field,2 we can derive the magnitude of the thermal-equilibrium
magnetization as:
for γ[hstrok]B0 « kT, where N is the total number of spins. Note that M0 increases with
the square of the gyromagnetic ratio. Another quantity of interest is the fractional excess
of spins, (Nup - Ndown)/N, which is called the nuclear polarization.
Now we know that a net magnetization aligned with the applied magnetic field is created in a subject's
body when it is placed in the magnet of an MRI scanner (see Fig. 2-4). With this and the concept of
precession in hand, our next step is to understand how we can manipulate the net magnetization to
move it away from alignment with the applied magnetic field. This is a necessary step in producing a
signal from the body that can be measured. Before proceeding, however, let us discuss briefly why the
phenomenon is called magnetic "resonance."
Magnetic Resonance
In an electrical or mechanical system (for example, the suspension system in a car), recall that
resonance is defined as a vibration of large amplitude caused by a relatively small periodic stimulus,
wherein this stimulus has approximately the same period as one of the so-called natural vibration
periods of the system. In other words, if the stimulus to an electrical or mechanical system is at the
"resonant" frequency (for instance, if you push down on the fender of a car at just the right frequency),
the transfer of energy between the stimulus (your hand) and the system (suspension of the car) is
enhanced. In analogy, nuclear magnetic resonance is the enhanced exchange of energy by the nuclei
of atoms within an externally applied magnetic field that occurs for a specific frequency of applied
radiant energy.
From our discussion in the previous section, we know that the energy difference corresponding to a
5 of 11 11-04-2010 23:37
transition between the low- and high-energy states (Fig. 2-3) is ∆E = γ[hstrok]B0. The frequency of a
photon (a quantum or packet of radiant energy) that possesses this energy is ∆E/[hstrok] = γB0. So,
photons with this frequency are absorbed by the spin system and stimulate transitions between the
two energy states; this is magnetic resonance. Note that γB0 is the Larmor frequency as defined in
Equation 2-1. Thus, the frequency of precession for a nuclear magnetic dipole moment in an applied
magnetic field, as viewed from the classical-physics perspective, is the same as the frequency of
radiation necessary to induce transitions of a nuclear magnetic moment between energy states in the
same applied field. In MRI, the required frequency corresponds to that for radiofrequency (RF)
electromagnetic radiation. The resonant frequencies for nuclei of medical interest can be calculated
from the gyromagnetic ratios listed in Table 2-1 by simply multiplying the values provided by the field
strength in tesla. For example, the resonant frequency for hydrogen nuclei (protons) at 1.5 tesla is 64
MHz. (In terms of precession, a resonant frequency of 64 MHz means that a magnetic moment would
rotate about the direction of the applied field 64 million times in one second.) Although the frequencies
of interest correspond to those for radio, the physics of MRI, at least for the field strengths in common
use today, does not involve the wave properties of electromagnetic radiation. In other words, MRI
does not use "radio waves," even though it has often been stated in the literature that transmission and
absorption of radio waves is the basis for MRI.
Up to this point, concepts from classical and quantum physics have been intermingled to form a
compact and easy to follow explanation of the principles at hand. Nonetheless, we caution the reader
that extrapolating this hybrid framework can lead to incorrect conclusions about more advanced
aspects of the behavior of spin systems. Fortunately, a classical-physics perspective is sufficient for
many of the remaining topics in this chapter.
page 26
page 27
Add to lightbox
Figure 2-5 A, The stationary (x, y, z) and rotating (xrot, yrot, zrot) frames of reference. The z and zrot
axes are coincident and the rotating frame of reference revolves about the z-axis with a constant
angular velocity ω 0. B, The time-varying magnetic field B1 that is produced when an RF pulse is
applied to a coil. B1 appears stationary when viewed from the rotating frame of reference. In this
example, the RF pulse is applied in a manner that causes the B1 field to lie along the xrot-axis.
Response of the Magnetization to a Radiofrequency Pulse

To help us understand the important principles of this section, we need to introduce a conceptual tool
called the rotating frame of reference. Figure 2-5A shows a standard Cartesian (rectangular) x-y-z
coordinate system, which we will call the stationary frame of reference, along with the rotating frame
of reference, which has axes labeled xrot, yrot, and zrot. The z and zrot axes are coincident; relative to
the stationary frame of reference, the rotating frame of reference revolves about the z-axis with a
constant angular velocity, typically equal to the Larmor frequency, ω0. Instead of xrot, yrot, and zrot, the
axes of the rotating frame are sometimes labeled x', y', and z'. The stationary frame of reference is
sometimes called the laboratory frame of reference.
A real-world situation analogous to our conceptual stationary and rotating frames of reference is a
merry-go-round or carousel. When standing on the ground, next to the merry-go-round, you are in the
6 of 11 11-04-2010 23:37
stationary frame of reference, whereas riders on the merry-go-round are in the rotating frame of
reference. From the perspective of one of the riders who is standing on the merry-go-round, another
standing rider is stationary, whereas from the perspective of a person standing on the ground, both of
these riders are rotating about the axis of the merry-go-round.
From this point forward, much of our discussion will pertain to events that are viewed in the rotating
frame of reference. The use of the rotating frame of reference in the explanation of MRI phenomena is
so pervasive that in much of the literature it is not explicitly stated that the rotating frame is being used.
The next concept we need to discuss is that of a radiofrequency, or RF, pulse. Consider a subject that
is placed within the magnet of an MRI scanner and surrounded by an RF coil. This RF coil may be as
simple as a loop of wire or as complicated as the sophisticated commercial designs with which you are
probably familiar from using an MRI scanner. If an alternating voltage is applied across the RF coil,
causing an alternating electric current to flow through the coil, a magnetic field will be produced within
the coil (namely, within the subject's body) that oscillates at the frequency of the applied voltage.
Specifically, if the voltage is applied at the Larmor frequency (which, as discussed above, is in the
range of frequencies associated with radio), it results in a magnetic field that, when viewed from our
rotating frame of reference, appears to be stationary (Fig. 2-5B). This magnetic field is typically
represented by a vector called B1. A radiofrequency voltage applied across the RF coil for a short
period of time, typically on the order of a millisecond, is called an RF pulse. As we will see later, it is
possible to vary the amplitude of the voltage over the duration of the RF pulse, thereby causing the
amplitude of the B1 vector to vary in a desired way during the RF pulse. Also, by manipulating the
characteristics of the applied voltage, it is possible to make the B1 field have any desired orientation
within the xrot-yrot plane (for example, the B1 field could be applied along the yrot-axis, instead of along
the xrot-axis as shown in Fig. 2-5B), or even change orientation during the RF pulse.
For the interested reader: A simple RF coil design produces what is called a linearly
polarized (LP) RF field, which is composed of two counter-rotating circularly polarized
(CP) components. One of these magnetic field components is the B1 field depicted in
Figure 2-5B. More sophisticated coil designs generate only one circularly polarized
component. This is desirable because the counter-rotating component deposits energy in
the tissue, but is not of use for manipulating the magnetization vector. The amount of
energy deposited is of particular concern for imaging at very high fields.
So how does an RF pulse affect the net magnetization that is generated in a subject's body when it is
placed in the magnet of an MRI scanner? Recall from our discussion above that a magnetic moment
precesses about an externally applied magnetic field. Just as a magnetic moment will precess about
the static field B0 that is produced by the main magnet of the MRI scanner (see Fig. 2-2), it will also
precess about the time-varying field B1 that is produced by applying an RF pulse. In the rotating frame
of reference, we do not "see" the precession about B0 because the frame of reference is rotating
about the axis of the applied static magnetic field at an angular frequency ω0, synchronous with the
precession. Thus, in the absence of B1, the net magnetization appears to be stationary in the rotating
frame. Once the RF pulse is turned on, the magnetization M begins to precess about B1. (Once the
magnetization is perturbed from alignment with the static magnetic field B0, it is called M instead of
M0.) The larger the magnitude of B1, the faster M precesses. By analogy to the Larmor equation (Eq.
2-1), the precessional frequency of M in the rotating frame is γB1. By varying the duration or strength
of the RF pulse (that is, by varying the time of application or the magnitude of B1), we can cause the
magnetization to precess a selected amount about the axis along which B1 is applied.
page 27
page 28
7 of 11 11-04-2010 23:37
Add to lightbox
Figure 2-6 The effect of an RF pulse on the magnetization. A, Viewed from the stationary frame of
reference, the magnetization M precesses about both the static magnetic field B0 and the time-varying
magnetic field B1. (B0 is directed along the z-axis; it is omitted to simplify the diagram.) The dotted
line shows the path that is traced out by the tip of the magnetization vector as it nutates toward the x-y
plane. B, Viewed from the rotating frame of reference, the magnetization M precesses only about B1.
The dashed line shows the path that is traced out by the tip of the magnetization vector as it rotates
toward the yrot-axis. The angle between the z-axis and the magnetization, measured at the end of the
RF pulse, is called the flip angle, α.
Figure 2-6 illustrates the effects of an RF pulse on the magnetization in both the stationary and rotating
frames of reference. In the stationary frame, M precesses about B0 and B1 simultaneously (Fig. 2-6A).
The resulting spiraling movement of M toward the x-y plane is called nutation. (For clarity, the
frequency of precession of M about B0 [γB0] is shown to be only a few times faster than that of M
about B1 [γB1]. In reality, γB0 is typically about a thousand times faster than γB1.) In the rotating
frame, only the precession about B1 is seen (Fig. 2-6B). The angle between the z-axis and M at the
end of the RF pulse is called the flip angle for the pulse. The symbol α is commonly used to denote the
flip angle.
Before proceeding further, it is useful to introduce some nomenclature associated with the effects of
RF pulses. Figure 2-7 shows a B1 field applied in the xrot-yrot plane (B1 is oriented about 30° toward
the negative yrot-axis) that causes the magnetization to precess through the flip angle α. The resulting
magnetization M can be represented by the two components Mz and Mxy. Mz is the projection of M
onto the z-axis, which is often called the longitudinal axis. Thus, Mz is called the longitudinal
component of the magnetization, or simply the longitudinal magnetization. Similarly, Mxy is the
projection of M onto the xrot-yrot plane, which is commonly called the transverse plane. Thus, Mxy is
called the transverse component of the magnetization or the transverse magnetization. Mxy can be
further decomposed into projections along the xrot and yrot axes, called Mx and My, respectively (Fig.
2-7).
Add to lightbox
8 of 11 11-04-2010 23:37
Figure 2-7 The magnetization M can be represented by a component parallel to the z-axis, called the
longitudinal magnetization, Mz, and a component that lies in the xrot-yrot plane, called the transverse
magnetization, Mxy. The transverse magnetization can be decomposed as shown into the
components Mx and My.
Add to lightbox
page 28
page 29
Figure 2-8 Common types of RF pulses used in MRI. A, A 90° excitation RF pulse converts
longitudinal magnetization into transverse magnetization. B, A 180° inversion RF pulse rotates the
longitudinal magnetization from the positive z-axis to the negative z-axis. C, A 180° refocusing RF
pulse flips transverse magnetization to the other side of the xrot-yrot plane. D, For this 90° excitation
RF pulse, the B1 field is applied along the yrot-axis, instead of along the xrot-axis as in A. That is, the
phase of this RF pulse is different than that for the pulse shown in A. The magnetization precesses to
the negative xrot-axis, instead of to the positive yrot-axis.
The most common RF pulses used in MRI are 90° and 180° pulses. A 90° RF pulse is typically used to
move the magnetization from alignment with the longitudinal axis into the transverse plane (Fig. 2-8A).
As will be discussed shortly, transverse magnetization is required to generate a magnetic resonance
signal. An RF pulse whose purpose is to convert longitudinal magnetization into transverse
magnetization is called an excitation RF pulse. In techniques designed for fast imaging, the flip angle of
an excitation RF pulse may be less than 90°. A 180° pulse is called an inversion RF pulse when it is
used to move the magnetization from the positive z-axis to the negative z-axis or, in other words, to
invert the longitudinal magnetization (Fig. 2-8B). In contrast to inversion, Figure 2-8C illustrates the use
of a 180° pulse to "flip" transverse magnetization about the axis along which the RF pulse is applied.
This form of 180° pulse is called a refocusing RF pulse. We will discuss the purpose of refocusing RF
pulses later in the chapter.
The reason for using the terms "excitation" and "inversion" to describe RF pulses can be appreciated
by referring back to Figure 2-3 and, for the moment, considering the quantum-physics perspective.
Following an excitation RF pulse, the longitudinal component of the magnetization is smaller than its
thermal-equilibrium value. Since the net magnetization is the sum of the individual magnetic moments,
the number of spins in the high-energy state must increase in order to achieve this reduction in the
9 of 11 11-04-2010 23:37
longitudinal component of the magnetization. The high-energy state is also called an "excited" state,
hence the origin of the term excitation. By the same reasoning, to invert the longitudinal magnetization,
an inversion RF pulse must reverse, or in other words cause an inversion of, the spin population.
Specifically, the number of spins that were in the low-energy state before the inversion pulse equals
the number that are in the high-energy state after the pulse, and vice versa.
As mentioned above, the B1 field may have any desired orientation within the xrot-yrot plane. The
orientation of the B1 field is called the phase of the RF pulse; the phase is often measured relative to
the xrot-axis. For example, the 90° RF pulse shown in Figure 2-8A has a phase of 0° whereas the one
shown in Figure 2-8D has a phase of 90°. Another common name for the RF pulse shown in Figure
2-8A is a pulse, denoting that it is applied along the xrot-axis. Similarly, the pulse in Figure 2-8D is
called a pulse.
We have seen how RF pulses permit us to manipulate the longitudinal magnetization that is generated
in a subject's body when it is placed in the magnet of an MRI scanner, and to thereby create
transverse magnetization. The final topic to be discussed in this section is how the magnetization can
be used to produce a measurable signal from the body.
Generation of the Magnetic Resonance Signal
Add to lightbox
Figure 2-9 A, Precessing transverse magnetization Mxy induces a radiofrequency voltage across a
nearby coil of wire. B, Conceptually, the precessing transverse magnetization produces the same
effect as a tiny spinning bar magnet.
In the frame of reference of the RF coil, that is, in the stationary frame, the transverse magnetization
created by an excitation RF pulse precesses about the axis of the applied static magnetic field at the
Larmor frequency. This precessing transverse magnetization creates a time-varying magnetic field. As
discovered by Michael Faraday in 1831, such a time-varying magnetic field induces a voltage across a
nearby coil of wire (Fig. 2-9A). This is Faraday's law of electromagnetic induction, which states that if
10 of 11 11-04-2010 23:37
a conducting wire loop or coil is in a magnetic field and the magnetic field changes relative to the coil,
then a voltage is induced across the coil. This voltage induces an electric current to flow through the
coil, if it is part of a closed circuit. This is the exact same principle that is used in an electric generator
to produce electricity. Conceptually, the precessing magnetization can be thought of as a tiny spinning
bar magnet (Fig. 2-9B). Note that transverse magnetization is required to generate the magnetic
resonance signal because it is the precession of this magnetization that results in the requisite
time-varying magnetic field. The coil, as shown in Figure 2-9, may be the same one that was used to
generate the RF pulse, or it may be a separate coil that is optimized for the region of anatomy being
imaged.
The signal that is measured just after an excitation RF pulse, and as a result of the precessing
transverse magnetization, is called a free induction decay, abbreviated FID. The term "free" refers to
the fact that after the B1 field is turned off at the end of the RF pulse, the magnetization precesses
freely in the applied static magnetic field. The term "induction" is used because the precessing
magnetization induces a voltage across the coil. The term "decay" is used because, as we will discuss
in the next section, the magnitude of the transverse magnetization, and thus the signal strength,
gradually decreases. The measured FID oscillates at the Larmor frequency, so the FID appears as a
sinusoidal waveform that slowly decays as illustrated in Figure 2-10A. Before processing the magnetic
resonance signal, for example to calculate an MR image, the oscillatory component of the signal at the
Larmor (resonant) frequency is removed by a procedure called demodulation. The resulting
demodulated FID exhibits the signal decay with time (Fig. 2-10B).
page 29
page 30
Add to lightbox
Figure 2-10 Original (A) and demodulated (B) free induction decay (FID) signals generated by the
precessing transverse magnetization (see Fig. 2-9). For clarity, the period of signal oscillation in A is
shown to be only a few times shorter than the decay time; in practice the period of oscillation is at least
one thousand times shorter.
We can now appreciate the fundamental symmetry involved in generating the magnetic resonance
signal. Summarizing what we have discussed, thermal-equilibrium magnetization is created from the
spins of hydrogen nuclei by placing a subject in the magnet of an MRI scanner, and a radiofrequency
voltage is applied across a coil surrounding the subject, thus resulting in a time-varying magnetic field
within the subject. This time-varying magnetic field perturbs the thermal-equilibrium magnetization,
thereby creating transverse magnetization, which precesses in the scanner's magnetic field and itself
generates a time-varying magnetic field. This second time-varying magnetic field then induces a
radiofrequency voltage across the coil surrounding the subject, and in doing so produces the magnetic
resonance signal that can be measured.
Additional information on the concepts discussed in this section can be found in references 2 to 5.
© 2010 Elsevier
11 of 11 11-04-2010 23:37
BASIC CHARACTERISTICS OF THE MAGNETIC RESONANCE SIGNAL

Longitudinal and Transverse Relaxation
Now that we know how an excitation RF pulse can create transverse magnetization, and how this
magnetization can generate a magnetic resonance signal, our next task is to discuss what happens to the
magnetization following the RF pulse. For example, after a 90° excitation RF pulse, the resulting transverse
magnetization precesses about the z-axis at the Larmor frequency. Can this go on indefinitely? The answer
of course is no, because to maintain the excited state would require a continual input of energy to the spin
system. If only the static magnetic field B0 is applied after the RF pulse, the original thermal equilibrium value
of the magnetization (that is, aligned parallel to the z-axis) must be eventually reestablished. Between the 90°
RF pulse and reestablishment of thermal equilibrium two things must occur: 1. the longitudinal component of
the magnetization Mz must grow back to its original value, and 2. the transverse component Mxy must die
out, or decay, to zero. The processes by which these events occur are termed longitudinal and transverse
relaxation.
First let us consider longitudinal relaxation as illustrated in Figure 2-11A. Immediately following a 90° RF
pulse, Mz is zero and the number of spins in the high-energy state equals that in the low-energy state (Fig.
2-11A, time = 0). As time passes, and the numbers of spins in the two energy states begin to approach the
values that correspond to thermal equilibrium, the longitudinal magnetization grows back toward M0. This
regrowth involves a loss of energy from the spin system. As discussed earlier in the chapter, the number of
spins in the high-energy state increases when the magnetization is disturbed from thermal equilibrium. As
longitudinal relaxation occurs, the number of spins in the high-energy state decreases and energy is
transferred from the spin system to its environment, which is referred to as the "lattice." Because it involves
energy transfer between the spin system and the lattice, longitudinal relaxation is also called spin-lattice
relaxation.
Longitudinal relaxation is characterized by an exponential regrowth of Mz with a time constant that is

traditionally called the T1 (or equivalently, T1) relaxation time. Hence, another common name for longitudinal
relaxation is T1 relaxation. The T1 time is an intrinsic property of any given tissue; different types of tissue
typically have different T1 values. For example, at 1.5 tesla, brain white matter has a T1 of roughly 600 ms
and brain gray matter has a T1 of roughly 1000 ms. The T1 value that a given tissue possesses depends on
the details of the tissue's molecular structure. (The factors involved in determining the T1 of a particular
tissue are beyond the scope of our discussion.) In addition, the T1 values for most tissues increase as the
6,7
field strength is increased (see Chapter 18, High-Field Imaging).
By plotting the length of the longitudinal magnetization vector (as shown in Fig. 2-11A) versus time, we can
appreciate the exponential time course of T1 relaxation as depicted in Figure 2-12A. When the time elapsed
-1
after the RF pulse equals T1, the ratio of Mz to its thermal equilibrium value is (1 - e ), which equals 0.63 or
-n
63%. Generalizing this statement, at n times T1 after the 90° RF pulse, the ratio equals (1 - e ). For
example, as illustrated in Figure 2-12A, at 2 T1 and 3 T1 after the 90° RF pulse, the longitudinal
magnetization has regrown to 86% and 95% of its thermal-equilibrium value, respectively.
page 30
page 31
1 of 14 11-04-2010 23:38
Add to lightbox
Figure 2-11 Schematic of the time course of the longitudinal and transverse components of the
magnetization as they relax toward their thermal equilibrium values following a 90° excitation RF pulse. A,
Behavior of the longitudinal magnetization (black vector aligned parallel to the z-axis). Longitudinal relaxation
involves the transfer of energy from the spin system to its surroundings. B, Behavior of the transverse
magnetization (black vector aligned parallel to the yrot-axis). Transverse relaxation involves the loss of phase
coherence among the spins.
Add to lightbox
2 of 14 11-04-2010 23:38
Add to lightbox
page 31
page 32
Figure 2-12 Plots of the time course of the longitudinal (A) and transverse (B) components of the
magnetization as they relax toward their thermal-equilibrium values following a 90° excitation RF pulse.
Add to lightbox
Figure 2-13 Time course of the longitudinal magnetization as it relaxes toward its thermal equilibrium value
following a 180° inversion RF pulse. At the "null" point for the tissue (arrows), the longitudinal magnetization
passes through zero.
When the flip angle of the RF pulse is greater than 90°, for example 180° as for an inversion RF pulse, the
initial longitudinal magnetization is aligned parallel to the negative z-axis (Fig. 2-8B). For this case, the
longitudinal magnetization exhibits an exponential regrowth toward its thermal-equilibrium value in the same
general manner as discussed above. However, in contrast to the behavior described for a 90° pulse, the
magnitude (length) of the longitudinal-magnetization vector passes through zero during the relaxation process
(Fig. 2-13). The regrowth of the longitudinal magnetization following an inversion RF pulse is called inversion
recovery. Since the rate of regrowth depends on the T1 for the specific tissue, the longitudinal magnetization
corresponding to different tissues passes through zero at different times as shown in Figure 2-13. As will be
discussed in later chapters, there are several important MR imaging techniques that take advantage of this
behavior to acquire an image at the time (sometimes called the tissue's "null" point) when the relaxing
longitudinal magnetization for a given tissue passes through zero. In such an image, the signal from the tissue
is suppressed. Specific examples include the "STIR" technique,8,9 for which the signal from fat is
10
suppressed, and the "FLAIR" technique, for which the signal from cerebrospinal fluid is suppressed.
For the interested reader: In the more general case of T1 relaxation following an RF pulse
having a flip angle other than 90°, the longitudinal magnetization that remains just after the
pulse is not zero, as indicated above for the case of a 180° RF pulse. The corresponding
regrowth of the longitudinal magnetization can be thought of as the sum of an exponential
regrowth from an Mz value of zero, just as if a 90° pulse had been applied, and an exponential
decay of the initial value of the longitudinal magnetization following the RF pulse. This is written:
3 of 14 11-04-2010 23:38
where t is time and Mz0 is the initial value of the longitudinal magnetization following the RF
pulse. The first term on the right-hand side is the exponential regrowth and the second term is
the exponential decay. For the case of T1 relaxation following an inversion RF pulse, Mz0
equals -M0 and the equation for relaxation becomes:
From Equation 2-5, the time for which Mz passes through zero is T1 times the natural logarithm
of 2 (ln 2 ≈ 0.69).
Add to lightbox
Figure 2-14 A, Magnetization vectors precessing with random phases. B, Magnetization vectors precessing
in synchrony. This second group of vectors exhibits phase coherence. (The dashed circles show the paths
that are traced out by the tips of the magnetization vectors as they precess.)
Next let us turn to transverse relaxation, as illustrated in Figure 2-11B. Immediately following a 90° RF pulse,
the magnitude of the transverse magnetization is equal to that of the longitudinal magnetization just before
the pulse (Fig. 2-11B, time = 0). This transverse magnetization, which, as we already know, precesses at
the Larmor frequency, exists because there is phase coherence among the magnetic moments associated
with the protons (hydrogen nuclei). The idea of phase coherence is central to understanding several concepts
in the remainder of the chapter, so we will pause for a moment to discuss it further.
In our context, phase coherence refers to the relationship among precessing transverse-magnetization
vectors. Recall that the phase of a vector is its orientation in space relative to a reference axis. The term
"coherence" means that the phases of the magnetization vectors under consideration have a definite, as
opposed to random, relationship with each other. In its simplest form, this relationship is that the phases of
all magnetization vectors are identical, and that the vectors are precessing in synchrony. As an example,
consider the precessing magnetization vectors illustrated in Figure 2-14. In Figure 2-14A, there is no definite
relationship among the phases of the vectors or, in other words, their phases are random. In contrast, in
Figure 2-14B, the vectors are shown precessing in synchrony. This group of magnetization vectors exhibits
phase coherence. Vectors precessing in synchrony are said to be "in phase." If some process disrupts the
phase relationship among precessing vectors that were originally in phase, the vectors are said to have been
dephased.
Returning to our description of transverse relaxation (see Fig. 2-11B), we see that the transverse
magnetization decays toward its thermal equilibrium value of zero as time passes following the 90° RF pulse.
This decay process involves an irreversible loss of phase coherence (that is, an increase in randomness of
the phases) among the magnetic moments. In other words, the spin system has undergone irreversible
dephasing. (Later in the chapter we will see that under certain conditions dephasing can also be reversible.)
Magnetic interactions between the spins are a mechanism for this irreversible loss of phase coherence, and
thus transverse relaxation is also called spin-spin relaxation.
page 32
page 33
Transverse relaxation is characterized by an exponential decay of Mxy with a time constant that is
traditionally called the T2 (or equivalently, T2) relaxation time. Hence, another common name for transverse
relaxation is T2 relaxation. Analogous to the situation for the T1 time, the T2 time is an intrinsic property of
any given tissue and different types of tissue typically have different T2 values. For example, at 1.5 tesla,
4 of 14 11-04-2010 23:38
brain gray matter has a T2 of roughly 100 ms and cerebrospinal fluid has a T2 of roughly 2000 ms. Also
analogous to T1, the T2 value that a given tissue possesses depends on the details of the tissue's molecular
structure. The T2 times for most tissues remain approximately constant as the field strength is increased up
6,7
to roughly 1.5 tesla. With further increases in field strength, the T2 values for many tissues decrease. For
example, the T2 values for brain matter at 7 tesla are markedly less than those at 2 tesla.11
For the interested reader: Representing T2 relaxation by using an exponential decay with a
single time constant is, in fact, an approximation. For many biological tissues the T2 relaxation
6
curve is multi-exponential. Specifically, the curve is represented by the sum of two or more
exponential decays with different time constants. The exponential decay approximation is not
valid for solids.
By plotting the length of the transverse magnetization vector (as shown in Fig. 2-11B) versus time, we can
appreciate the exponential time course of T2 relaxation as depicted in Figure 2-12B. When the time elapsed
after the RF pulse equals T2, the ratio of Mxy to its initial value is e-1, which equals 0.37 or 37%. Generalizing
this statement, at n times T2 after the 90° RF pulse, the ratio equals e-n. For example, as illustrated in Figure
2-12B, at 2 T2 and 3 T2 after the 90° RF pulse, the transverse magnetization has decayed to 14% and 5%
of its initial value, respectively.
Considering the longitudinal and transverse relaxation processes together as they act to return the spin
system to thermal equilibrium, note that transverse relaxation is ultimately limited by longitudinal relaxation.
Specifically, when the longitudinal magnetization is fully relaxed, there can be no transverse magnetization.
Therefore the T2 value is always less than or equal to the T1 value. In biological tissues, except fluids, the
T2 is typically five to ten times shorter than the T1.
In summary, putting T1 relaxation in its simplest terms, we can state: if you "knock down" the longitudinal
magnetization by using an RF pulse, it will grow back within a time period that is characteristic of the
particular tissue. Similarly, for T2 relaxation, we can state: once you create transverse magnetization by
using an RF pulse, it decays toward zero within a time period that is characteristic of the particular tissue.
Later in the chapter, we will discuss how a difference in either the T1 or T2 values among tissues can be
exploited to create contrast in the image. Disease processes often alter both T1 and T2, thus providing a
7,12,13
mechanism for the detection of disease. A discussion of fundamental mechanisms underlying relaxation
can be found in reference 14.
Dephasing and T2*
Add to lightbox
Add to lightbox
5 of 14 11-04-2010 23:38
Add to lightbox
Add to lightbox
Figure 2-15 A, The square represents a cross-section through a voxel of tissue that possesses a transverse
magnetization vector Mxy. B, The voxel can be divided into an arbitrary number of subvolumes, each of
which possesses a proportionate fraction of the total transverse magnetization. For this example, four
subvolumes are shown. C, The representation of transverse magnetization vector Mxy in the rotating frame
of reference. D, The sum of the transverse magnetization vectors from the four subvolumes equals the total
transverse magnetization. In D, we could have drawn all of the vectors superimposed at the origin of the
coordinate system.
We just discussed how T2 relaxation results in the decay of transverse magnetization through irreversible
dephasing among the magnetic moments associated with the spins. Now we will discuss another process
that results in the decay of transverse magnetization, but in this case the dephasing is reversible, meaning
that, under appropriate circumstances, the effects of the dephasing can be undone.
Consider a volume of tissue, for example a cubic centimeter of brain tissue, in which transverse
magnetization has been created by applying a 90° RF pulse. In practice, our volume of tissue might
correspond to an individual volume element, or voxel, of an MR image. The total transverse magnetization
vector Mxy corresponding to any chosen volume of tissue can be thought of as the sum of several smaller
transverse magnetization vectors, each of which corresponds to a subdivision of the volume. For example, in
Figure 2-15, we show that the magnetization vector for our voxel of brain tissue can be conceptually divided
into four smaller transverse magnetization vectors for which the sum equals the original vector.
page 33
page 34
So, why would we want to divide a volume of interest into several subvolumes? Recall from Equation 2-1 that
the frequency of precession for transverse magnetization is directly proportional to the applied magnetic-field
strength. Up to this point we have implicitly assumed that the static magnetic field B0 is perfectly uniform in
space. Consider, for example, a subject's head in an MR scanner-we have assumed that the magnetic field
within the cerebellum is exactly equal to that within the temporal lobes. This, however, is generally not the
case; for various reasons that are discussed later in the chapter, the magnetic-field strength within tissue
typically varies among locations. Thus, the frequency of precession for transverse magnetization within tissue
also varies among locations. In the context of Figure 2-15, we could just as easily have divided the voxel into
40 or 400 subvolumes. Into how many parts should a selected volume of tissue be divided? Conceptually, the
volume should be divided into increasingly smaller parts until all of the spins within each subvolume
experience the same magnetic-field strength, and hence all precess at the same frequency.
6 of 14 11-04-2010 23:38
Now, imagine that the magnetic field within our voxel of brain tissue varies so that at some positions it is
greater than that ideally produced by the main magnet of the MRI scanner, while at other positions it is less.
The magnetic field within this voxel is said to be inhomogeneous, and the tissue within the voxel therefore
experiences magnetic-field inhomogeneities. The spins within each subvolume precess at a frequency that
corresponds to the local magnetic-field strength. Thus, the transverse magnetization corresponding to some
subvolumes precesses faster than ω0 (see Eq. 2-1), while that corresponding to other subvolumes
precesses slower than ω0.
Add to lightbox
Figure 2-16 Dephasing of the transverse magnetization due to magnetic-field inhomogeneities. Just after a
90° RF pulse, the Mxy vectors that correspond to the various subvolumes that make up the voxel are in
phase. As time progresses, transverse magnetization that experiences a higher than average magnetic-field
strength precesses faster than ω0 (light gray vectors), while transverse magnetization that experiences a
lower than average field strength precesses slower than ω0 (dark gray vectors). The total transverse
magnetization (lower portion of diagram), which is the sum of the transverse magnetization vectors
corresponding to the subvolumes, gradually decays as dephasing progresses. (For simplicity, the effects of
T1 and T2 relaxation are not included. The scale for the magnetization vectors corresponding to the
subvolumes is different than that for the other vectors in the diagram. In the first time frame after the RF pulse,
the magnetization vectors corresponding to the subvolumes are superimposed such that only one vector is
visible.)
The behavior of the transverse magnetization that results from precession in an inhomogeneous magnetic
field is illustrated in Figure 2-16. Just after the RF pulse creates transverse magnetization, the Mxy vectors
that correspond to the voxel subvolumes are aligned or, in other words, are in phase. As time passes, the
spins within different subvolumes precess at different frequencies and the associated Mxy vectors get out of
phase with each other or, in other words, dephase. As shown in the lower part of Figure 2-16, the total
transverse magnetization for the voxel, which is the sum of the Mxy vectors corresponding to the subvolumes,
gradually decreases as dephasing progresses. When the Mxy vectors are fully dephased, that is, when they
point in directions that span 360° as shown on the far right of Figure 2-16, the total transverse magnetization
for the voxel has decayed to zero. Since the MR signal from our voxel is directly proportional to the total
transverse magnetization, it likewise decays following the RF pulse due to the effect of field inhomogeneities.
page 34
page 35
7 of 14 11-04-2010 23:38
Add to lightbox
Figure 2-17 In the presence of field inhomogeneities, the FID (shown in gray) generated by an excitation RF
pulse decays with the time constant T2*, which is shorter than T2. Dotted and solid lines depict the
exponential decays associated with the T2* and T2 relaxation times, respectively.
For simplicity, the effect of T2 relaxation was not included in Figure 2-16. Considering both T2 relaxation and
the effect of field inhomogeneities, we find that the FID generated by an RF pulse decays at a rate faster
than that which results from T2 relaxation alone (Fig. 2-17). This decay is characterized by a time constant
that is traditionally called the T2* (or equivalently, T2*) relaxation time, which incorporates both the tissue T2
value and the contribution from field inhomogeneities as:
Because the inhomogeneity term cannot be negative, Equation 2-6 requires that T2* be less than or equal to
T2. Unlike the T1 and T2 relaxation times, the T2* time is not necessarily an intrinsic property of a given
tissue; T2* may depend on factors other than the tissue structure itself.15,16 Often, the size of the
inhomogeneity contribution, and thus the value of T2*, depend on the magnitude and distribution of field
inhomogeneities as well as on the size and shape of the voxel. Common sources of field inhomogeneities will
be discussed later in the chapter.
As long as the field inhomogeneities are static (that is, as long as they do not change over time), the
dephasing, and thus the associated portion of the MR signal decay, is reversible. This reversibility of the
dephasing brings us to our next topic-the formation of what is called an echo.
Spin-Echoes
Let us continue where we left off on the far right of Figure 2-16, with a collection of dephased transverse
magnetization vectors that belong to a collection of subvolumes within a voxel. Now imagine that a refocusing
RF pulse, as introduced in Figure 2-8C, is applied to this set of transverse magnetization vectors such that
the B1 field is directed along the yrot-axis. This RF pulse "flips" the vectors such that those precessing slower
than ω0 (dark gray) swap positions with those precessing faster than ω0 (light gray) as illustrated in Figure
2-18. The vectors continue to precess in the same way that they did before application of the RF pulse.
Namely, the slower vectors continue to precess counterclockwise and the faster vectors continue to precess
clockwise. However, instead of driving the vectors further out of phase with each other, precession now
brings the vectors back toward their starting positions. In other words, the vectors rephase. When the time
period following the refocusing RF pulse matches that between the excitation and refocusing RF pulses, all of
the transverse magnetization vectors have returned to the positions that they had just after the excitation RF
pulse and thus a spin-echo is formed.
Figure 2-19 illustrates how the total signal from the voxel evolves in relation to the application of the
excitation (90°) and refocusing (180°) RF pulses, now including the effect of T2 relaxation. As already
described in Figures 2-16 and 2-17, application of the excitation RF pulse results in an FID that decays
toward zero with a time constant T2*. As seen in Figure 2-18, application of the refocusing RF pulse swaps
the positions of transverse magnetization vectors that precess slower than ω0 with those that precess faster,
and this leads to a regrowth of the signal and formation of the spin-echo at a time that is traditionally labeled
TE, standing for "time to echo" or "echo time." As time progresses beyond TE, the transverse magnetization
vectors again begin to dephase and the signal decays toward zero with a time constant T2*. The
8 of 14 11-04-2010 23:38
fundamental symmetry of the spin-echo process requires that the refocusing RF pulse be applied at time
TE/2, halfway between the excitation RF pulse and the echo. As shown in Figure 2-19, the signal strength at
the echo is decreased relative to that just after the excitation RF pulse due to T2 decay (that is, irreversible
dephasing). Note that at any time following TE, another refocusing RF pulse can be applied to cause a
second echo to form. Extending the concepts illustrated in Figure 2-19, we find that this second echo would
occur at a time equal to the first echo time plus twice the time between the first echo and the second
refocusing RF pulse. (In other words, the second refocusing RF pulse must be halfway between the first and
second echoes.) Likewise, additional refocusing RF pulses can be applied to form additional echoes; this
principle is discussed further in Chapter 5, Pulse Sequence Design.
In summary, two RF pulses are required to form a spinecho. The first RF pulse, the excitation pulse,
"excites" the spin system and creates transverse magnetization that dephases due to field inhomogeneities
as time progresses. The second RF pulse, the refocusing pulse, "flips" the precessing transverse
magnetization vectors so that they can rephase or, in other words, refocus to form the echo. A spin-echo is
often called an RF echo, since RF pulses are required to form the echo, or a Hahn echo, after Erwin Hahn
17
who discovered this phenomenon in 1949.
For the interested reader: In general, any pair of two RF pulses, not just a 90°, 180° pair of
pulses, creates a spin-echo. Excitation pulses less than 90° create a smaller transverse
magnetization vector than that produced by a 90° pulse, and refocusing pulses other than 180°
rephase only a fraction of the transverse magnetization. The dependence of the signal S on the
excitation flip angle α and the refocusing flip angle β is:
where the sin(α) term gives the relative amount of transverse magnetization that is created by
2
the excitation RF pulse and the sin (β/2) term gives the fraction of this transverse
magnetization that is refocused. Note that the right-hand side of Equation 2-7 is equal to one
when α is 90° and β is 180°.
Mechanisms of Dephasing
page 35
page 36
9 of 14 11-04-2010 23:38
Add to lightbox
Figure 2-18 Formation of a spin-echo. Just after a 90° RF pulse, the Mxy vectors that correspond to the
various subvolumes that make up the voxel are in phase. As time progresses, transverse magnetization that
experiences a higher than average magnetic-field strength precesses faster than ω0 (light gray vectors),
while transverse magnetization that experiences a lower than average field strength precesses slower than
ω0 (dark gray vectors). A 180° refocusing RF pulse flips the vectors such that those precessing slower swap
positions with those precessing faster. As time progresses further, the transverse magnetization vectors
rephase and form an echo when the time period following the 180° RF pulse matches that between the 90°
and 180° RF pulses. The total transverse magnetization (lower portion of diagram), which is the sum of the
transverse magnetization vectors corresponding to the subvolumes, gradually decays as dephasing
progresses and then grows back to its original length as the echo is formed. (For simplicity, the effects of T1
and T2 relaxation are not included. The scale for the magnetization vectors corresponding to the subvolumes
is different than that for the other vectors in the diagram. In the first time frame after the RF pulse, the
magnetization vectors corresponding to the subvolumes are superimposed such that only one vector is
visible.)
Our last topic for this section of the chapter is a brief discussion of the typical sources of dephasing in MRI.
Anything that causes the magnetic field experienced by the spins to vary from one position to another (or, in
other words, causes the magnetic field to be inhomogeneous) is a mechanism for dephasing. The primary
mechanisms of dephasing include: 1. inhomogeneities in the field produced by the magnet of the MRI
scanner-these are commonly called main-field inhomogeneities; 2. differences in magnetic susceptibility
among various tissues or materials in the body; 3. chemical shift; and 4. magnetic-field gradients that are
applied for spatial encoding. The last item, magnetic-field gradients applied for spatial encoding, is discussed
in detail in the next section.
Ideally, the field produced by the magnet of an MRI scanner would be perfectly homogeneous over the
10 of 14 11-04-2010 23:38
imaging volume, meaning that it would have exactly the same value at all positions within the magnet. In
practice, this cannot be achieved. The magnetic-field homogeneity is typically specified for a conceptual
spherical volume that is centered in the magnet. Quantities related to magnet-field (or frequency) variations
in MRI are often stated in terms of parts per million, abbreviated ppm; one ppm is one ten-thousandth of a
percent or, literally, one in a million. For example, over a 50-cm diameter spherical volume, a state-of-the-art
whole-body 1.5-tesla magnet has a homogeneity of a few parts per million.
page 36
page 37
Add to lightbox
Figure 2-19 Signal evolution during the formation of a spin-echo. Just after a 90° RF pulse, precessing
transverse magnetization creates an MR signal that decays with a time constant T2* due to dephasing in the
presence of field inhomogeneities. The 180° RF pulse applied at time TE/2 swaps the positions of
transverse magnetization vectors that precess slower than ω0 with those that precess faster, leading to a
regrowth of the signal and formation of a spin-echo at time TE. Due to T2 relaxation, the signal strength at
the echo is decreased relative to that just after the 90° RF pulse.
In general, main-field inhomogeneities result in signal loss (T2* decay) or geometric distortion within the
image. Geometric distortion, which causes signal from one physical location to appear in a different physical
location, can result in relative increases or decreases in signal intensity. The extent of signal-loss and
distortion effects depends on the type of imaging technique and its parameter values, although for a modern
MRI magnet the effects of main-field inhomogeneities are often negligible; as discussed next, the principal
cause of field inhomogeneity in clinical MRI is the interaction of the magnetic field with the human body when
it is placed in the scanner.
18
Magnetic-susceptibility differences are the most important source of field inhomogeneity in clinical MRI.
The magnetic susceptibility of a substance, such as your brain tissue, is a measure of the degree to which it
is magnetized when placed in an external magnetic field, such as that produced by the magnet of an MRI
scanner. In other words, when you place a substance within a magnetic field, the substance interacts with
the magnetic field and so the total magnetic field within the substance is different than the applied field. The
relationship between the magnetic field induced within a substance and the externally applied field strength is
given by a material property called the volume magnetic susceptibility. Three ranges of susceptibility values
are of interest: diamagnetic, paramagnetic, and ferromagnetic. Diamagnetic materials, which include water
and most biological tissues, have a small (on the order of a few ppm), negative magnetic susceptibility. A
negative susceptibility means that the magnetic field induced within the material is slightly less than the
externally applied magnetic-field strength, as illustrated in Figure 2-20. Paramagnetic materials, which include
11 of 14 11-04-2010 23:38
O2 and gadolinium-based contrast agents, have a small, positive magnetic susceptibility, and thus the
induced magnetic field is slightly higher than the applied field. Finally, ferromagnetic materials like iron have a
large, positive magnetic susceptibility. Note that the presence of ferromagnetic material of either biological
(for example, chronic hemorrhage) or nonbiological (for example, dental work, prostheses, surgical clips, or
clothing) origin produces pronounced local distortions in the magnetic field.
page 37
page 38
Add to lightbox
Figure 2-20 The magnetic field that is induced within a material when placed in an external magnetic field,
such as that produced by the magnet of an MRI scanner (B0), depends on the magnetic susceptibility of the
material. Brain tissue has a diamagnetic (negative) susceptibility, and thus the magnetic-field strength
induced in brain tissue (Bbrain) is slightly less than B0, and also slightly less than that induced in air (Bair),
which has a magnetic susceptibility close to zero. The magnetic-susceptibility difference at the interface
between brain tissue and air in the sinus creates a localized magnetic-field gradient that can cause
dephasing of the transverse magnetization (see Fig. 2-21). The difference between Bbrain and Bair is
exaggerated in this figure for clarity.
When two materials with different magnetic susceptibilities are adjacent to one another (for example, air in
the sinuses next to tissue [Fig. 2-20], or a metallic clip imbedded in tissue), the change in susceptibility at the
interface between the materials results in a localized magnetic-field gradient that causes dephasing of the
transverse magnetization. The term "magnetic-field gradient" means that the strength of the magnetic field
varies with position along a specific direction. The term "localized" denotes that the magnetic-field gradient
exists only in the region surrounding the susceptibility interface, and also differentiates magnetic-field
gradients generated by a susceptibility interface from those that we apply to perform spatial encoding. As an
example, Figure 2-20 illustrates that when your head is placed in the magnet of an MRI scanner, the
magnetic-field strength induced in brain tissue (Bbrain) is slightly smaller than that induced in air in the frontal
sinus (Bair). The difference between Bbrain and Bair results in a magnetic-field gradient at the interface of the
brain tissue with the air in the sinus. If you were to measure the magnetic-field strength along a line segment
that starts in the brain tissue and ends in the frontal sinus, you would find that the field strength increases
from Bbrain to Bair in the vicinity of the air-tissue interface.
12 of 14 11-04-2010 23:38
Add to lightbox
Figure 2-21 Axial spin-echo (A) and gradient-echo (B) images of the brain that illustrate the effect of
magnetic-field inhomogeneities created by air-tissue susceptibility interfaces. Gradient-echo imaging is
much more sensitive than spin-echo imaging to field inhomogeneities. As a result, localized magnetic-field
gradients in the vicinity of air-tissue interfaces cause dephasing of the transverse magnetization, and thus
signal loss, as indicated by the arrows. (From Mugler JP III: Overview of MR imaging pulse sequences.
Magn Reson Imaging Clin N Am 7:661-697, 1999)
Analogous to main-field inhomogeneities, localized magnetic-field gradients that result from susceptibility
differences can result in signal loss (T2* decay) or geometric distortion within the image. This effect is
illustrated in Figure 2-21, which shows axial images of the brain that were acquired using spin-echo
(abbreviated SE) and gradient-echo (abbreviated GRE or GE) techniques. The next section of this chapter
will explain the basic mechanics of these fundamental MRI methods, and additional information on their
characteristics can be found in Chapters 3 and 5. For the moment, we only need to know that GRE imaging
is much more sensitive than SE imaging to field inhomogeneity, such as that caused by a susceptibility
interface. (As its name implies, SE imaging is based on the spin-echo mechanism that was discussed above.
Recall that a spin-echo permits the effects of dephasing from magnetic-field inhomogeneities to be reversed
[see Fig. 2-18].) Comparing the GRE image (Fig. 2-21B) to the SE image (Fig. 2-21A), areas of signal loss
(marked by arrows) are seen in the GRE image in the vicinity of air-tissue interfaces. As discussed above,
the susceptibility differences at these interfaces result in localized magnetic-field gradients (that is, field
inhomogeneities) that cause the transverse magnetization in these regions to dephase (see Fig. 2-16).
Without a spin-echo to reverse this dephasing, signal from the affected regions is lost.
In practice, field-inhomogeneity effects can be either an advantage or a disadvantage. Signal loss secondary
19
to the ferromagnetic nature of hemorrhage byproducts can aid in their detection. The high sensitivity of
GRE imaging to susceptibility effects is fundamental to the implementation of several important techniques,
including those for perfusion imaging20 and brain functional MRI.21 On the other hand, the substantial signal-
18,22
to-noise loss and geometric distortion that can occur in GRE images, especially at high-field strengths,
can render them unusable.
page 38
page 39
Our final topic for this section is the concept of chemical shift. When placed in an external magnetic field, the
electrons in a molecule or compound partially "shield" the nuclei so that the field strength experienced by a
given nucleus is slightly modified, or "shifted," compared to the field strength that is applied. Hence, the
Larmor frequency for the nucleus is also shifted. This phenomenon is called chemical shift because the
magnetic field that a nucleus experiences is dependent on its chemical environment. Like field inhomogeneity,
chemical shift is expressed in ppm. Of particular relevance to clinical MRI, the chemical shift for hydrogen
nuclei (protons) in fat is 3.5 ppm less than that for hydrogen nuclei in water. Thus, in the magnetic field of an
MRI scanner, the frequency of precession for transverse magnetization associated with fat is slightly lower
than that for transverse magnetization associated with water. This slight difference in resonant frequency,
which leads to a time-dependent phase difference between the transverse magnetization associated with fat
and that associated with water, gives rise to a chemical-shift artifact as discussed in Chapter 22, Image
Artifacts and Solutions. The time-dependent phase shift causes periodic dephasing and rephasing of fat
magnetization relative to water magnetization as time progresses following an excitation RF pulse. If fat and
13 of 14 11-04-2010 23:38
water are present in the same voxel of a GRE image, this dephasing and rephasing leads to a periodic signal
modulation (see Chapters 7 and 22).23
© 2010 Elsevier
14 of 14 11-04-2010 23:38
SPATIAL LOCALIZATION OF THE MR SIGNAL TO FORM AN IMAGE

In this section we discuss spatial encoding-the process by which the spatial locations of MR signals
from the body are determined in order to create an image. Building upon our knowledge of how the
MR signal is generated and its basic characteristics, the key additional ingredient that is needed for
24
spatial encoding is an externally applied magnetic-field gradient.
Magnetic-Field Gradients
The concept of a magnetic-field gradient was introduced in the previous section when we discussed
magnetic-susceptibility interfaces. Recall that the term "magnetic-field gradient" refers to the situation
wherein the magnetic-field strength varies with position along a specific direction. To perform spatial
encoding, an MRI scanner uses three gradient coils, called the x-, y-, and z-gradient coils, to generate
linear magnetic-field gradients along each of the axes of a standard Cartesian (rectangular) x-y-z
coordinate system. Conceptually, these gradient coils are simply large coils of wire. Considering, as an
example, a conventional cylinder-shaped MRI magnet, these coils of wire are wound onto a tube
having a diameter slightly smaller than the internal diameter of the magnet, so that they can be
positioned inside the bore of the magnet.
Add to lightbox
Add to lightbox
Add to lightbox
Figure 2-22 Spatial variations in the total magnetic-field strength for linear magnetic-field gradients
applied along the x-axis (A), y-axis (B), or z-axis (C). The total magnetic-field strength includes B0 and
the magnetic field generated by the respective gradient coil. The diagrams depict cross-sections
through a conventional cylinder-shaped MRI magnet. The amount of field-strength variation is
exaggerated for clarity.
For a conventional cylinder-shaped MRI magnet, Figure 2-22A illustrates the variation in total
magnetic-field strength (specifically, the combination of B0, created by the magnet of the MRI scanner,
and the magnetic field produced by the gradient coil) that results when electric current flows through
the x-gradient coil. We see that the magnetic-field strength, represented by the length of the arrows in
1 of 24 11-04-2010 23:39
the figure, varies along the x-axis but is constant along the y- and z-axes. Similarly, for a y-gradient the
field strength varies along the y-axis but is constant along the x- and z-axes (Fig. 2-22B), and for a
z-gradient the field strength varies along the z-axis but is constant along the x- and y-axes (Fig.
2-22C). Two or all three of the gradient coils can be turned on simultaneously to create a
magnetic-field gradient along any direction; this capability is used to achieve arbitrary image
orientations. Figure 2-22 uses the standard convention for labeling the x-, y- and z-axes of the magnet;
when facing the front of the magnet, the positive x-axis points to the right, the positive y-axis points
upward, and the positive z-axis (the axis parallel to B0) points toward you. The amount of field-strength
variation that occurs with the application of a gradient is exaggerated in Figure 2-22 for clarity. In
practice, the difference between the minimum and maximum field strengths would be no more than
approximately 2% (20,000 ppm) of B0.
page 39
page 40
Add to lightbox
Add to lightbox
Add to lightbox
Figure 2-23 A, Plot of the total magnetic-field strength versus distance during the application of a
linear magnetic-field gradient. At the isocenter of the magnet, the applied gradient has no effect on the
magnetic-field strength. B, The slope of the line that depicts the relationship between field strength and
distance gives the strength of the magnetic-field gradient; a steeper slope corresponds to a stronger
2 of 24 11-04-2010 23:39
gradient. The slope of the line, and hence the polarity of the magnetic-field gradient, can be positive or
negative. C, A gradient timing diagram shows how the strength and polarity of a magnetic-field
gradient change over time. The strengths for the portions of the gradient waveform labeled "strong,"
"weak," and "negative" correspond to the three lines shown in B. In A and B, the amount of field-
strength variation compared to B0 is exaggerated for clarity.
Another way to represent the effect of a magnetic-field gradient is to plot the variation in field strength
versus position as shown in Figure 2-23A. Gradient coils are designed to yield a linear variation in field
strength with distance, as illustrated in the figure. At the center of the magnet, which is called its
isocenter, the magnetic field from the gradient coil is zero and hence, even with the gradient activated,
the total magnetic-field strength equals B0. By varying the amplitude and polarity of the electric current
applied to the gradient coil, the gradient strength, which is the slope of the line that relates field
strength to distance, can be controlled. In Figure 2-23B, the black line labeled "strong gradient" has the
steepest slope, and hence represents the highest gradient strength of the three lines shown. The dark
gray line represents a weaker gradient, and the light gray line illustrates that the gradient can also be
negative. Gradient strengths are typically specified in units of milliTesla per meter (abbreviated mT/m)
or gauss per centimeter (abbreviated G/cm); state-of-the-art clinical scanners have a maximum
gradient strength of approximately 40 mT/m, which equals 4 G/cm. High gradient strengths are useful
to achieve short acquisition times (that is, rapid imaging) and high spatial resolution.
The characteristics of an applied magnetic-field gradient are commonly represented by a timing

diagram, as shown in Figure 2-23C. This diagram depicts when the gradient is switched on and off,
how quickly it is switched on and off, and the strength of the gradient. The times required to switch the
gradient on and off are called its ramp up and ramp down times, respectively. Ideally, we would like to
switch the gradient on and off instantaneously, but this is not physically possible because a finite period
of time is required to increase the current flowing through the gradient coil to the desired level. The
slope of the ramp up (or ramp down) portion of a gradient waveform is called its slew rate. Slew rates
are typically specified in units of tesla per meter per second (abbreviated T/m/s) or milliTesla per
meter per millisecond (abbreviated mT/m/ms); state-of-the-art clinical scanners have a maximum slew
rate of approximately 200 T/m/s. To facilitate fast imaging techniques, it is desirable to have a high
slew rate. In the timing diagram, the height of the plateau relative to zero denotes the gradient
strength. Note that the strengths (and polarities) of the gradient waveforms labeled "strong," "weak,"
and "negative" in Figure 2-23C correspond to the three lines in Figure 2-23B that have the same labels.
Slice Selection
To form an image the MR signals from the body must be localized in three dimensions. Much of clinical
MRI is performed by using two-dimensional (abbreviated 2D) techniques wherein the first step is to
localize the signals along one of the dimensions by exciting the magnetization only within a thin slice of
tissue25,26-a process called slice selection. (In this context, the terms "slice" and "section" are
interchangeable; which term is used is simply a matter of preference.) We can understand how slice
selection works by considering the effect of a linear magnetic-field gradient together with the Larmor
equation (Eq. 2-1). As shown in Figure 2-23A, activating a given magnetic-field gradient creates a
linear correspondence between the magnetic-field strength and the position along the gradient. Since,
as given by the Larmor equation, the frequency of precession is directly proportional to the field
strength, the magnetic-field gradient likewise creates a linear correspondence between the resonant
frequency and the position along the gradient. Therefore, as indicated in Figure 2-24, an RF pulse that
is applied at a selected center frequency (ωC) affects only spins located at a specific position (zC),
called the slice position, because only these spins possess a resonant frequency that matches the
frequency of the RF pulse. By changing the center frequency of the RF pulse, the position of the slice
along the direction of the gradient can be chosen freely. When you select specific slice positions for an
MR exam, the scanner software calculates the required center frequencies. The gradient that is
applied during the RF pulse is called the slice-select (or section-select) gradient, typically abbreviated
GS or GSS, and may be the x-, y-, or z-gradient depending on the slice orientation that is desired, or
may be some combination of these gradients if an oblique slice is desired.
page 40
3 of 24 11-04-2010 23:39
page 41
Add to lightbox
Figure 2-24 Slice selection. When applied in the presence of a linear magnetic-field gradient, an RF
pulse with center frequency ω C excites spins located at position zC along the direction of the gradient.
The RF-pulse bandwidth ∆ω and the strength of the gradient determine the slice thickness ∆Z.
An RF pulse consists of not just a single frequency, but instead a range of frequencies; this range is
called the transmit bandwidth of the pulse and is denoted by the symbol ∆ω in Figure 2-24. Because
the RF pulse contains a range of frequencies, a range of positions is excited along the direction of the
gradient; this range is the slice thickness (∆Z in Fig. 2-24). The RF-pulse bandwidth can be used to
control the slice thickness-increasing the transmit bandwidth increases the slice thickness. For a given
shape of RF pulse, the transmit bandwidth is inversely proportional to the duration of the pulse.
Therefore, making the RF pulse longer decreases the transmit bandwidth and results in a thinner slice.
As is apparent from Figure 2-24, the gradient strength can also be used to control the slice thickness.
Increasing the gradient strength localizes the range of frequencies excited by a given RF pulse to a
smaller range of positions and thus results in a thinner slice.
Add to lightbox
Figure 2-25 Timing diagram for slice selection. A 90° excitation RF pulse is applied simultaneously
with the slice-select gradient (GS). Following the RF pulse, a rephasing gradient is applied along the
slice-select axis to restore phase coherence across the thickness of the slice.
4 of 24 11-04-2010 23:39
Add to lightbox
Add to lightbox
Figure 2-26 Plot of the flip angle generated by a 90°, slice-selective, excitation RF pulse versus
distance perpendicular to the slice for an ideal rectangular slice profile (A) and a realistic slice profile
(B).
Extending the concept of the timing diagram that was introduced in Figure 2-23C, the process of slice
selection can be depicted as an RF pulse applied simultaneously with a linear magnetic-field gradient
as illustrated in Figure 2-25. This figure highlights the fact that the RF-pulse waveform possesses a
particular shape. This shape depicts the time course of the amplitude of the magnetic-field B1 (see the
first section of this chapter), and determines the slice profile for the RF pulse which, in general, refers
to how well the pulse achieves its intended effect, measured as a function of distance across the
thickness of the slice. To put this in concrete terms, consider a 90° excitation RF pulse that we want to
use to select a 5-mm thick slice. Ideally, this RF pulse would produce a flip angle of exactly 90° within
the desired 5-mm thick slice and exactly 0° everywhere outside of the slice. Thus, a plot of the flip
angle for this pulse versus distance perpendicular to the slice would have the shape of a rectangle as
shown in Figure 2-26A. An infinitely long RF pulse is required to obtain a perfect rectangular slice
profile-obviously this is not practical. For a realistic slice profile (Fig. 2-26B), the flip angle varies from
the intended value to zero over some finite distance. In addition, the flip angle decreases to zero
beyond the slice thickness, meaning that the RF pulse has an effect on tissue outside of the desired
slice.
page 41
page 42
5 of 24 11-04-2010 23:39
Add to lightbox
Figure 2-27 The profiles of closely spaced slices overlap, which can result in slice-to-slice
interference, also known as slice crosstalk. This interference can cause decreased signal intensities,
decreased contrast, and other image artifacts.
In clinical MRI the slice profile is of particular importance because we typically want to acquire a set of
closely spaced slices. Therefore, the trade-offs between the RF-pulse duration, the RF power required
for the pulse (which directly affects the energy deposited in the patient), and the quality of the slice
profile are considered in the design of the RF-pulse waveform. Nonetheless, due to the inevitable
excitation of some tissue outside of the desired slice, the profiles of closely spaced slices always
overlap to some degree (Fig. 2-27), which results in slice-to-slice interference, often called slice
crosstalk. Crosstalk can cause decreased signal intensities, decreased contrast, and other image
artifacts27-29; this source of image degradation is one of the reasons that gaps between the slices are
often used. (The other reason for using gaps between slices is to increase anatomic coverage for a
given number of slices.) Nonetheless, ongoing improvements in RF pulse design continue to lessen the
impact of slice crosstalk on image quality.
Before concluding our description of slice selection, there is one last detail to be discussed. In Figure
2-25, there is a negative portion of the gradient waveform that is applied immediately following the RF
pulse. This part of the gradient waveform is called a rephasing gradient. Consider, as an example, that
the excitation RF pulse in Figure 2-25 is a pulse as discussed earlier in the chapter (Fig. 2-8A),
which is intended to convert longitudinal magnetization into transverse magnetization that is aligned
parallel to the y-axis of the rotating frame of reference. Because this RF pulse is applied in the
presence of a magnetic-field gradient to achieve slice selection, the frequency of precession for
magnetization located at any given position along the direction of the slice-select gradient will be
different than that for magnetization located at any other position. During the application of the RF
pulse, this effect results in dephasing of the transverse magnetization across the thickness of the slice,
which is clearly undesirable because it reduces the signal from the slice. At the end of the RF pulse,
only a small fraction of the magnetization remains in the desired configuration, aligned parallel to the
y-axis. The rephasing gradient reverses this dephasing and thereby restores phase coherence across
30
the thickness of the slice, permitting the maximum signal to be obtained. Note that only excitation RF
pulses require a rephasing gradient; this gradient is not required for a slice-selective refocusing RF
pulse (that is, a refocusing RF pulse applied in the presence of a magnetic-field gradient).
In closing, let us summarize the main features of slice selection: the slice-select gradient localizes the
effect of the RF pulse; the slice position is determined by the center frequency of the RF pulse and the
strength of the gradient; the slice thickness is determined by the bandwidth of the RF pulse and the
strength of the gradient; and the slice profile is determined by the shape of the RF-pulse waveform.
The Fourier Transform and an Introduction to k-Space

Slice selection provides the means to excite magnetization within a thin section of tissue; to form an
6 of 24 11-04-2010 23:39
image the next task is to spatially encode the MR signals from this slice in the remaining two
dimensions so that the locations from which the signals originate can be determined in all three
dimensions. Before describing the details of this process, we need to discuss the Fourier transform
and k-space-concepts that lay the foundation for us to understand how spatial encoding works.
page 42
page 43
Add to lightbox
Add to lightbox
Figure 2-28 The Fourier transform (FT) allows us to determine the frequency components contained in
a signal. A, When sunlight is refracted by a glass prism or by raindrops in the atmosphere to form a
rainbow, the prism or raindrop in essence physically performs a Fourier transform, letting us see the
various frequencies of electromagnetic radiation, and thus the various colors, of which white light is
composed. B, An MR image is composed of a collection of spatial intensity variations. The image's
spatial-frequency spectrum specifies the amplitudes, orientations, and spatial frequencies of these
intensity variations. Because the signals that are acquired during an MRI exam are the k-space
-1
components of the image, the image itself is obtained by applying an inverse Fourier transform (FT ).
The Fourier transform is a mathematical tool, typically implemented in the form of a computer program,
that allows us to calculate the frequency components contained in a signal. Here, "signal" is used as a
general term that includes many different types of information, including an MR image. Some examples
may be helpful to gain an intuitive understanding of the Fourier transform. When you listen to music on
the radio, the sound is reproduced by applying an electrical signal to the speaker. If you were to apply
a Fourier transform to this electrical signal, the result would be the spectrum of the music, which
indicates how much of each audio frequency is contained in the music. For example, if bass
instruments were dominant in a song, its spectrum would contain a large low-frequency component.
Note that the Fourier transform can be inverted, meaning that the inverse Fourier transform of the
music's spectrum yields the original electrical signal. As a second example, consider sunlight being
refracted by a glass prism, or by raindrops in the atmosphere, thus forming the familiar "rainbow" of
colors. White light is composed of a range of frequencies of electromagnetic radiation; these
frequencies correspond to the colors that we see in the rainbow. In essence, the prism or raindrop
physically performs a Fourier transform, letting us see these various frequencies (colors) that make up
white light (Fig. 2-28A). Finally, consider an MR image. In mathematical terms, an MR image is a set
of intensity values, each of which is assigned to a specific x, y co-ordinate in space. The Fourier
transform of an MR image is its spatial-frequency spectrum (Fig. 2-28B), which specifies the amount
of each of the various spatial-frequency components that make up the image. The term "spatial
frequency" is commonly abbreviated by using the letter k. The MR image is a representation of the
anatomy of interest in (physical) space, whereas its spectrum is a representation of the same
7 of 24 11-04-2010 23:39
information in spatial-frequency space, or k-space.
So, what does understanding MRI have to do with k-space? The signals that are received from the
body by an RF coil during the MR imaging process are in fact the k-space components of the
image-the image itself is obtained by applying an inverse Fourier transform to these k-space data, as
indicated in Figure 2-28 by the arrow pointing from the k-space spectrum to the image. Analogous to
the rainbow example above, the process of receiving MR signals in combination with an appropriate
configuration of magnetic-field gradients physically performs a Fourier transform on the transverse
magnetization vectors within the slice of interest. An inverse Fourier transform, performed by the
image-reconstruction computer that is part of the scanner, is thus required to obtain the image that we
desire.
Taking these concepts a step further, we can state that an MR image is composed of the sum of a
large number (typically equal to the number of points in the image, for example, 256 × 256 = 65,536)
of sinusoidal intensity oscillations with various spatial frequencies and orientations. Each point in the
k-space spectrum tells us the amplitude, spatial frequency, and orientation for one of these intensity
oscillations that make up the image. The amplitude of the intensity oscillation is given by the magnitude
of the k-space component. For example, in the spatial-frequency spectrum shown in Figure 2-28B,
different points in k-space have different levels of brightness, which reflect the different magnitudes of
these components. The frequency of a given intensity oscillation in the image is proportional to the
distance of the associated k-space component from the center of k-space. Thus, a component at the
center of k-space corresponds to a uniform intensity throughout the image, a component close to the
center corresponds to a low spatial-frequency oscillation, and a component far from the center
corresponds to a high spatial-frequency oscillation. The orientation of the intensity oscillation in the
image is determined by the position of the k-space component relative to the k-space axes.
Some image examples will help to provide a feeling for these concepts. Figure 2-29 shows a number
of individual k-space components (each marked by a white dot in k-space) and their associated
sinusoidal intensity oscillations in the image. In Figure 2-29A, we see that a k-space component
located on the kx-axis, very close to the center of k-space, results in a low-frequency intensity
oscillation whose pattern of light-dark bands is oriented perpendicular to the x-axis in the image. A
k-space component located on this same axis, but farther from the center, results in a higher-
frequency oscillation with the same orientation in the image (Figs. 2-29B and C). Keeping the k-space
component the same distance from the center of k-space, but locating it instead on the ky-axis, yields
an intensity-oscillation pattern that is oriented perpendicular to the y-axis in the image as shown in
Figure 2-29D. Finally, if the k-space component is located between the kx and ky axes, for example
along a line oriented 30° counterclockwise from the kx-axis, the corresponding intensity-oscillation
pattern in the image is oriented perpendicular to a line that is 30° counterclockwise from the x-axis.
Generalizing the concepts illustrated in Figure 2-29, we can state the following relationships between
the k-space data and the image, which are illustrated in Figure 2-30. The data in the central region of
k-space, namely, the low spatial-frequency components, represent the gross structure and contrast in
the image (Fig. 2-30B). On the other hand, the data in the outer regions of k-space, namely, the high
spatial-frequency components, represent the detailed structure and contrast in the image. The position
of the k-space data relative to the k-space axes corresponds to the orientation of the structures that
the data represent in the image (compare Fig. 2-30C to Fig. 2-30D). For example, in Figure 2-30C,
only high spatial-frequency k-space data in the vicinity of the horizontal (kx) axis are included. Thus,
only features with rapidly changing intensity values (that is, "edges," particularly in the vicinity of the
bright subcutaneous fat) oriented perpendicular to the horizontal (x) direction are visible in the
associated image.
page 43
page 44
8 of 24 11-04-2010 23:39
Add to lightbox
Figure 2-29 Examples of individual k-space components (denoted by small white dots in k-space and
highlighted by yellow arrows) and the corresponding sinusoidal intensity oscillations in the image. A, A
spatial-frequency component located on the kx-axis and very close to the center of k-space yields a
low-frequency intensity oscillation whose pattern of light-dark bands in the image is oriented
perpendicular to the x-axis. B and C, When the spatial-frequency component is located at increasingly
higher spatial frequencies, intensity oscillations with correspondingly higher frequencies result in the
image. D, A spatial-frequency component located on the ky-axis yields an intensity-oscillation pattern
that is oriented perpendicular to the y-axis. E, A spatial-frequency component located between the kx
and ky axes, at a particular angle relative to the kx-axis, yields an intensity oscillation that has a
matching angulation in the image.
9 of 24 11-04-2010 23:39
The final issue associated with k-space that needs to be addressed before turning our attention to the
details of spatial encoding concerns the amount of k-space data that must be collected to make an MR
image. Specifically, how close does any given data point in k-space need to be to its neighbor (in other
words, how densely must k-space be sampled) and how far out in k-space do we need to collect
data? The distance in k-space between adjacent data points (∆k) is inversely proportional to the extent
of the image (namely, its field-of-view, abbreviated FOV) in the corresponding direction as illustrated in
Figure 2-31A. (As you probably already know from practical experience, if the FOV is too small then
tissue from one side of the image wraps around to appear on the opposite side of the image-see
Chapter 22.) As shown in Figure 2-31B, the inverse relationship also holds, specifically, the distance in
the image between adjacent points (∆y in Fig. 2-31B), which gives the spatial resolution, is inversely
proportional to the extent of k-space that is collected in the corresponding direction. Conceptually this
makes sense-high spatial frequencies are required to represent fine details in the image and thus
achieve high spatial resolution.
In general, the k-space components on the positive side of k-space have a specific mathematical
relationship to those on the negative side. Thus, components on one side of k-space can be used to
calculate those on the other side, permitting an image to be reconstructed from a fraction of the "full"
(that is, containing equal portions of positive and negative spatial frequencies) data set. 31,32 Common
names for this type of acquisition include partial-Fourier, half-Fourier, fractional-NEX, half-NEX,
fractional-echo, half-echo, and asymmetric-echo imaging, depending on the details of the acquisition
and whether the intent is to shorten the imaging time or decrease the echo time. Although a fractional
acquisition is advantageous for specific applications, many common techniques acquire full k-space
data sets.
For the interested reader: Because k-space data represent the spatial distribution of
the transverse magnetization, which is a vector quantity, these data, and likewise the
inverse Fourier transform of these data (the image), are complex. For simplicity, the
magnitude of the k-space data was presented in Figures 2-28, 2-30, and 2-31. For most
clinical applications, the magnitude of the MR image is used.
Considering Figure 2-29, all of the real-valued intensity oscillations shown are even
because real-valued k-space components were used. By using such intensity oscillations
we can describe only objects that also possess this symmetry; obviously complex
k-space components are required to describe realistic objects. (In Fig. 2-29, complex-
valued k-space components would result in intensity oscillations that are displaced
relative to the center of k-space.) Nonetheless, in the ideal case, the MR image is real
valued, and so the corresponding spatial-frequency spectrum possesses Hermitian
(complex conjugate) symmetry. Specifically, a k-space component at co-ordinate (kx, ky)
is the complex conjugate of the component at (-kx, -ky). This symmetry is the basis for
the fractional-acquisition methods discussed above. In reality, the image data typically
possess some imaginary component, although in many cases this can be adequately
described by only low spatial-frequency data. Thus, fractional-acquisition methods collect
somewhat more than one half of the k-space data to permit accurate reconstruction of
the actual (non-real-valued) image data.
Spatial Encoding
page 44
page 45
10 of 24 11-04-2010 23:39
Add to lightbox
Add to lightbox
11 of 24 11-04-2010 23:39
Add to lightbox
Add to lightbox
Figure 2-30 The relationship between the data location in k-space and the resulting features in the MR
image. The k-space data (upper row) and corresponding images (lower row) are shown for: A, a
complete 256 × 256 data set; B, the central 64 × 64 values in this data set; C, two 96 × 32 data blocks
centered on the horizontal (kx) axis; and D, two 32 × 96 data blocks centered on the vertical (ky) axis.
The k-space data is displayed using a logarithmic intensity scale to provide an improved visualization
of the high spatial-frequency components. (From Mugler JP III: Overview of MR imaging pulse
sequences. Magn Reson Imaging Clin N Am 7:661-697, 1999)
As introduced during our discussion of k-space, the spatial-encoding process used in MRI involves
collecting k-space data that correspond to the desired image. The key element of this process is the
relationship between k-space and applying a linear magnetic-field gradient. Let us consider in detail
how the transverse magnetization along the direction of a magnetic-field gradient is affected during the
12 of 24 11-04-2010 23:39
application of this gradient. In particular, Figure 2-32 illustrates the behavior of transverse
magnetization vectors located along a line that is parallel to the x-axis (Fig. 2-32B) during the
application of a constant magnetic-field gradient (Gx) along this axis.
Just after application of the excitation RF pulse, the transverse magnetization vectors are in phase
(time point 1, Fig. 2-32D). Note that the labels 1, 2, 3, and 4 in Figures 2-32A, C, and D designate
successive points in time and emphasize the relationship between the application of the gradient (Fig.
2-32A), the state of the transverse magnetization vectors (Figs. 2-32D and E), and the associated
positions in k-space (Fig. 2-32C). Once the gradient is turned on (lower portion of Fig. 2-32B),
transverse magnetization vectors located at positions greater than zero along the x-axis (positions F
through I in Figs. 2-32B and D) precess faster than ω0 and, conversely, those located at positions less
than zero (positions A through D) precess slower than ω0. As a result, magnetization vectors located
at positions other than zero gradually accumulate phase relative to that located at zero (position E),
magnetization vectors at negative positions accumulate phase in a counterclockwise sense, and those
at positive positions accumulate phase in a clockwise sense (Fig. 2-32D).
Add to lightbox
Add to lightbox
page 45
page 46
Figure 2-31 Relationships between the density and extent of data in k-space, and the field-of-view and
spatial resolution in the image. A, The distance in k-space between adjacent data points (∆ky) is
inversely proportional to the field-of-view of the image in the corresponding direction (FOV y). B, The
distance in the image between adjacent points (∆y), which gives the spatial resolution, is inversely
proportional to the extent of k-space in the corresponding direction (2ky,max).
Add to lightbox
13 of 24 11-04-2010 23:39
Add to lightbox
Add to lightbox
Add to lightbox
Figure 2-32 The spatial-encoding effect of a linear magnetic-field gradient on the transverse
magnetization. A, An excitation RF pulse followed by a constant, linear magnetic-field gradient that is
applied along the x-axis. The effect of this gradient is evaluated at the four time points labeled 1, 2, 3,
and 4. B, A linear relationship between position (x) and frequency (ω) is created by the gradient
applied in A. The resulting behavior of the transverse magnetization along a line parallel to the x-axis is
considered at nine positions (A through I) that are equally spaced along the direction of the gradient.
C, The positions in k-space that correspond to the four time points. D, The positions of the transverse
magnetization at positions A through I as time progresses. Using position E as reference,
magnetization at positions F through I precesses faster than ω 0 (clockwise) while that at positions A
through D precesses slower than ω 0 (counterclockwise). The dashed circles show the paths traced
out by the tips of the precessing magnetization vectors. E, Plots of the tips of all magnetization vectors
between positions A and I versus distance along the x-axis for the four time points. The gradient twists
the magnetization vectors into a helical pattern that becomes tighter and tighter as time progresses.
14 of 24 11-04-2010 23:39
The reciprocal of the spatial wavelength for this helix gives the spatial frequency that is plotted in C.
The effects of field inhomogeneities, and of T1 and T2 relaxation, are neglected. Variations in the
relaxation times among the tissues would result in different lengths for the vectors associated with
these tissues.
It is instructive to consider the behavior for the complete set of transverse magnetization vectors
located between positions A and I, not just those at the selected nine points. By plotting the positions
of the tips of all magnetization vectors versus distance along the x-axis, we obtain the diagrams shown
in Figure 2-32E. At time point 1, when the magnetization vectors are in phase, the tips of the
magnetization vectors lie along a straight line. As time progresses, we see that the tips of the vectors
trace out a helix, which is twisted tighter and tighter as the effects of the gradient accumulate.
Adjacent equivalent positions on the helix define a spatial wavelength, which is the reciprocal of spatial
frequency. As the wavelength decreases from time points 1 to 4, the associated spatial frequency
increases as shown in Figure 2-32C; this situation is directly analogous to that depicted in Figures
2-29A through C. Thus, and this is the critical point, applying a linear magnetic-field gradient twists the
magnetization vectors along the direction of the gradient into a pattern in space that, as the effects of
the gradient accumulate, corresponds to increasingly higher spatial frequencies. By receiving the MR
signal when the magnetization exists in a configuration associated with a particular spatial frequency,
the value of the k-space component for that spatial frequency is obtained. Stated in a more general
way, the process of applying a linear magnetic-field gradient to transverse magnetization and receiving
the MR signal physically performs a Fourier transform on the spin system,33 and hence this method is
also known as Fourier encoding.
page 46
page 47
Since the accumulation of phase discussed above is, in other words, dephasing, you may wonder why
dephasing caused by the magnetic-field gradients allows us to perform spatial encoding while that
caused by field inhomogeneities, as discussed at the end of the previous section, results in image
artifacts such as signal loss. The essential difference between these two mechanisms is that
dephasing due to the application of a linear magnetic-field gradient occurs uniformly across the field-
of-view of the image and to an extent that is under our full control, while dephasing from undesired field
inhomogeneities, such as those created by a susceptibility interface, are localized to a particular region
and cannot be changed in magnitude or switched on and off.
Add to lightbox
15 of 24 11-04-2010 23:39
Add to lightbox
Add to lightbox
Add to lightbox
Figure 2-33 Relationship between applied magnetic-field gradients and the resulting path through
k-space. A, Timing diagram for the spatial-encoding configuration used for a GRE-type acquisition. B,
Path through k-space that corresponds to the timing diagram in A. C, Timing diagram for the spatial-
encoding configuration used for a SE-type acquisition. D, Path through k-space that corresponds to
the timing diagram in C. The labels 1, 2, etc., emphasize the association between the application of
the magnetic-field gradients and the resulting positions in k-space. ADC, analog-to-digital converter;
Gx, magnetic-field gradient applied along x-axis; Gy, magnetic-field gradient applied along y-axis; RF,
radiofrequency pulses.
Generalizing the concept presented in Figure 2-32, we find that by applying linear magnetic-field
gradients one can "navigate" through k-space and visit the desired spatial frequencies. Immediately
following an excitation RF pulse (and its associated rephasing gradient) the transverse magnetization
vectors are in phase throughout the slice, and we are thus positioned at the center of k-space. At any
future point in time, our position in k-space along a particular axis depends on the accumulated effect
of the gradient applied along the associated spatial axis. Specifically, for a constant gradient the
spatial frequency is directly proportional to the product of the strength of the gradient and its duration
and, in more general terms, for any arbitrary gradient waveform the spatial frequency is directly
16 of 24 11-04-2010 23:39
proportional to the area under the gradient waveform. Note that a negative gradient (Figs. 2-23B and
C) permits us to move toward negative spatial frequencies in k-space. As an example, consider the
timing diagram shown in Figure 2-33A. The negative portion of the gradient Gx moves our position in
k-space from the center (point 1 in Fig. 2-33B) along the negative kx-axis. At the same time, the
gradient Gy moves our position along the positive ky-axis. The net result is a movement from point 1 to
point 2 as illustrated in Figure 2-33B. Next, the positive portion of the gradient Gx moves our position
from left to right in k-space, parallel to the kx-axis. Since our fundamental goal is to collect the k-space
data needed to form an image, we could instruct the scanner to measure MR signals during the
positive portion of the gradient Gx, and thereby acquire the portion of the spatial-frequency spectrum
that lies along the "line" of k-space between points 2 and 4 in Figure 2-33B. These data are sampled
and converted to numbers that the reconstruction computer can process by an analog-to-digital
converter, abbreviated ADC. In Figure 2-33A, the raised portion of the ADC line that coincides with the
positive portion of the gradient Gx denotes the time period during which the ADC is turned on to
sample data. This data-sampling period typically ranges between 1 and 25 milliseconds.
For the interested reader: Considering, for example, the gradient Gx, the associated
value for the spatial frequency as a function of time, kx(t), can be written:
Generalizing this expression to three dimensions gives:
where k(t) = kxi + kyj + kzk (i, j and k are the unit vectors along the x, y and z directions,
respectively) and G(t) = Gxi + Gyj + Gzk. By using this expression for k, the signal S
received from the body can be written (neglecting relaxation):
where V is the volume of interest, r = xi + yj + zk, and FT stands for Fourier transform.
Thus, the received signal is the Fourier transform of the spatial distribution of the
transverse magnetization. The exponential term in Equation 2-10 indicates the effect of
the magnetic-field gradients, which twist the magnetization vectors relative to one
another to achieve the desired spatial frequencies, and the integral corresponds to
receiving the MR signal by using an RF coil.
The gradient configuration shown in Figure 2-33A uses a gradient-polarity reversal (that is, a negative
portion of Gx followed by a positive portion) to permit a complete line of k-space to be sampled. In the
context of dephasing and rephasing, the negative portion of Gx dephases the transverse magnetization
and its positive portion rephases the magnetization such that phase coherence (aside from the effects
of field inhomogeneities) along the direction of the gradient is restored at point 3 in Figure 2-33B. In
other words, an echo is formed at point 3 by the gradient reversal. For this reason, imaging techniques
that are based on this gradient configuration are typically called gradient-recalled echo or, more
simply, gradient-echo (GRE) methods.
page 47
page 48
The gradient that is active during the time period when data is sampled, or "read out," is commonly
called the readout gradient, abbreviated GR or GRO. Since the frequency of precession varies along
the direction in which it is applied, another common name for this gradient is the frequency-encoding
gradient, abbreviated GFE. The gradient that is applied in a direction perpendicular to that for the
readout gradient is called the phase-encoding gradient (Gy in Fig. 2-33A), abbreviated GP or GPE. The
term "phase-encoding" is used because this gradient does not affect the frequency of precession
during the time period when data are sampled; its effects are manifested through the phase shifts that
accumulate prior to the data-sampling period. Analogous to the situation for the slice-select gradient,
the readout and phase-encoding gradients may be any of the x-, y-, or z-gradients, depending on the
slice orientation that is desired, or may be some combination of these gradients if an oblique slice is
17 of 24 11-04-2010 23:39
desired.
To complete our basic understanding of how to navigate through k-space, we need to discuss the
effects of a 180° refocusing RF pulse as used to form a spin-echo. Consider the timing diagram shown
in Figure 2-33C, which is the same as that in Figure 2-33A except a 180° RF pulse has been added
and the polarities of Gy and the initial portion of Gx have been reversed. The combination of Gy and the
portion of Gx before the 180° RF pulse move our position from the center of k-space to point 2 in
Figure 2-33D. The effect of a refocusing RF pulse is to "flip" our position in k-space about the center.
Thus, as illustrated in Figure 2-33D, the 180° RF pulse moves our position from point 2 to point 3.
Next, the second portion of the Gx waveform moves us along a line in k-space parallel to the kx-axis,
as already seen in Figure 2-33B. At point 4 in Figure 2-33D, phase coherence is restored with respect
to the effects of both the gradient Gx and field inhomogeneities. In other words, a spin-echo is formed
at point 4. Therefore, imaging techniques that are based on this gradient and RF-pulse configuration
are called spin-echo (SE) methods.
Add to lightbox
Figure 2-34 Relationship between the amplitude and polarity of the phase-encoding gradient (GPE)
and the spatial frequency for which data are acquired. High gradient amplitudes with a positive polarity
correspond to high positive spatial frequencies. Likewise, high gradient amplitudes with a negative
polarity correspond to high negative spatial frequencies. Low gradient amplitudes correspond to the
central portion of k-space.
Now that we know how to navigate k-space, the remaining task necessary to complete the spatial-
encoding process is to choose the path in k-space that will be used to collect the required data, and
the order in which this data acquisition will occur. Collectively, the path and order of data acquisition is
called the k-space trajectory. As long as the requirements are met concerning the density and extent of
sampling that were outlined at the end of our discussion of k-space, any trajectory can, in principle, be
used.34,35 Many clinical MRI techniques use a simple k-space trajectory that consists of collecting the
data sequentially, line-by-line, moving from one extreme of k-space to the opposite extreme. A
trajectory that collects the data along parallel, straight lines in k-space is commonly called a Cartesian
or rectilinear trajectory. The rectilinear case will be discussed here; more complicated trajectories are
discussed in Chapter 7, Advanced Imaging Techniques. Consider again Figures 2-33A and B. To
sample the required area of k-space, we can simply repeat the configuration shown in Figure 2-33A
multiple times and, with each repetition, change the amplitude of the gradient Gy in equally-spaced
steps from a high positive value to a high negative value. In this way, parallel lines in k-space are
acquired, extending from high positive to high negative ky values. For example, if a 256 × 256 image
matrix is desired, the gradient Gy steps through 256 values and during each data-sampling period 256
k-space components are sampled. Figure 2-34 illustrates the relationship between the amplitude of the
phase-encoding gradient and the corresponding lines of k-space data. As will be discussed in the next
section, the time between successive repetitions (and thus the time between adjacent lines of k-space
in the phase-encoding direction), called the repetition time (abbreviated TR), is chosen based on the
18 of 24 11-04-2010 23:39
image contrast that is desired and may be as long as several seconds. Note that the time to acquire a
full set of k-space data is equal to the TR times the number of phase-encoding lines.
The fact that k-space data corresponding to the readout direction are collected as time evolves
following the excitation RF pulse, but over a relatively short period of a few milliseconds or less, while
the data corresponding to the phase-encoding direction are collected each repetition at the same time
relative to the excitation RF pulse, but over a total time that can be up to several minutes, leads to
different properties for the image artifacts that may appear due to motion, field inhomogeneities, or
chemical shift. The appearance of these artifacts and their relationship to the timing of the data
collection process are discussed in Chapter 22.
page 48
page 49
Table 2-2. Formulae for the Pixel Dimensions and Fields-of-View along the
Readout and Phase-Encoding Directions for Standard Fourier Encoding
Pixel Dimension Field-of-View
Readout
Phase-Encoding
∆GP, difference between consecutive phase-encoding gradient strengths; ∆t, time between
consecutive samples during TS; GP,max, maximum strength of phase-encoding gradient; GR, strength
of readout gradient; TP, duration of phase-encoding gradient; TS, data-sampling period.
It is useful to have an intuitive feel for the relationships between spatial resolution and the gradient
characteristics. We learned above that the spatial frequency value is directly proportional to the
product of the gradient strength and duration for a constant gradient. Since high spatial frequencies are
required to represent details in the image, it follows that increasing the strength or duration of a
spatial-encoding gradient, thus permitting higher spatial frequencies to be sampled, will result in higher
spatial resolution (that is, smaller pixel dimensions). It is beyond the scope of our discussion to derive
the associated mathematical relationships, but they are nonetheless presented in Table 2-2 for
reference, along with those governing the image field-of-view. The interested reader should find that
these formulae are straightforward to derive from the k-space relationships presented in Figure 2-31.
To wrap up our discussion, we describe another formalism that is commonly used to explain spatial
encoding along the readout direction. Recall from our discussion of slice selection that the application
of a linear magnetic-field gradient creates a linear relationship between resonant frequency and
position along the gradient. Therefore, as illustrated in Figure 2-35, if MR signals are received in the
presence of a gradient, the signal from any given position along the gradient corresponds to a unique
frequency. As a result, by applying a Fourier transform to these MR signals, the positions from which
they originated can be recovered. Although this explanation of spatial encoding along the readout
direction is easy to grasp, it is not particularly helpful for understanding phase-encoding or other
advanced techniques such as spiral imaging, or even for understanding the function of the initial
portions of the Gx gradients shown in Figures 2-33A and C.
The Pulse-Sequence Timing Diagram
19 of 24 11-04-2010 23:39
Add to lightbox
Figure 2-35 Frequency encoding. When MR signals are received in the presence of a linear
magnetic-field gradient, the signal from any given position along the gradient corresponds to a unique
frequency. For example, tissues located at position x1 resonate at frequency ω 1.
The temporal sequence of the RF, magnetic-field gradient, and data-sampling events that compose a
given technique for MR imaging is called a pulse sequence. A pulse-sequence timing diagram, which
depicts these events in a graphical format, is the most commonly used method for describing the
characteristics of pulse sequences. The primary elements of the pulse-sequence timing diagram were
already introduced during our discussions of the slice-selection and spatial-encoding processes (Figs.
2-25 and 2-33). We can combine these pulse-sequence elements to form the timing diagrams for two
fundamental techniques in MRI, gradient-echo and spin-echo imaging.
page 49
page 50
Add to lightbox
20 of 24 11-04-2010 23:39
Add to lightbox
Add to lightbox
Figure 2-36 Pulse-sequence timing diagrams for gradient-echo (A), single spin-echo (B), and double
spin-echo (C) techniques. ADC, analog-to-digital converter; GP, phase-encoding gradient; GR,
readout gradient; GS, slice-select gradient; NPE, number of times that the basic pulse-sequence timing
is repeated, which equals the number of phase-encoding lines; RF, radiofrequency pulses; TE, echo
time; TR, repetition time.
Figure 2-36A shows the pulse-sequence timing diagram for a basic GRE technique. The first line,
labeled RF, illustrates the time of application and the waveforms for the RF pulses-in this case an
excitation RF pulse with a flip angle α. The next three lines depict the time of application and the
waveforms for the linear magnetic-field gradients applied along three mutually-orthogonal axes. These
are labeled GS, GP, and GR, for the slice-select, phase-encoding, and readout gradients, respectively,
although other labels, for example Gx, Gy, and Gz, are often used to indicate the assignment of the
gradients for a particular slice orientation. As discussed above, the slice-select gradient is applied in
synchrony with the RF pulse to spatially localize its effects, thus "selecting" the slice of interest. As
also discussed above, the gradient waveforms applied along the phase-encoding and readout axes
spatially encode the magnetization within this slice along the remaining two directions. The gradient
waveform on the phase-encoding axis uses a symbolism that we have not yet discussed. The series of
closely spaced horizontal lines is the graphical symbol for what is called a gradient table. This
representation means that the strength of the gradient is stepped through a series of values (as shown
in Fig. 2-34) as the basic pulse-sequence timing is repeated to collect the k-space data required to
form the image. With each repetition, another line of data is collected. The fifth line in the timing
21 of 24 11-04-2010 23:39
diagram shows when the MR signals occur. Although some pulse sequences generate a number of
distinct signals, typically only the signals of interest are depicted. The last line of the timing diagram,
labeled ADC, illustrates the time period during which the data are collected. It is not uncommon for
either or both of the signal and ADC lines to be omitted. Finally, the square brackets around all of the
timing events with the NPE label in the lower right corner indicate that these events are repeated NPE
times, which is the number of phase-encoding lines or views.
The pulse-sequence timing diagram for a conventional single-SE technique is presented in Figure
2-36B. As explained when we discussed Figure 2-33, the polarity of the initial portions of the GR and
GP waveforms for the SE pulse sequence is reversed compared to those for the GRE pulse sequence
to account for the effect of the refocusing RF pulse. The other pulse-sequence events for the single-SE
technique are directly analogous to those for the GRE technique. As discussed earlier in the chapter,
successive spin-echoes can be formed following a given excitation RF pulse by applying successive
refocusing RF pulses. One of the most common implementations of this principle is the double-SE
pulse sequence, shown in Figure 2-36C, which is used to collect a pair of images-the first having a
relatively short echo time and the second having a relatively long echo time.
page 50
page 51
Add to lightbox
Figure 2-37 The three basic strategies used to obtain a set of MR images: two-dimensional (2D)
slices, three-dimensional (3D) single-slab, and 3D multi-slab. This example illustrates the possibilities
for acquiring axial images of the brain. Acquisitions using 2D slices or 3D multi-slab often employ a
gap between the slices or slabs, as shown, to increase the anatomic coverage and to decrease
crosstalk effects, although the use of contiguous slices or slabs (that is, no gaps) is appropriate for
some applications. (From Mugler JP III: Overview of MR imaging pulse sequences. Magn Reson
Imaging Clin N Am 7:661-697, 1999)
The double-SE timing diagram illustrates two important pulse-sequence features that are used for a
variety of applications. Comparing the slice-select and readout gradient waveforms for the second
echo to those of the first echo, several differences are seen. The gradient waveforms applied following
the first echo, aside from those during the refocusing RF pulse and the data-sampling period, perform
what is commonly called flow compensation. The goal of flow compensation is to manipulate the
22 of 24 11-04-2010 23:39
transverse magnetization such that the gradient-induced phase shift at the echo time for moving
material is equal to that which would have occurred if the material were stationary. 36,37 Thus,
associated motion artifacts are suppressed. This process, which is applicable to most types of MRI
36
pulse sequences, is also called motion artifact suppression technique (abbreviated MAST) or
gradient moment rephasing (abbreviated GMR). Additional information concerning motion artifacts and
the details of flow compensation can be found in Chapter 27. Another difference between the first and
second echoes in Figure 2-36C is that the data-sampling period for the second echo is longer.38 Pulse
sequences with data sampling periods of different durations are often called variable-bandwidth
techniques. In a double-SE pulse sequence, the variable-bandwidth strategy is implemented by using a
short data-sampling period for the first echo, which permits a short TE value, and a long data-sampling
period for the second echo, which increases the signal-to-noise ratio since the noise level varies
inversely as the square root of the data-sampling period. A lower noise level is beneficial for the later
echo because the signal intensity is, by design, attenuated by T2 decay. Optimization of the signal-
to-noise ratio is discussed further in Chapter 3.
Multislice, Sequential-Slice, and Three-Dimensional Imaging

To obtain MR images that cover the anatomic region of interest, either a set of 2D slices or a three-
dimensional (abbreviated 3D) volume is typically acquired as illustrated in Figure 2-37. Compared to
2D slice-selective imaging as discussed in detail above, a 3D acquisition uses an additional phase-
encoding gradient table that is applied along the third dimension. This additional gradient performs
spatial encoding along the slice-select direction, thereby generating a set of contiguous slices within
the volume. These slices are often referred to as partitions to distinguish them from image slices
derived from a fundamentally 2D acquisition. The motivations for performing a 3D acquisition include
the availability of thin, contiguous slices, which can be used to obtain high-resolution images of
39-41
arbitrary orientations through post-processing and, for equivalent pulse-sequence parameter
values in the 2D and 3D cases, an increase in the signal-to-noise ratio by a factor equal to the square
root of the number of partitions.42 Note, however, that the time to acquire a full set of k-space data for
3D imaging is equal to the TR times the product of the number of phase-encoding steps for the second
dimension and that for the third dimension. Thus, if a large number of slices are desired in the third
dimension, the imaging time will be quite long unless TR is relatively short. A hybrid of 2D and 3D
imaging is also sometimes used, called 3D multi-slab. 43,44 For this type of acquisition a set of slabs is
acquired, and each slab is in turn phase-encoded along the third dimension to yield a set of contiguous
partitions (Fig. 2-37, right). In this context, a "true" (single-volume) 3D acquisition is sometimes
referred to as 3D single-slab imaging.
page 51
page 52
Add to lightbox
Add to lightbox
23 of 24 11-04-2010 23:39
Add to lightbox
Figure 2-38 The order of data acquisition for 2D multislice and sequential-slice imaging. A
hypothetical case is illustrated where three images are to be acquired, and the k-space data
corresponding to each image are composed of three segments. One data segment of an image slice
is acquired following each application of the excitation RF pulse, and thus nine excitations are
required to collect the data for all three images. A, The k-space data layout corresponding to the three
images. B, The data acquisition order for 2D multislice imaging. C, The data acquisition order for 2D
sequential-slice imaging.
Two-dimensional acquisitions can be subdivided into two important forms: sequential-slice and
multislice45 (Fig. 2-38). In a sequential-slice acquisition, all of the data required for a given image are
collected before proceeding to the next image (Fig. 2-38C), analogous to the operation of conventional
X-ray computer tomography. In contrast, a multislice acquisition collects only a portion of the data
required for a given image, typically one or several phase-encoding lines, before collecting the
corresponding data for each of the other images. This process is repeated until all of the data for all of
the images are collected (Fig. 2-38B). Obviously, a given image can be acquired more rapidly by using
the sequential-slice method, and this approach is typically used when short acquisition times are
critical, for example in some implementations of MR angiography or for imaging rapidly moving
structures. On the other hand, a multislice acquisition is typically used when a relatively long time
period (on the order of 1 second) is required between the interrogations of a particular slice to permit
T1 relaxation to occur and thereby generate the desired image contrast, as discussed in the next
section. While waiting for the magnetization in a given slice to recover, slices at different positions can
be excited and encoded. Thus, for the multislice method, the imaging time for several slices is the
same as that for one slice, which is an important feature of this strategy.
Another practical aspect of sequential-slice and multislice acquisitions is the temporal order in which
the data for the slices is collected. This may be important, for example, if there is blood flow
perpendicular to the slices,46,47 or if there is a significant degree of crosstalk between slices.27,29 The
most common acquisition orders are consecutive, beginning at either end of the image set, and
interleaved, for example collecting all odd-numbered slices followed by all even-numbered slices,
although arbitrary acquisition orders can be used. Note that the sequential-slice and multislice concepts
are also applicable to 3D multi-slab acquisitions.
© 2010 Elsevier
24 of 24 11-04-2010 23:39
BASIC FORMS OF IMAGE CONTRAST

The contrast between two tissues is defined as the difference between their signal intensities divided
by their average signal intensity. For example, two tissues with very different signal intensities exhibit
high contrast, while two tissues with similar intensities exhibit low contrast. One of the principal
advantages of MRI over other imaging modalities, such as X-ray computed tomography, is the high
level of contrast among soft-tissues that can be obtained.
By using appropriate pulse-sequence designs, many different physical and physiologic parameters can
be encoded in the form of image contrast. The most basic form of contrast in MRI, and most widely
used, is that based on the tissue relaxation times T1 and T2. Due to the requirements of the spatial-
encoding process, it is not possible to generate contrast in a straightforward manner that depends
solely on one of these parameters. Instead, images are typically created whose contrast depends
largely on a given parameter, for example T1, and such images are said to be weighted by this
parameter. For instance, the contrast in a T1-weighted image primarily reflects differences in the T1
relaxation times among the tissues.
page 52
page 53
Add to lightbox
Add to lightbox
Figure 2-39 Plots of T1 (A) and T2 (B) relaxation following a 90° excitation RF pulse for three
relaxation times. Equal proton densities are assumed for the three curves in each plot.
In Figure 2-39 we review the effects of T1 and T2 relaxation on the magnetization, as discussed earlier
in the chapter. The longitudinal magnetization corresponding to a tissue with a short T1 relaxes rapidly
1 of 8 11-04-2010 23:39
toward thermal equilibrium (upper curve in Fig. 2-39A) in comparison to a tissue with a long T1, which
relaxes more slowly (lower curve). Similarly, the transverse magnetization corresponding to a tissue
with a short T2 decays rapidly toward zero (lower curve in Fig. 2-39B) in comparison to a tissue with a
long T2, which decays more slowly (upper curve). For both T1 and T2 relaxation, the amplitudes of the
magnetization vectors are the same at time equal to zero (assuming equal proton densities; see
below), before any relaxation has occurred, and when the elapsed time is several times the longest
relaxation-time value, such that the magnetization has returned to thermal equilibrium. As a result, for
times that are very short or very long compared to the values of T1 or T2, the difference between the
relaxation curves, and hence the corresponding image contrast due to differences in T1 or T2, is small.
Further, the maximum difference between the curves associated with two different relaxation times
occurs at an intermediate time, which falls between the two associated relaxation times.
The T1 and T2 relaxation curves for a given tissue can be used to calculate the associated signal
intensity in the image. A description of the evolution of the magnetization during a pulse sequence will
help us to understand how this calculation is performed. Consider, as an example, the single-SE pulse
sequence discussed in the previous section (see Fig. 2-36B). The pulse-sequence events, including the
RF pulses, are repeated many times to collect the k-space data required to form the image. Because
the time period (TR) between successive corresponding pulse-sequence events, such as two
successive 90° RF pulses, is constant, the magnetization is put into a steady state. This means that if
the longitudinal or transverse component of the magnetization associated with a given tissue is
measured at corresponding time points in any two repetitions of the pulse sequence, the same value
will be obtained. In other words, the time course of longitudinal and transverse relaxation for each
tissue is the same for each repetition of the pulse sequence. With this in mind, note that the transverse
magnetization created by a given 90° excitation RF pulse is equal to the longitudinal magnetization that
existed just before this pulse. In addition, recall that the MR signal is directly proportional to the
transverse magnetization. Thus, to determine the relative value of the signal intensity for a given tissue,
we can plot the T1 relaxation curve for the time period TR followed immediately by the T2 relaxation
curve for the time period TE, as illustrated in Figure 2-40. The T1 relaxation curve provides the value of
the longitudinal magnetization that is available for any given excitation RF pulse (except the first one) to
convert into transverse magnetization. The T2 relaxation curve indicates the degree to which this
transverse magnetization decays before the echo is acquired. Thus, the final value of the combined T1,
T2 relaxation curve is directly proportional to the signal intensity for the tissue.
page 53
page 54
Add to lightbox
Figure 2-40 Calculation of the tissue signal intensity for a single spin-echo pulse sequence by
combining the associated T1 and T2 relaxation curves. The T1 relaxation curve (gray) shows the
evolution of the longitudinal magnetization following a given excitation RF pulse. (The small
discontinuity in the curve near time zero represents the effect of the refocusing RF pulse.) At time TR,
the subsequent excitation RF pulse converts the available longitudinal magnetization into transverse
magnetization; the magnitude of the transverse magnetization is directly proportional to signal intensity
2 of 8 11-04-2010 23:39
and thus the final value of the T2 relaxation curve gives the relative signal intensity in the image.
Another important factor that influences the basic contrast behavior of MR images is the relative
fraction of signal-producing protons (that is, hydrogen nuclei) in one tissue compared to another. Only
the protons in water and lipids produce an MR signal that can be detected by the commonly used
pulse sequences; the number of these signal-producing protons per unit volume of tissue is called the
spin density or proton density. (The T2 times for signals from the protons in macromolecules are too
short to be detected, but these protons can nonetheless affect image contrast through the
magnetization transfer mechanism discussed in Chapter 7.) For example, gray matter has a proton
density that is roughly 15% larger than that for white matter. The MR signal from a given tissue, and
hence the corresponding image signal intensity, scales directly with the proton density. Typically,
proton-density values are stated as a percentage or fraction of that for cerebrospinal fluid (abbreviated
CSF).
For the interested reader: The signal intensity S for a single-SE pulse sequence can be
written:
where ρ is the proton density. The term in parentheses, just after ρ, corresponds to T1
relaxation and the last term accounts for T2 relaxation. This is the equation that was
used to generate the curves in Figures 2-40 and 2-41. If the effect of the refocusing RF
pulse is neglected, the term in parentheses simplifies to that in parentheses in Equation
2-4 for t = TR.
With these underlying concepts in hand, we will now discuss how relaxation-time-based forms of image
contrast are generated, including proton-density, T1, and T2 weighting for SE imaging, and T2*
weighting for GRE imaging. Three basic principles determine the degree to which proton-density, T1,
and T2 (or T2* for GRE imaging) contribute to the image contrast: 1. the proton-density contribution is
always present; 2. the amount of T1 weighting is controlled by varying the TR; and 3. the amount of T2
weighting (or T2* weighting for GRE) is controlled by varying the TE.
Proton-Density-Weighted Imaging
Add to lightbox
3 of 8 11-04-2010 23:39
Add to lightbox
Add to lightbox
Figure 2-41 Calculated evolutions of the magnetization, and the resulting image signal intensities, for
A, proton-density-weighted (TR/TE = 2500/15 ms), B, T1-weighted (TR/TE = 500/15 ms), and C,
T2-weighted (TR/TE = 2500/90 ms) spin-echo brain imaging. The curves for fat, gray matter (GM),
white matter (WM), and cerebrospinal fluid (CSF) were derived from estimated relaxation times (1.5
tesla) and proton densities. Simulated brain images, based on the calculated signal intensities, are
shown to the right of each plot. For clarity, the time scale for T2 decay is expanded relative to that for
T1 decay.
A proton-density-weighted image is obtained by minimizing the contributions from T1 and T2 relaxation,

thus yielding an image wherein tissues with high proton densities are relatively bright and those with
low proton densities are relatively dark. Differences in signal intensity due to differences in T1 are
minimized by using a TR that is long compared to the T1 values of interest; differences in signal
intensity that arise from differences in T2 are minimized by using the shortest TE permitted by the
pulse sequence. (The minimum TE depends on the durations chosen for the RF pulses and the
data-sampling period, on the capabilities of the scanner's gradient system, and, to some extent, on the
desired spatial resolution.) Thus, in general, proton-density weighting is obtained by using a long TR
and a short TE.
page 54
page 55
4 of 8 11-04-2010 23:39
Add to lightbox
Add to lightbox
5 of 8 11-04-2010 23:39
Add to lightbox
Figure 2-42 The image contrasts typically created using conventional spin-echo imaging. Axial brain
images are shown for T1 (A), proton-density (B), and T2 weighting (C). The image in A was obtained
by using a single-SE sequence (TR/TE 500/15 ms) and those in B and C were obtained by using a
double-SE sequence (TR/TE1/TE2 2500/15/90 ms). (From Mugler JP III: Overview of MR imaging
pulse sequences. Magn Reson Imaging Clin N Am 7:661-697, 1999)
By implementing for brain MRI the concept illustrated in Figure 2-40, Figure 2-41 depicts the calculated
evolutions of the magnetization and the resulting relative image signal intensities for fat, gray matter,
white matter, and CSF based on estimated relaxation times (1.5 tesla) and proton densities. For a TR
of 2500 ms and a TE of 15 ms, values that are typical for proton-density-weighted brain imaging at 1.5
tesla, Figure 2-41A shows that the calculated signal intensities for fat, gray matter, and white matter
correlate with their established proton densities. However, for this TR, the signal intensity for CSF
does not reflect its proton density because the T1 of CSF is approximately 4 seconds. Thus, the image
is proton-density weighted for fat, white matter, and gray matter, but T1 weighted for CSF. Despite
this ambiguity, images produced by using this parameter combination are traditionally called proton-
density weighted and are useful for the detection of various pathologies. Figure 2-42B shows an axial
brain image acquired using these parameter values.
T1-Weighted Imaging
A T1-weighted image is obtained by minimizing contributions from T2 relaxation while emphasizing
those from T1 relaxation, thus yielding an image wherein tissues with short T1 values are relatively
bright and those with long T1 values are relatively dark. As discussed above, differences in signal
intensity due to differences in T2 can be minimized by using the shortest TE permitted by the pulse
sequence. To emphasize differences in signal intensity secondary to differences in T1, a TR in the
vicinity of the T1 values of interest is chosen. The exact value of TR is not critical, and its choice is
often influenced by the number of slices that can be acquired within the TR as well as by the desired
contrast. For example, a TR between 400 and 700 ms is commonly used for T1-weighted brain
imaging at 1.5 tesla; the associated T1 values for white matter and gray matter are approximately 600
and 1000 ms, respectively. In general terms, T1 weighting is obtained by using a short TR and a short
TE.
For a TR of 500 ms and a TE of 15 ms, Figure 2-41B shows that the calculated signal intensities for
fat, white matter, gray matter, and CSF are inversely related to their T1 relaxation times, with fat
having the highest signal and CSF the lowest. Figure 2-42A shows an axial brain image acquired using
these parameter values.
6 of 8 11-04-2010 23:39
T2-Weighted Imaging
In contrast to a T1-weighted image, a T2-weighted image is obtained by minimizing contributions from
T1 relaxation while emphasizing those from T2 relaxation, thus yielding an image wherein tissues with
long T2 values are relatively bright and those with short T2 values are relatively dark. As discussed for
proton-density-weighted imaging, differences in signal intensity that arise from differences in T1 can be
minimized by using a TR that is long compared to the T1 values of interest. To emphasize differences
in signal intensity due to differences in T2, a relatively long TE is chosen. For example, a TE of
approximately 100 ms is commonly used for T2-weighted brain imaging at 1.5 tesla. The associated
T2 values for white matter and gray matter are approximately 90 and 100 ms, respectively, and those
for various pathologies are often substantially longer. The choice of TE represents a trade-off between
increased contrast at longer TE values and increased signal-to-noise ratios at shorter TE values. In
general terms, T2 weighting is obtained by using a long TR and a long TE.
page 55
page 56
For a TR of 2500 ms and a TE of 90 ms, Figure 2-41C shows that the calculated signal intensities for
the four tissues correlate with their T2 relaxation times. Even though this TR is short compared to the
T1 for CSF, its signal intensity ends up as the brightest in the image due to the very long T2 relaxation
time of CSF (approximately 2000 ms). Figure 2-42C shows an axial brain image acquired using these
parameter values. Note that proton-density-weighted and T2-weighted spin-echo images are typically
collected during the same acquisition by using a double-SE pulse sequence as described in Figure
2-36C.
T2*-Weighted Imaging
Add to lightbox
7 of 8 11-04-2010 23:39
Add to lightbox
Figure 2-43 Coronal gradient-echo (TE 4 ms) (A) and spin-echo (TE 36 ms) (B) images of the chest
demonstrating the short T2* value for lung tissue.
Conceptually, T2*-weighted imaging is identical to T2-weighted imaging; contributions from T1

relaxation are minimized while those from T2* are emphasized. The practical difference is that T2*
values are often much shorter than the underlying T2 values. Thus, what is considered a long TE for
T2*-weighted imaging is often much less than that used for T2-weighted imaging. For example, Figure
2-21B illustrates the signal loss that occurs in GRE images in the vicinity of air-tissue interfaces. The
tissue surrounding such an interface effectively has a very short T2* value. (The term "effectively" is
used because in this case the T2* value is not an intrinsic property of the tissue.) The TE for Figure
2-21B was only 12 ms, which, in the context of T2-weighted SE imaging, is a short TE. Figure 2-43
shows an example where the T2* value is an intrinsic property of the tissue. Lung parenchyma has a
very short T2* (~1 ms)48,49 due to microscopic susceptibility-induced field gradients, which arise from
the countless air-tissue interfaces that characterize the lung. Thus, even with a relatively short TE of 4
ms, lung tissue appears black in a GRE image (Fig. 2-43A). In contrast, in the SE image (TE 36 ms)
shown in Figure 2-43B, we can see the lung parenchyma (along with numerous pulmonary blood
50
vessels), demonstrating that the T2 value for lung tissue is much longer than its T2* value.
Nonetheless, the signal intensity for lung tissue in the SE image is low compared to that for other
tissues such as fat because the proton density of lung tissue is relatively low.
Mixed Weighting
The goal of many clinical MRI exams is to detect and characterize a pathologic process. In this
context, it does not matter what type of contrast weighting is used as long as the pathology can be
detected against a background of normal tissue. As a result, it is not uncommon for imaging parameter
combinations to be used that are optimized to detect a certain pathology, but that do not fit into one of
the categories described above. In the nomenclature of our discussion, such a parameter combination
can be considered to yield mixed weighting.
Finally, summarizing the general parameter choices for obtaining relaxation-time-based contrast,
proton-density weighting requires a long TR and short TE, T1 weighting is obtained with a short TR
and short TE, and T2 weighting is achieved with a long TR and long TE. Since T1 values generally
increase with field strength, the TR value appropriate for a particular type of image contrast must be
scaled accordingly.
© 2010 Elsevier
8 of 8 11-04-2010 23:39
CONCLUSION
In this chapter we have outlined the major issues involved in generating the magnetic-resonance signal
and in manipulating the underlying source of this signal, the magnetization, to create images that
convey medically useful information. The topics discussed should provide a solid foundation for
understanding subsequent chapters that deal with MRI physics or advanced clinical applications.
Hopefully, this introductory material has provided a flavor of the tremendous versatility that
characterizes MRI.
page 56
page 57
REFERENCES
1. Lide DR (ed): CRC Handbook of Chemistry and Physics, 84th ed. Boca Raton, Fla: CRC Press, 2003.
2. Slichter CP: Principles of Magnetic Resonance, 2nd ed. Berlin: Springer-Verlag, 1980.
3. Abragam A: The Principles of Nuclear Magnetism. London: Oxford University Press, 1961.
4. Andrew ER: Nuclear Magnetic Resonance. London: Cambridge University Press, 1955.
5. Farrar TC, Becker ED: Pulse and Fourier Transform NMR: Introduction to Theory and Methods. New York: Academic Press,
1971.
6. Bottomley PA, Foster TH, Argersinger RE, et al: A review of normal tissue hydrogen NMR relaxation times and relaxation
mechanisms from 1-100 MHz: Dependence on tissue type, NMR frequency, temperature, species, excision, and age. Med
Phys 11:425-448, 1984.
1
7. Bottomley PA, Hardy CJ, Argersinger RE, et al: A review of H nuclear magnetic resonance relaxation in pathology: Are T1
and T2 diagnostic? Med Phys 14:1-37, 1987.
8. Bydder GM, Pennock JM, Steiner RE, et al: The short TI inversion recovery sequence-an approach to MR imaging of the
abdomen. Magn Reson Imaging 3:251-254, 1985. Medline Similar articles
9. Bydder GM, Young IR: MR imaging: clinical use of the inversion recovery sequence. J Comput Assist Tomogr 9:659-675,
10. Hajnal JV, Bryant DJ, Kasuboski L, et al: Use of fluid attenuated inversion recovery (FLAIR) pulse sequences in MRI of the
brain. J Comput Assist Tomogr 16:841-844, 1992. Medline Similar articles
11. Malisch TW, Hedlund LW, Suddarth SA, et al: MR microscopy at 7.0 T: Effects of brain iron. J Magn Reson Imaging
1:301-305, 1991.
12. Damadian R: Tumor detection by nuclear magnetic resonance. Science 171:1151-1153, 1971. Medline
Similar articles
13. Eggleston JC, Saryan LA, Hollis DP: Nuclear magnetic resonance investigations of human neoplastic and abnormal
nonneoplastic tissues. Cancer Res 35:1326-1332, 1975. Medline Similar articles
14. Bloembergen N: Nuclear magnetic relaxation. New York: WA Benjamin, 1961.
15. Yablonskiy DA, Haacke EM: Theory of NMR signal behavior in magnetically inhomogeneous tissues: The static dephasing
regime. Magn Reson Med 32:749-763, 1994. Medline Similar articles
16. Sukstanskii AL, Yablonskiy DA: Gaussian approximation in the theory of MR signal formation in the presence of structure-
specific magnetic field inhomogeneities. J Magn Reson 163:236-247, 2003. Medline Similar articles
17. Hahn EL: Spin echoes. Phys Rev 80:580-594, 1950.
18. Schenck JF: The role of magnetic susceptibility in magnetic resonance imaging: MRI magnetic compatibility of the first and
second kinds. Med Phys 23:815, 1996. Medline Similar articles
19. Edelman RR, Johnson K, Buxton R, et al: MR of hemorrhage: A new approach. Am J Neuroradiol 7:751-756, 1986.
Medline Similar articles
20. Edelman RR, Mattle HP, Atkinson DJ, et al: Cerebral blood flow: Assessment with dynamic contrast-enhanced
T2*-weighted MR imaging at 1.5T. Radiology 176:211-220, 1990.
21. Ogawa S, Tank DW, Menon R, et al: Intrinsic signal changes accompanying sensory stimulation: Functional brain mapping
with magnetic resonance imaging. Proc Natl Acad Sci USA 89:5951-5955, 1992. Medline Similar articles
22. Posse S, Aue WP: Susceptibility artifacts in spin-echo and gradient-echo imaging. J Magn Reson 88:473-492, 1990.
23. Wehrli FW, Perkins TG, Shimakawa A, et al: Chemical shift-induced amplitude modulations in images obtained with
gradient refocusing. Magn Reson Imaging 5:157-158, 1987. Medline Similar articles
24. Lauterbur PC: Image formation by induced local interactions: Examples employing nuclear magnetic resonance. Nature
242:190-191, 1973.
25. Lauterbur PC, Kramer DM, House WV Jr, et al: Zeugmatographic high resolution nuclear magnetic resonance
spectroscopy. Images of chemical inhomogeneity within macroscopic objects. J Am Chem Soc 97:6866-6868, 1975.
1 of 2 11-04-2010 23:40
26. Mansfield P, Maudsley AA, Bains T: Fast scan proton density imaging by NMR. J Phys E - Sci Instrum 9:271-278, 1976.
27. Kneeland JB, Shimakawa A, Wehrli FW: Effect of intersection spacing on MR image contrast and study time. Radiology
28. Crawley AP, Henkelman RM: A stimulated echo artifact from slice interference in magnetic resonance imaging. Med Phys
29. Kucharczyk W, Crawley AP, Kelly WM, et al: Effect of multislice interference on image contrast in T2- and T1-weighted MR
images. Am J Neuroradiol 9:443-451, 1988. Medline Similar articles
30. Hoult DI: Zeugmatography: A criticism of the concept of a selective pulse in the presence of a field gradient. J Magn Reson
26:165-167, 1977.
31. Margosian P, Schmitt F, Purdy D: Faster MR imaging: Imaging with half the data. Health Care Instrum 1:195-197, 1986.
32. Feinberg DA, Hale JD, Watts JC, et al: Halving MR imaging time by conjugation: Demonstration at 3.5 kG. Radiology
161:527-531, 1986.
33. Brown TR, Kincaid BM, Ugurbil K: NMR chemical shift imaging in three dimensions. Proc Natl Acad Sci USA
34. Ljunggren S: A simple graphical representation of Fourier-based imaging methods. J Magn Reson 54:338-343, 1983.
35. Twieg DB: The k-trajectory formulation of the NMR imaging process with applications in analysis and synthesis of imaging
36. Pattany PM, Phillips JJ, Chiu LC, et al: Motion artifact suppression technique (MAST) for MR imaging. J Comput Assist
Tomogr 11:369-377, 1987. Medline Similar articles
37. Haacke EM, Lenz GW: Improving MR image quality in the presence of motion by using rephasing gradients. Am J
Roentgenol 148:1251-1258, 1987.
38. Mugler JP III, Brookeman JR: Implementation of mixed bandwidth MRI pulse sequences using a single analog lowpass
filter. Magn Reson Imaging 7:487-493, 1989. Medline Similar articles
39. Kramer DM, Schneider JS, Rudin AM, et al: True three-dimensional nuclear magnetic resonance zeugmatographic images
of a human brain. Neuroradiology 21:239-244, 1981. Medline Similar articles
40. Buonanno FS, Pykett IL, Brady TJ, et al: Clinical relevance of two different nuclear magnetic resonance (NMR) approaches
to imaging of a low-grade astrocytoma. J Comput Assist Tomogr 6:529-535, 1982. Medline Similar articles
41. Pykett IL, Buonanno FS, Brady TJ, et al: True three-dimensional nuclear magnetic resonance neuro-imaging in ischemic
stroke: correlation of NMR, X-ray CT and pathology. Stroke 14:173-177, 1983. Medline Similar articles
42. Parker DL, Gullberg GT: Signal-to-noise efficiency in magnetic resonance imaging. Med Phys 17:250-257, 1990.
43. Oshio K, Jolesz FA, Melki PS, et al: T2-weighted thin-section imaging with the multislab three-dimensional RARE
technique. J Magn Reson Imaging 1:695-700, 1991. Medline Similar articles
44. Yuan C, Schmiedl UP, Weinberger E, et al: Three-dimensional fast spin-echo imaging: pulse sequence and in vivo image
evaluation. J Magn Reson Imaging 3:894-899, 1993. Medline Similar articles
45. Crooks LE, Ortendahl DA, Kaufman L, et al: Clinical efficiency of nuclear magnetic resonance imaging. Radiology
46. Axel L: Blood flow effects in magnetic resonance imaging. Am J Roentgenol 143:1157-1166, 1984.
47. Bradley WG Jr, Waluch V: Blood flow: magnetic resonance imaging. Radiology 154:443-450, 1985. Medline
Similar articles
48. Bergin CJ, Glover GH, Pauly JM: Lung parenchyma: magnetic susceptibility in MR imaging. Radiology 180:845-848, 1991.
49. Stock KW, Chen Q, Hatabu H, et al: Magnetic resonance T2* measurements of the normal human lung in vivo with
ultra-short echo times. Magn Reson Imaging 17:997-1000, 1999. Medline Similar articles
50. Shioya S, Christman R, Ailion DC: An in vivo NMR imaging determination of multiexponential Hahn T2 of normal lung.
Magn Reson Med 16:49-56, 1990.
© 2010 Elsevier
2 of 2 11-04-2010 23:40
RACTICAL ONSIDERATIONS AND MAGE PTIMIZATION

Robert R. Edelman
Eugene E. Dunkle
Wei Li
Kraig V. Kissinger
Kathleen Thangaraj
Magnetic resonance imaging (MRI) is a dynamic field of study, continually evolving with an
ever-increasing range of clinical applications. Improvements in both software and hardware have
resulted in enhanced image quality and shorter scanning times, providing healthcare professionals with
information not previously available with this modality. Developments such as multichannel
phased-array coils, MR angiography, functional imaging (fMRI), parallel imaging, ultrafast imaging, and
spectroscopy give the technologist and imaging specialist a number of new and exciting approaches
that place MRI at the forefront of medical imaging. It is imperative, in these rapidly changing times, that
the imaging technologist keep informed of new developments in design and technology and
successfully apply these skills to provide both quality care of patients and diagnostic imaging. Several
regularly updated web sites are available to provide a variety of useful information to the
1-3
technologist.
This chapter serves as an overview of performing an examination in a typical MRI environment (Fig.
3-1). It is organized in the sequence of the patient and technologist experience, from the initial
encounter with the facility, basic safety issues, preparation and positioning on the table, choice of
imaging technique and means for image optimization, and specialized acquisition techniques.
© 2010 Elsevier
1 of 1 11-04-2010 23:41
THE FACILITY
Facility Design
The needs for equipment and facility organization will differ somewhat, depending on an MRI facility's
affiliation with a clinic, a hospital environment, or a freestanding imaging center. Facility location not
withstanding, most MRI centers have a mix of patients consisting of a larger percentage of outpatients
than inpatients. Decisions on size and scope of resources should be made with an understanding of
market demand for the MRI facility location and population served.
page 58
page 59
Add to lightbox
Figure 3-1 Flow chart for patient imaging.
A floor plan common to all facility types has several basic features (Fig. 3-2). The first is a reception
area, where patients register for an appointment and non-MRI healthcare providers may make
inquiries, receive information, and ask for assistance and direction from reception staff. A waiting area
should be located adjacent to the reception desk. This area should be large enough to serve as a
place for patients to complete questionnaires and for their companions to wait comfortably while a
patient undergoes the procedure. Rest rooms are necessary outside the suite for visitors or patients
not yet cleared by MRI staff for entrance into the MRI suite. Additional facilities should be located
within the suite to provide convenience and privacy for gowned patients.
1 of 9 11-04-2010 23:43
Add to lightbox
Figure 3-2 Typical MRI unit floor plan.
Every MRI department should provide a physical barrier between the reception area and the scanner
environment. Warning signs should be posted at or before the entrance to the MRI suite notifying all
persons entering the area of the potentially hazardous magnetic field. These signs should clearly state
that no ferromagnetic object can be brought into the area and that all patients, personnel, and visitors
are required to stop at the reception area and check with department personnel before proceeding.
Entry doors should be installed with an electronic, manual, or combination lock controlled by
appropriate department personnel. An informed staff member, such as a receptionist, should be
located near this physical barrier. The reception staff monitors the passage of patients and personnel
beyond the barrier. In addition, the boundary may aid in preventing unauthorized patients and personnel
from straying into the magnet's fringe field. The reception area should be located at a safe distance
4
from the 5 gauss (5-G) line, the defining line for safety purposes. The major concern of entering a
fringe field beyond the 5-G line is specific to individuals with a pacemaker, because its function may be
altered. This line may be defined by measuring the magnetic force with a magnetometer.
page 59
page 60
The location of the defining line depends on the magnet strength and the type of shielding.
Manufacturers offer actively shielded magnets, which consist of additional superconducting loops of
wire around the magnet. These coils partially negate the magnetic fringe field, bringing the 5-G line
closer to the magnet. Passive shielding is another way to bring the 5-G line closer to the magnet. This
method utilizes large pieces of iron placed in strategic locations around the magnet. A major drawback
with this system is that the quantity of iron needed greatly increases the overall weight of the system.
2 of 9 11-04-2010 23:43
The excess weight limits its placement in certain locations because it is more challenging and costly to
install such heavy equipment several floors above ground level. Permanent magnets tend to have much
more tightly confined fringe fields than superconductive ones.
Receptionists need to monitor pedestrian traffic continually. Communication and team support with
other team members such as technologists and radiologists are essential. Therefore, it is important to
make this area easily accessible to MRI healthcare workers as well. A technologist should be available
to reception staff to assist with scheduling and patient-related questions.
Dressing rooms and lockers are needed, because a patient undergoing MRI will need to remove
personal belongings such as credit cards with a magnetic strip; clothing with zippers, snaps, or hooks;
jewelry; hairpins; and any other metallic objects. A room designated for preparation of patients should
be provided for the following instances: interviewing patients, explanation of procedures, preparation
for intravenous (IV) access for examinations that require IV administration of contrast medium;
administration and pre- and post-examination monitoring of conscious sedation; and a pre-procedure
interview and physical examination of the patient by the radiologist. Ideally, the preparation room is
located close to the MRI examination room, to reduce movement of patients between the two areas.
The MRI examination room should be as esthetically pleasing to patients as possible. It must be
functional for the technologist's use, as well. Certain stringent technical requirements must be met so
the system may perform reliably and produce artifact-free images. Equipment and supplies specifically
designed for the MRI environment should be stored within the MRI examination room, including all
radiofrequency (RF) coils, MR-compatible vital sign monitoring equipment, earplugs, prism glasses,
blindfolds, positioning sponges and sandbags, and injector kits. Other items that are not unique to the
MRI examination requirements but are universal to the hospital or clinic setting should also be stocked
in this room. Examples include clean and dirty linen stores, blankets, antibacterial cleaning fluids,
emesis basins, glucagon, IV supplies, disposable needle containers, contrast agent reaction kits, and
denture cups. A technologist should need to exert minimal effort in gaining access to these items to
perform an examination adequately in a comfortable, clean, and safe environment without delay. A
well-organized examination room is more reassuring to a patient because it is less technically
foreboding. Another advantage is that a technologist familiar with the location of equipment can more
easily provide timely, competent, and professional care.
Some surface coils are awkward and heavy. Therefore, proper storage location is an important
consideration. If coils are stored in an inconvenient location, they may be subject to damage, or worse,
the technologist may suffer injury. Heavy coils should be stored as nearby as possible and at waist
height. There should be no hindrance between the coil storage area and the examination table.
Adequate shelving for coil storage is also important in preventing coil damage. Coils should not be
stacked on top of each other in any way, nor should other equipment be stored on top of the coils, or
vice versa. When planning for shelf space, it is wise to allow for more than what seems needed in the
immediate future to allow for growth of services and changes in technology.
Room decor also plays a role in reducing a patient's level of anxiety. Indirect room lighting has a softer,
less abrasive tone. Plants, skylight windows, or wall hangings depicting relaxing scenery may assist to
focus and calm the anxious patient. Assisting a patient to perceive the environment as spacious and
nonconfining enhances the patient's overall MRI examination experience.
An MR-compatible stereo system provides the patient with relaxing music and better communication
capabilities. Patients reported lower levels of anxiety after the MRI scan when able to listen to music
of choice during the procedure compared with patients who had no music during MRI. 5 It is also helpful
to tell patients ahead of time to bring their own music so they can listen to music of choice. Some
facilities actually create a music library with a wide range of musical choices to choose from. In
addition, stereo systems that provide ear protection are an important consideration, because some
scanning techniques are loud enough to cause temporary hearing loss. 6
7
Room ventilation is an important factor to consider for comfort of patients. Air exchange is impeded
3 of 9 11-04-2010 23:43
by the room's copper shielding, which prevents RF interference from the outside environment. It is
helpful to have a separate dedicated temperature control for the examination room. In addition, most
manufacturers' equipment specifications call for cooler temperatures and higher humidity than in other
areas of the hospital. Collaborating with architects and engineers about such requirements during the
facility planning stages will prevent future problems in meeting the demands on the heat, ventilation,
and air conditioning system, which also controls humidification, required in the MRI environment. The
humidity required is higher than most other environments. A scan room with humidity below 50% will
likely produce unpredictable image artifacts. This point should be stressed with planning engineers
and vendor service engineers because each system's specification may be slightly different and
correcting an inadequate system after it is implemented is costly and disruptive to this environment.
Monitoring Equipment
page 60
page 61
A sedated patient's vital signs should be monitored by reliable equipment. Critical care/intensive care
patients must also be monitored. The nature of the MRI environment makes it difficult, if not
impossible, to have proximity to a patient to check for level of responsiveness. Also, each scan
typically lasts from a few seconds to 10 minutes or more, with complete examination times running 30
to 60 minutes or longer. This is too long to wait to monitor a patient's level of consciousness. The Joint
8
Commission on Accreditation of Healthcare Organizations has recommended a protocol for conscious
sedation. To keep within compliance of these guidelines, the MRI facility must have the capability to
monitor pulse rate, oxygen saturation, respiration, blood pressure, and electrocardiogram (ECG).
MR-compatible monitoring devices come equipped with fiberoptic cables from the equipment to the
patient's contact areas, thus preventing skin burns that might result from monitoring equipment used in
other areas of the hospital. In addition, the use of monitoring equipment not specifically designed for
the MRI environment during image acquisition may result in gross image artifact. Display screens for
the measured vital signs must also be designed for this environment to avoid a distorted ECG and
other displayed information. MR-compatible anesthesia machines and ventilators are safer and more
reliable than standard equipment.9 These are highly recommended for institutions that perform any
general anesthesia during MRI.
Safety
Safety issues are reviewed in depth in Chapter 24, along with sample screening forms. We will only
consider some of the key issues here. There are many potential risks in the MRI environment. The first
concern is related to the high magnetic field strengths used in the clinical and research settings. Strong
magnetic fields are measured in tesla (T), whereas the unit of measure for a small field is gauss (G).
One tesla is equal to 10,000 G. Field strengths of magnets used for clinical purposes vary from 0.2 to
3 T, though systems up to 9.4 T are being used for research purposes in human subjects. Magnetic
field strengths of 8 T or less are considered nonsignificant risk by the US Food and Drug
Administration (FDA) for anyone over the age of 1 month.10-12
A 1.5 T magnet has more than 45,000 times the Earth's magnetic pull. Attention must be given to
keeping all ferromagnetic objects away from the magnet. In general, the force on the object is
proportional to the product of magnetic field strength and the gradient of magnetic field strength or, for
magnetically saturated ferromagnetic materials, just the gradient of magnetic field strength. This
gradient is most severe near the entrance to the magnet bore; therefore, the attraction on a
ferromagnetic object may not be noticeable until it is brought relatively close to the magnet. While
reducing the extent of the fringe field, magnetic shielding actually increases the gradient in the
magnetic field and therefore the potential acceleration of a ferromagnetic object. The risk may be
worse for some high-field short-bore systems.
Personnel should be educated on the types of objects that are potential threats to patients, visitors, or
healthcare workers in the examination room, not to mention potential damage to the system. A variety
of sources, including books and web sites, are available to assist in determining if an object poses any
risk.13 However, in all cases common sense must be used. All metallic objects should be scrutinized
4 of 9 11-04-2010 23:43
carefully regardless of size or location.
If a patient has an implant and the hazard potential in the MRI environment is unknown, the technologist
should postpone the examination until safety can be confirmed. To determine the safety of a metallic
implant, it is best to contact the manufacturer of the device to ascertain if it is safe to scan. Many
patients will have ID cards indicating where to call for information regarding the device. A technologist
may tie a string to a small implant and suspend it while slowly bringing it toward the bore of the
14
magnet as another method for testing magnetic attraction. If the implant in question is pulled to align
with the magnetic field by more than 45° from vertical (i.e., indicating the magnetic attraction is
stronger than the pull of gravity), the implant should be considered a contraindication. An implant or
other device with wires may also be considered a contraindication, because of the possibility of
inducing a current in a wire and causing a burn on a patient. The risk of burns is highest when imaging
sequences are used that are RF intensive, such as fast spin-echo (SE).
Cardiac pacemakers are generally categorized as a contraindication to MRI. One study found that
17% of patients with pacemakers were denied an MRI study in a one year period. 15 Risks arise from
the effects of the magnetic field on pacemaker function, electrical stimulation of the heart, and tissue
16
heating around the pacemaker leads from the RF pulses. Several deaths associated with scanning
pacemaker patients have been reported. In one case, an elderly patient died after he failed to tell the
technologist that he had a pacemaker, despite being asked twice.15 Nonetheless, hundreds of patients
with pacemakers have safely undergone MRI with physician supervision.
MRI should only be considered if the risk-benefit ratio justifies the test. In a recent editorial by
Martin,17 the following recommendations were made for MRI of patients with pacemakers: (i)
Document that a clinically necessary MR study is warranted and obtain informed consent; (ii) maintain
SAR levels below 2 W/kg; (iii) have emergency equipment and Advanced Cardiac Life Support
(ACLS)-trained personnel readily available; (iv) scan only non-pacemaker-dependent patients; (v)
interrogate the pulse generator immediately before and after MRI and reprogram if necessary; (vi)
disable the minute ventilation feature; (vii) maintain voice contact throughout the procedure and
continuously monitor heart rhythm and rate. Pulse oximetry monitoring is not necessary but can be
used concomitantly with rhythm monitoring to provide an additional level of safety; (viii) a physician
adept in pacemaker programming needs to be present during the MRI; (ix) sub-threshold output
programming is reasonable but has not been shown to be necessary if the above guidelines are
followed; (x) scan modern pacemakers (manufactured after 2000).
page 61
page 62
The use of transdermal patches to deliver medication is increasing. Several reports have indicated that
transdermal patches that contain aluminum foil or a similar metallic component may cause excessive
heating or a burn in a patient undergoing an MRI procedure. It is recommended that any patient
wearing a transdermal patch that has a metallic component be identified prior to undergoing MRI. The
patient's physician should be contacted to determine if it is possible to temporarily remove the
medication patch in order to prevent excessive heating. The Institute for Safe Medical Practices
recently stated that medication patches such as Androderm, Transderm-Nitro, Deponit, Nicoderm,
Nicotrol, Catapres-TTS, and possibly others, should be removed prior to an MRI examination. Other
patches to be aware of include the nicotine patch marketed as Habitrol and its "private label"
equivalents and hyoscine bromide, marketed as TrasDerm Scop. Not all medication patches contain a
18
metallic component. Accordingly, these patches do not need to be removed for the MRI examination.
Medical devices that have moving parts, such as a pump, may malfunction in a strong magnetic field
causing potentially serious complications for a patient. Therefore, these devices should be tested for
magnetic attraction as well as malfunction in the MRI environment. IV infusion control pumps are
commonly used for heparin or other medications that require regulated continual infusion rates. If an
MR-compatible pump is not available, the referring physician may opt to disconnect the device for the
duration of the MRI examination and arrange for another method of administering the medication to the
patient in the interim. Alternatively, the medical team can prepare the pump with a sufficiently long
5 of 9 11-04-2010 23:43
piece of extension tubing between the patient and the pump so that the pump mechanism remains
outside the examination room. Under these circumstances, the IV tubing is unusually long and the flow
rate may decrease. This would also require that the examination room door remain open during image
acquisition, thus risking image artifact from outside RF sources. Clearly this is not an ideal approach in
producing high-quality images, yet it may suffice in obtaining a diagnosis. It is desirable to have a
wave-guide between the examination room and control room, so that the IV tubing can go through the
wave-guide without suffering RF interference. The wave-guide can also be used for passage of other
items such as tubing for IV sedation, and projector cables for fMRI.
Some patients are nonambulatory and require assistance with transportation. If the system does not
have a table that can be undocked from the gantry and wheeled out of the MRI room, it may be
necessary to purchase an MRI-compatible wheelchair and stretcher so patients can enter the magnetic
field without risk of injury. In addition, the facility should be equipped with nonferrous IV poles for
patients requiring IV fluids. Many equipment suppliers now provide medical equipment specifically
designed for the MRI environment.
Attention should also be given to the arrangement of the MRI scanner in relationship to the scan
console. It is best for a technologist to have visual contact with a patient in the bore of the magnet to
detect more easily if a patient is in distress or in need of assistance. One scenario for seeing into the
bore is by direct viewing through a window between the control room and the examination room.
Another way to see a patient in the bore is through a video camera at one end of the magnet bore with
an image display located at the scan console. Also, a patient should be given a call button device to
contact a healthcare worker at any time during the scan. This provides a method of communication for
a patient who is not able to communicate verbally. All patients may use this as an emergency call
button, if claustrophobic, anxious, or in pain.
Another concern arises with regard to the cryogens used to cool the superconducting magnets. These
systems use liquid helium and possibly liquid nitrogen. During the catastrophic event known as a
quench, these supercooled liquids vaporize nearly instantaneously. When a quench occurs, it is usually
unmistakable. The cryogen burn-off rate is so rapid that a loud rumbling sound is heard, similar to
thunder. The room air pressure may increase rapidly as well if the quench vent is not properly
designed to handle a large output of gases. Therefore, it could be difficult to open the examination
room door. Technologists should be aware of the potential hazard of a magnet quench or the
possibility of a leaking dewar when the magnet cryostat is being refilled during routine maintenance. In
both situations, these gases replace oxygen and can cause a brisk decline of oxygen levels in the
room. Patients, healthcare workers, or service personnel in the room may lose consciousness rapidly
and unexpectedly.19 Therefore, the scan room should be equipped with sensors to measure oxygen
levels, with an alarm display panel outside of the scan room accessible to maintenance staff. Because
helium rises when released into room air, one sensor should be placed high on a wall. If the
superconducting magnet also uses nitrogen, a second sensor should be placed lower on a wall,
because the cooled nitrogen is heavier than room air and will fall.
In the event of a quench, the MRI examination room should be vacated of all patients and employees
as quickly as possible, and the room should be securely closed until sufficient time has passed to allow
the oxygen levels to return to normal. Room air normally contains approximately 20% oxygen.
MRI facilities should have protocols established as a framework for the MRI team to respond quickly
and skillfully in emergency situations such as a quench. Other crisis situations include medical
emergencies. These situations present additional challenges to staff because most items of equipment
used elsewhere in the hospital for emergency medical care, such as stethoscopes, Kelly clamps,
laryngoscopes, and non-MRI-compatible oxygen tanks become harmful projectiles when introduced
into high magnetic fields. It has been reported that a 6-year-old boy was killed during a routine MRI
procedure when the powerful magnetic field accelerated a non-MRI-compatible oxygen tank into the
center of the magnet bore, crushing the child's head.20,21
Other equipment, such as defibrillators and non-MRI-compatible ECG recorders, may malfunction in a
6 of 9 11-04-2010 23:43
magnetic field. In the event of a medical emergency, the safest environment for implementation of
resuscitation or treatment of unexpected medical complications is outside the 5-G line. The MRI
receptionist should place a call to a code team or emergency medical technician for assistance. Each
healthcare team member in MRI should be trained in basic cardiac life support or advanced cardiac life
support so that resuscitation may begin while a patient is being transported away from the potentially
harmful fringe field of the magnet. The patient should be transported immediately from the MRI
examination room to an area furnished with all necessary emergency supplies and equipment. This
may be in the preparation room or immediately outside the MRI examination room. Cardiopulmonary
resuscitation should continue until the specially trained team summoned for this event responds.
page 62
page 63
Another potential emergency situation may occur if a large ferromagnetic object is brought into an MRI
examination room. If the object were brought close to the magnetic field, it would quite likely become a
projectile mass and move rapidly toward the magnet, striking any person or object in its path. It may
also become bound to the magnet with a person or object pinned between the two. If an object is
pinned to the magnet and cannot safely be removed while the main magnetic field is active, a
technologist may need to call a service technician to ramp down the magnetic field. This can be
performed safely under controlled supervision. However, if a person is constrained by the object to the
magnet, there is a need for rapid response on the part of the MRI team to free the individual and call
for emergency medical care due to the trauma likely incurred by the victim. The fastest action would
be to quench the magnet, although this may cost tens of thousands of dollars worth of cryogens (all
MRI systems have an emergency quench button, and all MRI personnel should know its location),
and remove the victim immediately. This is an extremely dangerous circumstance. Precautions should
be strictly enforced and measures taken seriously by all MRI staff to avoid accidents.
MR Safety for Pregnant Patients and Technical Staff

MRI is believed to be a safe imaging modality because there is no ionizing radiation as is used in
general radiology or computed tomography. However, it is a relatively new method, which means that
long-term effects are yet to be determined. There are two populations to consider: the pregnant
patient and the pregnant healthcare provider.
A patient and a healthcare worker have different exposures to different factors in the MRI environment.
The patient is placed in the center of the bore of the magnet for approximately 45 minutes. She may
also be given IV contrast medium. The potential risks a patient faces are from the static magnetic field,
changing electromagnetic fields, RF, and contrast agents. The patient's exposure is usually a one-time
occurrence during a pregnancy, whereas a healthcare worker is constantly exposed to the static
magnetic field.
It is difficult to determine what effects MRI may have on a pregnant woman and her fetus for several
reasons. First, the spontaneous abortion rate in the normal pregnant population is approximately 30%
during the first trimester. The population of pregnant patients and healthcare providers exposed to MRI
is small. A survey of women of childbearing years working in the MRI environment was conducted in
1990.22 The results show no statistically significant variations in reproductive health or menstrual cycles
of the respondents compared with the general population. Facilities differ on a policy of restriction of
work responsibilities for the pregnant healthcare provider. It has been suggested by Kanal and
colleagues in the aforementioned study that activities should be the same for all employees. The
pregnant healthcare workers may perform examinations, which include entering the MRI examination
room in the absence of scanning and attending to the patient's needs, without concern of harmful
effects.
Because there is no conclusive evidence that undergoing MRI is completely safe, it is wise to act
conservatively in scanning a pregnant patient. Before performing an examination on a pregnant
patient, a discussion should take place between a referring physician and a diagnosing radiologist to
determine the effect the results would have on treatment of the patient and if MRI is the best modality
for the clinical indications. Also, the patient should be made aware that, to date, there has been no
7 of 9 11-04-2010 23:43
indication that the use of clinical MR imaging during pregnancy has produced deleterious effects.
However, as noted by the FDA, the safety of MRI during pregnancy has not been proved.23
Scheduling
Studies are typically scheduled into time slots of between 20 minutes and 1 hour, depending on the
complexity of the study and capabilities of the MR system. Certain examinations, such as pediatric
sedation and fMRI studies, may require even more time. Once the study is requested, the MRI staff
efficiently schedules the study, ensures the appropriateness of the requested examination, and obtains
key information about the patient. The scheduler or receptionist staff begins the process by obtaining
and validating a significant amount of information from the clinician's office, such as clinical history and
reason for examination. If the scheduler is not sure what examination is being requested, he or she
should have the ability to ask either the technologist or the radiologist for assistance. This will help
avoid problems at the time of the patient's appointment. The demographic and insurance information
may be completely or partially procured from the referring clinician's office or from the patient.
Either at the time of the initial appointment or when confirming the appointment time with a patient,
staff should interview patients briefly for a prescreening of metallic implants, pacemakers, or other
electronic implants, claustrophobia, allergies, and whether a patient needs special assistance. Staff
may also give some information to a patient about the length of the visit, how a patient should dress
(e.g., no jewelry or eye make-up), bringing family members, or directions to the facility. This is a good
opportunity for a patient to ask questions on topics that may have been causing some anxiety. A
knowledgeable and skilled receptionist will be able to allay some fears a patient may have or have the
patient talk to a technologist or nurse for technical questions or questions regarding oral and IV
sedation. Some sites may offer sedation, in which case it must be scheduled ahead of time and a
nurse should contact the patient to set up the specifics. At most sites where sedation is not offered, it
is up to the patient to speak to the ordering physician to prescribe oral medication that they can bring
to the examination with them. This should also be mentioned at the time of scheduling so that the
scheduler can tell the patient to arrive at least 45 minutes early to take their medication and bring
someone with them to take them home. Also, if a patient requires special assistance in ambulating, the
appointment time should coincide with adequate staffing.
page 63
page 64
It is quite costly to a facility to have an open time slot if a patient does not arrive for an appointment.
Because there is seldom advance notice, it is nearly impossible to arrange for another patient to come
in for the opening. Therefore, a receptionist or scheduler should call all patients scheduled for the MRI
examination approximately 2 days in advance to confirm the appointment date and time. Speaking to a
patient directly helps to reduce any potential "no-shows."
After the examination is scheduled, the radiologist reviews the indications for the request and decides
on a scanning protocol. The team determines if the examination is scheduled appropriately based on
such factors as the time of day, available staff, and the length of the examination. Most scheduled
examinations will require few alterations when handled by experienced personnel. Applications for MRI
are becoming more complex, and certain examinations require detailed clinical information for the MRI
team to plan properly for the scanning protocol. Continuous flow of communication among all team
members is essential. An adequately prepared team reduces the chance of unanticipated problems,
allowing for smooth execution of the examination.
Staffing
Technical staffing needs vary from site to site. Factors such as hospital-based versus outpatient center
and the type and volume of referrals received help to determine the number of imaging technologists
required for various shifts. A hospital-based facility will refer more inpatients, thereby increasing the
nonambulatory population and patients requiring more medical attention. In a busy hospital setting it is
beneficial to have dedicated nursing assistance. Some of the advantages include improved care of
patients by having a nurse monitor vital signs as needed; ability to administer sedation to a
claustrophobic or physically distressed patient; and ability to start an IV line when technologists are not
8 of 9 11-04-2010 23:43
trained to do so. It is a facility's obligation to provide an inpatient with the same level of care that would
normally be received in the inpatient unit. A patient's needs should be met not only during the MRI
examination but also before and after the examination while in the department. With these
responsibilities removed, the technologist is able to focus on performing the examination in an
organized fashion, quickly, and precisely. This is helpful not only for throughput but also for comfort and
safety of patients, particularly when a patient's condition is unstable.
Regardless of the facility siting, most MRI centers perform more technically challenging examinations
during the weekday, while a radiologist is available to monitor the examination. These examinations
usually require more involved protocols for preparation of patients. For example, to perform a study of
the pelvis or prostate, a radiologist may request that glucagon be prepared for an intramuscular
injection. Many examinations require contrast medium administration; therefore, it may be beneficial to
have an IV line in place before the examination commences. Abdominal and thoracic cavity imaging
usually requires breath-holding by patients. The radiologist may also wish to examine the patient and
possibly place a marker over an area of point tenderness or palpable mass. Other examinations may
require the use of ECG leads for gated acquisitions, and still other examinations may require unusual
positioning of the patient and coil. Each of these steps takes time to execute properly. The MRI
examination and all of the previously mentioned steps must be explained to the patient.
When scanning has concluded, the technologist may have image processing, transfer to archive and/or
a picture archiving and communication system (PACS), and filming yet to complete. Some of this may
begin while the scan is active, but for many monitored examinations a second technologist is required
to accomplish this task. During a typical weekday shift at a busy MRI center, many other
responsibilities and duties arise, as this is the time of day when referring physicians' offices and other
services are open and fully staffed. This time may become easily filled with a wide variety of tasks,
such as filming requests for referring physicians; 3D reconstructions; testing new scanning protocols;
assisting in scheduling of future or emergent examinations; and speaking to patients or referring
physicians about examination-related questions. Therefore, two or three technologists should be
available to manage a fully scheduled MRI system on a weekday shift. Facilities that have two MRI
systems within proximity to each other may require five technologists between the two systems.
Technical staffing for evening, night, and weekend shifts will vary. The majority of the examinations
performed on these shifts are unmonitored by a physician, resulting in more straightforward
examination protocols. Because examination length is more predictable during these time periods,
appointment times may be scheduled with more regularity and closer together, thereby increasing
examination throughput. However, because the technical staff is always responsible for final screening
of the patient for contraindications and for explaining the procedure as well as for filming, image
processing, and image archival, if increased examination throughput is expected, then technical staffing
should be maintained at a level of two technologists per shift. One technologist working alone will not
likely be able to perform more than one routine outpatient examination per hour.
Provision should be made on all shifts for the likely possibility of urgent or emergent add-on
examinations to the daily schedule of patients. Openings throughout the day may also be used as a
buffer for times when the actual schedule is delayed compared with the planned schedule. This may
happen for a variety of reasons, such as when a patient arrives late for an appointment; a patient
unexpectedly experiences a claustrophobic or anxiety attack; or a monitored examination requires
additional series of images for diagnosis.
© 2010 Elsevier
9 of 9 11-04-2010 23:43
PREPARING A PATIENT FOR MRI

Screening the Patient
page 64
page 65
When a patient arrives for the scheduled MRI examination, a receptionist should instruct the patient to
complete a detailed screening checklist. The checklist is designed specifically for detecting potentially
dangerous ferromagnetic implants or objects outside the body that could also be considered potentially
harmful to the patient or anyone else in the area. This checklist should also screen for metallic objects
that may degrade image quality, such as dentures or braces. Women of childbearing age should be
asked if they are pregnant. It is necessary to document the screening process to ensure consistency in
quality of care. Before bringing the patient into the fringe field, a technologist should carefully review
the completed checklist and obtain verbal confirmation directly from the patient as well. Repetition in
questioning the patient is often necessary because a patient may have either skimmed through the
forms or misunderstood a specific question. Literature has shown that incomplete or incorrect
information has resulted in serious consequences for the patient and staff.24,25 Reception staff also
needs to receive visitors and non-MRI healthcare workers and possibly direct them into areas near the
magnetic field. For this reason reception staff needs to be well versed on the MRI environment and
areas to avoid physical harm or equipment damage.
The screening should also ask for a patient's weight. This is needed for several reasons. First, all MRI
systems have a weight limit defined by the manufacturer for each model to avoid damage to the
mechanics that move the patient table. Second, when a patient is exposed to RF, the body tissue is
prone to warming. The patient's weight determines the allowable RF deposition and must be entered
accurately into the MR system at the beginning of the examination. Third, contrast material dosage is
based on bodyweight, as recommended by the manufacturer. For a cardiac examination, a patient's
height is also needed for calculation of ventricular function.
Education of the Patient

As with any medical procedure, providing the patient with information about the examination is
essential. In doing so, the healthcare provider must address the patient's needs and anxieties.
According to Devine and Cook26,27 three psychoeducational interventional domains should be
addressed: procedural, sensory, and psychosocial. The healthcare provider must be knowledgeable in
the subject matter to inform the patient with clear and concise communication.
Addressing the procedural aspect of education of patients entails informing a patient what the
procedure is, why it is being performed, and that the examination is not painful. Technical staff should
educate a patient about what is required of the patient, such as the specific positioning required for an
examination. In some examinations, as in abdominal imaging, a patient may also need to follow
breath-hold instructions, preferably explained before the patient is placed into the bore. A patient
should also know that the technologist will be in constant communication, giving information on scan
lengths and periodic verification of the patient's comfort.
The sensory aspect of educating the patient includes information on the physical environment of the
machine and the bore size. Some patients consider it confining. Additional sensory information should
include the noise associated with the scanning and that it may be considered annoying. Also, a patient
should be aware of the expected duration of the procedure.
28
Studies have shown that patients may experience high levels of anxiety while undergoing MRI. The
technologist should ascertain specific concerns, using open-ended questions to address the sensory
and psychosocial aspects. It may be necessary to use prompts when trying to determine the source of
the anxiety. Adverse psychologic reactions to MRI are often lumped together as claustrophobia.
However, research has shown that anxiety may also result from other factors.29 A patient may be
concerned about the pending diagnosis and the availability of the results. In addition to physical
discomfort, noise from the scanner may induce emotional distress. Another common factor may be
1 of 4 11-04-2010 23:45
that a patient feels a lack of control. A patient may have already experienced significant testing or
other medical procedures and may be overwhelmed by yet another test about which he or she knows
little. The MRI environment may be even more distressing owing to a feeling of confinement and
restrictiveness. Once the patient's concerns are discovered, they must be addressed before the
patient can reliably comply with the requirements of the procedure.
The technologist should direct efforts to a patient's physical and emotional comfort by asking for
suggestions from the patient. The technologist may further assist the patient in alleviating anxieties by
using methods described in the section on claustrophobia. Employing psychological methods in addition
to education has shown a reduction in anxiety levels, compared with giving procedural information
only.30 Once a patient's physical and psychological needs are satisfied, the healthcare provider may
focus on performing the MRI.
When the patient is in the examination room, certain important information and instructions should be
reviewed with the patient for reinforcement. Examples include the importance of remaining motionless
during image acquisition; methods of communication available to the patient to reach the technologist;
reassurance of the technologist's immediate availability; expected examination duration; and, when
applicable, a review of breath-hold instructions with a practice session. In short, both the technologist
and patient benefit when the healthcare worker enlists proactive methods in preparation for the MRI
examination.
Physical Preparation of the Patient

page 65
page 66
Patients should be asked to change from their personal clothing into a gown and robe. Patients who
are allowed to wear personal clothing during a scan have been known to harbor ferromagnetic objects
in the examination room. Attempts to remove the object free it from restraint, and it then becomes a
projectile mass. This is not only a threat to the physical safety of patients and staff but also a means of
potential damage to the MRI scanner. A ferromagnetic object could hide in the magnet and may go
undetected for a time, resulting in image degradation. Additionally, bra straps commonly contain metal
hooks that cause severe artifacts and also may interfere with proper shimming.
If a protocol includes IV contrast medium administration, a technologist, nurse, or physician should

consider establishing IV access before the examination while the patient is in a preparation room. By
starting the IV line in advance, some potential problems may be avoided. For instance, starting an IV
line can be time-consuming, depending on the condition of a patient's peripheral veins. It places
additional emotional and physical stress on a patient who may already be experiencing some duress.
Another potential problem occurs when a patient is taken out of the bore to obtain IV access,
potentially decreasing the likelihood of obtaining images at the same anatomic location before and
after contrast medium administration. Last, if a healthcare provider, such as an IV nurse, attempting to
gain IV access is not familiar with the MRI environment, additional delays and potential safety issues
may arise.
Before starting the scan, several details should be considered. A marker may be used to aid in
31,32
definition of a mass or as a reference point in relation to anatomic structures. The best materials
to use for a marker in MRI are those that have a bright signal on a T1-weighted sequence. Because
fat appears bright on a T1-weighted scan, many markers are made of a substance that contains
animal or vegetable oil. Some examples are cod liver oil, soybean oil, and various nonroasted nuts
(e.g., almonds). Vitamin E capsules are also used. It is helpful to use more than one marker. Placing
two or three markers side by side may save scanning time because acquired images often have an
interslice gap and may not well visualize a single marker.
When the structures in the abdomen and pelvis are imaged, motion of the small intestine can be a
significant detractor in image quality. It may be reasonable for the radiologist to give a patient an
intramuscular injection of glucagon (1 mg) to slow peristalsis, thereby reducing motion artifact.
Alternatively, glucagon may be given IV shortly before the most critical images are acquired, because
2 of 4 11-04-2010 23:45
the effects of an intramuscular injection given before initial positioning of the patient might not last for a
sufficiently long time. Contraindications include patients with a known hypersensitivity to glucagon or
history of pheochromocytoma or insulinoma.33 Healthcare workers should exercise caution if a patient
has a medical history of diabetes. Some patients have experienced a short duration of slight nausea up
to 6 hours after completion of the examination. The patient should be informed of this so as not to be
alarmed. If the nausea continues, the patient should contact his or her referring physician or the
radiologist supervising the MRI examination.
Claustrophobia
Due to the structure of the MRI machine, claustrophobic reactions severe enough to cancel or
postpone the study occur in approximately 5% of patients.34,35 Of the patients attempting to undergo a
scan, as many as 65% experience some level of anxiety or discomfort. Some manufacturers have
designed systems that have a magnet architecture that is shorter or more open. These systems are
less threatening for those affected with mild claustrophobia.
page 66
page 67
Some suggestions for the MRI technologist to assist claustrophobic or anxious patients through an MRI
examination are listed here:
1. Reassure patients that they will not be kept in the machine against their will. The greatest fear
patients may have is a feeling of loss of control.36
2. Maintain close, two-way verbal contact with patients throughout the examination, so they do not
feel abandoned.
3. Give patients a device that allows them to establish contact at any time, such as a panic button.
4. Whenever possible, try to have patients enter the bore of the magnet feet first, thus giving them
the sensation of being farther out of the magnet.
5. Offer the patients a pair of prism glasses, which may be used to look out of the machine
through a reflection in the prism.
6. Give patients a blindfold to cover the eyes so they are less aware of the environment.
7. Guide patients through relaxation techniques such as deep breathing exercises immediately
37
before the scan or guided imagery before and during the scan.
8. Give patients either headphones designed specifically for use in the MRI system to play relaxing
music or earplugs to reduce the noise level they encounter during imaging.
9. Ask a family member to accompany the patient through the examination, touching and talking to
the patient as permitted through the constraints of examination execution. Patients usually find it
reassuring to have physical contact with the environment outside of the "tube" of the magnet.
10. Provide good ventilation to patients while in the bore of the magnet.
11. Advise patients to try to "condition" themselves to the MRI environment by resting in a quiet
environment at home and practice lying down with a large box over their head for a few minutes
at a time.
For obese or physically extra large patients, claustrophobia reaction happens even more often. This is
because the feeling of confinement and restrictiveness is more serious for them than for average-sized
patients. For these patients, in addition to the suggestions listed above, some other means may need
to apply. Increasing the space for the patient in the magnet can be obtained by changing the table
cushion to a thinner one (e.g., a blanket) or the type of the coil (e.g., torso coil to body coil) when it is
acceptable for the type of the study being performed. Asking patients to place arms over the head
rather than rest them at the sides is also helpful. Although the amount of space increase with these
methods is limited, it can be a key to successfully performing the study. Sometimes applying a body
position different from routine use but more comfortable for the patient can help patients stay longer in
the magnet. In body imaging, for instance, patients with a protuberant abdomen may feel more
comfortable in the lateral position than the routinely used supine position, because patients in lateral
position often feel less pressure to their chest so that it is easier to breathe. Providing 2 to 3 L/min of
100% oxygen will reduce the feeling of breathing difficulty, and help patients hold their breath longer. A
technologist should acquire the most critical pulse sequences at the beginning of the examination in
3 of 4 11-04-2010 23:45
case the patient cannot tolerate the full study.
If these techniques fail, the patient may need oral or IV sedation.
Sedation
A patient could receive sedation for several reasons including: 1. prior history of uncompensated
claustrophobia when undergoing MRI; 2. previous reaction of claustrophobia in unrelated situations
(e.g., cannot enter an elevator); 3. acute or chronic pain rendering proper positioning intolerable for the
duration of the procedure; 4. infant and pediatric patients; and 5. mental incapacitation.
Oral sedation is usually given for mild claustrophobia. The referring physician usually prescribes oral
medication for a claustrophobic patient. Instructions may be given to the patient about when to take a
prescribed oral sedative by either the prescribing physician or a pharmacist. Arrival time of the patient
and the actual examination time do not necessarily coincide, owing to examination-related paperwork,
removal of worn metallic objects, obtaining IV access for contrast medium administration, or a
pre-procedure interview with the radiologist for examination-related medical history. Therefore, the
MRI staff may wish to advise the patient on the timing of ingestion of the prescribed oral medication.
It is important for the patient to bring a responsible adult with them to drive them home after the
procedure.
IV sedation may be arranged for a severely claustrophobic or anxious patient. If IV sedation is

planned, it is necessary to obtain a more extensive medical history for a history of respiratory
problems. It is also necessary to monitor a patient's vital signs during the MRI examination. Due to the
strong magnetic field, MRI-compatible equipment is needed to carry out the monitoring. When IV
sedation is needed, the examination should be scheduled with a nurse or physician who can administer
IV sedation and monitor a patient's response. A patient's vital signs should be recorded before the
administration of IV sedatives for a baseline comparison. Vital signs should be continually monitored
throughout the procedure and for a period after the examination is completed, ensuring vital signs have
returned to baseline. This may take 2 to 3 hours. A patient scheduled for IV sedation should be
instructed not to eat for 8 hours before the examination, and a responsible adult should accompany a
patient to the procedure and drive the patient home after being released. The effects of the medication
may inhibit motor skills for several hours after the sedation has been given. Therefore, a healthcare
worker should advise the patient that it is unsafe for the patient to drive or go home alone.
© 2010 Elsevier
4 of 4 11-04-2010 23:45
SELECTING AN IMAGING PROTOCOL

Outline of Relevant Imaging Principles
Several steps are involved in the production of an MR image. Each one relates to the choice of the hardware and
software used during the imaging procedure. The technologist will need to choose the most appropriate RF coil;
select a pulse sequence; set the imaging parameters such as TR, TE, flip angle (for GRE sequences), sampling
bandwidth, echo train length and echo spacing (for fast SE sequences); geometric factors such as slice thickness,
matrix size, field-of-view, and orientation of the phase- or frequency-encoding gradient; along with a variety of
other imaging options such as ECG gating, partial Fourier, parallel imaging, phase oversampling, and image
filtering.
page 67
page 68
We will briefly outline how the underlying MR physics relate to the work done by the technologist in preparing the
patient and optimizing the data acquisition, and then go into more specific details for the remainder of the chapter.
1. Randomly oriented tissue nuclei are aligned by a powerful, uniform magnetic field, producing equilibrium
"magnetization" of the tissue.
The magnetic axes of the tissue protons align with the main magnetic field within a few seconds after the
patient is placed within the magnet bore. By the time the imaging procedure starts, the tissue spins are fully
equilibrated.
2. Properly tuned RF pulses then disrupt the magnetization. As the nuclei recover ("relax") to equilibrium
after the application of the RF pulses, they produce RF signals that are proportional to the magnitude of
the initial alignment. Tissue contrast (i.e., differences in signal) develops as a result of the different rates
at which nuclei relax with the magnetic field.
The technologist chooses a pulse sequence in order to generate measurable MRI signals from the region of
interest. The pulse sequence is repeated at an interval equal to the repetition time (TR), which largely
determines the amount of T1 weighting. The choice of pulse sequence helps to ensure that one obtains
appropriate tissue contrast (i.e., differences in tissue signal intensity, which permit tissue characterization
and lesion detection). Typical pulse sequences include T1-weighted gradient-echo (GRE), T2-weighted fast
spin-echo (SE), and diffusion-sensitive echo-planar imaging (EPI). Additionally, the tissue signals may be
enhanced by the administration of a paramagnetic contrast agent.
3. Positions of the nuclei emitting the MR signals are localized during this process by purposely distorting
the magnetic field with spatially dependent magnetic fields, called gradients. The gradients encode spatial
information into the amplitudes, frequencies, and phases of the MR signals.
As addressed in Chapter 2, the magnetic field gradients are the main mechanism for spatial localization.
The area under the gradient waveform determines spatial resolution. Generally speaking, the technologist
does not directly choose the gradient amplitude (measured in milliTesla/meter) and slew rate (rate of
change in gradient amplitude from zero to maximum or vice versa, measured in milliTesla/meter/millisecond
or tesla/meter/second). Instead, the amplitude and slew rate are automatically determined from parameters
set by the technologist including the pulse sequence, slice thickness, field-of-view, matrix, and sampling
bandwidth.
Of note, some MR systems permit the use of stronger and faster gradients in a special "research" mode,
as compared with the standard clinical mode. In systems with "twin" gradients or insert gradient coils, the
gradient capabilities also depend on which gradient mode is selected.
4. MR signals are measured, or read out, after a user-determined time has elapsed from the initial RF
excitation. The computer transforms the signal into an image using a mathematical process called the
Fourier transform (FT).
The time between the RF excitation of the tissue protons and signal readout is the echo time (TE). Echo
time largely determines the amount of T2 contrast in the image. In reality, the MR signal is not read out
instantaneously, but rather over a period of time lasting at least a few milliseconds. To be precise, the TE is
actually the time between the center of the RF excitation and the point at which the signal intensity is
maximal. This time point may occur at the middle of the sampling period, in which case the echo is said to
be symmetrical or "full" (generally used for SE and balanced steady-state free precession [SSFP]
sequences). Alternatively, the echo may deliberately be shifted to occur earlier in the sampling period, in
which case the echo is said to be asymmetric or "partial" (often used with MR angiography to minimize
flow-related artifacts).
The readout period over which the MR signal is measured, along with the desired in-plane spatial resolution,
determines the sampling bandwidth. The bandwidth is given in hertz (cycles per second, abbreviated as
Hz) or hertz/pixel and is inversely related to the duration of the readout period. In some circumstances, the
use of a lower (or equivalently narrower) bandwidth improves image quality, since the SNR is inversely
proportional to the square root of the sampling bandwidth. There are drawbacks to the use of lower
1 of 22 11-04-2010 23:46
bandwidths, such as certain artifacts, that will be considered later on.
Selecting a Pulse Sequence

The clinical chapters in this text address the specific imaging protocols for each organ system. In this section, we
will now consider the general principles underlying the choice of imaging technique and parameters.
The basic structure of a pulse sequence consists of RF pulses used to tip the magnetization and generate a signal,
and gradients used to spatially localize the signal. As reviewed in Chapter 5, pulse sequences do not measure the
free induction decay (FID) that is produced immediately after an RF pulse; instead they produce and measure an
"echo" called a gradient-echo or gradient-recalled echo (GRE) or spin-echo (SE) depending on the type of pulse
sequence. This echo signal decays away like T2* for GRE, and like T2 for SE.
Commonly, additional RF pulses are applied before the standard pulse sequences in order to alter the longitudinal
magnetization of the tissue protons and hence image contrast; this process is called "magnetization preparation."
For example, a 180° preparation is used for inversion recovery (IR) sequences. Other clinically relevant types of
magnetization preparation include chemical shift-selective fat suppression, spatial presaturation, and magnetization
transfer.
A multitude of pulse sequences and magnetization preparations are available on MR scanners. A summary of the
acronyms for various imaging techniques and vendors is given in Table 3-1. In selecting the most appropriate
imaging techniques, the technologist and radiologist or other imaging specialists need to consider the anatomy of
interest, desired tissue contrast, and level of spatial and temporal resolution. For instance, echo-planar is
appropriate for diffusion-weighted imaging of the brain, but would not be appropriate for imaging of the knee or
abdomen because of its sensitivity to susceptibility artifact and poor spatial resolution. A fast SE sequence would
be ideal for brain imaging, but would be too slow and have inappropriate contrast for MR angiography.
Spin-Echo
Spin-echo imaging used to be the "bread and butter" of pulse sequences, but has since been largely supplanted by
faster GRE and fast SE sequences. In its most basic form and as elaborated in Chapters 2 and 5, the SE pulse
sequence consists of two RF pulses, 90° and 180°, separated in time by equal intervals of TE/2:
The 90° RF pulse tips the longitudinal magnetization into the transverse plane. The 180° RF pulse refocuses the
transverse magnetization so that dephasing effects resulting from static magnetic field inhomogeneities, caused by
the magnet or local differences in magnetic susceptibility, are canceled when the echo peaks at time TE. As a
result, an image acquired with long TR and long TE is T2 weighted rather than T2* weighted.
Spin-echo scans are usually acquired as sets of 2D slices. The maximal number of slices that can be acquired
practically depends on several variables and can be summarized by
where ∆ is a factor that is related to the pulse sequence structure and performance constraints of the gradients,
RF, and measurement systems. For a given TR, more slices can be acquired if TE is short rather than long and if
a higher sampling bandwidth is applied.
page 68
page 69
Table 3-1. Acronyms for Pulse Sequences and Options Used by the Various
Manufacturers*
A. Pulse Sequences
Sequence GE Philips Siemens Picker Elscint Hitachi Shima
Spin-echo MEMP,VEMP Spin-Echo Spin-Echo Spin-Echo Spin-Echo Spin-Echo Spin-E
Fast spin-echo FSE TSE TSE FSE FSE
Single-shot SSFSE Single Shot HASTE EXPRESS
technique TSE
Coherent GRASS, GRE, FFE FISP, ROAST FAST F SHORT GFEC SSFP
gradient-echo FGR, FMPGR
Incoherent SPGR, T1 FFE RF spoiled GE/GFE STAG
gradient-echo FSPGR FAST T1W
(RF spoiled)
2 of 22 11-04-2010 23:46
Incoherent MPGR FLASH T1-FAST, SHORT GRE STAG

gradient-echo NOSE
(gradient
spoiled)
Contrast- SSFP, DE T2 FFE PSIF CE FAST, E SHORT GFEC STER
enhanced FGR FADE Contrast
gradient-echo
sequence
Balanced FIESTA, Balanced-FFE True FISP SARGE, STER
coherent SSFP BASG
gradient-echo
Ultrafast FAST, TFE Turbo FLASH, RAM FAST V-SHORT, RS, SPGR, SMAS
gradient-echo GRASS, 3D MP RAGE Turbo-SHORT FAST SPGR
SPGR (IR/DE
prep), IR FGR
Gradient and GRACE GRACE GSE GSE
spin-echo
Inversion MPIR,TIR IR, IR-TSE, IR,TIR IR IR IR IR
recovery IR-TFE
Short T1 STIR STIR STIR STIR STIR STIR STIR
inversion
recovery
Phase-contrast Phase Phase Phase VENC VENC
sequence Contrast Contrast Contrast
Parallel ASSET SENSE IPAT
imaging
technique
B. Options
Sequence GE Philips Siemens Picker Elscint Hitachi Shim
Signal NEX NSA AC NSA NSA
averaging
Partial Fractional NEX Half Scan Half Fourier Phase Single Side Half Fourier
averaging Conjugate Encoding
Symmetry
Partial echo Fractional Partial Echo Asymmetric Read Single Side Half E
Echo Echo Conjugate View
Symmetry
Rectangular RFOV RFOV HFI HFI (under RFOV RFOV
field-of-view sampling)
Off-center Off Center Off Center Shift, Offset FOV Offset Off Center
shifting slices FOV Shift FOV
Spacing Spacing Slice Gap Distance Gap Slice Interval
between slices Factor in %
Presaturation Spatial SAT REST SAT PRE-SAT Spatial SAT SAT
PRE-SAT
Fat saturation FAT SAT, SPIR, SPAIR, FAT SAT FAT SAT FAT SAT
CHEM SAT WaterSEL
Moving Walking SAT Travel REST Travel SAT Walking PSAT
saturation
pulse
Gradient FC FC GMR MAST STILL GR SMAR
moment
rephasing
Respiratory Respiratory PEAR, Respiratory Respiratory FREEZE Phase
compensation Compensation, Respiratory Gated Gating, Reordering,
Respiratory Trigger PRIZE MAR
3 of 22 11-04-2010 23:46
Triggering
ECG Cardiac ECG ECG ECG ECG ECG, FPG ECG
synchronization Gated, Triggered Triggered Gated
Triggering
Delay after R Trigger Delay Trigger Delay Trigger Delay Trigger Delay Time
wave Window
Automatic Smart Prep Bolus Track Care Bolus
bolus detection
Number of ETL TSE TF ETL, Turbo ETL ETL
echoes Factor
Time between Echo Spacing Echo Spacing Echo Spacing IES Echo Train
echoes Internal
Oversampling Always On Always On Over-sampling Anti Aliasing Anti Aliasing Frequency
in frequency Over-sampling
direction
Oversampling No Phase Fold Over Over-sampling Over-sampling Anti Aliasing Anti Wrap
in phase Wrap Suppression
direction
Bandwidth Received Water/Fat Bandwidth BW Bandwidth
Bandwidth Shift
Variable VB Optimized Optimized Variable BW Variable BW
bandwidth Water/Fat Bandwidth
Shift
Segmented Views per Views, Lines, PG
k-space data Segment Segments Segments
acquisition
Multislice Multi Slice Multiple Slice Multi Slice 2D 2D
imaging
3D Imaging 3D 3D 3D Volume 3D 3D 3D 3D
Orientation Localizer Plan Scan, Localizer, Scout Scanogram
scan Survey Scout
page 70
page 71
*Adapted with permission from the Magnetic Resonance-Technology Information Portal.
4 of 22 11-04-2010 23:46
Add to lightbox
Add to lightbox
Figure 3-3 Spin-echo images acquired at 1.0 T in a patient with periventricular multiple sclerosis plaques. A,
TR/TE = 2500/30. B, TR/TE = 2500/80. Note that the plaques appear brighter than cerebrospinal fluid on the
proton density-weighted image but are seen less clearly on the T2-weighted image because of isointensity with
cerebrospinal fluid.
In an SE pulse sequence, additional 180° RF pulses can be used to generate multiple echoes, with little or no
increase in scanning time. Commonly, two echoes are acquired resulting in proton density (early echo) and
5 of 22 11-04-2010 23:46
T2-weighted (late echo) images. For example, a tissue such as liver, which appears moderately bright with an
early echo but dark with later echoes, can be inferred to have a short T2. Conversely, a tissue such as
cerebrospinal fluid, which appears bright with late as well as early echoes, must have a long T2. Images obtained
with a long TE can help discriminate, for instance, cystic from solid components of a tumor. On the other hand, the
bright signal from cerebrospinal fluid on a long TE image may obscure lesions near the ventricles, which are better
shown on proton density-weighted (Fig. 3-3) or FLAIR images.
The appearance of various tissues with T1- and T2-weighted pulse sequences is summarized in Table 3-2. On
T1-weighted images, tissues with short T1, such as fat, appear bright, whereas tissues with long T1, such as
tumor and edema, appear dark. On T2-weighted images, tissues with long T2, such as tumor, edema, and cyst,
appear bright, whereas tissues with short T2, such as muscle, tendon, and liver, appear dark. On proton density-
weighted images, tissues with increased proton density such as CSF appear moderately bright. Both T1- and
T2-weighted images are always partly weighted toward proton density as well, although the weighting is usually
subtle compared with the weighting toward relaxation time.
page 71
page 72
Spin-echo images have contrast that is manipulated primarily through adjusting TR and TE (Fig. 3-4) as follows:
1. Images acquired with short TR (TR approximates to T1) and short TE (TE is much less than T2) are T1
weighted (Fig. 3-5A).
2. Images acquired with long TR (TR is much greater than T1) and short TE (TE is less than T2) are proton
density weighted or balanced (Fig. 3-5B).
3. Images acquired with long TR and long TE (TE approximates to T2) are T2 weighted (Fig. 3-5C).
To produce a T1-weighted image with a SE sequence, the TR typically ranges between 300 and 800 ms (Fig. 3-6)
and the TE ranges between 5 and 40 ms. Because the TR is one of the factors determining scan time, the
technologist should try to minimize the TR as much as possible within the constraints of having to accommodate an
adequate number of slices to span the region of interest. If an insufficient number of slices can be obtained within
the TR, one can attempt to use a slightly longer TR or fewer slices. Alternatively, one can use a higher sampling
bandwidth and shorter TE, which will allow more slices. As a last resort, one can simply concatenate a second
measurement for the remaining slices but at the expense of additional scan time.
Table 3-2. Appearances of Tissue on MR Images

Tissue T1-Weighted Image T2-Weighted Image
Fat* Very bright Intermediate to dark
Cysts
Watery fluid Very dark Very bright
Proteinaceous fluid Intermediate to bright Very bright
Brain
White matter Bright Dark
Gray matter Dark Bright
Cerebrospinal fluid Very dark Very bright
Bone marrow
Yellow* Very bright Intermediate to dark
Red† Intermediate Dark
Cortical bone Very dark Very dark
Cartilage
Fibrocartilage Very dark Very dark
Tear of fibrocartilage‡ Intermediate Intermediate to bright
Hyaline‡ Intermediate Intermediate
Intervertebral disk
Normal‡ Intermediate Bright
Degenerated Intermediate to dark Dark
Osteophyte
6 of 22 11-04-2010 23:46
Marrow containing Bright Intermediate to dark

Calcified only Dark Dark
Tendons/ligaments§ Very dark Very dark
Muscle Dark Dark
Lung[Verbar] Very dark Very dark
Liver
Normal parenchyma Bright Dark
Metastasis¶ Dark Usually bright, may have dark center
Hemangioma Dark Bright
Fatty infiltration† Bright Dark
Iron overload Intermediate to dark Very dark
Pancreas Bright Dark
Spleen Dark Bright
Contrast-enhanced tissue
Gadolinium chelate
Low concentration Very bright Bright
High concentration Intermediate to dark Very dark
SPIO Dark Very dark
Ultrasmall SPIO Bright Very dark
Fluosol Very dark Very dark
Hematoma**
Hyperacute (<6 h) Intermediate Intermediate
Acute (6-24 h) Intermediate to dark Dark
Subacute (1 d-1 mo) Bright rim Bright
Chronic (>1 mo) Dark rim ± bright center Dark rim ± bright center
*Bright on T2-weighted images acquired using fast spin-echo sequence with short interecho interval.
†Dark on out-of-phase image.
‡Bright on proton density-weighted image.
§Increased signal, particularly on proton density-weighted images, when tissue is oriented at 55° to B 0 (magic
angle effect).
[Verbar]Increased parenchymal signal with TE <2 ms or fast spin-echo sequence with very short interecho interval.
¶Hepatocellular carcinoma, melanoma, hemorrhagic lesion may appear bright on T1-weighted image.
**Low signal because of magnetic susceptibility effects predominantly seen on high-field images, more
pronounced on gradient-echo than spin-echo images. Time frames indicated for hematoma evolution are
approximations; in reality the time frames may vary widely from the values shown.
For T1 weighting, the TE should be kept to a minimum to reduce any T2 weighting in the image. This tactic may
not be possible for low-field imaging, where low-bandwidth sequences are used to maximize the SNR. If TE is
substantially lengthened due to the use of a very low sampling bandwidth, then there may be a loss of lesion
conspicuity due to a counterbalancing of T1 and T2 weighting (Fig. 3-7).
page 72
page 73
7 of 22 11-04-2010 23:46
Add to lightbox
Figure 3-4 Combined impact of TR and TE on tissue contrast. A, For a short TR of 500 ms, tissue contrast is
primarily T1-dependent. The effect of lengthening TE is to introduce T2-weighting and diminish overall tissue
contrast. For instance, in this example, tissue contrast disappears at a TE of approximately 35 ms despite good
contrast at shorter TE. B, For a longer TR of 2000 ms, tissue contrast is primarily proton-density at short TE or
T2-dependent at long TE.
Add to lightbox
8 of 22 11-04-2010 23:46
Add to lightbox
Add to lightbox
page 73
page 74
Figure 3-5 Spin-echo images of a glioblastoma of the right cerebral hemisphere, acquired at 1.5 T. A,
T2-weighted image. TR/TE = 500/20. B, Proton density-weighted image. TR/TE = 2500/20. C, T2-weighted
image. TR/TE = 2500/80. Note signal changes of fat, cerebrospinal fluid, normal brain, cystic tumor, and
surrounding edema. Also note that signal is absent from the cortical bone in the inner and outer tables of the skull
but that signal is present in the marrow-containing diploic space.
9 of 22 11-04-2010 23:46
Add to lightbox
Figure 3-6 T1 contrast between gray and white matter and TR. At short TR, there is little signal except from short
T1 species (fat). Increasing TR to 300 and 500 ms increases both signal and contrast. As TR increases further,
however, even though the signal increases, contrast between gray and white does not. (Courtesy of D. Kaufman,
Siemens Medical Systems, Iselin, NJ)
Add to lightbox
page 74
page 75
Figure 3-7 Comparison of low-bandwidth imaging at 0.2 T (left) using a TE of 30 ms compared with
high-bandwidth imaging at 1.5 T (right) using a TE of 15 ms. Conspicuity of the L5 vertebral metastasis is better
10 of 22 11-04-2010 23:46
on the high-field image acquired with the shorter TE.
Fast Spin-Echo
Fast SE (also called RARE or turboSE) images can be acquired much more rapidly than conventional SE images,
but show similar tissue contrast. A series of spin-echoes (called the echo train) is generated by the application of
180° refocusing RF pulses at regular intervals of several milliseconds or longer. Rather than reconstruct a
separate image for each echo, as is done with conventional SE, all echoes are collectively phase encoded for a
single image. Therefore, data are accumulated much more efficiently and scan time is greatly reduced.
ECHO TIME.
Unlike a standard SE sequence where the meaning of TE is obvious, it is less obvious in fast SE since multiple
echoes are generated. The nominal TE selected by the technologist for a fast SE sequence represents the echo
time when the centerline of k-space (which largely determines image contrast) is acquired. The choice of TE thus
alters the degree of T2 contrast, as with standard SE.
REPETITION TIME.
Compared with standard SE, one typically uses a somewhat longer TR with fast SE. There are two reasons for
this. First, the longer TR tends to improve the degree of T2 contrast. The longer TR would unacceptably lengthen
scan times with standard SE, but is not a problem with the more efficient fast SE sequence. Another consideration
is that the time consumed by the fast spin-echo train reduces the number of slices that can be acquired within a
given TR interval. Therefore, a longer TR is needed to compensate.
ECHO TRAIN LENGTH AND SPACING.

For fast SE, one selects the number of echoes or echo train length (ETL), sampling bandwidth (typically 32 to 64
kHz), and echo train spacing (ETS). Scan times are reduced by a factor equal to the ETL, which typically ranges
from 3 to 16 or more. The reduction in scan time is only approximately half as much if the data are shared to
reconstruct dual echoes (i.e., proton density-and T2-weighted images).
Because of the use of a Fourier transform to reconstruct the data, signal loss from T2 relaxation over the duration
of the echo train manifests as image blurring. The blurring worsens as the ETL is increased. In the spine, ETL of
16 or more may be acceptable because CSF has a very long T2, so that the rate of signal loss is modest. On the
other hand, such a long ETL might produce unacceptable image quality in the brain, where the T2 of white matter
is relatively short. T1-weighted fast SE sequences rarely use ETL much longer than 3.
The ETS is on the order of 4 to 10 ms. In some MR systems the technologist may not have direct control over
ETS. Instead, it may be automatically determined based on sampling bandwidth, field-of-view, and other
user-selected factors.
TISSUE CONTRAST.
Although the tissue contrast is generally similar to standard SE images, there can be noticeable differences. For
instance, fat often appears brighter with fast than conventional SE, particularly with short ETS, due to the reduction
in J-coupling effects. Magnetization transfer effects produced by the repeated application of the 180° refocusing
pulses modify tissue contrast compared with standard SE, tending to make white matter appear relatively darker
than in standard SE images whereas the signal intensity of CSF is unaffected. Susceptibility effects are less
apparent with fast SE (especially with short ETS), which is advantageous for minimizing artifact from metallic
implants but disadvantageous for detecting hemorrhage. Fast SE images are also more sensitive to eddy current-
induced imperfections in the magnetic field gradients. As a consequence, fast SE sequences often incorporate an
automated phase correction to compensate.
3D FAST SE.
Like 3D GRE, 3D fast SE permits the acquisition of thin, contiguous slices. However, the need for long TR means
that scan times cannot be nearly as short as with 3D GRE. Three-dimensional fast SE may have utility for
thin-section imaging of the brain, spine, and biliary system.
FAST SPIN-ECHO AND SPECIFIC ABSORPTION RATE.

The specific absorption rate (SAR) refers to the amount of RF power deposition per unit mass of biological tissue
measured in watts per kilogram. Limits on SAR vary among countries. For instance, current SAR and other
MR-related safety restrictions in the European Union tend to be more conservative than in the United States and
revisions are under consideration.38,39 In the United States, the FDA has set specific limits on the amount of RF
11 of 22 11-04-2010 23:46
power a patient may receive during each series of images. The current upper limits are 4 W/kg averaged over the
whole body for 15 minutes, 3 W/kg averaged over the head for 10 minutes, 8 W/kg per gram of tissue for the head
or torso for 5 minutes, and 12 W/kg per gram of tissue for the extremities for 5 minutes. The purpose of setting
limitations is so that a patient's core body temperature will increase no more than 1°C.
Factors influencing the amount of RF power deposited in the patient include magnetic field strength, type of pulse
sequence, type of RF transmitter coil, and patient weight. Stronger magnets transmit at higher radiofrequencies
and therefore deposit more power in tissues; lower field systems transmit at lower radiofrequencies, deposit less
RF power, and are less likely to encounter SAR restrictions. A 3 T system deposits approximately four times as
much RF power as a 1.5 T system for the same pulse sequence, so that SAR limits are a routine consideration.
Small transmitter coils (e.g., head coil) deposit less RF energy than larger coils (e.g., body coil).
Certain pulse sequences and imaging techniques, in particular fast SE, balanced SSFP, and contrast-enhanced
MRA, tend to push the SAR limits more often than others. Fast SE utilizes rapid RF pulses of high power, thus
increasing the RF deposition. In order to reduce RF power deposition with fast SE, one can use refocusing RF
pulses of less than 180° or fewer refocusing pulses. For instance, one can reduce the refocusing flip angle from
180° to approximately 130° or less and thereby cut RF power deposition by 50% or more. Image quality is largely
unaffected. Parallel imaging (see Chapter 8), hyperechoes (see Chapter 5), and gradient and spin-echo (GRASE)
are also effective means to reduce RF power deposition. Conversely, RF power deposition is worse with certain
imaging options such as magnetization transfer.
page 75
page 76
Inversion Recovery
One way to produce images that are highly sensitive to T1 relaxation times is to prepare the magnetization before
the body of the fast SE pulse sequence with an additional 180° RF pulse as follows:
which is repeated at intervals of TR. The technique is called inversion recovery (IR).
Because the 180° RF pulse acts on the longitudinally aligned spins, it rotates them from the +z-axis all the way to
the -z-axis. This "inversion" of the magnetization and its subsequent recovery dictates the contrast in the images.
The waiting period, called the inversion time (TI), allows the protons to realign with the main magnetic field, which
they do at rates that depend on T1. After this waiting period, an MR signal is generated by the fast SE pulse
sequence. In certain circumstances (e.g., delayed contrast-enhanced imaging of myocardial viability), a GRE
sequence is used to generate the signal instead of fast SE.
Unlike standard T1-weighted SE sequences, T1 contrast develops during, and is primarily dependent upon, the
interval TI (Fig. 3-8) rather than TR. Depending on the choice of TI and the particular T1 relaxation times, the
signal from certain tissues will be emphasized, and the signal from other tissues suppressed.
It should be noted that the choice of TI has a modest dependency on TR. The TI needed to suppress a particular
tissue's signal will be reduced if the TR is shorter and, conversely, increased if the TR is longer. In general, long
TR is desirable to maximize tissue contrast and the SNR (within the constraints of an acceptable scan time, which
is proportional to TR).
Short-Tau Inversion Recovery (STIR)

One particularly useful version of IR is the STIR sequence; the distinguishing feature is the use of a short TI value,
typically on the order of 150 to 250 ms. Tissue contrast in STIR images is unusual, in that it appears reversed from
the contrast typically seen in T1-weighted SE images. Moreover, STIR offers two general benefits: 1. improved
lesion conspicuity through the additive effect of T1 and T2 contrast, and 2. effective fat suppression independent of
B0 homogeneity.
TISSUE CONTRAST IN STIR IMAGES.

In STIR images, tissues with long T1 relaxation times (e.g., gray matter) appear brighter than tissues with shorter
T1 relaxation times (e.g., white matter). This point is illustrated in Figure 3-8 for the images obtained with TI values
of 150 ms and 250 ms. In these images, gray matter and CSF appear bright, whereas fat and white matter
appear dark. By comparison, gray matter is darker than white matter in a T1-weighted SE image.
The contrast reversal is a direct consequence of the use of magnitude-sensitive, rather than phase-sensitive,
reconstructions of the images.
Aside from applications involving flow quantification, MR images are usually reconstructed in a way that makes
them insensitive to the phase of the signal. On such magnitude images, information about whether the longitudinal
12 of 22 11-04-2010 23:46
component of M is positive or negative is lost. As a result, Mz can only vary between 0 and 1. (It is also feasible in
certain circumstances to reconstruct phase-sensitive images, which can depict whether Mz is positive or negative.
As a result, Mz can vary from -1 to 0 to 1; that is, the dynamic range for T1 contrast is doubled (Fig. 3-9).
However, phase-sensitive images are more sensitive to image distortions produced by motion or magnetic field
inhomogeneity. Although phase-sensitive reconstructions are seldom used, they may be helpful for certain
applications such as imaging of delayed myocardial enhancement as addressed in Chapter 35.)
How does the use of the magnitude reconstruction cause a reversal of tissue contrast in STIR images? As an
example, consider the contrast between a hepatic tumor and normal liver parenchyma (Fig. 3-10). The longitudinal
magnetization of a tumor is more negative, i.e., smaller, than that of liver. However, it has a larger absolute value,
and hence the signal is greater, when only the magnitude is considered (Fig. 3-10A). The tumor therefore appears
brighter than liver (Fig. 3-10B). The contrast reversal only applies with relatively short TI values; with longer TI
values, the contrast relationships are similar to standard T1-weighted sequences (Fig. 3-10C).
In using a moderately long TE with STIR, the contrast between the lesion and normal tissue is further increased.
Thus, T1 contrast is additive with T2 contrast in STIR images, which is beneficial for lesion detection.
FAT SUPPRESSION WITH STIR.

Fat suppression with STIR relies on the principle that the longitudinal magnetization of any tissue, in crossing from
a negative to a positive value, must pass through zero (see Fig. 3-10A). If the MR signal is read out when the
magnetization is near zero, little or no signal should be produced. This condition allows signal from a selected
tissue to be suppressed based on its T1 relaxation time. The appropriate choice of TI depends on TR and T1. If a
tissue is fully magnetized (i.e., TR is long), the signal minimum (null point) occurs at TI = 0.69 × T1. For fat, which
has the shortest T1 of any commonly imaged tissue constituent, this TI is quite short, on the order of 150 ms.
However, the precise value depends on the choice of TR and the field strength-dependent value of T1.
By suppressing the normally intense signal from fat, STIR greatly increases the conspicuity of tissues and
lesions that are embedded in fat, such as lymph nodes (Fig. 3-10D), structures within the orbital cone, edema,
and metastasis involving the bone marrow. In the abdomen, the method reduces artifacts caused by respiratory
motion by suppressing the bright signal from moving abdominal fat.
page 76
page 77
13 of 22 11-04-2010 23:46
Add to lightbox
page 77
page 78
Figure 3-8 Inversion recovery images. By progressively increasing the inversion time TI, signals from increasingly
long T1 are first attenuated (null) and then recover. (Courtesy of D. Kaufman, Siemens Medical Systems, Iselin,
NJ)
14 of 22 11-04-2010 23:46
Add to lightbox
Figure 3-9 Longitudinal magnetization versus time after 180° RF pulse in an inversion recovery pulse sequence.
For each tissue, there is an inversion time (A and B), which results in negligible signal from that tissue. Phase-
sensitive image (dashed line) retains sign of magnetization and may result in improved tissue contrast.
Add to lightbox
Add to lightbox
15 of 22 11-04-2010 23:46
Add to lightbox
Add to lightbox
page 78
page 79
Figure 3-10 A, Illustration of STIR contrast for fat, liver, and a metastasis (met). Immediately after the 180° pulse in
an IR sequence, Mz for all tissues is negative. The longitudinal magnetization recovers according to the T1 of each
tissue. If a short T1 is used (e.g., 150 ms), then Mz of fat is near zero and its signal is nulled. Mz for other tissues is
still negative. However, a magnitude reconstruction displays only positive signals, so the curves are reflected
above the zero line. As a result, the contrast relationships are reversed and the metastasis appears brighter than
the liver. The lesion-liver signal difference is further improved by the addition of T2 weighting. B, STIR image of
patient with a large hepatic tumor (arrow). The tumor and surrounding edema appear brighter than the normal liver
parenchyma, as does the spleen. C, Compared with (B), liver-lesion and liver-spleen contrast are reversed on this
T1-weighted GRE image. D, Coronal STIR fast SE image (TR/TE/TI = 6000/90/150) of a patient with
lymphoepithelial cysts. This sequence is highly sensitive for lymphadenopathy because the lymph nodes appear
bright against a dark background of fat.
Unlike other fat-suppression techniques in which frequency-selective pulses are applied (see below), STIR is not
dependent on a high degree of B0 homogeneity for effective fat suppression. On the downside, STIR cannot
readily distinguish fat from blood products, as both tissues may have short T1 and the signal intensities are
suppressed in both cases. STIR sequences should also be avoided in the setting of contrast enhancement. In a
STIR image, the shortened T1 of the enhanced tissue can cause it to paradoxically decrease in signal intensity,
rather than increase. In these situations, one is better off using chemical shift-selective fat-suppression methods.
Fluid-Attenuated Inversion Recovery (FLAIR)

Fluids tend to have the longest T1 values of the visible tissue constituents. For example, the T1 of cerebrospinal
fluid is several seconds. Using an IR-prepared fast SE sequence, one can suppress fluid signal by using a long TI
(e.g., 750 to 2000 ms, depending on the TR). This technique is called FLAIR.
16 of 22 11-04-2010 23:46
Add to lightbox
Add to lightbox
Add to lightbox
17 of 22 11-04-2010 23:46
Add to lightbox
Add to lightbox
Figure 3-11 Comparison of T1- and T2-FLAIR with standard sequences in a patient who has undergone resection
of an intracerebral glioma. A, T1-weighted SE, TR/TE 500/10. B, T1-FLAIR, TR/TE/TI 2162/10/750. Contrast
between gray and white matter is slightly better than in A. C, T1-FLAIR after gadolinium administration shows
sensitivity of the sequence to contrast enhancement. D, T2-weighted SE, TR/TE 7500/116. E, T2-FLAIR, TR/TE/TI
10,000/168/1750. The periventricular edema is better shown than in D because of the suppression of CSF signal.
There are two versions of FLAIR in routine use: T1-FLAIR and T2-FLAIR (Fig. 3-11). Sample parameters for the
T1-FLAIR sequence would be TR/TE/TI of 2000 ms/10 ms/750 ms, and for T2-FLAIR 10,000 ms/170 ms/1750
ms. Compared with T1-weighted SE for imaging of the brain, T1-FLAIR produces better contrast between gray
and white matter and better suppression of CSF signal. T2-FLAIR is helpful because bright periventricular lesions
appear bright whereas the signal from the ventricles is suppressed. By comparison, these lesions can be obscured
by adjacent CSF signal on T2-weighted fast SE images.
Gradient-Echo
Gradient-echo pulse sequences (bearing acronyms such as FLASH, SPGR, FFE, trueFISP, FIESTA, and
balanced-FFE) use only a single RF pulse as represented by
where α is an RF pulse with a flip angle that is usually less than 90°. Because there is no 180° pulse, much shorter
TR (on the order of 3 to 250 ms) can be used than with SE, so that scan times are dramatically lessened.
page 79
page 80
18 of 22 11-04-2010 23:46
Despite their simple structure, GRE pulse sequences produce images with complex contrast behavior. For
instance, tissue contrast relates more to the choice of flip angle and to the presence or absence of spoiling than to
TR and TE.
Incoherent versus Coherent Gradient-Echo

As discussed in greater depth in Chapter 5, there are two general types of GRE pulse sequences: incoherent (or
spoiled) and coherent (or steady-state free precession, SSFP). With the incoherent sequences (e.g., FLASH,
SPGR, T1FFE), a combination of a strong gradient and/or an appropriately randomized RF phase serves to
disperse the transverse magnetization after signal readout. Incoherent sequences are used primarily for T1- and
proton density-weighted acquisitions.
With coherent GRE, the transverse magnetization is refocused and reutilized after each excitation and signal
readout. Tissue contrast is a function of the T1/T2 ratio. The sequences can be further categorized into the less
commonly used contrast-enhanced version (e.g., PSIF) and the widely used balanced steady-state SSFP version
(e.g., trueFISP, FIESTA, balanced-FFE). The latter method provides efficient motion artifact suppression. It also
provides the highest SNR per unit time and is therefore particularly beneficial for ultrafast acquisitions. With
balanced SSFP sequences, the flip angle is kept large (e.g. 50° to 90°) and the TE is minimized to avoid artifacts
(see Chapter 5). Fluids (including blood and cerebrospinal fluid) and fat appear bright.
Balanced SSFP can be used as a scout sequence, since the images can be acquired in less than half a second
and are motion insensitive. It is also the preferred sequence for cine cardiac imaging given its insensitivity to
motion artifact, high SNR, and excellent contrast between myocardium and blood in the cardiac chambers. Since
the contrast is neither truly T1 nor T2 weighted, it does not substitute for standard pulse sequences in routine
clinical applications outside of the heart.
SLICE ACQUISITION MODES.

There are three commonly used slice acquisition modes for imaging with incoherent GRE sequences: sequential
2D, multislice 2D, and 3D. Sequential 2D GRE (short TR on the order of 5 to 40 ms) is primarily used for time-of-
flight angiography. Multislice 2D GRE (moderate TR on the order of 100 to 250 ms) is used for breath-hold
imaging of the upper abdomen. 3D GRE (very short TR on the order of 3 to 6 ms) is also used for breath-hold
imaging of the upper abdomen, for MR angiography, and for thin slice acquisitions in general.
With coherent GRE, the TR must be kept very short in order to maintain the steady state. Consequently, only
sequential mode is used, never multislice mode.
SUSCEPTIBILITY EFFECTS.
Because GRE sequences do not use 180° pulses, GRE images are more sensitive than SE images to signal loss,
geometric distortion, and other artifacts (Fig. 3-12) caused by magnetic field inhomogeneities (magnetic
susceptibility effects). Sources of such distortions include ferromagnetic implants, air-soft-tissue or bone-soft-
tissue interfaces, or just a poor shim.
Add to lightbox
19 of 22 11-04-2010 23:46
Figure 3-12 Coronal GRE image obtained with TR/TE = 21/12 and a 40-cm FOV. Note the zebra-striped patterns
of signal loss at the periphery of the FOV caused by the arms and shoulders lying in an inhomogeneous portion of
the static magnetic field. This artifact would not usually be encountered in an SE image.
At times the sensitivity to magnetic field inhomogeneities can be helpful. For instance, GRE images acquired with
long TE (e.g., 10 to 30 ms) and low flip angle (e.g., 20°) show heightened contrast between hemorrhagic lesions
(short T2*) and brain tissue, even at low magnetic field strengths where SE images are insensitive to blood
products (Fig. 3-13).
Three-Dimensional GRE
Three-dimensional methods involve the volumetric acquisition of a data set that is reconstructed into thin contiguous
slices, also called 3D partitions. The thickness of the region (or 3D slab) encompassed by the 3D acquisition is
equal to the product of the number of slices and slice thickness that are selected. Unlike 2D, the slices are truly
contiguous and motion artifacts are reduced.
Two types of 3D acquisitions are possible: isotropic (i.e., the voxel is a cube) and anisotropic (i.e., the voxel is
elongated in one dimension, usually along the slice-selection axis). With isotropically acquired 3D data sets,
multiplanar reformations are of uniform quality irrespective of orientation. With anisotropic data, there are some
orientations for multiplanar reformations that produce better image quality than others. Compared with isotropic
acquisitions, anisotropic ones are less time-consuming.
LIMITATIONS OF 3D GRE.
page 80
page 81
Add to lightbox
Figure 3-13 Comparison of sequences for sensitivity to hemorrhage. Patient with presumed amyloid angiopathy.
Upper left, Proton density-weighted fast SE, 5-mm slice thickness, TR/TE = 4500/22; upper right, T2-weighted
fast SE, 5-mm slice thickness, TR/TE = 4500/85; lower left, 3D GRE with 0.8-mm slice thickness, TE = 10 ms;
lower right, 2D GRE with 5-mm slice thickness, TE = 28 ms. The GRE sequences are most sensitive to
hemorrhage, with 2D more sensitive than 3D because of the longer TE and thicker slice.
Despite the many benefits, there are some drawbacks to 3D acquisitions. For instance, wraparound artifact may
20 of 22 11-04-2010 23:46
occur in the slice direction in addition to the usual in-plane phase-encoding direction. Also, for reasons relating to
the imperfect slab profile produced by the RF pulse, the number of useful images may be at least one or two
fewer than the number specified. A subtle additional issue is that slice profiles may be degraded by ringing artifact
if only a few slices are specified. Image quality tends to be better with a larger number of partitions (e.g., 32 or
more).
MAGNETIZATION-PREPARED 3D GRE.
An inversion recovery, magnetization-prepared 3D GRE acquisition, called MP-RAGE, is a useful alternative to
spoiled 3D GRE images. Compared with a typical spoiled 3D GRE sequence, it offers much better contrast
between gray and white matter in the brain (Fig. 3-14A). The improved contrast simplifies the task of image
segmentation required for fMRI studies of the brain, and is generally helpful for evaluation of cerebral anatomy.
The improved T1 contrast is particularly helpful at 3 T, where T1 relaxation times are longer and T1 contrast
diminished compared with 1.5 T.
Single-Shot Imaging
Single-shot imaging offers the benefits of motion insensitivity and the ability to depict dynamic processes with a
temporal resolution of one second or less, though at the expense of worsened spatial resolution.
SINGLE-SHOT GRE.
An inversion recovery, magnetization-prepared 2D GRE acquisition, called turboFLASH, snapshot FLASH, or
FIRM, gives excellent T1 contrast in a scan time on the order of one second (Fig. 3-14B). Although image quality is
not competitive with standard but lengthier acquisitions, it is especially useful for applications that require
time-resolved imaging of contrast enhancement (e.g., dynamic gadolinium-enhanced perfusion imaging of the
heart, liver, or kidney). It is also commonly used with a small test bolus of contrast agent to determine the proper
delay time for contrast-enhanced MR angiography, and as a primary T1-weighted imaging technique in patients
who cannot cooperate for breath-hold studies of the upper abdomen.
SINGLE-SHOT FAST SE.

page 81
page 82
Add to lightbox
Add to lightbox
21 of 22 11-04-2010 23:46
Figure 3-14 A, Comparison of a single slice from a 3D SPGR acquisition (left) with an IR-EFGRE3D (MP-RAGE)
acquisition (right). All images were acquired at 3 T. The T1 contrast and gray-white differentiation are superior with
the MP-RAGE method. Scan time was 2 minutes 51 seconds for 3D SPGR and 4 minutes 49 seconds for
MP-RAGE. B, TurboFLASH imaging at 3 T of a patient with a liver metastasis. Each image was acquired in
approximately one second. The TI was 1200 ms (left), 800 ms (middle), and 550 ms (right). The best combination
of image quality and tissue contrast was obtained at the TI of 800 ms in this case. Also note the absence of
respiratory motion artifact. (A, Courtesy of Dr. Bob Lenkinski, Beth Israel Deaconess Medical Center, Boston,
MA)
By using a fast SE sequence with long ETL of 80 or more, along with partial-Fourier reconstruction in certain
implementations (in which case the sequence is known as half-Fourier acquisition single-shot turbo spin-echo or
HASTE), one can acquire all the data needed for an image after a single RF excitation. Scan times are on the
order of a second or less, and the method is insensitive to cardiac or respiratory motion. Single-shot fast SE is
routinely used for MR cholangiography (see Chapter 79), and imaging of the abdomen and other regions.
However, single-shot fast SE shows noticeable blurring related to T2 decay over the duration of the echo train,
particularly for tissues with short T2 relaxation times. Blurring can be minimized by the use of a short ETS as well
as parallel imaging methods (see Chapter 8). Blurring is less of a concern for imaging of bile (MR cholangiography)
and urine (MR urography) where the T2 relaxation times are prolonged.
ECHO-PLANAR IMAGING.
Echo-planar imaging is another commonly used single-shot method. It acquires a series of gradient-echoes after a
single RF excitation so that scans are completed in as little as a tenth of a second or less. It requires especially
powerful, fast gradients compared with other imaging methods and thus is not available on all MR systems.
Echo-planar imaging has several problems that preclude its routine use: limited spatial resolution, marked
sensitivity to artifacts from a poor shim or imperfect fat suppression, as well as hardware-related concerns such
as gradient-induced eddy currents. Nonetheless, it is the technique of choice for various functional techniques in the
brain, such as diffusion imaging and fMRI (see Chapters 9 to 10).
Other fast imaging techniques (see Chapter 7) are too numerous to mention here, but few are in routine use yet.
Radial acquisition methods such as PROPELLOR and VIPR show particular clinical promise. They have diminished
sensitivity to motion artifact compared with standard Cartesian acquisition methods due to oversampling of the
center of k-space. Spiral imaging also manifests diminished motion sensitivity due to oversampling of the center of
k-space. It is used in some centers, particularly for cardiac, breast, and functional brain imaging studies. However,
it is highly sensitive to artifacts from off-resonance effects if the shim is imperfect, in which case fat signals are not
adequately suppressed.
© 2010 Elsevier
22 of 22 11-04-2010 23:46
IMAGE ACQUISITION
In order to obtain images, a technologist must first select an RF coil and position the patient, perform
prescan adjustments in a specific order, and then acquire and reconstruct the data. Without the proper
implementation, a variety of problems may occur, such as unusable image data or equipment
malfunction or breakdown. The following is an overview of the steps involved in acquiring the MR
images once the patient has been registered and the imaging protocol selected at the operator's
console.
page 82
page 83
Positioning of Patients
The patient should be made as comfortable as possible to achieve cooperation. For most MRI
examinations, the patient lies supine, although other positions are used if supine positioning is not
tolerable for the patient. Prone positioning may reduce feelings of claustrophobia since the patient can
more easily look out of the bore, but it is less comfortable than supine positioning. Centering is done to
the region of interest using a laser cross-hair system; this region will then be moved automatically to
the center of the magnet. With certain open MR systems, the patient can actually be outside of the
magnet while the structure of interest (e.g., wrist or knee) is positioned within the magnet bore.
Phased-array receiver coils with one or more posterior elements must be positioned on the table
before the patient lies down. Some systems have the posterior elements built into the table, which
simplifies the positioning process. The anterior elements are applied afterwards and are typically
stabilized using Velcro straps or by locking into the posterior elements. Adequate cushioning must be
placed between the patient and coil to ensure that the patient will be comfortable for the duration of
the study.
Receiver Coil
Coil Selection
The first step to consider is receiver coil selection. A receiver coil acts like an antenna for a radio
signal tuned to a specific frequency. There are many receiver coils available, each having a different
function determined by both shape and design. Coil designs vary according to the type of magnet as
well (see Chapter 4). For instance, solenoid coils are sometimes used in permanent magnet systems,
owing to a different orientation of the main magnetic field. These coils are more commonly used in
lower field strength magnets.
Surface coils, also known as local coils, improve image quality but are limited to a smaller FOV than
the body coil. Surface coils are most sensitive to signals arising from tissues close to the coil. The
improved image quality in the sensitive region is a result of a better SNR and, compared with the body
coil, reduced sensitivity to motion of structures far from the coil. The range of the coil sensitivity directly
relates to the size of the surface coil. The SNR typically remains superior to the body coil up to at
least the depth of one coil radius. Surface coils are used to image small superficial structures such as
the temporomandibular joint or wrist.
Flex coils are designed to be flexible, as the name would suggest. The flexible material that encases
the receiver coil protects the coil from damage yet allows it to be wrapped around such anatomic
structures as the upper and lower extremities. These coils provide more uniform signal than flat
surface coils because of their capability of conforming to the shape of the structure.
Intracavity coils are placed inside a body cavity such as the rectum to image nearby structures such as
the prostate, cervix, or rectal wall. Because of the small radius of the coil, the sensitive region is limited
to tissues immediately adjacent to the coil. The sensitive region may be oriented in one particular
direction (e.g., rectal coil) or be relatively uniform in all directions (e.g., prostate coil). Given the limited
extent of the sensitive region, such coils are commonly combined with a larger external phased-array
coil (e.g., torso coil) to provide a more uniform and extensive field-of-view. Also, experimental
1 of 11 11-04-2010 23:47
intravascular coils are under development to study plaque characteristics and help guide intravascular
procedures (see Chapter 20).
A whole-volume coil is used to acquire a uniform signal from a large volume. The body, head,
body-phased array, and some knee coils are examples of whole-volume coils. These may function as
both transmit and receive coils or as receive-only coils.
Coils are designed as linear or quadrature, also known as circularly polarized coils, the latter consisting
of two sets of magnetically decoupled coils. Advantages of a quadrature coil over a linear coil design
include: 1. a more spatially uniform flip angle from the RF pulse; 2. a reduction in power deposition of
up to 50%; 3. increased SNR; and 4. increased homogeneity of the signal. Most head and body coils,
as well as many phased-array coils, are quadrature coils.
When choosing a surface coil, one should keep in mind that a smaller coil yields a better SNR.
Therefore, it is advantageous to select a coil that will cover the area of interest and can be placed at a
distance away from the area of interest that is less than the radius of the receiver coil. For larger
areas or deeper anatomic structures, volume coils may give better signal uniformity, but they are also
more sensitive to noise produced from the patient's motion. Another variable to consider is the filling
factor, which refers to the ratio of the volume of the area being sampled to the volume of the coil. It is
most desirable to have as high a filling factor as possible to maximize the SNR from a given coil.
Therefore, one should not choose a large coil, such as a body coil, for imaging a small structure, such
as the head.
Phased-Array Coils
Even though they provide a high SNR, one problem with small surface coils is that they cover a limited
FOV. However, a series of small surface coils can be combined to form a coil array and thus provide
sensitivity for imaging over a large volume. In such coil sets, called phased-array coils or multicoil
40
arrays, the individual coils are magnetically decoupled. Each coil element is typically connected to its
own receiver channel (see Chapter 4). Because each coil in the array is magnetically isolated from the
other elements, large phased-array coils provide about the same SNR over a large FOV as would be
provided by each individual element over its much smaller FOV. Compared with a large, single-channel
RF coil covering the same volume as the phased-array coil, the SNR is much better for the
phased-array coil.
page 83
page 84
With phased-array coils, high-resolution images of the spine, abdomen, pelvis, 41 and other organs can
be acquired in just a few seconds to minutes. Drawbacks of phased-array coils are their greater
expense and several-fold longer reconstruction times because of the need to combine data from
multiple receiver channels. Phased-array coils nowadays may have as few as 4 to as many as 32 or
more elements. The development of phased-array coils with such large numbers of elements is partly
driven by the development of parallel imaging methods (see Chapter 8). Generally speaking, parallel
imaging methods work most effectively (best image quality, highest acceleration factors) with coils that
have larger numbers of elements as well as an appropriate geometry.
In using phased-array coils that have anterior elements in addition to posterior elements (e.g., torso
array coil), breath-holding, respiratory compensation, and/or fat suppression can be essential for
optimal chest or abdominal imaging. Otherwise, moving fat in the near field of the anterior elements is
very bright and generates substantial ghost artifact. The problem is less severe in the pelvis.
By contrast, phased-array coils designed for spine imaging in horizontal-field superconducting magnets
have only posteriorly located elements (with the exception in some designs of an anterior neck
element) so that sensitivity to motion of the anterior chest and abdominal walls is minimized. For
vertical-field magnets, the use of a circumferential, solenoidal design for the spine coil (see
"Restrictions in Receiver Coil Design and Positioning") means that the images are more sensitive to the
motion of anterior fat, and the imaging technique must take this into account.
2 of 11 11-04-2010 23:47
With certain phased-array coils, there may be more coil elements available than receiver channels.
Consequently, one should only activate those coil elements that are needed to image a specific region
of interest. If too many coil elements are activated, multiple elements are combined into a single
receiver channel. Moreover, some coil elements might be located far outside of the magnet isocenter.
The result is a worsening of the SNR, along with the potential to introduce artifacts into the image.
Moreover, image reconstruction time increases with the number of elements, as does the amount of
storage taken up by the raw data. Long reconstruction times can limit the practical application of some
advanced imaging techniques, particularly when a large amount of data is rapidly acquired in
conjunction with the use of a phased-array coil having numerous elements.
Restrictions in Receiver Coil Design and Positioning

There are times when an RF coil designed for a specific anatomic structure should not be used. For
instance, a patient with severe scoliosis or with a halo for a cervical fracture or stereotactic frame may
not be positioned easily within a head or neck coil. In such instances, the body coil may be used to
obtain adequate images by increasing the FOV, increasing the number of excitations, and using thicker
slices to improve the SNR.
Coil design depends on the magnetic field strength. Low- to middle-field magnets may employ saddle-
shaped and spherical coils. These coils are incompatible with high-field systems because of
self-resonance, which is a physical limitation, preventing the coil from tuning and matching at higher
frequencies. The receiver coil is also restricted by design with regard to the orientation of the main
magnetic field. The coil will have highest sensitivity to the signal when its magnetic axis (B1) is
positioned perpendicular to the main magnetic field. Therefore, different RF coils must be used with
different types of magnets.
In systems with the main magnetic field, B0, oriented horizontally, the magnetic axis of a coil is oriented
vertically. For example, a surface coil, such as a spine coil, is positioned horizontally in the center of
the main magnetic field, which results in B1 being positioned perpendicularly to B0 (Fig. 3-15). This is
the most desirable position, because tilting the coil away from the perpendicular reduces its sensitivity
and effectiveness, thereby reducing the SNR.
When the main magnetic field is oriented vertically, such as a permanent magnet, solenoid coils are
sometimes used (Fig. 3-16). The magnetic field of these coils is oriented horizontally, which is
perpendicular to the main magnetic field and therefore most desirable.
Coil Centering
The magnetic field is most uniform at the magnet isocenter. The magnetic field uniformity and linearity
of the gradients decline as one moves from isocenter. Therefore, surface coils and the body part of
interest should be placed near the center of the magnet to obtain the best image quality possible.
Some systems require imaging to be performed at isocenter, whereas this is not a requirement with
others. Generally, image quality will not suffer if the body part is a few centimeters away from
isocenter. However, with large displacements, SNR may suffer and the shim may worsen, resulting in
poor fat suppression. In addition, gradient nonlinearity (which results in stretching, warping, or
telescoping of the image) becomes most apparent for structures located the farthest distance from
isocenter. The gradient nonlinearity artifact is most apparent in MR systems with short gradient coils.
However, imaging away from the isocenter is sometimes unavoidable, e.g., for MRI of the shoulder or
wrist owing to physical limitations in positioning the patient.
Coil Tuning and Matching

page 84
page 85
3 of 11 11-04-2010 23:47
Add to lightbox
Add to lightbox
Figure 3-15 Positioning of a flat "license plate" receiver coil in a superconducting magnet. A, With
horizontal positioning of a receiver coil, its B1 RF field is perpendicular to the main magnetic field (B0),
resulting in optimal sensitivity for the tissue MR signal. In addition, its B1 field is perpendicular to that of
a linear transmitter coil (i.e., physically decoupled), which is required for safety reasons if the receiver
and transmitter coils are not electronically decoupled. Note that the B1 field of a quadrature (circular
polarized) transmitter coil continually changes (i.e., rotates within the x-y plane). Therefore, it is not
possible to decouple the receiver coil physically from the transmitter coil, and electronic decoupling is
required. B, The receiver coil can be rotated to a sagittal orientation with no loss of sensitivity.
Electronic decoupling of the coils is now required because the B1 field of the receiver coil is aligned
with that of the transmitter coil. C, Tilting a receiver coil toward a coronal orientation results in a loss of
sensitivity owing to partial alignment of the B1 field of the coil with the B0 field. Optimal sensitivity is
attained when the two are perpendicular (θ = 90°), as in A or B.
In order to maximize the sensitivity with which the receiver coil detects the MR signal, the coil
impedance must match that of the preamplifier and the coil circuitry be tuned to resonate at the tissue
resonance frequency. The MR system measures the RF power reflected from the coil as the feedback
to enable coil tuning and matching. The coil is optimally tuned and matched when the reflected power is
near zero. Since the RF coil electronically couples to tissue and the coil properties are altered
differently for each patient, the coil may be tuned and matched for each individual. This process
happens automatically as soon as the patient is positioned within the magnet bore. Some
manufacturers design the body and receiver coils so that they do not require tuning for each individual
patient. This can save some time during the prescan process. The disadvantage in some cases is a
lower quality factor for the coil, which results in a slightly reduced SNR.
Rarely, a coil will not tune to a patient. This may be due to incorrect positioning, a loose coil connector,
a blown coil diode, or the patient's size being either too small or too large. If the patient is small (e.g.,
an infant or child), the coil may have difficulty tuning because there is not enough signal returning to the
receiver coil. Placing bags of saline within the surface coil area may help load the coil so that it can
4 of 11 11-04-2010 23:47
tune. Patients who are too large may need to be slightly adjusted in position to have the coil tune
successfully. Another scenario that may create difficulty for coil tuning is a patient with a significantly
large metallic implant, such as a hip replacement. Slightly adjusting the patient's position in this
situation may be all that is required. Another possible remedy is to use the body coil instead of a
surface coil, although the SNR will worsen.
Frequency Adjustment and Shimming
Add to lightbox
Figure 3-16 Flexible solenoid coil used for spine imaging in a permanent magnet system with a
vertical main field (B0) perpendicular to B1. (Courtesy of Fonar Medical Systems, Melville, NY)
Introducing a patient into the magnetic field alters its homogeneity. Because each patient is unique in
size and shape, it is necessary to make an adjustment of the transmitted frequency. This is
accomplished by emitting a wide range of frequencies, and the result is Fourier transformed. The
transformed information is displayed as a spectrum on a graph, with peaks representing various signal
intensities. Ideally, two peaks are displayed, with the higher frequency peak (right) representing water
and the lower frequency peak (left) representing fat. The adjustment process takes this information
and adapts the system's RF frequency to match that of water.
page 85
page 86
Occasionally, the fat peak is so large or the shim so poor that the system incorrectly centers the
frequency on the fat peak, in which case fat suppression will not work properly. In this case, one can
add a value of 220 Hz at 1.5 T, or proportionally less at lower field strengths, to center correctly at
the water peak. Some systems offer an option to use a STIR-type sequence during the frequency
adjustment procedure to suppress the fat signal intensity. Typically, the separation of fat and water
peaks is easier at 3 T, since the separation of the peaks is twice that at 1.5 T.
Shimming is a process of maximizing the homogeneity of the static magnetic field using active or
passive methods. Most systems today have an automated shimming procedure that is used to optimize
the uniformity of the magnetic field while the patient is in the bore. A good shim is particularly difficult to
achieve with certain body parts such as the foot or neck. It can also be problematic for bilateral breast
imaging. Some pulse sequences, such as echo-planar, spiral, and balanced steady-state free
precession, are very sensitive to off-resonance effects from magnetic field nonuniformities. As a result,
artifacts such as ghosting, dark stripe, or dark flow artifacts may occur. 42 In this case, a better shim
setting usually can be achieved by repeating the automated shimming procedure for a volume just
larger than the region of interest. The center frequency should be manually adjusted both to ensure
that the shim is adequate and that the proper peak is selected.
5 of 11 11-04-2010 23:47
In the case of bilateral breast imaging, the shim values can be determined separately for each breast.
The values are then averaged to set the shim for bilateral imaging.
Some systems contain dedicated shim coils that permit higher-order shimming (in addition to the
standard linear shimming that just uses the three gradient coils). Higher-order shimming allows the user
to improve B0 uniformity to a value better than 1 ppm over a specified volume. This feature can be
helpful for applications that are particularly sensitive to B0 uniformity such as spectroscopy, spiral
scanning, and functional brain imaging. Although linear shimming is sufficient for cardiac imaging at 1.5
T, susceptibility artifacts may occur at 3 T especially along the infero-posterior margin of the heart at
the interface with the lung. Although time-consuming to perform, the use of higher-order shimming can
ameliorate the artifacts.
Repeated inability to obtain an adequate shim or proper frequency adjustment may indicate a serious
problem, such as magnetic field drift caused by hardware instability or a major change in the magnet
environment. The frequency adjustment process is especially important on mobile systems, as the
magnetic environment varies from site to site so that the shim is inconsistent. It is always desirable to
keep large ferromagnetic objects away from the mobile magnet, and the problem is even worse if the
object moves.
Transmitter Adjustment
Adjustment of the transmitter is performed to calibrate how much RF power must be applied to
produce any desired flip angle. It is necessary to readjust the transmitter each time a patient is
repositioned or when scanning a new patient.
Some difficulties may arise when adjusting the transmitter. For instance, the RF coil may not tune and
match properly because of patient size, presence of a ferromagnetic implant, or electronic failure such
as a blown diode or improper decoupling of the transmitter and receiver coils. Under these
circumstances, the transmitter subsystem may not be capable of emitting enough power to produce
the desired flip angle, or the resultant RF power deposition may be excessive. The service engineer
may need to be contacted to evaluate the situation.
Receiver Gain Adjustment

During MR imaging, the weak signal from the tissue protons is amplified and then translated from an
analogue signal to a digital output by the analogue-to-digital converter (ADC). Digitization makes it
possible for the computer to process the data. In certain MR systems, the ADC has a limited capacity
for digitizing the received MR signal, so that the receiver gain must be automatically adjusted. Receiver
gain adjustment will keep the range of numbers manageable for the ADC; otherwise, the peak
amplitude components in the MR signal would be clipped resulting in image artifacts. The duration of
the receiver adjustment process typically varies from a few seconds to a minute or longer, and is highly
dependent on the particular MR system, pulse sequence, and number of slices.
In MR systems with a higher digitization depth for the ADC, the receiver adjustment may not be
needed, thereby saving time during the prescan process.
Image Measurement
Preparation Scans
Most systems have a slight computational delay between when the user activates the measurement
process and when the measurement actually begins. In addition, there may be a delay from the
application of preparation scans. For certain pulse sequences, a number of preparation scans are
applied during the first few seconds after the scan is initiated, but no data are collected. The purpose
is to avoid an anomalous change in signal intensity over the first few repetitions of the sequence (Fig.
3-17). Various sequence types employ differing preparation schemes. For instance, with a SE
sequence equilibrium is achieved after the first 90° excitation pulse. The first data set is not collected in
this case, but all subsequent data are used for image reconstruction. With GRE, the number of
6 of 11 11-04-2010 23:47
preparation pulses needed to reach a steady state depends on the particular version, TR, and flip
angle. For instance, dozens of RF pulses are needed to reach equilibrium with coherent GRE
sequences. However, there are methods to shorten the preparation (e.g., α/2 preparation pulse, see
Chapter 7).
Measurement Period
page 86
page 87
Add to lightbox
Figure 3-17 Example of the effect of preparation scans. At the start of the measurement process, a
number of preparation scans may be applied to establish an equilibrium magnetization, or steady
state. The data acquisition is started only after the preparation scans are finished.
A jackhammer-like knocking sound indicates the onset of image measurement. This sound represents
the mechanical vibrations of the gradient coils produced by changing magnetic fluxes during gradient
pulsing. Gradient-produced noise is most severe with pulse sequences that use rapid switching of
high-amplitude gradients (e.g., MR angiography). At the start of each measurement, a technologist
should forewarn a patient of this noise to prepare the patient psychologically and to prevent motion
due to involuntary reaction. As mentioned earlier, it is advisable to give all patients ear protection
against the noise created by the scanning.
Some patients may become agitated after the measurement is under way. Several options are then
available:
1. On some systems, selection of a "pause" option temporarily interrupts the current acquisition.
This allows the technologist to speak with the patient to determine if the request for a pause is
of an urgent nature or if the patient needs a moment to cough, for instance. This option is of
value only if the patient is able to continue with the procedure without having moved from the
original position. Otherwise, the acquisition will need to be restarted from the beginning.
2. Some systems can reconstruct images from incomplete data sets if more than 50% of the
information has been collected. The quality of these images improves as the percentage of data
acquired approaches 100%. If the patient starts to move toward the end of the measurement
period, it may be prudent to abort the acquisition prematurely and use these images
reconstructed from the incomplete data set.
3. If the patient moves transiently during the acquisition, the resultant image degradation may be
mild, particularly if the movement occurs when the outer portions of k-space are being acquired.
Although some blurring will occur, it may not be necessary to abort the acquisition.
4. The acquisition may be repeated using a smaller matrix, fewer excitations, and shorter TR for
faster imaging to improve the likelihood of obtaining a motion-free image. Alternatively, a
single-shot acquisition may be employed.
5. Motion-suppression techniques such as PROPELLOR, or ultrafast acquisitions such as echo
7 of 11 11-04-2010 23:47
planar can be used.
Breath-Hold Imaging
Respiration may be voluntarily suspended for a limited time in certain clinical applications. In particular,
breath-holding markedly improves the quality of GRE and fast spin-echo images in the chest and
abdomen. These techniques can be used to evaluate vessel patency or to detect a lesion in the upper
abdomen or heart.
Fairly consistent results can be obtained when the duration of the breath-hold is short, perhaps 10 to
20 seconds. Breath-holding at end-expiration tends to produce the most consistent diaphragm position
across breath-holds. However, some technologists prefer to have the patient suspend respiration at
end-inspiration because the patient may be able to suspend respiration for a longer period of time.
A technologist should review the breath-hold instructions with a patient before the start of the
examination. If the patient is prepared for the breath-hold instructions, there will be less chance of
confusion, especially with systems that may have an inferior quality intercom or if a patient has hearing
loss. A technologist should also review the instructions to obtain consistency in breath-holding (e.g.,
inhale and exhale the same amount each time the patient is directed to do so). The patient is instructed
to breathe in (wait approximately 3 seconds), breathe out (again wait approximately 3 seconds), and
stop breathing. The instructions should be given more slowly than would be needed outside of an MRI
environment; otherwise the patient may not be able to keep up with the instructions.
Hyperventilation, commonly used before breath-hold spiral CT acquisitions, is equally helpful for MRI.
One can use respiratory bellows to show the respiratory waveform while performing breath-hold
imaging. By reading the respiratory waveform, technologists are able to more accurately instruct
patients to hold their breath at the end-expiration phase, and to monitor if the patient is really following
the instruction correctly. In difficult cases, breath-hold capability can be improved by nasal
administration of 100% oxygen.43 The command to stop breathing should be given at the last possible
moment, taking into account the intrinsic delays that occur before the start of data acquisition.
page 87
page 88
New respiratory compensation techniques such as navigator gating may obviate the need for breath-
holding in some cases.
Use of Contrast Agents

Many different kinds of contrast agents are involved in MR imaging. The following is a brief overview of
the administration of gadolinium-based extracellular paramagnetic agents. The details of different
contrast agents are discussed in Chapters 13 and 14, as well as in the various clinical chapters.
The typical dosage for extracellular gadolinium chelates is 0.1 mmol/kg, but higher or lower dosage
can be applied. Double-dose contrast medium may be beneficial for improved detection of small
lesions. Also, double- or even triple-dose contrast medium can be used for 3D MR angiography. In
certain applications, the dose of contrast agent is much lower since the contrast agent is diluted (e.g.,
44
MR arthrography, direct MR venography ).
Extracellular gadolinium chelates are typically given as a bolus IV injection over 5 to 30 seconds
followed by 10 to 20 mL of normal saline to flush the tubing. The injection rate varies depending on the
exam type-usually from 1 mL/s to 5 mL/s with approximately 2 mL/s being typical. The injection can be
implemented by hand or a MR-compatible power injector. If a power injector is used, the volume of
contrast agent and flushing saline, along with the rate of injection are selected. If contrast agent is to
be injected by hand, in addition to preparing the syringe, the technologist may wish to prepare a
second syringe of normal saline as a bolus to follow the contrast agent, using a three-way stopcock on
the line, and completely flush the tubing.
8 of 11 11-04-2010 23:47
Another technique is to draw up 10 to 20 mL of saline into a 60-mL syringe and then carefully draw up
the needed contrast agent so that the two substances are layered one over the other (the contrast
agent is heavier than the saline so it will stay closer to the end of the tip of the syringe as long as the
syringe is kept in a vertical position with the tip down). This particular method works well with
gadopentetate dimeglumine , but not with some nonionic contrast media that are less dense and mix
with saline.
In brain or spine scanning, post-contrast scanning should begin soon (usually within 1 minute) after the
IV contrast agent has been administered. When imaging liver or kidneys, a technologist or radiologist
may wish to begin breath-hold imaging within 20 seconds of the bolus of contrast medium. Multiphase
imaging of contrast enhancement is particularly important for lesion characterization (e.g., cavernous
hemangioma of the liver, breast cancer). In 3D MR angiography, care must be taken to start the
acquisition after an appropriate delay to avoid missing the arrival of the contrast bolus. Test bolus and
automated triggering methods are reviewed in Chapter 29.
A note of caution should be made that a rapid bolus injection sometimes causes a patient to feel
nauseated (a nonspecific effect resulting from administering a hypertonic solution, though perhaps
more common with some contrast agents than others). The consequences of nausea developing into
vomiting are hazardous, as the patient's position in the magnet is supine in most cases and the bore of
the magnet is limiting, thus putting a patient at higher risk of aspiration.
The technologist should record the volume and name of contrast agent. This information is critical in
case of allergic reaction or other complication. This information is also important for the hospital
laboratory to know, since certain contrast agents can be associated with spurious hypocalcemia. 45
Image Filtering and Reconstruction

Reconstruction times depend on matrix size, as well as on the capabilities of the computer and array
processor. Images acquired with a larger matrix will require longer reconstruction times.
A variety of filters are available for image smoothing, signal intensity normalization, and reduction of
artifacts. The availability of various image filters is system dependent. Some of these filters are
selected prospectively and are applied to the data during reconstruction, whereas others are selected
after the measurement and are applied directly to the reconstructed image. Noise reduction filters
reduce the background noise in the image but may cause artifacts such as excessive smoothing.
Renormalization filters correct for the signal nonuniformity typical with surface coils. Caution must be
used with all image filters not to introduce clinically confusing artifacts into the image.
Troubleshooting
Quality control is a more complex process for MRI systems than for other imaging modalities such as
CT. Artifacts may occur in MR images from a variety of causes, as reviewed in detail in Chapter 22.
Although many of these artifacts require the attention of service personnel, it is helpful to know when
image quality problems are related to hardware rather than to inappropriate selection of pulse
sequence or scan parameters or to the patient. Therefore, the technologist and physician should be
aware of basic troubleshooting procedures.
A consistent downward trend in SNR measurements warrants a call for service. One should note
whether the noise exhibits any "structure." For instance, discrete vertical lines oriented to the
frequency-encoding axis represent RF noise. RF noise may be caused by a defective RF enclosure for
the magnet room, and one should check to make sure that the door is tightly closed and that the
copper tabs surrounding the door are intact. One should also check for flickering within the magnet
room.
page 88
page 89
Repeated artifacts propagating along the phase-encoding axis may represent ghost artifact from
respiration or pulsatile flow. If this cause is excluded, the artifacts likely relate to hardware problems,
9 of 11 11-04-2010 23:47
such as RF or gradient instability or inadequate eddy current compensation. Intermittent diagonal lines
or "snowstorm" artifacts may relate to hardware problems or to static discharges, which tend to occur
during conditions of low humidity. One should check the humidity levels in both the magnet room and
the room containing the gradient subsystems and should consult service personnel. Declines in humidity
occur more frequently during periods of extremely cold weather. Artifacts that occur only intermittently
may be caused by temperature instability in the magnet room or the room housing the computer and
other hardware. Ambient temperatures greater than 72°F (25° C) are not usually well tolerated and
may result in damage to electronic components. Also, intermittent artifacts can be encountered from a
loose screw or other hardware adjoining the body coil or a gradient coil, which results in abnormal
mechanical vibration and arcing. These artifacts may be intermittent because the vibrations are
problematic only with particular pulse sequences. Direct inspection by service personnel may be
required to troubleshoot this problem.
Unstructured noise may be due to a variety of causes. Although service is probably required, one can
check the integrity of the connectors and cables for the receiver coils, because these, on occasion, are
mangled during positioning of the patient. Artifacts may also be caused by malfunctioning monitoring
equipment. Metallic implants such as Harrington rods or dental plates produce localized regions of
geometric distortion and may also cause poor SNR, owing to improper coil tuning. More generalized
evidence of geometric distortion suggests problems with magnet shimming or gradient linearity. If
intermittent, this artifact might be caused by the movement of large metal objects, such as cars or
elevators, in the vicinity of the magnet. Mobile MR systems may be more affected by this type of
incident, as the siting of the magnet is usually in a parking lot adjacent to a building.
For any of these artifacts, it is important to save raw data (Fig. 3-18), which often provide additional
information about the source of the artifacts. However, in most commercial systems, the raw data
must be saved immediately after the scans or even when the scan is prescribed, because the data
may be deleted when the next scan is initiated.
At times it may be necessary to continue scanning despite problems with image quality. In this case,
acquiring thicker slices, larger fields-of-view, or more excitations may be used to improve image
quality. If artifacts are particularly severe with certain pulse sequences or planes of section, these
should obviously be avoided.
Add to lightbox
10 of 11 11-04-2010 23:47
Add to lightbox
Figure 3-18 Diagonal snowstorm artifacts usually suggest a hardware problem or static discharges. In
this case, the problem was a loose screw in the body coil, resulting in abnormal vibrations and arcing.
A, Sagittal knee imaging is degraded by these artifacts. B, Raw data set corresponding to A. Each
column corresponds to a specific time interval or frequency sample; each row or "line" corresponds to
data acquired from one phase-encoding step. The semiperiodic nature of the artifacts, suggesting the
presence of abnormal vibrations, is evident in the raw data but not the image.
© 2010 Elsevier
11 of 11 11-04-2010 23:47
HOW TO OPTIMIZE IMAGE QUALITY

Signal-to-Noise Ratio
The quality of an MR image is partly determined by the SNR. For clinical MRI, a good SNR is essential for
diagnostic image quality, and just about everything affects the SNR. A higher SNR means the image will appear to
be less grainy and fine details will be more apparent. Conversely, a low SNR results in a grainy image in which fine
details are obscured. Although a high SNR is desirable, increasing the SNR may require more scan time. Given
that the human eye is able to detect differences only up to a certain point, it is best to maximize the SNR to an
acceptable level for making a diagnosis without increasing scan time to a length that is unacceptable to a patient.
A useful (though simplified) expression for signal to noise (S/N) is:
Since the SNR is roughly proportional to magnetic field strength, it makes sense that the choice of a higher field
system (e.g., 1.5 to 3 T) is likely to give better image quality (Fig. 3-19). On the other hand, low-field systems
(e.g., 0.2 to 0.5 T) may have advantages in terms of convenience, cost and, if open in design, patient acceptance.
Intermediate-field (e.g., 0.7 T) open systems provide good image quality and patient acceptance, though the cost
can be close to that of a high-field system.
Many factors aside from magnetic field strength impact the SNR, such as the choice of pulse sequence, RF coil,
quality of the shim, and so forth. For instance, low-bandwidth pulse sequences produce images with a better SNR
than high-bandwidth sequences. Coherent noise (e.g., respiratory and pulsatile motion causing ghost artifact) and
incoherent noise (e.g., RF leak) reduce the SNR. Receiver coil sensitivity along with the distance from the region of
interest to the coil will alter the SNR. As spatial resolution-which is determined by field-of-view, matrix (number of
pixels along the frequency and phase-encoding directions), and slice thickness-is improved, the SNR decreases in
direct proportion to the volume of the voxel.
Image Contrast
1 of 13 11-04-2010 23:48
Add to lightbox
Figure 3-19 Using a thin slice and small field-of-view, contrast-enhanced MRI of the pancreas at 1.5 T (top) and 3 T
(bottom) was performed in the same patient. The 3 T image is obviously less grainy than the one acquired at 1.5 T.
Image contrast refers to differences in signal intensity between tissues. High tissue contrast, or more precisely a
high contrast-to-noise ratio (in consideration of both contrast and SNR), is essential for lesion detection. Tissue
contrast may be manipulated through selection of scan parameters and pulse sequences. Some factors are choice
of pulse sequence, magnetization preparation, TR, TE, and use of a contrast agent. A technologist and radiologist
may maximize either the T1 or T2 contrast of different tissues through appropriate choices of scan parameters for
the area of interest.
Sampling Bandwidth
Low-bandwidth, or narrow-bandwidth, pulse sequences use a weak frequency-encoding gradient and a prolonged
readout period to improve SNR (Fig. 3-20). The SNR is inversely proportional to the square root of the sampling
bandwidth. Low-bandwidth methods are particularly useful on low-field systems in which the intrinsic SNR is worse
than on high-field systems. Another benefit of low-bandwidth sequences is that smaller FOVs can be acquired,
since less gradient amplitude is needed than with high-bandwidth sequences.
page 90
page 91
2 of 13 11-04-2010 23:48
Add to lightbox
Figure 3-20 Comparison of image acquired with a high sampling bandwidth (125 kHz, left) and one acquired with
a low sampling bandwidth (15.6 kHz, right). The lower-bandwidth image demonstrates a better SNR, but shows
more flow artifacts from the aorta.
Add to lightbox
Figure 3-21 Abdominal MR images acquired using a 2 mm slice (left), 4 mm slice (middle), and 8 mm slice (right).
The images become less noisy and are more easily interpreted as the slice thickness is increased, but partial
volume averaging worsens.
As previously addressed, low-bandwidth sequences should not be used when a short TE is desired. The prolonged
readout period of these sequences necessitates an increase in the minimal TE. Images obtained with
low-bandwidth methods may also be more susceptible to image artifact from the patient's motion, magnetic
susceptibility artifacts, and any instrument instabilities such as eddy current effects. Additional drawbacks of
low-bandwidth methods include reduced multislice capability and greater chemical-shift artifact, which may be
troublesome on high-field systems.
Spatial Resolution
Spatial resolution determines the viewer's ability to discern two points as separate and distinct. Spatial resolution is
determined by the dimensions of the voxel, so that a small voxel (i.e., thin slice, large matrix, small FOV) is
generally desirable. However, it is important to realize that a high level of spatial resolution is useless if there is
insufficient SNR to support it.
If one increases slice thickness, which is one dimension of the voxel, then the SNR is proportionately improved
(Fig. 3-21). The SNR increases according to the square of the FOV, so that small increases in FOV can produce a
large increase in the SNR.
Slice Thickness
page 91
3 of 13 11-04-2010 23:48
page 92
Thin (e.g., 1- to 2-mm) slices should be used when scanning small structures such as a pituitary gland or the
internal auditory canal. However, thinner slices yield a lower SNR. To compensate, the user may increase the
number of signals averaged (NEX/NSA). Thicker slices (5 mm) may be used for a routine screening brain study.
Even thicker slices (up to 10 mm) may be used for abdominal or thoracic cavity surveys. A low-field magnet may
require the use of thicker slices for an adequate SNR. Gradient strength and the RF pulse characteristics will
determine the minimal slice thickness, which is usually approximately 2 mm for two-dimensional imaging
techniques. Three-dimensional imaging techniques allow acquisition of thinner slices of 1 mm or less.
Interslice Gap, Slice Profile, and Crosstalk

It would be ideal to obtain contiguous images in MRI. In practice, it is necessary to insert a gap between slices.
The reason for this practice relates to imperfections in slice profiles. A mathematic function known as sinc [sin
(x)/x] is commonly used to modulate the waveform of the RF pulse. The Fourier transform of this function is a
rectangle. However, a perfectly rectangular slice profile is produced only by an RF pulse of infinite duration, which
is impractical. RF pulses of shorter duration always produce imperfect slice profiles, so that the flip angle varies
across the thickness of the slice. If narrow gaps between the slices are used, then each slice is contaminated by
RF excitations from the edges of the adjacent slices (crosstalk) and tissue contrast is degraded (Fig. 3-22).
On T2-weighted images, crosstalk is manifested as a reduction in the signal from structures with long T1 relaxation
times, such as CSF and neoplasms (Fig. 3-23). A gap of 20% to 50% will substantially reduce crosstalk.
T1-weighted images are less degraded by crosstalk than are T2-weighted images, so that slice gaps of 10% to
30% may be adequate.
Crosstalk can be reduced by the use of longer RF pulses at the expense of longer TR and TE. Slice profile can be
improved by use of specially tailored RF pulses even if the pulse duration is relatively short. Alternatively, one can
use 3D acquisitions, which have no crosstalk between adjacent slices.
Use of small gaps may make it difficult to obtain adequate coverage of the region of interest. Total coverage,
defined as the space between the centers of the outer slices, may be expressed as
For instance, a 20-slice axial acquisition using 5-mm thick slices and no gap would produce coverage of 95 mm,
which is less than the vertical extent of the average adult brain. Increasing the interslice gap to 40% of the slice
thickness, or 2 mm, will enable coverage of the entire area.
Field-of-View
Add to lightbox
Add to lightbox
4 of 13 11-04-2010 23:48
Add to lightbox
Figure 3-22 Slice profiles. A, Sinc pulse of moderate duration. B, Sinc pulse of shorter duration. Crosstalk
(shaded area) 1 and 2 is worse in B than A, owing to the worsened slice profile in the latter case. This
necessitates increasing the interslice gap. C, Long sinc pulses or computer-optimized RF pulses of shorter
duration produce nearly rectangular slice profiles, permitting interslice gaps less than or equal to 10% of the
normal slice thickness.
The FOV is defined as the horizontal or vertical distance across an image. The minimal FOV is primarily
determined by the peak gradient amplitude and gradient duration. Decreases in the FOV are brought about by
increasing the strength of the frequency-encoding and phase-encoding gradients. Spatial resolution improves as
the FOV is decreased, within the constraints of the SNR limitations. Reductions in FOV produce a rapid decline in
the SNR in proportion to the square of the FOV. Reducing slice thickness also reduces the SNR, but in a
one-to-one proportion.
page 92
page 93
Add to lightbox
Add to lightbox
5 of 13 11-04-2010 23:48
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
6 of 13 11-04-2010 23:48
Add to lightbox
Add to lightbox
Add to lightbox
7 of 13 11-04-2010 23:48
Add to lightbox
Add to lightbox
Add to lightbox
page 93
page 94
Figure 3-23 Effects of crosstalk on image contrast and SNR. T2-weighted spin-echo images (A to F) and
T1-weighted images (G to L) with decrease in interslice gap from 100% (A and G) to 0% (F and L). Note marked
worsening in T2 contrast and SNR, even with moderate gaps (e.g., 20% in D), whereas T1-weighted images are
less severely degraded at the same gap (J). (From Kucharczyk W, Crawley AP, Kelly WM, et al: Effect of
multislice interference on image contrast in T2- and T1-weighted MR images. Am J Neuroradiol 9:443-451,
1988, © by American Roentgen Ray Society)
8 of 13 11-04-2010 23:48
Add to lightbox
Figure 3-24 Pelvic image acquired with a small (28-cm) FOV in the coronal plane. There is severe wrap-around
artifact that could be avoided by the use of phase oversampling or a larger FOV. Wrap-around artifact is less
commonly a problem in axial or sagittal images.
Small FOVs (e.g., 8 to 12 cm) are used for detailed anatomic evaluation, as needed in the temporomandibular joint
or the wrist. Small FOVs may produce wraparound artifacts unless oversampling methods are used (Fig. 3-24).
It is common for the minimum TE to lengthen when imaging with small FOVs, since the available gradient is used
up and the sampling bandwidth must be reduced. Low-bandwidth sequences, because of the weaker readout
gradient, typically permit smaller FOVs than do higher-bandwidth sequences, but chemical-shift artifact is worse.
Larger FOVs (e.g., 30 to 40 cm) are used when more spatial coverage is required, such as in chest, abdominal,
pelvic, and spine imaging. Good spatial resolution can be maintained despite the large FOV by using a
512-acquisition matrix.
Matrix
The image matrix represents the number of pixels along the frequency-encoding and phase-encoding axes. Imaging
time is proportional to the number of phase-encoding steps. Most systems allow the user to select from several
matrix sizes.
Images can be acquired using a 128 × 256 (phase × frequency) matrix in half the time of a 256 × 256 matrix, at
the expense of a 50% reduction in spatial resolution. Because of the lower spatial resolution, the SNR improves.
However, objectionable truncation artifacts may result from small matrix sizes so that a compromise may be made
between spatial resolution and imaging time by using a larger matrix (e.g., 192 × 256).
As discussed in "Rectangular Field-of-View", spatial resolution can be improved, despite a reduced matrix, by using
an asymmetric (rectangular) FOV. This technique improves resolution along the phase-encoding axis by increasing
the amplitude of the phase-encoding gradient, while keeping constant the number of pixels in the matrix.
Tradeoffs among Imaging Parameters

There are several ways to adjust the imaging parameters so as to maximize the SNR or spatial resolution, but
there may also be undesirable side-effects that need to be understood.
Choosing the Number of Excitations

The parameter NEX (number of excitations) refers to the number of times the pulse sequence is repeated in each
projection (i.e., phase encode) during the process of acquiring an image. The SNR is improved according to:
The use of multiple excitations reduces both incoherent thermal noise (Fig. 3-25) and coherent noise from repetitive
motion (e.g., ghost artifacts from respiration; Fig. 3-26). Acquiring multiple excitations is common at low field
strengths, whereas a single excitation is typically adequate at 1.5 T and higher fields.
Scan time is proportional to NEX, so that using a single excitation cuts the scan time in half compared with that for
two excitations. However, the penalty is a reduction in SNR. For instance, if a 2 NEX acquisition had an SNR of
9 of 13 11-04-2010 23:48
141, then a 1 NEX acquisition would only have a SNR of 100 (a reduction of 29%).
Partial Fourier
The partial-Fourier method reconstructs an image from only part of the available data. In the specific case of
half-Fourier (also called 1/2 NEX), slightly more than half of the data are acquired while the remaining data are
mathematically synthesized in mirror-image fashion (conjugate synthesis). Half-Fourier is most robust when used
with SE sequences, but can be used with GRE (including balanced SSFP) if the echo is relatively centered and the
TE kept short.
This technique provides a reduction of nearly 50% in imaging time. However, with the faster acquisition, SNR will
be reduced by about 29% (i.e., as if the NEX were cut in half), so the method should be employed only when the
SNR is not a limiting factor. Moreover, half-Fourier images are prone to artifacts produced by magnetic field
inhomogeneity or motion of the patient; the artifacts worsen with higher field strength (e.g., 3 T). Better image
quality is obtained by acquiring somewhat more data with a 3/4 NEX acquisition, at the expense of slightly greater
scan time.
Reduced Matrix Size

Scan time is proportional to the number of phase-encoding steps, so scan time can be decreased by using a
smaller matrix size along the phase-encoding direction (changing the matrix size along the frequency-encoding
direction has no effect on scan time). Decreasing the matrix size blurs the image and exacerbates truncation
(Gibbs) artifact, which manifests as multiple ghost images of certain features. Truncation artifacts are difficult to
discern in images acquired with 256 or more phase encodes but are more apparent in images made with 128
phase encodes. Truncation artifacts can be reduced by utilizing appropriate image filtering and reconstruction
algorithms.
page 94
page 95
Add to lightbox
Figure 3-25 Effect of averaging on image signal-to-noise and contrast-to-noise ratios. NEX (number of averages)
increases from 1 to 8 going from upper left to lower right. (Courtesy of D. Kaufman, Siemens Medical Systems,
Iselin, NJ)
10 of 13 11-04-2010 23:48
Add to lightbox
page 95
page 96
Figure 3-26 Signal averaging: issues for abdominal imaging. Abdominal images of a normal subject acquired in a
multicoil array. Upper left, SE image (TR/TE/NEX = 500/12/1) obtained in 1 minute during breathing shows
blurring and ghost artifacts from respiratory motion. Upper right, Same with 4 NEX obtained in 4 minutes shows
improved SNR and a reduction in ghost artifacts, but the image remains blurred. Lower left, Breath-hold FLASH
image (TR/TE/flip angle/NEX = 120/4.8/80°/1) obtained in 15 seconds shows no ghost artifacts and greatly
improved detail. The SNR is better for the FLASH than the SE images despite the much shorter scan time,
because of the reduction of coherent noise from respiratory ghost artifacts. Lower right, Breath-holding is not
required with such fast acquisitions, but is helpful to ensure proper registration of all the slices. HASTE image
(TE/NEX = 60/1) acquired in only 370 ms also shows good SNR and good detail.
For a given FOV, a 256 × 256 matrix has twice the spatial resolution of a 128 × 256 matrix. However, 128 phase
encodes can be acquired in half as much time as 256 phase encodes. In many applications, a 128 × 256 matrix
provides adequate spatial resolution. Furthermore, despite the shorter scan time, an image acquired with a 128 ×
256 matrix will have a higher SNR than one acquired with a 256 × 256 matrix. For instance, if the SNR of the 256
× 256 matrix image were 100, then the SNR of the 128 × 256 image would be 141. The reason is that the increase
in SNR resulting from the twofold larger pixel more than compensates for the loss from acquiring half as many
lines of data. If the amount of ringing and blurring artifact is clinically unacceptable, then an intermediate matrix
(e.g., 160 × 256 or 192 × 256) can be acquired. It is more time-consuming than a 128 × 256 matrix but produces
sharper images with fewer artifacts.
Rectangular Field-of-View
11 of 13 11-04-2010 23:48
Add to lightbox
Add to lightbox
Add to lightbox
Figure 3-27 Schematic comparison of the effects of acquisition with different matrices on pixel size and image
appearance. Note that subtle differences, such as truncation artifacts, are not shown here. A, A 256 × 256
(frequency × phase) image has square pixels (0.9 × 0.9 mm). B, A 256 × 128 image has rectangular pixels (0.9 ×
1.8 mm) with half the spatial resolution of A, but the image is acquired in half as much time. The area under the
phase-encoding gradient is half of that in A. C, A 256 × 128 image acquired using asymmetric FOV along the
phase-encoding direction with one half the FOV in the frequency-encoding direction (24 × 12 cm). The resolution is
the same as that in A, but the image is acquired in half as much time. It may not always be possible to reduce the
FOV by such a large amount in the phase-encoding direction because of wraparound.
There is one situation in which the matrix size can be reduced with minimal or no loss in spatial resolution. If the
anatomy being imaged is elliptical in shape (e.g., head or abdomen), a rectangular or asymmetric FOV can be
used to recover some or all of the resolution lost by using an asymmetric matrix (Fig. 3-27). For instance, consider
1. a 256 × 256 image with a symmetric 24 × 24 cm FOV, versus 2. an image with 128 phase-encoding steps and
12 of 13 11-04-2010 23:48
a rectangular FOV of 12 cm (phase-encoding direction) × 24 cm (frequency-encoding direction). These images

have equal pixel resolutions of 0.9 × 0.9 mm, but the image with the smaller matrix is acquired in half the time.
By permitting smaller acquisition matrices to be used without unacceptable compromises of spatial resolution,
rectangular FOVs are routinely useful for reducing scan time, provided that the FOV accommodates the anatomy.
If the anatomy does not fit in the rectangular FOV, substantial wrap-around artifacts may occur along the phase-
encoding direction. Wrap-around is not a problem along the frequency-encoding direction because oversampling in
this direction does not require extra time and, in conjunction with the application of a band-pass filter, cuts off
signals arising from outside the prescribed FOV.
The SNR is a consideration in the use of rectangular FOVs but is not a limiting factor as long as the FOV is not too
asymmetric. Consider 1. an image acquired with a 256 × 256 matrix and a symmetric 24 × 24 cm FOV, and 2. an
image acquired in half the time with a 128 × 256 matrix and a 17 × 24 cm (29% smaller) FOV along the phase-
encoding direction. Surprisingly, these two images have the same SNR. The loss of SNR resulting from the
reduced FOV in the second image is precisely compensated by the gain in SNR from the larger pixel size.
page 96
page 97
Low-Field MR Imaging
MR systems with magnetic field strengths on the order of 0.2 to 0.5 T are commonly referred to as low-field MRI
systems, although the nomenclature varies by user and vendor. When imaging at low fields, one should keep in
mind that, for given scan parameters, the image has an intrinsically lower SNR than at high field. This disadvantage
usually can be compensated by increasing slice thickness, increasing FOV, decreasing matrix, using lower
bandwidth, and increasing the number of acquisitions, at the expense of decreasing the spatial resolution or/and
46
increasing the scan time.
Users of low-field instruments should also recognize that contrast enhancement with gadolinium-based agents is
dependent on the field strength. Contrast enhancement of certain lesions may not be as vivid on the low-field
images as on those obtained at high field. Some lesions that enhance only weakly at high field may appear not to
enhance at all on low-field images. Therefore, users of a low-field magnet always should endeavor to use
post-contrast pulse sequences that are as T1-weighted as possible, and try to use higher dosage of contrast
47
agents for better enhancement.
Scanning at 3 Tesla
In the recent past, it was thought that scanning at "high field" meant scanning at 1.5 T. With the introduction of the
3 T scanner for clinical imaging, it is quickly being realized that 3 T imaging is the new "high field" (see Chapter 18).
Improved SNR can be used to decrease the scan time and/or increase the spatial resolution of the image. 3 T is
already the field strength of choice for clinical brain imaging as well as functional MRI of the brain, spectroscopy,
and high-resolution intracranial MR angiography.
The increase in SNR and superior fat suppression are beneficial for abdominal and pelvic MRI as well as contrast-
enhanced MRA in the body. Musculoskeletal imaging also has an advantage at 3 T because of the ability to
increase the spatial resolution and more robust fat suppression.
A given sampling bandwidth produces double the chemical-shift artifact at 3 T as at 1.5 T. To compensate, higher
bandwidths may be used, which lowers the SNR. At 3 T, power deposition is drastically increased, making it
necessary to adjust protocols. For example, the user may need to increase the TR to allow time for the heat to
dissipate. Smaller flip angles may be required for RF-intensive pulse sequences, such as SSFP, contrast-enhanced
3DMRA, and fast SE. One can set up the protocols so as to alternate high (e.g., fast SE) and low (e.g., GRE)
power deposition scans to give the patient time to "cool down."
T1 relaxation times are longer at 3 T, which may cause T1 contrast to be slightly diminished. To compensate,
slightly longer TR may be desirable or one may choose to use an MP-RAGE or other magnetization-prepared
sequence. On the other hand, the longer T1 relaxation times tend to enhance the effects of paramagnetic contrast
48
agents. As a consequence, the dose of contrast agent can be reduced (e.g. to half-dose from single dose).
© 2010 Elsevier
13 of 13 11-04-2010 23:48
SPECIAL TECHNIQUES
Parallel Imaging
The individual elements of a phased-array coil are more sensitive to signal sources that are nearer the
coil than those further away. Parallel imaging methodologies use this phenomenon to mathematically
generate additional phase-encoding lines independent from the magnetic field gradients, thereby
reducing scan time by a large factor (2 to 4 or more). Among the numerous potential and realized clinical
applications, the method can be used to reduce magnetic susceptibility artifacts for echo-planar
acquisitions, to accelerate contrast-enhanced MR angiography, and to improve the time resolution of
realtime cardiac imaging. The main limitations include the need for phased-array coils with proper
geometry and large numbers of elements, reduced SNR, and certain kinds of image artifacts. In
addition, one may need to acquire an additional calibration scan, though some parallel imaging methods
are self-calibrating. Parallel imaging methods are reviewed in detail in Chapter 8.
Fat-, Water-, and Silicone-Selective Imaging

In some cases, one chemical species is clinically more interesting and the elimination of other moieties
would improve the diagnostic image quality. These techniques are generally selective based on either
the T1 relaxation time (STIR) or the chemical shift (chemical-shift-selective imaging, phase-contrast,
two- and three-point Dixon techniques). 49 These methods will be considered in greater depth in Chapter
7.
STIR
As described previously, STIR can provide robust fat suppression. However, there are several
limitations to the use of IR for fat suppression. First, image contrast is substantially altered compared
with that for typical sequences, and the SNR tends to be lower for the same scan time. Second, as
already addressed, the technique is problematic for contrast-enhanced studies. Finally, the technique
does not readily distinguish between fat and tissues with short T1 associated with blood products.
Chemical-Shift-Selective Fat Suppression

page 97
page 98
Other techniques are based on chemical-shift differences. For example, fat and water protons have
different resonance frequencies. There is a dominant fat peak associated with methylene (-CH2-)
protons at a frequency approximately 3.5 ppm lower than that of water protons, corresponding to a
frequency shift of 220 Hz at 1.5 T. ("Fat" actually has a complex spectrum, with contributions from
methyl, vinyl, and other lipid constituents. 50) As another example, frequencies of the dimethylsiloxy units
[-Si(CH3)2-Oy-]n of silicone breast implants are shifted even lower, approximately 80 Hz lower than that
of fat at 1.5 T.
By using an appropriate pulse sequence one can selectively suppress the signals from water, fat, or
silicone.51 The signal intensity of fat, water, or silicone can be selectively suppressed by applying an RF
pulse, called a chemical-shift-selective pulse, tuned to the proper frequency. For instance, a
narrow-band RF pulse at 220 Hz below the water frequency followed by gradient spoiling will, at 1.5 T,
suppress the fat resonance.
The main limitations of fat suppression are related to B0 and B1 field inhomogeneities. Because the
chemical-shift difference simply makes the spins resonate at the slightly different magnetic field,
imperfections in the static field (B0) can mimic the frequency shifts caused by chemistry. That is,
nonuniformity of the order of 3.5 ppm will cause the fat suppression pulses to saturate fat protons in
some regions and water protons in other regions. This is especially a problem away from the isocenter
of the magnet or near air-tissue or bone-tissue interfaces (e.g., skull base, near paranasal sinuses). 52
In addition to static field inhomogeneity, nonuniformity of the RF excitation field (B1) causes the flip angle
of the fat saturation pulse to vary across the image, resulting in variable degrees of fat suppression.
Even modest B1 inhomogeneity can dramatically limit the effectiveness of fat suppression; for example,
1 of 13 11-04-2010 23:48
a 20% variation in B1 limits fat suppression to about a factor of 3. Consequently, it may be helpful to use
a chemical-shift-selective inversion pulse to suppress the fat signal (e.g., "SPIR" technique).
Selective Water Excitation

Instead of trying to suppress the signal intensity of fat, some MR systems may offer the capability to
selectively excite water protons using a spatial-spectral pulse53 (Fig. 3-28). Selective water excitation
often gives more uniform results than fat suppression, because the fat signal is absent regardless of the
B1 field inhomogeneities. However, it is less widely available than fat saturation techniques, and is only
applicable to a few types of pulse sequences.
Phase-Contrast Imaging
Instead of selectively suppressing or exciting the spins of interest, it is also possible to use phase
information to distinguish species. Phase-contrast (not to be confused with the entirely different phase-
contrast methodology used for flow quantification or angiography) GRE imaging is an example of this
strategy (Fig. 3-29). To understand these methods, imagine first that one is imaging with a
gradient-echo pulse sequence and the signal is measured immediately after the RF excitation (i.e., TE =
0); the transverse magnetization vectors of fat and water signals would then be precessing coherently,
that is, in phase. In this case, the fat and water signals from protons residing in the same voxel would be
additive. Because water protons precess 3.5 ppm more rapidly than fat protons, over time the phase of
the water protons would advance with respect to the fat protons. At a certain time after excitation, the
transverse magnetization vectors of the water and fat protons would be 180° out of phase, resulting in
signal cancellation and an out-of-phase or opposed image. (At 1.0 T, signal cancellation occurs at 1/(3.5
× 42.6 Hz)/2 = 3.4 ms.) The water and fat signals thus oscillate in and out of phase with a periodicity
(∆TE) determined by the magnetic field strength according to
When TE is equal to an even multiple of ∆TE, the fat and water signals are in phase, whereas images
acquired at odd multiples of ∆TE show signal loss caused by fat-water phase contrast. For instance, at
1.5 T a TE of 2.2 ms gives an opposed image and a TE of 4.4 ms an in-phase image. The cycling at 3 T
is twice as fast (i.e., opposed image at TE of 1.1 ms and opposed image at TE of 2.2 ms). At 1.0 T the
corresponding out-of-phase and in-phase TE values are 3.4 and 6.8 ms. Fat-water signal modulation
does not occur with SE pulse sequences because of the refocusing effect of the 180° pulse.
Opposed images are readily identified by dark borders surrounding organs at fat-water interfaces. This
artifact is not limited to the frequency-encoding direction, as is the case for the usual chemical-shift
artifact seen with both spin-echo and in-phase gradient-echo sequences.
These phase-contrast images can be acquired in just a few seconds. Use of phase-contrast imaging is
helpful for detection of fat-water mixtures, as in fatty infiltration of the liver, fat in an adrenal adenoma,
and differentiation of normal red bone marrow (fat-water mixture) from infiltrated marrow (mostly
water). Phase-contrast imaging would not be useful for detection of fat occupying an entire voxel, such
as fat in a lipoma or teratoma, because the modulation must come from the destructive interference of
fat and water in the same voxel.
The phase-contrast method is at the heart of the more quantitative, but less commonly used, two- and
three-point Dixon techniques,54,55 which are discussed in Chapter 7.
Magnetization Transfer Contrast

In many tissues, there can be a substantial quantity of MR "invisible" spins. These spins may be bound
or closely associated with large proteins, membranes, and so forth. Because of their very slow motion,
these spins have a very short T2 (<0.1 ms) and cannot be seen at any typically used TE. However,
these spins can nonetheless have a visible effect by a process called magnetization transfer (MT)
contrast.56
page 98
page 99
2 of 13 11-04-2010 23:48
Add to lightbox
Add to lightbox
page 99
page 100
Figure 3-28 Fat suppression. A, Abdominal MR image without (upper left, gradient-echo, TR/TE/FA=
40/8/30°) and with (upper right) fat saturation, both in body phased-array coil. In an unsuppressed
image, the aortic signal is bright because of rapid inflow of unsaturated blood. For the image at the
upper right binomial fat suppression was used. Suppression in plane is effective, but fat saturation
behaves like water saturation in some other planes. For example, inadvertent water saturation near the
heart due to field inhomogeneity causes the aortic signal to be suppressed. The image at the lower left
shows the improvement obtained by selectively exciting water (rather than fat suppression), in this case
using a spatial-spectral excitation for slice selection, improving in-plane fat suppression and preserving
the inflow signal in the aorta. B, Use of fat suppression to differentiate blood products from fat, both of
which can be bright on T1-weighted images. Left, T1-weighted axial image showing two hemorrhagic
3 of 13 11-04-2010 23:48
endometriomas (arrows) that appear bright. Right, On a fat-suppressed coronal image, the hemorrhagic
lesion appears bright (arrow). In contrast, the fatty signal from a dermoid cyst would have been
suppressed.
Add to lightbox
Add to lightbox
Add to lightbox
Figure 3-29 A, Calculated signal intensity versus TE at 1.5 T in a gradient-echo image of simulated
marrow consisting of 75% fat and 25% water. The oscillations in the signal intensity are due to a slight
difference between the resonance frequencies of fat and water (i.e., chemical shift). B and C, A patient
with oat cell metastasis to the right humeral shaft. Gradient-echo images are sensitive to bone marrow
abnormalities such as metastatic lesions. On T1-weighted SE image (B), normal marrow appears
bright and metastasis appears dark. On an out-of-phase proton density-weighted gradient-echo image
4 of 13 11-04-2010 23:48
(C), marrow appears dark and metastasis appears bright.
Magnetization transfer contrast is enhanced by saturating the tissue magnetization with RF pulses
applied at a frequency that is far from the center frequency of water (e.g., frequency shift of 1.5 kHz).
With such a large frequency shift, one might expect no effect on tissue signal. In reality, the "invisible"
spins resonate over a very broad range of frequencies and are saturated by the MT pulses. They then
exchange magnetization with the mobile spins, and thereby reduce the signal intensity of the latter spins.
MT RF pulses primarily affect tissues that contain both mobile and invisible protons (Fig. 3-30). Such
tissues include brain, liver, and muscle, among others. Tissues that have a paucity of restricted protons,
such as cerebrospinal fluid, urine, and blood, show minimal if any signal change. In addition, as this
effect is essentially an additional path of T1 relaxation, MT RF pulses only minimally affect tissues with
short T1. Thus, MT pulses have a minor effect on the signal intensity of fat or that of gadolinium-
enhanced tissue.
The use of MT RF pulses involves substantial additional RF power deposition, which is more of a
limitation in the body than in the head. It becomes a more significant issue at 3 T, where power
deposition is so much higher. Some systems reduce power deposition by providing an option to apply
the MT RF pulses only during the center lines of k-space.
There are several potential clinical applications for MT contrast. 57 It has proved clinically useful for
intracranial MR angiography.58 The signal from brain tissue is reduced much more than blood by the MT
RF pulses. Contrast between vessels and brain is improved and small vessels in particular are better
delineated. MT is also helpful for contrast-enhanced MRI of the brain. The MT pulses only slightly
reduce the signal intensity of tissues with short T1 relaxation times, such as enhancing metastases. 59 On
the other hand, the signal intensity of brain tissue is substantially reduced, so that the conspicuity of the
metastases is improved. Moreover, the amount of MT is affected by the physical and chemical state of
the tissue. Thus, the MT rate is apparently lowered in multiple sclerosis and thus may be of value in
characterizing the degree of demyelination in plaques. 60 As previously mentioned, MT effects can cause
substantial alterations of tissue contrast even in routine imaging61 especially for fast SE where a large
number of 180° pulses are applied.
Flow Compensation
Flow compensation, also known as "flow comp," gradient motion rephasing (GMR), gradient moment
nulling (GMN), or motion artifact suppression technique (MAST), is a method for reducing flow and
motion artifacts (see Chapter 27). The method incorporates into the pulse sequence additional gradient
pulses that reduce motion-related phase shifts.
The effect of flow compensation is to increase signal intensity from flowing blood and CSF. This method
is particularly useful for improving the signal uniformity of, and suppressing ghost artifacts arising
from, CSF on T2-weighted fast SE images of the spine. When flow compensation is used for
nonbreath-hold abdominal imaging, signal intensity from moving organs, such as the liver, is increased
and ghost artifacts are reduced.
One major limitation of flow compensation is that the minimal TE, slice thickness, and FOV may be
increased compared with standard pulse sequences. As a result, it may not be possible to use flow
compensation in applications that require extremely high spatial resolution, such as imaging the
temporomandibular joint. In addition, in some patients flow compensation may not provide adequate
suppression of ghost artifacts from CSF pulsation. In these instances, the combination of flow
compensation and ECG or pulse gating may provide better artifact suppression.
page 100
page 101
5 of 13 11-04-2010 23:48
Add to lightbox
Add to lightbox
6 of 13 11-04-2010 23:48
Add to lightbox
Figure 3-30 Principle of magnetization transfer. A, The mobile protons have a relatively long T2 and
resonate over a narrow range of resonance frequencies. This pool of protons continually exchanges
magnetization with a pool of protons having restricted motion, short T2, and a wide range of resonance
frequencies. Application of an off-resonance MT pulse (black rectangle in lower diagram) saturates the
restricted pool. The mobile pool also becomes saturated because of the exchange of magnetization
between the two pools. MT contrast can help visualization of small lesions, especially after gadolinium.
T1-weighted images without (B) and with (C) MT after gadolinium. Although the image in B does not
show a definite metastasis, that in C with MT shows a marked reduction in brain signal, providing much
better visualization of this metastasis (arrow).
Flow compensation is routinely used with time-of-flight angiography (though not with contrast-enhanced
MR angiography, where a very short TE is more important). On the other hand, one should be aware
that the use of flow compensation will reduce or eliminate the CSF flow void sign in the aqueduct of
Sylvius. This sign is occasionally used to characterize hydrocephalus. Similarly, flow compensation may
not be desirable on T1-weighted images of the spine, because it increases the signal intensity of CSF
and may reduce contrast between CSF and spinal cord or cauda equina.
Presaturation
Presaturation incorporates additional RF pulses, applied just before the body of the pulse sequence,
that saturate tissue magnetization and thereby reduce the signal intensity over user-defined regions.
Presaturation applied along the slice-select direction suppresses flow artifacts and is commonly used for
imaging the brain, neck, chest, abdomen, and extremities, wherever flow artifacts may limit image
quality.
page 101
page 102
The effectiveness of flow presaturation depends on the velocity and direction of flow. Therefore, one
should not be surprised if the technique fails to eliminate signal from CSF within the thecal sac, where
the flow is to and fro rather than unidirectional, or in certain patterns of blood flow. For instance,
presaturation does not typically affect artifacts caused by vascular pulsation in the cerebral venous
sinuses, because the effects of the presaturation pulses on arteries do not carry through the capillary
system into the veins. It may also fail to suppress intravascular signal adequately within large aortic
aneurysms, which have slow flow. In the heart, special dual inversion techniques are required to
suppress the signal intensity of blood since presaturation is ineffective. Additional limitations of the
7 of 13 11-04-2010 23:48
method include substantially reduced multislice capability and increased RF power deposition.
Consequently it is typical in commercial MR systems that the presaturation pulses are applied
intermittently (e.g., one presaturation pulse for every four slices).
Presaturation becomes ineffective after IV administration of gadolinium chelates, because the T1 of

blood becomes too short.
One can apply presaturation along directions other than slice. For instance, presaturation applied along
the phase-encoding direction anterior to the vertebral bodies may be useful for spine imaging to
suppress ghost artifacts arising from vascular pulsation and respiratory motion. Presaturation may also
be used to suppress the high signal from subcutaneous fat in the near field of a surface coil, which
reduces motion artifacts, and to reduce the severity of wraparound artifact.
Respiration Motion Compensation

A simple way of reducing respiratory artifact when imaging the abdominal area is to use a piece of
fabric, available through vendors selling positioning devices, with Velcro for keeping a snug fit around a
patient's abdomen. This limits the up-and-down motion of a patient's abdomen without increasing
image acquisition time and therefore reduces ghost artifact from that type of motion. However, this
procedure may not be comfortable for the patient.
Depending on the MR system, available respiratory gating methods include using a mechanical bellows
and navigator gating. While highly effective in some patients, the results vary according to the regularity
and depth of the patient's breathing. Extra time is required to position the bellows optimally and to
choose the thresholds for accepting the data, or in the case of navigator echoes, to position the
navigator region over the region of interest (e.g., lung-diaphragm interface) and obtain baseline data
about respiratory patterns. Moreover, with respiratory gating methods, some data must be rejected so
that scan times are increased, typically by a factor of 2 to 4.
Cardiac Gating
Synchronization of the data acquisition to the cardiac cycle reduces flow artifact and blurring for the
heart and nearby structures including the ascending aorta. It can be used to define cardiac and vessel
anatomy more precisely, or to permit dynamic display of cardiac function using cine techniques. Cardiac
gating and imaging methods are reviewed in Chapter 32.
There is some confusion of terminology with respect to the terms triggering and gating. Although these
terms are often used synonymously, some manufacturers use the term triggering specifically to refer to
"prospective" methods of cardiac gating (Fig. 3-31).
For prospective cardiac gating, data acquisition is initiated after a user-determined time interval following
the R-wave of the cardiac cycle. This "trigger delay" can be varied so that an image (e.g., of myocardial
delayed enhancement) is obtained within a particular phase of the cardiac cycle. Note that for a
multislice acquisition, the images are acquired over a finite period corresponding to the user-selected
TR.
For retrospective gating (standard for cine imaging), the data are accumulated according to the TR
irrespective of the cardiac cycle, while the ECG signals are recorded. The data are subsequently
processed using the recorded ECG signals to assign data to the appropriate phases of the cardiac
cycle. Retrospective gating has the advantage that the entire cardiac cycle is depicted (essential for
measurements of left ventricular function). By comparison, prospective gating may miss the final portion
of the R-R interval; triggering is inaccurate if data are acquired too close to the end of the cardiac cycle
due to overlap of the RF and gradient pulses with the R-wave.
8 of 13 11-04-2010 23:48
Add to lightbox
page 102
page 103
Figure 3-31 Comparison of prospective cardiac gating (A) and retrospective cardiac gating (B) for
five-slice acquisition. In prospective gating, the time interval between sequence repetitions is
determined by the R-R interval, not the TR. The TR determines the duration of the cardiac cycle during
which slices are acquired. In retrospective gating, the time interval between sequence repetitions is
determined by the TR, as it would be for an ungated acquisition. Retrospective gating provides more
consistent image quality for cine imaging than does prospective gating.
For several reasons, difficulties may be encountered in obtaining satisfactory ECG gating:
1. Poor lead contact may result from inadequate skin preparation. The skin should always be
cleansed of oil and debris by using an alcohol swab or by gently scraping the skin with a piece of
thin plastic (the cover from some ECG electrodes will suffice) before applying the electrodes. In
some instances, the hair should be shaved where lead contact is to be made. Standard ECG
electrodes have caused severe burns in some cases; therefore only MR-compatible ECG
electrodes should be used.
2. Respiratory motion causes a variation in the ECG baseline, making it difficult for the system to
find and follow the R-wave of the cardiac cycle. Attaching ECG leads to the back of a patient or
using breath-hold acquisitions may prevent this problem.
3. Gradient and RF pulsing during imaging induces electrical currents in ECG leads, which
(depending on the pulse sequence and slice orientation) may completely obscure the ECG
tracing. In most cases, such interference can be electronically filtered out of the ECG signal.
4. Magnetohydrodynamic effect. Flow of blood (which is a conducting solution) through the static
magnetic field alters the ECG by simulating an augmented T-wave or other nonspecific ECG
changes. These distortions worsen with increasing magnetic field strength.
5. Arrhythmias. These cause two problems: 1. the gating system may confuse these extra
potentials with R-waves and gate incorrectly; 2. the volume of the heart is different after a
premature beat than a normal one, which may cause blurring and ghost artifacts when these data
are combined with data acquired during normal sinus rhythm.
Before using ECG gating, it is wise to consult applications personnel for the specific MR system. Gating
methods vary widely among different manufacturers.
Vectorcardiographic gating is an ECG gating method that analyzes the direction of the electrical axis of
the heart in order to distinguish the true ECG signal from interference. 62 It is proving to be a particularly
robust means for ECG gating and can be used to identify and reject arrhythmias.
In some cases, gating using finger plethysmography may be preferred to ECG gating, given its ease of
use. However, problems may then be encountered from finger motion or in patients with severe
peripheral artery disease and weak pulses. Moreover, one must consider the additional delay in the
9 of 13 11-04-2010 23:48
gating signal with respect to the R-wave (typically a few hundred milliseconds). For pulse gating, it is
helpful to place a warm towel on the patient's hand to prevent vasoconstriction, and to instruct the
patient not to move the fingers.
Some pulse sequences used for the heart are uniquely useful for that organ. They tend to be
implemented in a vendor- and application-specific manner, so it is difficult to make generalizations about
optimal imaging parameters. Imaging protocols are reviewed in detail in the pertinent clinical chapters.
Breath-holding is used for most cardiac imaging sequences. A notable exception is coronary artery MR
angiography, where respiratory-gated navigator methods may be preferred.
Compared with standard sequences, one must decide whether to acquire cine (e.g., FIESTA, trueFISP,
balanced-SSFP) or single-phase acquisitions. Both are useful. Cine is the method of choice for studies
of ventricular function, valve motion, and flow patterns and quantification. For cine, the number of phase-
encoding lines per cardiac phase must be selected (e.g., 4 to 12). The use of more lines shortens the
scan duration (which depends on the number of R-R intervals required to accumulate all the data) at the
expense of temporal resolution (equal to the TR × number of lines).
For single-phase techniques, one must select a trigger delay. A delay on the order of 60% to 80% of the
R-R interval permits imaging during end-diastole, when the motion of the heart is least apparent. In
certain circumstances, imaging at end-systole may be preferred. An example would be in the setting of
suspected right ventricular dysplasia. Since the right ventricular free wall is extremely thin and difficult to
evaluate during diastole, end-systolic imaging may be advantageous since the myocardium is thicker. A
similar argument applies for imaging of myocardial delayed enhancement. However, imaging near
end-systole entails more risk of motion artifact degrading the image than imaging during end-diastole.
Dynamic imaging methods permit evaluation of myocardial perfusion during the first pass of a
paramagnetic contrast agent. Stress pharmaceutical agents such as adenosine may also be required.
Various fast imaging techniques are employed including echo train gradient-echo (FGRET), inversion
recovery-prepared single-shot trueFISP, and others.
Despite the obvious benefits and necessity of ECG gating for cardiac MRI, outside of the heart it should
only be used when needed. Drawbacks include: 1. more cumbersome and time-consuming patient
preparation, 2. lengthened data acquisition, 3. difficulty obtaining a reliable gating signal, and 4.
dependence of T1 weighting on the R-R interval instead of TR.
Oversampling to Eliminate Wrap-around

Oversampling methods acquire an expanded matrix (typically up to 512 pixels) along either or both the
frequency and the phase directions to eliminate wraparound artifact. The extra pixels that contain the
wrapped portion of the image are discarded, so that only the central 256 pixels are displayed. In the
frequency direction, there is no time penalty for using oversampling. In the phase direction, the scan time
increases in direct proportion to the percentage of oversampling, because additional phase encodes
must be acquired.
Gradient Rotation (Phase Frequency Swap)

page 103
page 104
The user determines the orientation of the phase-encoding axis. By reorienting the phase-encoding axis
by 90°, it may be possible to keep ghost artifacts off the region of interest. For axial brain imaging, the
phase-encoding axis is typically oriented horizontally (swapped) to direct artifacts from eye motion away
from the brain. For spine imaging, the phase-encoding axis may be oriented parallel to the long axis of
the spine to prevent ghost artifacts from respiratory motion and vascular pulsation. Axial images of the
upper abdomen usually have the phase-encoding axis oriented from anterior to posterior of the patient
so that blood flow artifacts from the inferior vena cava and aorta are away from the liver and kidneys.
However, if the left lobe of the liver is the area of interest, the user may wish to swap the phase-
encoding direction so that it is oriented from left to right, thus moving aortic ghost artifact off the left lobe
of the liver.
10 of 13 11-04-2010 23:48
A major limitation of gradient rotation is wrap-around artifact when the phase-encoding axis is oriented
along the long axis of the body. Suppression of wrap-around requires concomitant use of oversampling
methods.
In conclusion, we see that there are numerous practical considerations involved in the production of
high-quality MR images. Although this chapter offers recommendations for optimizing MRI examinations,
there are no panaceas. One must experiment with different imaging parameters and decide for oneself
what best suits the needs of a particular MRI site and the physicians involved there.
REFERENCES
1. http://www.ismrm.org/mr_sites.htm
2. http://www.users.on.net/~vision/#education
3. http://www.mr-tip.com/serv1.php
4. Saunders RD, Smith H: Safety aspects of NMR clinical imaging. Br Med Bull 40:148-154, 1984. Medline
Similar articles
5. Slifer KJ, Penn-Jones K, Cataldo MF, et al: Music enhances patients' comfort during MR imaging. Am J Roentgenol 156:403,
1991. Letter.
6. Brummett RE, Talbot JM, Charuhas P: Potential hearing loss resulting from MR imaging. Radiology 169:539-540, 1988.
7. Freinhar JP, Alvarez WA: Organic claustrophobia: an association between panic and carbon dioxide? Int J Psychosom
8. Accreditation Manual for Hospitals, Volume I-Standards. Oakbrook Terrace, IL: Joint Commission on Accreditation of
Healthcare Organizations, 1994.
9. Holshouser BA, Hinshaw DB, Shellock F: Sedation, anesthesia, and physiologic monitoring during MR imaging: evaluation of
procedures and equipment. J Magn Reson Imaging 3:553-558, 1993. Medline Similar articles
10. http://www.fda.gov/cdrh/ode/guidance/793.pdf
11. Kangarlu A, Burgess RE, Zu H, et al: Cognitive, cardiac and physiological studies in ultra high field magnetic resonance
imaging. Magn Reson Imaging 17:1407-1416, 1999. Medline Similar articles
12. Schenck JF: Safety of strong, static magnetic fields. J Magn Reson Imaging 12:2-19, 2000. Medline Similar articles
13. http://www.mrisafety.com/
14. http://www.fda.gov/cdrh/ode/odeclip.pdf
15. Sakakibara Y, Mitsui T: Concerns about sources of electromagnetic interference in patients with pacemakers. Jpn Heart J
16. Luechinger R, Zeijlemaker VA, Pedersen EM, et al: In vivo heating of pacemaker leads during magnetic resonance imaging.
Eur Heart J 26:376-383, 2005. Medline Similar articles
17. Martin ET: Can cardiac pacemakers and magnetic resonance imaging systems co-exist? Eur Heart J 26:325-327, 2005.
18. Institute for Safe Medical Practices, Medical Safety Alert! Burns in MRI patients wearing transdermal patches.
www.ismp.org/msarticles/burnsprint.htm December 7, 2004.
19. Shellock FG, Kanal E: Policies, guidelines, and recommendations for MR imaging safety and patient management. J Magn
Reson Imaging 1:97-101, 1991. Medline Similar articles
20. http://www.fda.gov/cdrh/safety/mrisafety.html
21. http://www.mrireview.com/docs/mrideath.pdf
22. Kanal E, Gillen J, Evans J, et al: Survey of reproductive health among female MR workers. Radiology 187:395-399, 1993.
23. U.S. Food and Drug Administration: Magnetic resonance diagnostic device: panel recommendation and report on petitions
for MR reclassification. Final rule. Fed Regist 54:5077-5078, 1988.
24. Stephenson GM, Freiherr G: Indifference to safety heightens MRI risks. Diagn Imag June:79-83, 1990.
25. Kanal E, Shellock FG, Talagala L: Safety considerations in MR imaging. Radiology 176:593-606, 1990. Medline
Similar articles
26. Devine EC, Cook TD: Clinical and cost-saving effects of psychoeducational interventions with surgical patients: a
meta-analysis. Res Nurs Health 9:89-105, 1986. Medline Similar articles
27. Devine EC, Cook TD: A meta-analytic analysis of effects of psychoeducational interventions on length of postsurgical
hospital stay. Nurs Res 32:267-273, 1983. Medline Similar articles
28. Flaherty JA, Hoskinson K: Emotional distress during magnetic resonance imaging. N Engl J Med 320:467-468, 1989.
11 of 13 11-04-2010 23:48
29. Melendez JC, McCrank E: Anxiety related reactions associated with magnetic resonance imaging examination. JAMA
30. Quirk ME, Letendre AJ, Coittone RA, et al: Evaluation of three psychologic interventions to reduce anxiety during MR
imaging. Radiology 173:759-762, 1989. Medline Similar articles
31. Abrahams JJ, Lange RC: Coil holder and marker system for MR imaging of the total spine. Radiology 172:869-871, 1989.
32. Henck ME, Simpson EL: Superiority of cod liver oil as a marker for lesions in MR imaging of the extremities. Am J
Roentgenol 161:904-905, 1993.
33. Chernish SM, Maglinte DD: Glucagon: common untoward reactions-review and recommendations. Radiology 177:145-146,
34. Brennan S, Redd W, Jacobsen P, et al: Anxiety and panic during magnetic resonance scans. Lancet 2:512, 1988.
35. Quirk ME, Letendre AJ, Coittone RA, et al: Anxiety in patients undergoing MR imaging. Radiology 170:463-466, 1989.
36. Beck A, Emery G: Anxiety Disorders and Phobias. New York: Basic Books, 1985.
37. Thompson MB, Coppens NM: The effects of guided imagery on anxiety levels and movement of clients undergoing magnetic
resonance imaging. Holistic Nurse Pract 8:59-69, 1994.
38. http://www.esmrmb.org/
39. http://www.esmrmb.org/download/03-12-18-EUR-EMFW-st13599r-text-en.pdf
40. Roemer PB, Edelstein WA, Hayes CE, et al: The NMR phased array. Magn Reson Med 16:192-225, 1990. Medline
Similar articles
41. McCauley TR, McCarthy S, Lange R: Pelvic phases array coil: image quality assessment for spin-echo MR imaging. Magn
42. Li W, Storey P, Chen Q, et al: Dark flow artifacts in steady-state free precession cine technique: causes and implications for
cardiac MR imaging. Radiology 230:569-575, 2004. Medline Similar articles
43. Danias PG, Stuber M, Botnar RM, et al: Navigator assessment of breath-hold duration: impact of supplemental oxygen and
hyperventilation. Am J Roentgenol 171:395-397, 1998.
44. Li W, David V, Kaplan R, et al: Three-dimensional low-dose gadolinium-enhanced peripheral magnetic resonance
venography. J Magn Reson Imaging 8:630-633, 1998. Medline Similar articles
45. Prince MR, Erel HE, Lent RW, et al: Gadodiamide administration causes spurious hypocalcemia. Radiology 227:639-649,
46. Rutt BK, Lee DH: The impact of field strength on image quality in MRI. J Magn Reson Imaging 1:57-62, 1996.
47. Alster AD: Field-strength dependence of gadolinium enhancement: Theory and implications. Am J Neuroradiol
15:1420-1423, 1994.
48. Trattnig S, Ba-Ssalamah A, Noebauer-Huhmann I, et al: MR contrast agent in high-field MRI (3 Tesla). Top Magn Reson
Imaging 14:365-375, 2003. Medline Similar articles
49. Brateman L: Chemical shift imaging: a review. Am J Roentgenol 146:971-980, 1986.
50. Brix G, Heiland S, Bellemann SE, et al: MR imaging of fat-containing tissues: valuation of two quantitative imaging
techniques in comparison with localized proton spectroscopy. Magn Reson Imaging 11:977-991, 1993. Medline
Similar articles
51. Pfleiderer B, Ackerman JL, Garrido L: In vivo 1H chemical shift imaging of silicone implants. Magn Reson Med 29:656-659,
1993.
52. Anzai Y, Lufkin RB, Jabour BA, et al: Fat-suppression failure artifacts simulating pathology on frequency-selective
fat-suppression MR images of the head and neck. Am J Neuroradiol 13:879-884, 1992. Medline Similar articles
53. Spielman D, Meyer C, Macovski A, et al: 1H spectroscopic imaging using a spectral-spatial excitation pulse. Magn Reson
Med 18:269-279, 1991. Medline Similar articles
54. Dixon WT: Simple proton spectroscopic imaging. Radiology 153:189-194, 1984. Medline Similar articles
55. Schneider E, Chan TW: Selective MR imaging of silicone with the three-point Dixon technique. Radiology 187:89-93, 1993.
56. Wolff SD, Balaban RS: Magnetization transfer contrast (MTC) and tissue water proton relaxation in vivo. Magn Reson Med
57. Balaban RS, Ceckler TL: Magnetization transfer contrast in magnetic resonance imaging. Magn Reson Q 8:116-137, 1992.
58. Edelman RR, Ahn SS, Chien D, et al: Improved time-of-flight MR angiography of the brain using magnetization transfer
contrast. Radiology 184:359-399, 1992.
59. Tan Hu JI, Sepponen RE, Lipton MJ, et al: Synergistic enhancement of MRI with Gd-DTPA and magnetization transfer. J
Comput Assist Tomogr 16:19-24, 1992. Medline Similar articles
60. Dousset V, Grossman RI, Ramer KN, et al: Experimental allergic encephalomyelitis and multiple sclerosis: lesion
12 of 13 11-04-2010 23:48
characterization with magnetization transfer imaging. Radiology 182:483-491, 1992. Medline Similar articles
61. Dixon WT, Engels H, Castillo M, et al: Incidental magnetization transfer contrast in standard multislice imaging. Magn Reson
62. Chia JM, Fischer SE, Wickline SA, et al: Performance of QRS detection for cardiac magnetic resonance imaging with a
novel vectorcardiographic triggering method. J Magn Reson Imaging 12:678-688, 2000. Medline Similar articles
© 2010 Elsevier
13 of 13 11-04-2010 23:48
NSTRUMENTATION
Ricardo J. Becerra
Michael J. Harsh
Eddy B. Boskamp
Ceylan C. Guclu
Tomas Duby
Timothy J. Havens
R. Scott Hinks
Jeffrey P. Noonan
William T. Peterson
Stanley J. Piepenburg
Robert S. Stormont
Robert M. Vavrek
INTRODUCTION
The discovery of magnetic resonance imaging was a spectacular event in the history of medical imaging. In
1946, two scientists in the United States reported the phenomenon called "Nuclear Magnetic Resonance," or
1-4
"NMR" for short. Felix Bloch and Edward M. Purcell were awarded the Nobel Prize in Physics in 1952. Since
then, the NMR technique has been extensively used to study molecular chemical structure.
In the early 1970s, Raymond Damadian showed the relaxation time differences between normal and cancerous
tissues. In 1972, he patented a method of obtaining a localized MR signal using a so-called "single-point"
5
technique.
In 1973, Paul Lauterbur demonstrated that the NMR technique, when combined with field gradients, can be
6
utilized to image an object. He called this NMR imaging technique "zeugmatography" derived from Greek word
"zeugma," meaning "joins together," due to simultaneous use of static magnetic and radiofrequency (RF) fields.
Shortly after this invention, Mansfield's group in Nottingham devised the selective excitation method with
7,8
gradients, which is frequently used in today's MRI scanners.
Later the term NMR Imaging was changed to "MRI" (Magnetic Resonance Imaging, sometimes abbreviated
further to MR). The word "nuclear" was removed to avoid confusion with other medical imaging modalities
using ionizing radiation.
9
From 1974 to 1976, MRI was mainly used for imaging animals and phantoms. In 1977 the first in vivo cross-
sectional image of a human finger was acquired, followed by an abdominal image by Peter Mansfield and his
10
team.
Since its conception, MRI has witnessed tremendous technical and clinical advancements. This trend continues
today as new technologies and clinical applications are discovered. See the reference section of this chapter
for additional literature about the history of MR imaging.
page 105
page 106
When analyzing an MRI scanner, it should be kept in mind that it is hard to separate the description of MRI
hardware from that of MR physics. Hence, it is recommended that the reader first review the MR physics
chapter to have an understanding of the origin of the MRI signal, and how the phase and frequency of the MRI
signal are controlled. There are three basic field components in MRI: static magnetic field (B0), radiofrequency
field (B1), and the gradient field (Gx, Gy, and Gz). These fields put specific sets of requirements on the
environment where the MRI system is installed. These requirements are known as "siting" and ensure the safe
and proper operation of the MRI system.
The hardware structure of an MRI scanner naturally follows a similar pattern. Figure 4-1 shows a typical
top-level breakdown of the MRI system. The magnet (1) is the source of the B0 static magnetic field. Although
a cylindrical magnet is shown in Figure 4-1 as an example, other configurations are possible as described in
the Magnets section.
1 of 3 11-04-2010 23:49
Add to lightbox
Figure 4-1 Simplified block diagram of a magnetic resonance imaging (MRI) system. 1, Cylindrical magnet; 2,
Gradient coil; 3, RF body coil; 4, Patient; 5, Patient table; 6, Head coil; 7, Transmit/receive chain; 8, imaging
display.
The RF field B1 is either transmitted by the body RF coil (3) or by an anatomy-specific transmit-receive (T/R)
coil. Recent advances in surface coil technology and reconstruction techniques introduced a variety of new coil
11
designs that improve both image quality and reduce scan time, respectively. The gradient fields that are used
for spatial encoding, are generated by the gradient coils (2) in the magnet bore. The echo (or FID) detected by
the receive coil is linked to and sampled by the receiver. It is then directed to a file called the "raw data" file. All
these events happen in synchrony with the pulse sequence prescribed for the exam.
Here is a brief functional overview of the sequence of events for a patient (4) scan in an MRI scanner. The
patient is prepared first by the technologist who takes care of patient safety, and places the coils on the
patient. Once the landmark (origin of the scan location) is set, and the patient is in the scanner, the prescribed
protocol is entered. The protocol is a set of parameter settings for the pulse sequence being prescribed. When
the protocol is loaded, the "pulse sequence generation and control unit" generates the gradient, and RF
waveforms, and prepares the system for data acquisition. The RF pulse generated goes through the RF
amplifier and feeds the transmit RF coil. Similarly, the x, y, and z gradient waveforms are amplified by the
gradient amplifiers (also known as "gradient drivers") and feed the gradient coils in accordance with the
waveforms prescribed by the current protocol. The receive portion of the RF subsystem, consisting of the
receive coil, preamplifier, T/R switch (7), and receiver, samples the echo (or FID) to be written into the raw
data file. Finally at the end of the scan, the raw data is reconstructed and processed to form the final images
displayed on the monitor (8).
page 106
page 107
In the following sections, all of the above subsystems shall be explained, including magnet, RF, and gradient. In
addition, we will elaborate on siting, pulse sequence generator and system calibration.
2 of 3 11-04-2010 23:49
© 2010 Elsevier
3 of 3 11-04-2010 23:49
INTRODUCTION
The discovery of magnetic resonance imaging was a spectacular event in the history of medical imaging. In
1946, two scientists in the United States reported the phenomenon called "Nuclear Magnetic Resonance," or
1-4
"NMR" for short. Felix Bloch and Edward M. Purcell were awarded the Nobel Prize in Physics in 1952. Since
then, the NMR technique has been extensively used to study molecular chemical structure.
In the early 1970s, Raymond Damadian showed the relaxation time differences between normal and cancerous
tissues. In 1972, he patented a method of obtaining a localized MR signal using a so-called "single-point"
5
technique.
In 1973, Paul Lauterbur demonstrated that the NMR technique, when combined with field gradients, can be
6
utilized to image an object. He called this NMR imaging technique "zeugmatography" derived from Greek word
"zeugma," meaning "joins together," due to simultaneous use of static magnetic and radiofrequency (RF) fields.
Shortly after this invention, Mansfield's group in Nottingham devised the selective excitation method with
7,8
gradients, which is frequently used in today's MRI scanners.
Later the term NMR Imaging was changed to "MRI" (Magnetic Resonance Imaging, sometimes abbreviated
further to MR). The word "nuclear" was removed to avoid confusion with other medical imaging modalities
using ionizing radiation.
9
From 1974 to 1976, MRI was mainly used for imaging animals and phantoms. In 1977 the first in vivo cross-
sectional image of a human finger was acquired, followed by an abdominal image by Peter Mansfield and his
10
team.
Since its conception, MRI has witnessed tremendous technical and clinical advancements. This trend continues
today as new technologies and clinical applications are discovered. See the reference section of this chapter
for additional literature about the history of MR imaging.
page 105
page 106
When analyzing an MRI scanner, it should be kept in mind that it is hard to separate the description of MRI
hardware from that of MR physics. Hence, it is recommended that the reader first review the MR physics
chapter to have an understanding of the origin of the MRI signal, and how the phase and frequency of the MRI
signal are controlled. There are three basic field components in MRI: static magnetic field (B0), radiofrequency
field (B1), and the gradient field (Gx, Gy, and Gz). These fields put specific sets of requirements on the
environment where the MRI system is installed. These requirements are known as "siting" and ensure the safe
and proper operation of the MRI system.
The hardware structure of an MRI scanner naturally follows a similar pattern. Figure 4-1 shows a typical
top-level breakdown of the MRI system. The magnet (1) is the source of the B0 static magnetic field. Although
a cylindrical magnet is shown in Figure 4-1 as an example, other configurations are possible as described in
the Magnets section.
1 of 3 11-04-2010 23:51
Add to lightbox
Figure 4-1 Simplified block diagram of a magnetic resonance imaging (MRI) system. 1, Cylindrical magnet; 2,
Gradient coil; 3, RF body coil; 4, Patient; 5, Patient table; 6, Head coil; 7, Transmit/receive chain; 8, imaging
display.
The RF field B1 is either transmitted by the body RF coil (3) or by an anatomy-specific transmit-receive (T/R)
coil. Recent advances in surface coil technology and reconstruction techniques introduced a variety of new coil
11
designs that improve both image quality and reduce scan time, respectively. The gradient fields that are used
for spatial encoding, are generated by the gradient coils (2) in the magnet bore. The echo (or FID) detected by
the receive coil is linked to and sampled by the receiver. It is then directed to a file called the "raw data" file. All
these events happen in synchrony with the pulse sequence prescribed for the exam.
Here is a brief functional overview of the sequence of events for a patient (4) scan in an MRI scanner. The
patient is prepared first by the technologist who takes care of patient safety, and places the coils on the
patient. Once the landmark (origin of the scan location) is set, and the patient is in the scanner, the prescribed
protocol is entered. The protocol is a set of parameter settings for the pulse sequence being prescribed. When
the protocol is loaded, the "pulse sequence generation and control unit" generates the gradient, and RF
waveforms, and prepares the system for data acquisition. The RF pulse generated goes through the RF
amplifier and feeds the transmit RF coil. Similarly, the x, y, and z gradient waveforms are amplified by the
gradient amplifiers (also known as "gradient drivers") and feed the gradient coils in accordance with the
waveforms prescribed by the current protocol. The receive portion of the RF subsystem, consisting of the
receive coil, preamplifier, T/R switch (7), and receiver, samples the echo (or FID) to be written into the raw
data file. Finally at the end of the scan, the raw data is reconstructed and processed to form the final images
displayed on the monitor (8).
page 106
page 107
In the following sections, all of the above subsystems shall be explained, including magnet, RF, and gradient. In
addition, we will elaborate on siting, pulse sequence generator and system calibration.
2 of 3 11-04-2010 23:51
© 2010 Elsevier
3 of 3 11-04-2010 23:51
SITING
Overview
The MRI scanner is a very sensitive device. The process of detecting the minuscule signal emitted
from hydrogen nuclei precessing in a magnetic field creates conditions for sensitivity to external
disturbances. Likewise, many devices in the vicinity of an MRI magnet may be sensitive to the
magnet's fringe field. Siting considers the impact of the site on the scanner, and the impact of the
scanner on the site.
Magnetic Field-Impact of the Scanner on the Site

The first considerations deal with the magnetic fringe field. For safety reasons, an exclusion zone is
defined around the magnet. The exclusion zone is an area where the magnetic flux density is greater
than five gauss. Personnel with cardiac pacemakers, neurostimulators and other biostimulation devices
must not enter this zone. Since the magnetic field is three-dimensional, the floors above and below the
magnet room would have exclusion zones if the five gauss line intruded into these spaces. This was a
problem with early superconductive cylindrical magnets, which lacked the actively shielded coil. Actively
shielded magnets have considerably reduced fringe fields.
The fringe field may adversely affect equipment in close proximity to the scanner. Likewise, equipment
in close proximity may disrupt the performance of the scanner. Proximity limits are distance limitations
placed on equipment to minimize the magnet's fringe field effect on certain equipment. A few examples
of typical proximity limits are listed in Box 4-1.
Magnetic Field-Impact of the Site on the Scanner

Structural steel within proximity of the magnet may have an impact on the homogeneity or uniformity of
the field within the magnet bore. The magnet's field homogeneity is an important criterion that impacts
the image quality of the scanner. Homogeneity requirements for imaging are just as stringent as the
homogeneity requirements for chemical shift analysis (spectroscopy). Since the structural steel usually
cannot be removed, its effect must be countered. One approach is to add additional iron to
magnetically balance out the structural steel. If the iron is placed within the magnet bore, substantially
smaller amounts of iron than that of the structural steel are needed to balance out the disturbance.
The MRI scanner may also be sensitive to a changing magnetic environment, such as large ferrous
objects. When large metal objects are close to the magnet, the magnet's field homogeneity is affected,
sometimes severely. As the large metal object is moved away from the magnet, the severity and
complexity of the magnetic field inhomogeneity decreases. At a far enough distance, only a linear
gradient across the magnetic field may be detected. Further out, this linear gradient disturbance
disappears, but the large metal object may still have an impact on the magnetic field by shifting the
static field strength (B0) slightly. If the object doesn't move, the typical scanner prescan self-calibration
will adjust for this frequency shift. But if the object moves during a scan, there will be a shift in B0,
which may cause some degradation of the image quality.
Moving metal refers to metal objects that move inside of the moving metal sensitivity line during system
scan. Typically, the moving metal sensitivity line is the 3 gauss fringe field line. For example, cars being
driven inside the moving metal sensitivity line are moving metal. However, if a car or truck is within the
moving metal sensitivity line and does not move during a scan, it may not be an issue. An MRI suite
immediately adjacent to a road requires extra scrutiny with regard to siting.
Most alternating current (AC) equipment in sites is not an issue if kept outside the 5 gauss line.
Electrical currents flowing in high voltage power lines, transformers, and large generators or motors
near the magnet can affect the magnetic field homogeneity that is essential to the proper performance
of the MRI system. Magnetic field interference at 50 or 60 Hz usually must not exceed a value in the
milligauss range at the magnet location.
Box 4-1 Proximity Limits
1 of 4 11-04-2010 23:51
0.5 gauss (0.05 mT) or less*

Nuclear cameras
1 gauss (0.1 mT) or less

Positron emission tomography (PET) scanner
Image intensifiers
CRT video display (color, B/W, monochrome)
CT (computed tomography) scanner
Ultrasound
Electron microscope
5 gauss (0.5 mT) or less

Cardiac pacemakers
Neurostimulators
Biostimulators
10 gauss (1 mT) or less

Magnetic tapes and floppy discs
Hard copy imagers
Line printers
Video cassette recorder (VCR)
Film processor
Credit cards, watches, and clocks
X-ray tubes
50 gauss (5 mT) or less

Telephones
Metal detector for screening
LCD color monitor
*At this low level of magnetic field, the earth's magnetic field (~0.5 gauss) will
significantly influence the shape of the 0.5 gauss line.
page 107
page 108
Acoustics and Vibrations-Impact of the Scanner on the Site

A typical MR suite has two types of acoustic noise issues. The first is the acoustics of the room, where
patients and technicians are affected by the noise of the MRI system as the gradients are pulsed. The
second is noise transmitted to surrounding spaces via airborne and structure-borne paths.
Airborne transmission is defined as air excitation within the magnet room. The magnet and gradient coil
generate acoustic noise. Current must pass through the gradient coil to create the necessary imaging
gradients and the gradient coil lies within the magnet bore. The result of the interaction between
current and magnetic field is a force (Lorentz force law) on the coil. This is the same fundamental
principle that makes a loudspeaker work. Noise generated by the MRI system is inherent to the design
of the system. The human ear has a large dynamic range covering orders of magnitude of sound
pressure levels. A logarithmic scale in decibels (dBA) is often used. The sound pressure level at 120
dBA is one million times greater than the threshold of hearing at 0 dBA. Normal speech is about 60
dBA. It is typically a challenge to reduce the sound pressure level by a factor of 10 (20 dBA).
The airborne noise passes through walls via any openings (i.e., small holes, cracks, HVAC ducts, and
wave-guides) into surrounding spaces within, and possibly beyond, the confines of the building.
Acoustic energy can transmit across significant distances. Acoustic noise control to mitigate noise from
being transmitted to other spaces often amounts to paying attention to small details while working with
2 of 4 11-04-2010 23:51
ordinary construction materials. The key objectives are eliminating all cracks and gaps in the wall
construction while making sure that doors, walls, floor, and ceiling have adequate transmission loss via
mass or special double wall construction along with good fitting massive doors.
The structure-borne transmission path is the result of mechanical excitation of the floor/building
structure causing the building to vibrate. The vibration of the surfaces at surrounding spaces then
radiates as acoustic noise.
Upper floor MRI installations represent the largest number of sites that experience structure-borne
acoustic issues. Typically, an architect, acoustic engineer, or project engineer might suggest placing
the magnet on a soft vibration isolator. Although this solution may reduce the magnet-generated
vibration from transmitting into the building structure, this solution can produce image quality issues if
the magnet "bounces" on the soft isolator.
Acoustics and Vibrations-Impact of the Site on the Scanner

Environmental site vibration has the ability to affect the magnet phase stability and ultimately image
quality. Certain MRI procedures require a stable environment to achieve high-resolution image quality.
The vibration effects on image quality can be minimized early in MRI suite site planning. The magnet
may be sensitive to vibration in the frequency range of 0.5 to 100 Hz depending on the amplitude of the
vibration. In the physical area where the MRI system is located, every precaution must be taken to
ensure that the vibration is minimized. In the magnet siting area, the structural stability and behavioral
characteristics should be assessed.
Vibration capable of affecting MR image quality can be classified as either steady-state or transient
vibration. Steady-state vibration typically refers to disturbances caused by rotating machinery.
Examples of machinery known to have generated vibration image quality problems are exhaust fans,
air-conditioning blower units, compressors, pumps (air and water), etc. Steady-state vibration would
be measurable during long time intervals.
Transient vibrations are typically a function of the building structure or the building foundation. They are
associated with vehicular traffic, pedestrian motion, patient transport, door slamming, etc. Transient
vibration would usually occur during short periods of time, and it typically decays from high vibration
amplitude to lower levels. Magnets improperly mechanically coupled to the floor may experience
steady-state and/or transient vibration system stability issues.
Radio Frequency Interference

The MRI system utilizes RF signals from the patient to create the MR image and is vulnerable to
ambient Radio Frequency Interference (RFI). For a 1.5 T scanner, this would be any unwanted RF
signal around 64 MHz. There are many electrical devices that create RF noise across a broad range of
frequencies in the Megahertz range. Spectroscopy may require multiple frequency bands to be clean
for any given field strength. To protect the MRI system from ambient RFI, and to protect any local
equipment from MRI RF transmission, sites require an RF enclosure, which is commonly called an RF
shielded room or an RF screen room. Radiofrequency signals from sources outside of the RF shielded
room are attenuated as they pass into it so they do not interfere with the MRI system. Likewise, RF
signals produced by the MRI scanner are kept from interfering with external RF devices. A typical
shield attenuation value is 100 dB at 64 MHz. This RF quiet environment is necessary for the MRI
system to produce quality images.
Radiofrequency shielded rooms come in a variety of shapes and sizes but one common feature is a
total RF shield produced by one continuous ground plane. This is achieved by making the walls, floor,
doors and ceiling out of an electrically conductive material such as copper, aluminum, brass, or steel.
All the room components (walls, floor, ceiling, etc.) must be electrically bonded together to form one
solid, common ground plane. Once this is established the ground plane is then tied to earth ground to
create the RF shield. This ground point is known as the primary ground. The primary grounding
technique works well for any battery powered devices that may be operated within the RF shield room.
But since the scanner requires the introduction of facility power into the screen room, a change is
3 of 4 11-04-2010 23:51
required. The RF shielded room must now be grounded back to the facility power ground.
page 108
page 109
The addition of water systems or other grounds will connect the RF shield room ground to earth
ground. These additional grounds are called secondary grounds. It is the secondary grounding that
needs to be controlled. If the secondary ground introduces any current to the RF shield, indicating the
RF shield is a better ground path than the secondary ground, then the current can set up electrical
fields on the surface of the RF shield, which may cause image artifacts.
Other Concerns
In the case of a superconducting magnet used in the MRI system, the magnet contains a large amount
of liquid helium at 4 kelvin (K) = -452º fahrenheit (F) or -269º celsius (C). During a quench, the magnet
quickly boils off 100% of this liquid. Consequently, a very large volume of extremely cold helium gas
must be safely vented outside of the building. Failure of the cryogenic vent can result in this cold gas
entering the magnet room or another portion of the building, and lowering the ambient oxygen supply to
an unsafe and potentially lethal level. When a quench occurs, the helium gas will immediately begin to
warm up and expand rapidly. As the gas escapes from the magnet, the helium gas will increase in
temperature from 4.5 K to approximately 10 K to 150 K depending on the location along the pipe. As a
result, the gas will expand to between 25 and 400 times its original volume. Therefore, it is critical that
designers/contractors familiar with industrial piping systems be responsible for the vent system design
and installation.
Temperature and humidity within the scan room must be controlled, and the MRI manufacturer is
responsible for providing the requirements. The air conditioning system must be sufficient to remove
any heat created by the scanner to maintain an optimal temperature for the patient. This is not only a
comfort issue for the patient, but it is a safety issue that relates to specific absorption rate (SAR). The
maximum SAR allowed is a function of the temperature and humidity of the patient environment. The
higher either of these are, the more the peak SAR must be reduced.
Finally, siting an MRI scanner involves some other issues, such as where to place the equipment. The
manufacturer provides the physical, magnetic, electrical data and cryogenic and plumbing (in the case
of superconducting magnet) necessary for planning and preparing a site for a magnetic resonance
imaging system. Work that may be necessary for site preparation is shown in Box 4-2.
© 2010 Elsevier
4 of 4 11-04-2010 23:51
MAGNET TECHNOLOGIES
Overview
The magnet is the heart of the MRI system, and its function is to provide a strong, uniform and stable
magnetic field (B0) for polarizing the nuclei in the patient. Magnetic resonance imaging has been
performed in B0 fields lower than 0.1 T, and as high as 9.4 T; however, most diagnostic imaging is
performed at B0 fields between 0.2 and 3.0 T. The signal-to-noise ratio (SNR) in the image increases
approximately linearly with increasing magnetic field strength. This increased SNR of higher field
scanners can be used to improve image quality or image resolution, or to increase scanning speed.
These benefits have led to an increase in the use of magnets with field strengths of 1.5 T and higher.
Homogeneity
Today, magnet and MRI manufacturers use several methods to specify their products. These
methods, although equivalent, express the homogeneity differently. One of the popular methods is
simply to specify the maximum variation in the field over the imaging volume. This homogeneity
measure is called "peak-to-peak" homogeneity and is typically measured at a number of points on the
surface of the imaging volume with a mechanical probe called a "magnetometer" or an array of such
probes called a "probe array." Since the homogeneity is measured at a finite set of points on the
surface of the imaging volume, it may not measure the absolute minimum and maximum of the field
over the entire volume. The measured peak-to-peak value will approach the true peak-to-peak
homogeneity value as more points on the surface of the imaging volume are included.
A second method of specifying the homogeneity is to take the root-mean-squared (rms) value of the
magnetic field values obtained in the above process. This is called the "surface rms" homogeneity, or
simply the rms homogeneity.
Box 4-2 Site Preparation Work
Site construction or renovation.

Installation of the electrical conduit, junction boxes, ducts, surface
raceways, outlets, and line safety switches.
Installation of nonelectrical lines: water plumbing, helium venting, and air
conditioning.
Installation of air, vacuum and oxygen lines into the magnet room.
Installation of RF shielding in magnet room.
Installation of structural reinforcements.
Installation of selected magnetic equipment.
Installation of water treatment equipment.
page 109
page 110
1 of 15 11-04-2010 23:52
Add to lightbox
Figure 4-2 Homogeneous magnetic field (B0) in a cylindrical magnet. Established by using several
electrical coils. The direction of the resultant magnetic field (B0) is determined by the direction of the
current in the coils, as shown by the arrows. This concept applies to both resistive and
superconducting magnets.
A third method of specifying the homogeneity involves computing the rms value of the homogeneity
across the whole imaging volume. This is called the "volume rms" or "Vrms homogeneity." Since it is not
possible to measure the homogeneity across the entire imaging volume with a mechanical probe, some
manufacturers have devised methods for computing the rms value across the imaging volume with the
imaging system and a large phantom. However, the Vrms can also be computed from the mechanical
field measurements and the mathematical expansion of the magnetic field described in the shimming
section at the end of the magnet section.12 Until all manufacturers adopt a standard measurement
technique, one must examine manufacturers' homogeneity specifications with care. The method used
to express homogeneity in the rest of this chapter is "peak-to-peak."
Image Quality
The uniformity or homogeneity of the B0 field, affects image quality via T2* decay, spectral resolution
and geometric distortion. For example, water and fat molecules have slightly different resonant
frequencies in the same magnetic field, separated by approximately 3.5 parts per million (ppm). In
high-field systems (>0.5 T) this frequency separation is high enough to be resolved and used to
distinguish between fat and water. Thus most high-field fat saturation techniques, used in high-field
systems, for eliminating fat from images require field homogeneity to be less than 3 ppm. Otherwise,
the fat suppression will fail as the homogeneity over the image approaches/exceeds this 3.5 ppm field
variation. Fortunately, most imaging techniques can be performed adequately within field homogeneity
of 10 ppm.
The other property of the static magnetic field critical for image quality is the stability. Image artifacts
such as ghosting are created by variations in the main magnetic field as low as 0.002 μT.
Consequently, the stability of the magnetic field required for imaging is on the order of parts-per-billion
(ppb) in the 1 Hz to 100 Hz frequency range.
2 of 15 11-04-2010 23:52
Add to lightbox
Figure 4-3 Static magnetic field (B0) produced by permanently magnetized pole faces. Orientation of
the magnetic field is perpendicular to the pole faces, vertical in this example.
There are several magnet technologies used in the design and manufacture of MRI magnets that
attempt to meet these requirements for very high, uniform, and stable magnetic fields. Listed in order
of increasing popularity, these technologies are resistive coils (resistive), permanent magnetic material
(permanent), and superconducting coils (superconducting). Resistive and superconducting magnets
both use electrical coils with circulating direct currents (DC) to establish the B0 field (Fig. 4-2).
Permanent magnets on the other hand, use pole faces made of permanently magnetized material to
establish the B0 field (Fig. 4-3).
Resistive Magnet Technology

page 110
page 111
Technically, resistive coil magnets are the simplest to understand and manufacture. A large and stable
DC power supply is required to keep circulating currents in the resistive coils to establish and maintain
the B0 field. In addition, a large cooling system is used to remove the large amount of heat dissipated
by the resistive coils. Typical stability of the power supply is about 100 ppm and the power required
ranges from 40 to 100 kW of electrical power, which is primarily dissipated by the resistive coils. Thus,
the stability of the B0 field depends directly on the stability of the power supply and, therefore, it is
usually poor. In addition, the large heat dissipated by the resistive coils limit the strength of the B0 field
in whole-body MRI systems. The resulting poor B0 field stability, together with the limitations of
resistive heating in the magnet winding, severely restricts the usefulness of resistive magnets. In some
cases a resistive magnet's field is amplified by the addition of "soft" magnetic material such as iron.
The iron is called "soft" because the material is magnetized while a background magnetic material is
supplied, but decays after the background magnetic field is removed. Iron can be used in the "poles" of
a resistive magnet in a manner similar to that used in a permanent magnet. Resistive magnets are
typically smaller and of lower strength than superconducting magnets.
Permanent Magnet
Next to superconducting magnet technology, the second most popular technology for MRI magnets is
permanent magnetic material. This technology can be used to produce MRI magnets up to
approximately 0.5 T. Permanent magnetic material is also called "hard" magnetic material, because,
3 of 15 11-04-2010 23:52
once magnetized, it maintains its magnetization almost indefinitely. While several permanent magnetic
materials have been known for some time, such as Alnico, barium ferrite (BaF12O19), and the
rare-earth samarium cobalt (SmCo5), it wasn't until June of 1983 when Sumitomo Special Metals
(SSM) of Japan announced the development of a new rare-earth permanent magnetic material,
neodymium-iron boron (NdFeB) that a permanent magnet technology for MRI systems became real, as
it is known today.
Sagawa et al of SSM subsequently published a paper describing the compositional and processing
details of NdFeB.13 This new material can be magnetized to a higher field per unit volume ("maximum
energy product" BHmax) enabling higher field strength magnets with lower weight. One disadvantage of
this rare-earth material is that its temperature sensitivity is much higher than the other materials listed
above, but its ability to produce higher field per unit volume is considered to out-weigh this
disadvantage. Therefore, most magnets manufactured for use in MRI systems today use
neodymium-iron boron. Permanent magnetic material manufacturers are continually striving to improve
the properties of this material to enable MRI magnet developers to design magnets with higher and
more stable magnetic fields and lower weights.
Add to lightbox
Figure 4-4 Permanent magnet and its associated components. The iron yoke and posts serve as
mechanical structure and flux return. The pole pieces shape the B0 field. The permanently magnetized
material is the source of magnetic flux, and the heaters keep the magnet at a constant temperature.
MRI magnets made of NdFeB typically have iron "pole pieces" for shaping the magnetic field, and iron
yokes to support the pole pieces. The layout of a typical permanent MRI magnet is shown in Figure
4-4. These iron yokes serve not only as mechanical support, but also as flux return paths for the
magnetic field. These flux return paths help increase magnetic field in the imaging volume, while
reducing stray field from the magnet. The magnetic properties of both the iron and permanent
magnetic material are temperature dependent. The magnetization of the iron changes roughly 100 ppm
per degree celsius, and the magnetization of NeFeB has a temperature coefficient an order of
magnitude higher. Thus, to produce an MRI system with a stable field, the temperature of the magnet
must be controlled. Temperature control is accomplished by a closed-loop temperature control system
composed of heaters and thermostats mounted on the magnet, and a precise controller. This
temperature controller maintains the temperature of the magnet above the magnet room temperature
and keeps the temperature constant within 0.1° C. The heaters of the permanent magnet MRI system
are also shown in Figure 4-4. This figure depicts typical permanent magnet geometry, which is, in
4 of 15 11-04-2010 23:52
general, more "open" than a cylindrical magnet. Thus, permanent magnet MRI systems are also
frequently referred to as "open MRI" systems. Most of the permanent magnet geometries produce a
vertical B0 field, compared with cylindrical magnets that produce a horizontal B0 field. This B0 field
orientation causes a significant impact in the design of RF coils, which will be discussed later in the RF
section.
Fields achieved in the imaging region of commercial permanent magnet MRI systems range up to 0.4
T, with the most popular models in the 0.2 T to 0.35 T range. Benefits of the permanent magnets
include low initial purchase cost, lower maintenance costs, and its inherent openness, which some
patients prefer. The typical vertical patient gaps in whole-body systems are on the order of 45 cm,
with imaging volumes on the order of 40 cm diameter spherical volumes. Design geometries include,
C-shaped, 4 posts, and two-legged systems with posts at various angles between 90 and 180
degrees. The magnet shown in Figure 4-4 is a two-legged system with posts at 180 degrees. The
weight of a typical 0.2 T whole-body system is approximately 10 tons, and the weight of a 0.35 T
system is approximately 20 tons. The 5 G stray field region of these whole-body systems is normally
constrained to an area within approximately 3 m of the center of the magnet.
page 111
page 112
Add to lightbox
Figure 4-5 Example of a dedicated MRI system optimized for orthopedic applications. The system has
field strength of 0.2 T and a field-of-view (FOV) in the order of 15 cm. The weight of this magnet is only
2 tons, and the 5 gauss line is only about 1.5 m from iso-center. This MRI product is the E-SCAN
manufactured by Esaote.
In addition to the whole-body systems discussed above, smaller orthopedic systems have recently
been introduced. One example of these extremity systems is shown in Figure 4-5. These systems offer
the benefits of even lower cost and ease of siting. The B0 field of these systems is around 0.2 T and
the homogeneity volumes approximately 15 cm diameter spherical volumes. The weights for these
systems are on the order of 2 tons, and 5 G stray field lines are approximately 1.5 m from the magnet
center.
Superconducting Magnet Technology

Superconducting magnet technology is by far the most popular technology used in the MRI industry
today, especially in high-field MRI systems. In 1911, the Dutch physicist Kamerlingh Onnes discovered
that the electrical resistance of mercury fell to zero as it was cooled below 4.15 kelvin. This complete
loss of resistance was termed "superconductivity," and many more materials have since been identified
that also exhibit superconductivity. Each of these superconductors exhibits zero electrical resistance
below a certain temperature, called its critical temperature. In addition to temperature, the background
5 of 15 11-04-2010 23:52
magnetic field also affects the superconductivity of the superconductors. As the background field
increases above a certain value called its "critical field," a superconductor becomes resistive and
looses its superconductivity properties. Table 4-1 lists critical temperatures and critical fields for a few
elements.
Table 4-1. Critical Temperatures and Critical Fields

Element Critical Temperature (K) Critical Field (mT)
Indium (In) 3.41 28.7
Tin (Sn) 3.72 30.9
Mercury (Hg) 4.15 41.2
Tantalum (Ta) 4.48 82.9
Lead (Pb) 7.18 80.4
The critical fields listed in Table 4-1 for the pure elements are all well below the field strengths required
for MRI magnets. These pure elements are labeled Type I superconductors. There is a second class
of superconductors, called "Type II superconductors" that maintain their superconducting properties in
high magnetic fields. Two commercially relevant Type II superconductors are niobium-titanium (NbTi)
and niobium-tin (NbSn). The critical temperature of NbTi is 9 K and the critical field at 4.2 K is
14
approximately 11 T.
Other types of superconductors have been discovered with critical temperatures above the
temperature of liquid nitrogen (77 K). These types of superconductors are called "high temperature
superconductors" (HTS). Today, the wire manufacturers are actively working to develop commercially
viable high temperature superconducting wires, but there are still significant cost and manufacturing
challenges. There are high hopes in the magnet community that this type of superconducting wire might
be manufactured at a cost low enough to be competitive for use in MRI magnets. While it would seem
easier to manufacture magnets using the superconductor with the highest critical temperature, most
MRI magnets are manufactured using NbTi wire due to cost and reliability.
The superconducting wire is made by embedding many small strands of superconductor (also known
as filaments) approximately 0.1 mm diameter in a copper matrix approximately 2 mm in diameter.
Figure 4-6 shows the relative size of the superconducting filaments and the copper matrix.
When the superconducting wire is below its critical temperature, all of the current flows through the
superconducting filaments. On the other hand, if the temperature of any section of the superconducting
wire is raised above its critical temperature, the current flow transfers to the copper matrix avoiding
damage to the superconducting filaments. This "copper stabilizer" significantly reduces a magnet's
propensity to quench due to small disturbances of the coils.15
One exception to the use of NbTi is the magnet used in the GE SP interventional system shown in
Figure 4-7. The windings in this magnet are made of Nb3Sn, allowing the magnet to operate at 10 K.
At this temperature, it is possible to eliminate the use of cryogens and use conduction cooled technique
to maintain the coils below their critical temperature. As the technology of manufacturing high
temperature superconductors advances, one can expect to see more conduction cooled MRI magnets
manufactured from these newer materials.
page 112
page 113
6 of 15 11-04-2010 23:52
Add to lightbox
Figure 4-6 Cross-section of a superconducting wire showing the superconducting filaments
embedded in a copper matrix. The copper is the primary current conduction path for temperatures
above the critical temperature of the superconducting wire. For temperatures below the critical
temperature, the current flows in the filaments.
During the 1970s, significant advancements were made in the manufacturing of superconducting wire.
In the 1980s, superconducting whole-body MRI magnets were developed. However, these magnets
had large fringe fields and large amounts of iron were needed around the magnet or in the magnet
room walls in order to contain the 5 G line inside the magnet room. This method for containing the
fringe fields is known as "passive magnetic shielding." In the 1990s, a secondary set of
superconducting coils was added to the magnet to contain the fringe fields as close to the magnet as
possible. This technology is known as "actively shielded." Figure 4-8 shows the relative location of the
primary coils, also known as "main coils" and the secondary coils, also known as "shield coils." Actively
shielded magnet technology has eliminated or greatly reduced the need for passive shielding in high
field MRI systems.
Today, superconducting magnet technology is the only technology used for producing whole-body MRI
systems above 0.7 T. Commercially available whole-body MRI superconducting magnets are available
in field strengths from 0.3 T to 3 T. In addition, custom systems are available at even higher field
strengths, limited only by the critical field of the superconducting wire.
7 of 15 11-04-2010 23:52
Add to lightbox
Add to lightbox
Figure 4-7 Dedicated MRI system for MRI-guided interventional and surgical procedures. The
system's imaging volume is located between the two halves of the magnet (A). The surgeon has
access to the patient from both sides of the magnet to perform the intervention (B). This magnet
operates at 0.5 T and uses conduction cooling technology, requiring no helium.
Shown in Figure 4-9 is a picture of a 9.4 T magnet produced by GE Medical Systems to perform
high-field MRI research at the University of Illinois at Chicago. The total length of wire required to wind
this magnet was 540 km or 335 miles. Note that while a superconductor does not have resistance, it is
impossible to manufacture wire 300 miles long. Thus magnet manufacturers must join shorter wire
lengths with superconducting joints. If any of these joints are not completely superconducting, the
magnet will exhibit some downward drift. A large number of joints will increase the probability of
downward field drift; therefore, long lengths of wire are very critical. Manufacturers typically specify
that the magnet will not drift downward in field strength by more than 0.1 ppm/hour. This extraordinary
field stability is achieved because of the "zero resistance" property of superconductivity, and is one of
the reasons why superconducting magnet technology dominates the MRI industry. Once current is put
into the magnet, and the superconducting switch is closed, the power supply is removed. Thus the
magnet's field stability is not governed by a power supply as it was in the case of resistive magnets.
Another benefit of the "zero resistance" property of superconductors is its ability to react to external
field disturbances. As mentioned previously, field variations of less than 1 μT cause image quality
issues. Moving metal objects near a magnet can perturb the field enough to cause image quality
problems. However, special superconducting coils can be wound in the superconducting magnet to
eliminate more than 90% of this external field fluctuation. These coils are frequently called "moving
8 of 15 11-04-2010 23:52
metal" coils.
page 113
page 114
Add to lightbox
Figure 4-8 Concept of actively shielded superconducting magnet. The shield coils carry current in the
opposite direction to the main coils. The field produced by the shield coils opposes the main field
resulting in much smaller fringe field.
Most use of superconducting magnets has been applied to magnets of cylindrical geometry. However,
recently, the need for combining the benefits of high-field superconductivity capability in open magnets
has been explored. Today, it has been shown that open superconductiving magnets with field strengths
up to 1 T are achievable. Several manufacturers now sell superconducting open magnets with
strengths above 0.5 T. Shown in Figure 4-10 is a GE OpenSpeed system, which has a field strength of
0.7 T. This system provides the openness of a permanent magnet system but with much higher SNR.
Cryogenics
As stated above, NbTi is the superconductor used in most MRI magnets, and this superconductor must
be maintained below 9 K to remain superconducting. In fact, most manufacturers immerse the
superconducting windings in a liquid helium bath at 4 K. The requirement to keep the superconductor at
4 K adds another demand on the design of the superconducting magnet. Liquid helium is a limited
resource, and quite costly in many parts of the world. To reduce ongoing operating costs,
manufacturers design superconducting magnets to minimize the loss of liquid helium. Typical boil-off
rates range from 0.03 liters/hour to 0.1 liters/hour, although some manufacturers are incorporating
re-condensing technology into their designs. This technology virtually eliminates the loss of liquid helium
while the magnet is in standard operating mode.
9 of 15 11-04-2010 23:52
Add to lightbox
Figure 4-9 Ultrahigh-field magnet for whole-body MRI. This one of a kind magnet operates at 9.4 T.
The system was installed at University of Chicago Illinois in 2003.
Add to lightbox
Figure 4-10 High-field open MRI system. This system uses an open magnet operating at 0.7 T. The
field strength is generated by superconducting coils, which are helium cooled.
The cryogenic design that achieves these low helium losses includes features that minimize the heat
transfer, such as convection, conduction and radiation, from the environment to the inner magnet
cartridge. A typical magnet includes an outer vacuum vessel, one or more thermal shields, multi-layer
insulation, a helium vessel, the magnet cartridge, a suspension system that supports the thermal
shields and helium vessel, and a cryocooler for removing heat from the magnet. A simplified cross-
section of a typical magnet is shown in Figure 4-11.
page 114
page 115
10 of 15 11-04-2010 23:52
Add to lightbox
Figure 4-11 Cross-section of a high-field superconducting magnet. The cross-section shows the
location of the shield coils relative to the main coils. The location of the gradient coil and RF body coil,
in the magnet, are also shown.
The space between the outer vacuum vessel and the helium vessel is evacuated to eliminate heat
transfer by convection. The thermal shields and multi-layer insulation intercept the thermal radiation
before it reaches the helium vessel from the environment and transfers the heat to the first stage of the
cryocooler, which operates at approximately 50 K. These thermal shields are typically made from a
conductor with high thermal conductivity, such as aluminum. The high thermal conductivity is required to
minimize the temperature difference across the thermal shield. Unfortunately, high thermal conductivity
also implies high electrical conductivity, and any leakage field from the gradient coil induces eddy
currents in these thermal shields. The effects of these eddy currents are reduced during the calibration
phase when eddy current compensation is set.
The suspension system is designed to minimize heat transfer by conduction, while providing enough
stiffness to eliminate vibration. The suspension system consists of thin straps or rods, made from
materials high in strength, yet low in thermal conductivity.
Finally, the second stage of the cryocooler is connected to a second thermal shield (in some
manufacturer's designs) or to the helium re-condenser in "zero boil-off" systems. This cryocooler
removes the heat from the magnet that cannot be eliminated by the design steps listed above. A
recent advance in cryocooler technology is the "pulse tube." Two advantages of the pulse tube are
reduced vibration and reduced acoustic noise.
Ramping
11 of 15 11-04-2010 23:52
Add to lightbox
Figure 4-12 Superconducting switch used to ramp-up and ramp-down the magnet. During ramp-up or
ramp-down, the heater is on to open the switch and allow current from the power supply to reach the
magnet coils. Once the magnet is up to field, the heater is turned off and the switch closes, effectively
shorting out the power supply. At this point, the power supply can be removed.
As stated above, a superconducting magnet only needs a power supply to install current into the
magnet. Once the current is installed in the magnet, the power supply can be removed, and the
magnet will maintain its field. The act of installing current into the magnet is called "ramping" the
magnet. A circuit diagram of a typical superconducting magnet is shown in Figure 4-12. Installing
current in the magnet requires a switch that can be "opened" while ramping the magnet, then "closed"
after current is installed in the magnet. Closing the switch allows the current to circulate in a closed
superconducting loop.
This switch must be made of superconductor material so that the closed circuit has no resistance when
in operation. The switch is actually a small winding of superconducting wire imbedded in an insulator.
The switch is opened by adding heat so that the temperature of the switch winding is increased above
the superconducting transition temperature. While the switch is above the superconducting transition
temperature, it is resistive, and the current from the power supply will preferentially flow into the
superconducting coil windings rather than the resistive switch. After the magnet has been energized,
the switch is closed by allowing the switch to cool below the superconducting transition temperature.
At this point the switch again has zero resistance and the power supply can be removed. A similar
operation is used for putting current into the superconducting shim coils used by some manufacturers.
Quench
page 115
page 116
If at any time during operation, any section of the windings of a superconducting magnet are disturbed
enough to raise its temperature above the superconducting transition temperature, that winding section
will become resistive. Once resistive, this section of the superconducting winding will generate heat,
heating more of the winding until all of the energy stored in the magnet is turned into heat. This process
happens very rapidly, over a period of seconds and is called a "quench." Surprisingly, an energy
disturbance as small as 2 μJ is enough to cause a quench, yet the total energy released can be as
high as 130 MJ for the 9.4 T whole-body magnet cited above. The forces on the large coils of a typical
1.5 T actively shielded magnet are on the order of 900,000 Newtons and large coil forces on a 3 T
actively shielded magnet are on the order of 3.5 million Newtons.
As the superconducting winding becomes resistive and heats up, it causes the liquid helium to boil,
expand, and exit the magnet in a burst of helium vapor. Superconducting magnets are designed to
expel this helium vapor safely through vents to the outside of the hospital. During a quench, it is
12 of 15 11-04-2010 23:52
possible for thousands of liters of liquid helium to turn rapidly to gas and gush from the magnet. As
described in the siting section, it is very important that the vents carrying the helium gas from the
magnet room are installed properly, and not modified or obstructed, as this could be a life-threatening
situation.
Shimming
After the magnet is installed and ramped, it is "shimmed." Shimming is the act of bringing the magnet's
homogeneity into specification. While a magnet may have been shimmed to specification in the factory,
there is always some magnetic material in the hospital building that modifies the magnetic field of the
magnet. Examples of this magnetic material are steel structural beams and concrete re-inforcing bars
in the building. This environmental steel can modify the homogeneity of the magnet from 10 ppm over
the imaging volume to 100 ppm or more. Some manufacturers complete this shimming operation with
superconducting correction coils incorporated in the magnet, this technique is called "super-con"
shimming; other manufacturers accomplish the shimming operation using small pieces of ferromagnetic
material, this technique is called "passive" shimming. In either case, the ramping and shimming phase
of the magnet installation is completed over a period of a few days, depending on the type of magnet.
The magnetic field of a magnet can be written as an expansion of mathematical functions as shown
below:
The summation over n and m is from 0 to some upper limit, typically on the order of 20. The terms with
lowest n,m are the terms with the largest contribution to the magnet inhomogeneity that can be
modified during the shimming process, and the first few terms are shown below:
This expansion can also be written in Cartesian co-ordinates, and the first few terms can be expressed
as follows:
In either formulation, the first term is a constant and represents the magnet's base field, such as 1.5 T.
If possible, all terms but the first term should be zero, leaving a perfectly uniform field. In practice, all
terms higher than the first are lowered to meet specifications during the installation "shimming"
process. However, the human body has magnetic properties and each patient modifies the
homogeneity of the magnet to some extent. This effect increases with field strength. Manufacturers
have designed their MRI systems to allow some correction of this effect for each patient as will be
explained below.
The first parentheses [lcub ][rcub ] in Equation 4-3 groups a set of three terms that vary linearly over
the homogeneity volume. The gradient coils also vary linearly over the imaging volume, thus the
gradient coils can be used to reduce these terms for improved homogeneity for each patient. The next
five terms grouped in the second set of parentheses [lcub ][rcub ] have quadratic variation over the
imaging volume. Some imaging systems include resistive coils to reduce this field error for each patient
also. These coils are usually called "resistive shims" or "high-order shims." Normally there is one coil
for each of the five second-order terms, and they are frequently labeled by the term they correct. Thus
they are labeled XZ, YZ, Z2, XY and (X2-Y2) in the Cartesian system. Alternatively, they are labeled
(2,0), (2,1), (2,-1), (2,-2), and (2,2) in the spherical expansion system. The terms with n equal to three
and higher are reduced as much as possible during the shimming process at installation and in general
are not adjusted on a per-patient basis.
Magnet Technology Summary
13 of 15 11-04-2010 23:52
Table 4-2. Magnet Type Comparison for MRI Systems

Resistive Permanent Superconducting
Field strength Low-field Low and mid-field Broad range of fields
<0.3 T ≤0.5T Mid: 0.5-1.0 T
High: 1.5-3.0 T
Ultrahigh: 7-10 T
Homogeneity Medium Medium High
≈40 ppm ≈40 ppm <10 ppm
Imaging volume ≈35 cm DSV ≈40 cm DSV ≥50 cm DSV
Stability Low Medium High
Limited by the Dependent on the Inherent to the technology
power supply temperature controller
Fringe fields Large Low Low with active shield
Requires Inherent to the Large without active shield
magnetic technology
shielding
Weight Low Very large Mid to large
<2 tons 10-35 tons 1.5 T ≈3 tons
3.0 T ≈10 tons
Emergency Very fast Not possible Fast
shutdown
After power is Can not be tusrned off Emergency quench <1
removed minute
Electrical power High Low Low
40-100 kW Temperature controller Air or water cooled 5-10
kW compressor
Initial costs Low Medium High
Operating costs High Low Medium
Electrical power None Helium refills
page 116
page 117
Three main magnet technologies are used to manufacture magnets for MRI systems: Resistive,
Permanent and Superconducting. Each magnet technology has its own pros and cons. The right
selection often depends on the application and scope of the MRI product. Table 4-2 summarizes the
advantages and disadvantage of each technology. Today's MRI industry is being dominated by the
superconducting technology, particularly in the high-field systems. Superconducting magnet technology
has a high potential to move into high temperature technology if and when the cost and manufacturing
(reliability) issues are resolved. The second growing technology is permanent magnet with potential
usage of higher "maximum energy product" (BHmax) magnetic materials, which could enable higher field
strengths while keeping weight down. The challenge permanent magnet technology faces is controlling
the high temperature sensitivity and higher cost associated with higher BHmax materials. Resistive
magnet technology is the least used technology today for whole-body MRI systems. The inherent poor
stability and high electrical power consumption severely limit the application of resistive magnet
technology in today's whole-body MRI products.
14 of 15 11-04-2010 23:52
© 2010 Elsevier
15 of 15 11-04-2010 23:52
GRADIENT SUBSYSTEM
Overview
The gradient subsystem enables spatial resolution of the MRI signal. The two main components of the
gradient subsystem are the gradient coil and the gradient driver. Slew rate, peak amplitude, duty
cycle, acoustic noise, and heat are determined by the combination of the gradient coil and driver. The
gradient coil and driver should be designed as one subsystem for optimal performance and to avoid
problems controlling waveform fidelity or with excessive heat dissipation in the gradient subsystem.
Gradient amplitude defines the maximum field strength of a gradient system assembly. It is measured
in mT/m (millitesla per meter) or G/cm (gauss/cm, where 1 G/cm = 10 mT/m). Gradient slew rate is
the rate of change of the gradient strength. It is expressed in T/m/s. For example, the slew rate of a
gradient subsystem that can reach gradient amplitude of 50 mT/m in 0.5 ms is:
Gradient Coils
The gradient coil assembly is comprised of three independent gradient coils. They are usually referred
to as X, Y, and Z gradient coils. Each coil generates a magnetic field gradient that is constant within
the imaging volume and proportional to currents Ix, Iy,, and Iz. Mathematically, the z component of the
magnetic flux density, Bz, at point (x, y, z) inside the imaging volume is given by:
where: Gx, Gy, and Gz are the strength or gain of X, Y, and Z gradient coils, respectively, in units of
mT/m/A or G/cm/A. B0 is the z component of the magnetic flux density at the isocenter of the magnet.
The Z gradient varies the magnitude of the Bz field along the z-direction while the X or Y gradient
varies the Bz magnitude along the x- or y-direction. The variation of the Bz field along an orthogonal
axis (x or y) is why the X and Y gradient coils have a different geometry than the Z gradient coil.
page 117
page 118
Linearity is the gradient field uniformity inside the field-of-view (FOV). It has a direct impact on image
geometrical distortion. The following two linearity definitions are widely used:
1. Linearity or integral linearity affects the perceived location of a voxel. If at some point inside the
imaging volume the gradient generates a magnetic field of Bz(x, y, z) per ampere, while a field
of Gzz per ampere was expected, then integral linearity is defined as the relative difference of
these two fields:
2. Differential linearity affects MRI signal intensity due to the change in gradient strength across a
voxel. It is defined for the X gradient, for example, at point (x, y, z) inside the imaging volume
as:
where: Bz(x, y, z) is the magnetic field generated by the X gradient coil per 1 A at this point.
Cylindrical Gradient Coils

As mentioned earlier, the Z coil has a different geometry than the X and Y coils. The axial or
longitudinal gradient coil, the Z coil, consists of a number of current loops wound on the surface of a
cylinder, usually connected anti-symmetrically with respect to the Z = 0 plane (for a loop at location z1
that carries current I, there is also a loop at location - z1 that carries current - I).
1 of 9 11-04-2010 23:53
The transversal gradient coils, the X and Y coils, consist of a number of saddle-shaped coils mounted
on the surface of a cylinder. These coils are almost identical but mounted at a 90° angle relative to
each other.
Add to lightbox
Figure 4-13 A Maxwell pair of coils on radius a. The optimum linearity occurs at a spacing of 1.31a
between the two coils. This coil pair is a rudimentary axial, or Z, gradient coil.
Since the appearance of the first cylindrical MRI systems in the early 1980s, the gradient coil
assembly has undergone major changes. The first cylindrical gradient coils consisted of a Maxwell coil
pair acting as Z gradient coils and two sets of two arc Golay coils for X and Y gradient coils.16 These
coils, coupled with the gradient drivers of that time, typically had a 5 mT/m maximum strength with a
rise time of 1 ms (slew rate = 5 T/m/s). In addition, their fringe field was not contained, and they
required a significant pre-emphasis in the current waveform to overcome the adverse effect of eddy
currents in conductive surfaces of the magnet. Eddy currents are electric currents induced in metallic
objects by time varying magnetic fields, such as gradient waveforms. These currents generate
magnetic fields that oppose the gradient field change. Figure 4-13 shows a Maxwell pair of coils on
radius a. Mathematically, it can be determined that the optimum linearity occurs at a spacing of 1.31a
between the two coils. Note the asymmetry in the current direction with respect to the Z = 0 plane.
Figure 4-14 is an example of a two arc gradient coil. Such a coil generates a linearly varying Bz field in
the Y direction. The X gradient is rotated by 90° about the Z axis. This gradient coil consists of four
"window frame" shaped saddle coils mounted on the surface of the cylinder.
A significant improvement came with the introduction of continuous surface current density patterns for
longitudinal and transversal gradient coils17 and the invention of actively shielded gradient coils.18 The
actively shielded gradient coil is designed to minimize eddy currents by surrounding the outside of the
gradient coil with additional coil windings and creating a field counter to that of the gradient. The shield
is designed to cancel the external or fringe field of the gradient coil, with optimal cancellation at
conductive surfaces within the magnet. However, the shield coil with its reversed polarity decreases
the gradient strength in the patient bore and more current is needed to overcome this effect. The
benefit of substantially reduced eddy currents more than offset this one disadvantage. Figure 4-15
depicts a view of a transversal gradient coil with distributed current density on a cylindrical surface.
The third generation of gradient coils perfected the shielded gradient coil design to better match human
anatomy. Several gradient coils were grouped in one assembly: a large FOV whole-body gradient coil
and a smaller FOV gradient coil with a higher slew rate.19,20 Figure 4-16 is an example of a dual FOV
gradient coil.
2 of 9 11-04-2010 23:53
Add to lightbox
page 118
page 119
Figure 4-14 The physical shape of transverse gradient coils, such as the four "window frame" shaped
saddle coils mounted on the surface of the cylinder, are quite different from the axial gradient coil.
Add to lightbox
Figure 4-15 This flat copper sheet, when rolled into a cylinder, forms a transverse gradient coil. Nearly
the entire surface is conductive, resulting in a distributed current density on a cylindrical surface.
3 of 9 11-04-2010 23:53
Add to lightbox
Figure 4-16 These two simple cylinders represent a large field-of-view whole-body gradient coil and a
smaller field-of-view specialty gradient coil.
Planar Gradient Coils
Add to lightbox
Figure 4-17 One half of a planar square-shaped Z-gradient coil. The other half is located on the
opposite magnet pole but with the current directions reversed.
4 of 9 11-04-2010 23:53
Add to lightbox
Figure 4-18 One half of a planar square shaped X-gradient coil pair. The other half is located on the
other pole face with the same winding pattern and current direction.
Planar Gradient Coils are used in open type magnets. Their topology is therefore quite different from
cylindrical coils21 as listed below:
1. Typically the Z gradient coil consists of two disks or square-shaped windings placed on
opposite sides of the open magnet pole pieces. To achieve the Z gradient coil effect, the
current directions are reversed. Figure 4-17 is an example of a planar square-shaped Z gradient
coil. The figure shows one of the boards. The other board, located on the other pole face, has
the same winding pattern with reversed current polarity.
2. X and Y gradients consists of two sets of D shaped coils rotated by 90° to each other. Figure
4-18 shows an example of a planar square shaped X gradient coil. The figure shows one of the
boards. The other board, located on the other pole face, has the same winding pattern and
current direction. Y gradient coil is rotated by 90°.
page 119
page 120
Gradient Driver
The gradient driver is usually discussed in terms of volts and amperes. But the control of this power
must be precise and accurate. The gradient controller takes on this task.
Gradient Controller
The gradient controller provides signals to three gradient amplifiers to provide an accurate, linear
current output to vary the gradient amplitude over time. How well the gradient in the imaging volume
follows the current waveform is referred to as "waveform fidelity." Achieving instantaneous current
accuracy requires high resolution, speed, and accuracy of the digital-to-analog converter and the
current-sensing circuit. Because of the spatial and temporal manipulation of the magnetic field, with its
corresponding impact on the proton precession phase, it is necessary to also accurately control the
product of gradient strength vs. time. This corresponds to the area inside of a gradient amplitude-
over-time gradient waveform. Gradient control methodologies use an integrator to accumulate and
correct for any error in this area.
Gradient drivers achieve highly accurate control using a closed-loop, current feedback control
architecture (Figure 4-19). A digital representation of the desired gradient command is passed through
a highly accurate digital-to-analog converter. The analog command, IREF, is compared with the actual
sensed current in the gradient coil, IFEEDBACK. The difference, IERROR, is amplified and integrated. The
5 of 9 11-04-2010 23:53
sum of the error and the integral become the control signal for the gradient amplifier, VCONTROL. The
amplifier then generates a voltage proportional to the control signal to drive the gradient coil. The
voltage is the electromotive force to drive the current into the coil.
Both the integrator and the digital-to-analog converter must be accurate at DC levels, and must also
respond accurately and quickly to transient events. Typically, the two devices must be accurate and
linear to a few parts per million over their full-scale range. This level of control accuracy is not common
in other industries and is a significant design challenge for MRI gradient controllers.
Add to lightbox
Figure 4-19 A typical closed-loop, current feedback control functional block diagram.
Another limitation lies in the bandwidth of the gradient feedback control loop. Bandwidth determines the
controller's ability to correct for error accumulated at the IERROR signal. At low frequencies, where the
closed-loop current control has very high gain, the error between the commanded current and detected
current is minimal. At frequencies near the bandwidth of the closed-loop control, the gain of the
closed-loop current control is much lower, which results in a relatively slow response to errors.
Feedback systems cannot accurately follow commands with frequency content above their bandwidth.
This circumstance has a significant impact on the gradient driver's ability to manage instantaneous
accuracy during transient current commands. To reduce the effects of this fundamental limitation, the
frequency content of the gradient waveform can be limited.
As mentioned before, eddy currents generate magnetic fields that oppose the gradient field change.
The opposing magnetic field has multiple time dependencies unique from the gradient driver because
the induced currents decay in time as a function of the different electrical resistances of the various
magnet components. There is no impact to the current through the gradient coil or the current sensed
by the controller. So the controller is unaware of the error in the gradient field. Figure 4-20 shows the
relationship of gradient coil current and the resulting gradient magnetic field in the presence of eddy
currents.
Uncorrected eddy currents have a detrimental effect on the MRI image quality. Eddy currents are
predictable since they are generated in the invariant magnet structure. Once the impact on imaging has
been measured, the three gradient current waveforms can be pre-emphasized with the extra current
necessary to produce the desired magnetic flux shape within the patient bore of the magnet. This
method is effective only if the opposing eddy current magnetic fields are spatially similar to those of the
gradient coil. Due to the variable location and proximity of various metallic components of the magnet
that surround the gradient coil, the eddy-current field spatial response is often more complicated than
the gradient field.
6 of 9 11-04-2010 23:53
Add to lightbox
page 120
page 121
Figure 4-20 The actual gradient magnetic field can be significantly distorted from the intended field in
the presence of eddy currents.
Gradient Amplifier
The gradient amplifier boosts the current and voltage delivered to the gradient coil. It converts the
error signal of the closed-loop current control into a high voltage. This voltage is applied to the gradient
coil to force the current to the required amount. The gradient coil inductance, resistance, and the
requested peak slew rate determine the maximum voltage needed from the amplifier. The maximum
current required is determined by the requested peak gradient amplitude and gradient coil gain. The
voltage bandwidth of the amplifier is one of the elements that determines the maximum closed-loop
bandwidth of the current control.
Since the frequency content of typical MRI pulse sequences is mostly in the audio range, the first
systems used existing linear audio amplifiers. Most linear power amplifiers use power bipolar junction
transistors (BJT) biased into their linear region. These amplifiers have the advantage of high voltage
bandwidths in the 100 kHz range, and are capable of very high fidelity with almost any type of
waveform. However, they are very inefficient and have limited voltage and current capability. Many of
these devices have to be connected in parallel in order to get a useable peak current. Their inefficiency
arises from their variable resistor behavior. Considerable power can be dissipated across this
resistance from the high voltages. Linear amplifiers modules are limited to about 150 V and 250 A.
To overcome the efficiency issues of linear amplifiers, switching amplifiers were developed. These
amplifiers drive the power transistors into either a low resistance "on" state or very high resistance
"off" state. Neither state dissipates as much power as bipolar junction transistors. In order to control
the amplitude, the transistors are switched continuously at a very high rate between on and off. The
amount of dwell time on vs. off determines the amplitude. This approach is known as pulse width
modulation (PWM). Switching transistors such as insulated gate bipolar transistors (IGBT) and metal
oxide silicon field effect transistors (MOSFET) are well adapted to this type of operation, and are the
devices now used for switching gradient amplifiers. Switching amplifiers for MRI have been designed
for voltages in excess of 2000 V and currents exceeding 500 A. The on/off nature of a switching
amplifier requires the switching frequency to be as high as possible to limit the current ripple on the
gradient coil. Ripple frequencies in the 50 kHz to 100 kHz range are typical, and provide acceptable
voltage bandwidths with minimal degradation of MR image quality.
Gradient Duty Cycle

The term "gradient duty cycle" describes the percentage of time the gradient subsystem can provide
maximum gradient strength pulse sequence waveforms. MRI systems typically have a duty cycle of
100%. This duty cycle applies to the pulse sequences provided on the system. A more strenuous
7 of 9 11-04-2010 23:53
criterion is the ability of the gradient driver to sustain maximum amplitude. This is important for diffusion
imaging where high amplitude diffusion pulses reduce the echo time. Often, the pulse sequence must
be modified to limit the diffusion pulse amplitude at the expense of longer echo times. Duty cycle is
often limited by the ability to remove the heat generated by pulsing gradient waveforms, whether in the
gradient coil or in the gradient driver.
Transient heating and cooling is characterized by the thermal time constant, τ, in the expression:
where 63% of a steady-state temperature is reached when the amount of elapsed time, t, is equal to
one time constant, τ. After three time constants a system is considered at steady state. For example,
if the thermal time constant of a gradient coil is 15 minutes, the coil will reach 95% of steady-state
temperature at 45 minutes (3 time constants).
Components in an MRI system have thermal time constants ranging from very short (milliseconds) to
long (>15 minutes). Peak gradient strength and duty cycle limits are calculated based on where the
heat is generated, and the time constant of the respective heat-generating components. Gradient
driver and coil heating are two important thermal considerations as both generate heat during scanning
that must be removed.
Components in the driver used to convert facility electricity into current waveforms have thermal time
constants ranging from 30 milliseconds for power electronics to minutes for other components.
Gradient coil heating is on a much longer time scale than the gradient driver. Typically, the thermal time
constant ranges from 15 to 30 minutes, where steady-state temperatures are reached in over an hour.
The amount of heat generated in an axis can be approximated as:
where Irms is the root-mean-square current on the axis and Raxis is the resistance of the coil axis, which
is a function of the wire temperature (T) and the frequency content (f) of the scan sequence. Skin
depth effects reduce the effective cross-sectional area of the current conducting wire, thus increasing
the coil resistance as frequency increases. Skin depth effects are proportional to the square root of
frequency. Additional heating can come from eddy currents.
The amount of heat removed from the gradient coil depends on the capability of the coil cooling system
and the temperature limits of the materials used in the coil. The properties of copper with respect to
temperature and frequency are given in Table 4-3.
Table 4-3. Copper Properties (Baseline is 20° C)

Temperature (°C) Resistance Increase (%) Frequency (Hz) Resistance Increase (fold)
40 7.8 1 (≈DC) 1.0
60 15.6 10 3.2
80 23.4 100 10
100 31.2 1000 32
120 39.0 10000 100
page 121
page 122
8 of 9 11-04-2010 23:53
Add to lightbox
Figure 4-21 The peak specific absorption rate (SAR) limit as a function of bore temperature and
humidity.
Finally, for patient safety reasons, the FDA limits peak SAR as a function of bore temperature and
humidity. A cooler bore temperature allows higher peak SAR (Fig. 4-21).
Acoustics
Gradient coils, when pulsed, generate acoustic noise. The source of the noise is the Lorenz force,
which is the vector product of the current and the static magnetic field at the conductor carrying the
current. Gradient waveform ramp times are often in the hundreds of microseconds. The major
frequency content of a waveform often lies within the most sensitive region of human hearing at about
1000 Hz.
Various methods have been proposed to lower the acoustic noise level of the gradient coil
assemblies.22 Attenuation and absorption of acoustic energy is always fundamental to any method.
Future Developments
The evolution path of further developments seems to be the continuation of the historical trends: the
quest for stronger and faster gradient coils surely will challenge a number of currently accepted
technological limits. As the physical and regulatory limits are reached with today's gradient coils, a
more revolutionary approach in coil design is needed. This approach, whatever form it takes, would
define the next generation gradient coils.
© 2010 Elsevier
9 of 9 11-04-2010 23:53
RADIOFREQUENCY SUBSYSTEM
Overview
The main functions of the RF subsystem are the spins excitation (transmit phase) and the "echo" reception
(receive phase). The excitation function is performed by the MRI system transmit chain, and the reception is
performed by the receive chain. During the excitation, the spins of the desired imaging region are flipped into
the transverse plane. During the receive phase, as the flipped spins relax back to align themselves in the
direction of the B0 field, they generate a small RF signal (MRI signal) which is detected, amplified and
processed by the receive chain components. In this section, the terms f0 and ω0 are used interchangeably: f0
refers to the Larmor-frequency in Hz, and ω0 refers to the Larmor-frequency in radians.
Transmit chain
The function of the transmit chain is to produce an RF magnetic field (B1) to excite the spins in the sample
being imaged. The polarization of the B1 field can be either linear or circular. In linear polarization, the B1 field
has a fixed orientation, orthogonal to the B0 field. On the other hand, in circular polarization the B1 field
rotates orthogonal to the B0 field at a rate equal to the Larmor-frequency, f0 (64 MHz for 1.5 T). Typically,
linear polarization requires the RF power to be delivered to the transmit coil as a single component, whereas
the circular polarization requires the RF power to be delivered in two components of equal amplitude and 90°
phase difference (I and Q components). Feeding the power in this manner is usually referred to as
"quadrature excitation."
Linear polarized B1 fields require twice the amount of RF power than circular polarized B1 fields to flip the
spins to a given angle. This is an important advantage of circular polarization over linear. The transmit RF
coils used to produce linear B1 fields are usually referred to as "linear" transmit coils and the coils used to
produce circular B1 fields are called "quadrature" transmit coils. Because circular polarization is twice as
efficient as linear, it is the most widely used by all MRI manufacturers. The transmit chain discussion in this
section is based on circular polarization of the B1 field produced by quadrature transmit RF coils also known
as "resonators."
page 122
page 123
1 of 16 11-04-2010 23:54
Add to lightbox
page 123
page 124
Figure 4-22 A high-level functional block diagram of the transmit chain: 1, RF synthesizer; 2, the mixer; 3,
power amplifier; 4, T/R switch; 5, 3 dB hybrid splitter; 6, transmit RF coil.
The primary characteristics of the B1 field include a stable frequency (f0), finite bandwidth, high uniformity
across the imaging volume, and the field intensity measured in micro-tesla (μT). To meet these requirements,
the transmit chain uses the following components (shown in Fig. 4-22): 1. A highly stable frequency
synthesizer to generate a continuous wave signal (carrier) at the Larmor-frequency (ω0). 2. An RF envelope
2 of 16 11-04-2010 23:54
modulator to convert the carrier signal into an RF pulse with finite bandwidth. 3. A power amplifier to elevate
the RF pulse from mW to several kW and a means to monitor the energy content of the transmitted RF
power. The power monitor measures the forward power going into the coil to ensure safe operating levels. 4.
A transmit/receive switch (T/R SW) to couple the power amplifier to the transmit coil during the transmit
phase while protecting the receive chain components, and to couple the receive coil(s) to the receive chain
while decoupling the power amplifier from the transmit coil. 5. A power splitter and phase shifter to split the
amplified RF pulse into two components of equal amplitude and 90° phase difference (I and Q); this type of
power splitter is know as "3 dB 90° hybrid splitter." 6. A quadrature whole-body RF transmit coil, which is fed
by the I and Q components to generate the B1 field. Figure 4-22 shows the primary functional blocks of the
transmit chain and their corresponding MRI functions.
Most MRI manufacturers use a whole-body size resonator operating in quadrature to generate the B1 field.
The transmit field has to be highly homogeneous across the imaging volume to provide a uniform flip angle (α)
to a selected slice or slab. The flip angle (α) is proportional to the integral of the RF pulse.
During the excitation pulse, linear gradients are applied such that they are superimposed on the static
magnetic field. Since the Larmor-frequency is proportional to the amplitude of the static magnetic field,
exciting a slice or a slab requires that RF pulses have a finite bandwidth.
A uniform flip angle over the width of the selected slice is needed to ensure a rectangular slice profile. To
achieve this, the RF carrier is modulated with a sinc function (defined as sinx/x) in the time domain,
corresponding to a rectangular pulse in the frequency domain. Since a sinc pulse is infinitely long (extending
from -∞ to +∞), it is usually truncated after one or two side-lobes, and convolved with a filter to minimize
ringing in the slice profile.
Frequency Synthesizer
The radio frequency synthesizer generates a sine wave at or near the center of the MRI frequency band of
interest (e.g. 64 MHz at 1.5 T). Its function is to provide the highly stable local oscillator for frequency
translation of the exciter modulation, and to provide the demodulating local oscillator required to translate the
MRI information to a suitable frequency for digitization. In addition, the frequency synthesizer has to be able
to provide fast and accurate phase switching and frequency offsets to the center frequency during multislice
scans and off-center FOV imaging.
Add to lightbox
Figure 4-23 A functional block diagram of an MRI direct digital synthesizer. The function of the synthesizer is
to generate an accurate reference frequency for a given set of phase (ϕ), center frequency (ωcenter), and
offset frequency (ω offset). The carrier signal, generated by the synthesizer, is modulated with the RF envelope
from the pulse sequence generator.
3 of 16 11-04-2010 23:54
Most current MRI systems employ a high-resolution direct digital synthesis technique to allow rapid coherent
frequency and phase switching. A coarse synthesizer is typically used to provide gross frequency setting
depending on the field strength and nuclei of interest. The high-resolution synthesizer is used during the MRI
sequence for dynamic phase and frequency control (e.g. during multislice excitation and off-center FOV
reception). Figure 4-23 shows a functional block diagram of a typical MRI frequency synthesizer.
Modulator
An amplitude modulator is used to impart a finite bandwidth onto the fixed frequency sine wave provided by
the synthesizer. This is accomplished by multiplying the synthesized sine wave by the RF pulse envelope (e.g.
sinc pulse). The spectra of the modulation and the sine wave are convolved as a result of this amplitude
multiplication, resulting in translation of the modulation spectrum to the center radio frequency of interest (see
Fig. 4-22). A common modulation waveform is the sinc (sinx/x) pulse, which provides a uniform band of RF
excitation (for slice or slab selection, as an example). A digital modulator is typically employed, which
multiplies a digitized version of a sine wave with the digital modulation waveform before digital to analog
conversion. This can avoid many of the imperfections in the analog multiplier circuits and provide high
waveform fidelity. Frequency and phase modulation can be accomplished in the direct digital synthesizer by
adding a recurring phase increment and a fixed phase step respectively (see Fig. 4-23).
The frequency multiplier and mixer are fundamental functional blocks commonly used in communication
systems. A brief description is included in Figure 4-22. For details on the operation of these functional blocks,
refer to the material in the references.23-27
page 124
page 125
Power Amplifier
The function of the RF power amplifier is to amplify the modulated MRI excitation signal to a level sufficient to
generate the required circularly polarized RF magnetic field (B1). Peak B1 amplitudes between 15 and 30 μT
inside a whole-body transmit coil require a peak power of 15 to 20 kW in a typical 1.5 T whole-body MRI
system. The RF power requirement is proportional to the square of the B0 strength. This becomes a very
important consideration for higher field strengths (e.g. 3 T) whole-body MRI systems.
Other significant requirements for the MRI RF power amplifier include the linearity, stability, and efficiency.
The linearity is very important to preserve the fidelity of the modulated pulse, since the accuracy of the slice
profile depends on it. The amplifier stability contributes to the repeatability of the MRI pulse sequences. The
efficiency of the amplifier is important to maintain the cooling requirements, packaging size and cost at a
practical level. Class A amplifiers are linear but inefficient (typically 25%). Class B amplifiers are more
efficient (up to 60%), but linearity suffers from signal "crossover distortion," which is not acceptable for
high-quality MRI applications. In class AB amplifiers, the operating point is adjusted to eliminate the crossover
distortion, yet provide efficiency in the 50% range. Therefore, class AB amplifiers are commonly used for MRI
systems.26,28,29
Power Monitor
The function of the power monitor is to ensure that the amount of average power delivered to any patient
during an MRI scan is within the safety limits set by the FDA. Particularly, at 1.5 T and higher field strengths,
most of the RF power going into the body coil will end up being dissipated in the patient. Therefore, the FDA
has set limits on the specific absorption rate (SAR), specified in watts per kg. The power monitor uses the
patient's weight and the efficiency of the transmit chain as inputs to calculate the maximum allowable average
power for each patient.
T/R Switch
The T/R switch is part of both the transmit chain and the receive chain. It has two functions: 1. During the
transmit phase, it couples the RF power amplifier to the input of the 3 dB hybrid splitter ("hybrid splitter") and
at the same time it provides protection to the delicate receive chain components. 2. During the receive phase,
the T/R switch couples the hybrid splitter to the receive chain and decouples the RF power amplifier from the
hybrid splitter. Figure 4-24 shows a typical implementation of a T/R SW.
4 of 16 11-04-2010 23:54
Add to lightbox
Figure 4-24 A basic circuit for a transmit/receive (T/R) switch. During the transmit phase, D1 and D2 are on
(switch closed), coupling the RF power to the in out ports of the hybrid splitter. During the receive phase both
diodes are turned off (switch open) by applying appropriate bias, coupling the hybrid output to the input of the
receive preamplifier.
Referring to Figure 4-24, PIN diodes D1 and D2 are used as electronic switches. PIN diodes are diodes with
an additional intrinsic (I) layer, which gives the PIN diode its characteristic of being able to transfer high
amounts of RF power when a small DC current is applied to forward bias (switch closed) the diode. To turn
off the PIN diode (switch open), it requires a bias voltage to reverse the diode. During the transmit phase, D1
and D2 are turned on and both diodes become short circuits and D1 couples the RF power amplifier to the
hybrid splitter. Similarly D2 shorted to ground protects the receive preamplifier from the high transmit power.
During the receive phase, both diodes are open circuit and thus D1 decouples the RF power amplifier from the
hybrid splitter. Similarly, D2 open results in effectively coupling the receive preamplifier to the hybrid splitter.
Hybrid Splitter
The 3 dB hybrid splitter has two functions: 1. During the transmit phase, it converts the output of the RF
power amplifier into two components (I, Q) of equal amplitude and 90° phase difference. These signals drive
two input networks on the transmit coil that are 90° apart, thus generating a circular polarized B1 field. 2.
During the receive phase, the hybrid reverses its operation; it takes the two MRI signals from the I and Q
ports of the body coil and combines them into a single component of higher amplitude and feeds this signal to
the receive chain components for signal amplification and imaging processing.
The hybrid splitter can be built using transmission line segments or lumped elements. 25,30 Figure 4-25 shows
a typical hybrid splitter using 4 quarter-wavelength (λ/4) sections implemented with lumped elements.
Quadrature Transmit Radiofrequency Coils

The transmit RF coil (RF body coil) takes the I and Q power components from the hybrid splitter and converts
them into a homogeneous circular polarized RF Magnetic field (B1) with sufficient volume to cover most of the
population range.
page 125
page 126
5 of 16 11-04-2010 23:54
Add to lightbox
Figure 4-25 A lumped-element diagram of a typical 3 dB-hybrid splitter circuit. The hybrid can also be built
with transmission lines.
Homogeneity is the most important design parameter in the RF body coil, since an inhomogeneous transmit
field would result in contrast variation over the FOV. Homogeneity can be calculated using the Biot-Savart
law, or from the vector potential which is the integration over the current paths in all conductors that make
24
up the body coil.
In Equation 4-12, are the current elements, and is the distance to the matrix point (ds [Lt ] r). The B1
field can also be calculated from the vector potential by using . This is a DC approximation, since
the field is calculated for a direct current in vacuum. However, it is fairly accurate for frequencies up to 64
MHz. An example of Biot-Savart calculation of a 16-rod birdcage resonator is shown in Figure 4-26. Above
1.5 T the homogeneity of the B1 field is affected by the electrical properties of the patient's tissue. Therefore,
other methods such as finite-difference time domain (FDTD) or finite-element method (FEM) must be applied
31,32
to calculate the B1 field map.
In FDTD and FEM methods the load characteristics, electrical permittivity and conductivity of the tissue, can
be included in the calculation. In most cases these techniques are approximations to the solution of Maxwell's
equations, but because they include the effect of the load, they yield more accurate field calculations than a
pseudo-static application of the Biot-Savart method, especially at field strength above 1.5 T. For instance
loading the birdcage head coil with a head-phantom, the momentary B1 field map looks similar to that shown
in Figure 4-27 for 1.5 T and Figures 4-28 and 4-29 for 3 T and 7 T respectively.
6 of 16 11-04-2010 23:54
Add to lightbox
Figure 4-26 An example of a Biot-Savart calculation of a 16-rod birdcage resonator is shown. This method is
a good approximation for 64 MHz (1.5 T). For higher frequencies, field maps are generally computed using
numerical techniques, such as finite-difference time domain (FDTD) and finite-element method (FEM).
Add to lightbox
page 126
page 127
Figure 4-27 The B1 field plots simulated from a birdcage head coil loaded with a human head at 1.5 T
showing excellent B 1 uniformity.
Add to lightbox
Figure 4-28 The B1 field plots simulated from a birdcage head coil loaded with a human head at 3 T showing
B 1 inhomogeneity.
Birdcage Resonator (Transmit Phase)
7 of 16 11-04-2010 23:54
Add to lightbox
Figure 4-29 The B1 field plots simulated from a birdcage head coil loaded with a human head at 7 T showing
significant B1 inhomogeneity. Note the gradual increase of the swirling at the center due to the dielectric
resonance.
Theoretically, perfect transverse field homogeneity can be obtained for coils used in cylindrical magnets by
having an axial current on the surface of an infinitely long cylinder, with amplitude that varies as the sine of the
azimuth angle θ.33 The birdcage is an approximation in which the current in the rungs (Fig. 4-30) varies as:
Since the coil is not infinitely long, the homogeneity in the Z direction (axial) is not as good as in the XY plane
(see Fig. 4-26). To make this structure resonant, capacitors can be placed either in the rungs for "low pass"
configuration, or in the end-rings for "high pass" (see Fig. 4-30). Capacitor placement affects the E field
34
generated by the coil, which is important when minimizing local power deposition.
Add to lightbox
8 of 16 11-04-2010 23:54
Add to lightbox
page 127
page 128
Figure 4-30 The structure of a typical birdcage RF coil. The coil can be designed to be either (A) low pass
(capacitors on the rungs), or (B) high pass (capacitors on the end-rings). When excited, this coil generates a
B1 field on the transverse-plane (xy plane) because of the sinusoidal current distribution along the azimuthal
direction (T).
Vertical open magnets require a totally different type of resonator.35 Since the B1 field has to be orthogonal
to the B0 field, the coil geometries are different than that for horizontal B0. In addition, since in either B0
orientation (horizontal or vertical), the patient lies in a horizontal plane (ZX), the RF coil design becomes more
challenging, particularly for quadrature coils. A popular geometry for vertical B0 field is the "solenoid." A
solenoid RF coil is normally placed in the magnet to produce a linear B1 field in the Z direction. A significant
disadvantage of the solenoid coil is that it can't be operated in quadrature and, therefore, to generate a
circular polarized B1 field, usually a secondary coil must be combined with the solenoid to produce a second
B1 field in the X direction. These two fields must be of equal amplitude and must have 90° of phase difference
to generate the circularly polarized B1 field rotating around the vertical B0. These design challenges also apply
to receive-only coils, such as surface coils and multi-coil arrays. Manufacturers of RF coils are creating very
innovative geometries to overcome these difficulties and to take advantage of the quadrature benefits.
Radiofrequency Shield
The primary function of the RF shield is to provide an isolated environment (free from coupling) for the RF
coils to operate. The RF shield separates the transmit coil from the gradient coil environment. Its purpose is
to prevent RF energy from being dissipated in the gradient coil, which is a high loss environment at radio
frequencies. At the same time, the RF shield has to be transparent for the gradient fields. Since gradient
fields are switched with rise-times on the order of 0.5 ms, the shield has to be transparent for frequencies of
several kHz. Otherwise eddy currents may be induced in the shield, causing distortion to the gradient field and
giving rise to artifacts in the image.
Some RF shield designs contain cuts that are carefully placed to break large copper surfaces to prevent eddy
currents, while leaving the RF current distribution intact. Other RF shield designs consist of ultra thin copper
sheets or copper wire mesh. An important consideration during the design of an RF coil and shield, is the
distance between the coil and the shield "gap" since the B1 field homogeneity and the coil efficiency depend
36
greatly on it.
Tuning and Matching

To optimize the power transfer between networks in a RF subsystem, all network interfaces and transmission
lines have to have the same characteristic impedance (Z0), typically 50 ohms. In general, when designing a
transmit coil, attention is focused on the geometry and B1 performance required of the coil structure. The coil
elements are "tuned" or resonated utilizing additional reactive components (see Fig. 4-30), allowing current to
flow efficiently in the paths required to generate the desired B1 field geometry. This often leaves a network
with a different drive-point impedance than the coaxial and equipment interfaces that provide power to the
coil. Unless proper impedance "matching" is provided, much of the power provided to the coil for B1 field
generation would be reflected at this "mismatched" interface and the desired B1 field amplitude would not be
achieved.
9 of 16 11-04-2010 23:54
Add to lightbox
Figure 4-31 A surface coil loop with a blocking network circuit. When the diode is on (short circuit), L2
resonates at the Larmor frequency with the capacitor to which it is connected. This introduces a high
impedance in the loop, effectively decoupling the receive coil loop.
The coil is usually "matched" to the interface impedance of 50 ohms for best transmit chain efficiency. This is
generally accomplished by the application of additional loss-less reactive components between the tuned coil
structure and the 50-ohm interfaces. Figure 4-30 shows a typical birdcage resonator with the placement of
the tuning capacitors (C), at the end-rings for "high pass" configuration, and in the rungs for "low pass"
configuration. A simplified high pass resonator circuit is shown in Figure 4-31. Several distributed capacitors
(CT) along the coil inductance (L1) are used to tune the coil. The impedance matching is done with a single
capacitor CM. However, other impedance matching techniques are possible. 29,37,38
Other impedance issues often encountered in the receive chain come from the fact that RF coils typically do
not have a well-defined ground. This is especially true for receive coils that need to be moved within the
patient bore. The only well-defined ground is at the end of the cable, where it is plugged into the system. High
voltage may occur on the coil side of the cable shield, setting up a standing wave in the coaxial shield. To
prevent this, a balanced to unbalanced transformer can be put in series with the cable. This device, typically a
parallel resonant circuit, puts high impedance on the cable shield but lets the signal through unaffected. 39
Transmit Coil Disabling

During the reception phase, the transmit coil must be disabled, if receive-only coils are used, to avoid coupling
between transmit and receive coils. If the transmit body coil were not disabled during the receive phase,
energy would transfer between the two coils severely impacting the SNR.
Typically, resonators are enabled and disabled by means of PIN diode networks that are part of a parallel
resonant circuit, as in Figure 4-31. When the diode is forward biased, the parallel circuit becomes resonant
(the inductance L2 is chosen such that the resonance frequency of the blocking network is the same as that of
the imaging coil) and will, therefore, create high impedance at its insertion points, effectively disabling the
resonator. During imaging with receive only coils, the enabling and disabling of the RF body coil has to occur
very rapidly, this function is usually called "dynamic disable."
page 128
page 129
Receive chain
The function of the receive chain is to detect and digitize the MRI signal for processing and image
reconstruction. The MRI signal is coupled into the RF receive coil. Next the MRI signal is amplified by a high
gain, low noise amplifier typically known as a "preamplifier." Once the signal is amplified, it is fed into the
receiver-demodulator for signal normalization, demodulation, and digitization. The output from the demodulator
is the complex raw data for the reconstruction process. Figure 4-32 is a simplified block diagram of the
functions of the receive chain.
Receive Radiofrequency Coils

Radiofrequency coils can be classified in many ways depending on their function. One category is based on
their T/R properties: 1. transmit, 2. receive, and 3. transmit-receive. Another category is based on their
design technology and shape: 1. volume coil, 2. surface coil, 3. phased-array (PA) coils, and 4. parallel
imaging optimized PA coils. The last two types are actually multi-coil arrays using the surface coil technology.
In the previous section, transmit RF coils were explained from a transmit chain point of view. In some cases,
the same coil may be used to transmit and receive (T/R coils). This section discusses T/R coils in receive
10 of 16 11-04-2010 23:54
mode and receive only coils.
In an RF coil functioning as a receive coil, there are two important design factors: SNR and receive field
homogeneity. These are the equivalent to transmit power efficiency, and transmit field homogeneity for the
transmit coils, respectively. In some cases global homogeneity is sacrificed at the expense of higher local
SNR in special anatomy-specific coils such as the cervical-thoracic-lumbar (CTL), and shoulder coils.
One of the most important characteristics of receive coils, is the quality factor (Q), since it correlates directly
to the SNR. Q is a measure of the efficiency of the coil, and it is typically expressed as the ratio of the
resonant frequency of the coil (f0) to the bandwidth (BW) in unloaded coil conditions (Q = f0/BW3dB).
Considering the coil parameters only, a very general proportionality expression for SNR can be written as:
where: 1/QL = total losses; 1/Q0 = loss in the empty coil; 1/Qpat = losses in the patient; Veff is effective
volume of coil; and ROI is region of interest.
The SNR is proportional to the square root of the loaded Q (QL). Optimum SNR results when the patient
noise is dominant-e.g., for 1.5 T-the loaded QL shall be more than a factor of five smaller than the unloaded
Q0. The effective volume of the coil, Veff is the volume integral of the magnetic energy in the entire volume
normalized on that of the region of interest (ROI), and is not the same as geometric volume. However, in
general this means that the coil has to fit the anatomy of interest. When the coil is loaded with a patient, there
are four types of loss that affect the loaded Q: 1. Coil losses: if high Q coil components are used, these
losses are usually small; 2. Dielectric losses: these are caused by parasitic capacitance between the patient
and the coil, and can generally be avoided by choosing the tuning capacitance an order of magnitude higher
than the parasites; 3. Radiation losses: these are usually small except for frequencies much higher than 60
MHz; 4. RF field induced losses: this is by far the most important loss contributor above 0.3 T, and is caused
by the RF magnetic fields. In a typical RF coil at 1.5 T, >90% of all losses are due to this mechanism.
Birdcage Resonator (Receive Phase)

Most whole-body MRI scanners use a transmit-receive coil. The most popular design is the birdcage coil
operated in quadrature. Similar to the transmit phase, in the receive phase the quadrature coil offers
significant performance benefit over coils operating in linear mode. The birdcage coil operating in quadrature,
provides over 40% higher ( ) SNR than the coil operating in linear mode.
Surface and Multi-Coil Arrays

page 129
page 130
11 of 16 11-04-2010 23:54
Add to lightbox
page 130
page 131
Figure 4-32 A basic functional block diagram of the receive chain showing the primary processing steps of
the MRI signal in the time domain and its corresponding frequency domain representation. The output of the
quadrature receiver is a set of complex numbers that are written in file called "raw-data" for a subsequent
image reconstruction.
12 of 16 11-04-2010 23:54
Add to lightbox
Figure 4-33 In phased-array (multi-coil array) coils, to minimize the next nearest coupling, the preamplifiers
used on each coil element should have a very low input impedance (<5 ohms). Therefore, the inductor shown
at the input of the preamplifier resonates with the capacitor shown to effectively decouple the loop from its
neighboring elements.
Individual single loop surface coils are still used by MRI manufacturers, and they are mostly the building
blocks of the multi-coil, also known as "phased-arrays." These arrays consist of a large number of small coil
loops together covering a large field-of-view. 40,41 Instead of a single large receive coil, these arrays combine
the SNR advantage of the smaller coil with the large FOV of a large coil. Data from the individual receive coils
are read out simultaneously, and the image is reconstructed using an algorithm that weighs each coil's
contribution depending on its individual SNR. Most of the multi-coil arrays are receive-only coils. In other
words, the body coil or a larger coil is used to transmit the RF pulses, whereas the multi-array coils receive
the MRI signal. Since both transmit and receive coils are tuned at the same frequency, the receive coils must
be decoupled from the transmit coil during the excitation. Otherwise a large current will be induced in the
receive coil, which might damage the delicate front-end of the preamplifier, and distort the transmit field.
Protection networks, also known as "blocking networks" are used to protect the preamplifier during the
transmit phase. In addition, in multi-coil arrays, the individual receive-coil loops have to be decoupled from
each other during the receive phase. The receive-decoupling from neighboring receive coils is done by using
low impedance preamplifiers in the coils, as shown in Figure 4-33. This method minimizes coupling among the
next-nearest neighboring coils.
Similarly, overlapping the coil elements until the net flux from one loop through the other loop become zero
can minimize the coupling between the neighboring coil loops. The upper plot in Figure 4-34 shows an
overlapping configuration.
Using the decoupling technique shown in Figure 4-34, and minimum coupling technique, the individual coils do
not couple inductively and there is minimal energy transfer between them, i.e., no noise is transferred. This
means that each coil maintains the SNR as though it was operating by itself.
Special Coil Designs for Faster Imaging (Parallel Imaging)
13 of 16 11-04-2010 23:54
Add to lightbox
Figure 4-34 One technique to decouple neighboring coil elements is to overlap them such that the flux from
one cancels the other. This overlapping design is used very frequently in phased-array coil designs. However,
it is not the optimal design for SENSE/ASSET applications (i.e., parallel imaging).
Recently new parallel imaging techniques have been introduced for speeding up the imaging process. One of
42,43
the techniques, called SENSE (SENSitivity Encoding) skips lines in k-space to accelerate the acquisition
time. This process shrinks the FOV, which may now be smaller than the sensitivity volume of the coil
elements. As a result, aliasing of the signal occurs, and the signal from a certain element may be confounded
with that of other elements. In SENSE reconstruction technique, the knowledge of the coil sensitivity profiles
and the noise correlation matrix are used to unfold the individual images and reconstruct the final image.
Special optimized coil arrays are needed for this technique in which the coil elements are designed and
positioned such that signal received from the same area by each coil element is significantly different in
amplitude and phase from that received by the other coil elements. The SNR in the accelerated image is
degraded compared with the original image as given in Equation 4-16.
where R is the reduction factor (acceleration factor) and g is the

geometry factor. The geometry factor is very dependent on coil design, and it is found that coils that are
separated in the L-R and A-P direction give better g-factor maps than coils that are partially overlapped to
eliminate mutual induction. The g-factor maps can be calculated from a matrix product of the sensitivity matrix
and the noise correlation matrix as follows:
Here S is the sensitivity matrix and the noise correlation matrix. Note that the g-factor map is dependent on
the reduction factor R. As an example, see Figure 4-34 for a four-channel coil in which the elements are
overlapped in the acceleration direction, and Figure 4-35 for a case where they are separated. Both cases
show a g factor map for R = 3 in the axial plane inside an elliptical phantom with phase-encoding vertical. This
indicates that the separated configuration gives better g maps.
page 131
14 of 16 11-04-2010 23:54
page 132
Add to lightbox
Figure 4-35 The coil elements optimized for parallel imaging result in no coil-coil overlap for the best g-factor.
This is quite different to the overlap required for minimum coupling in Figure 4-34. The acceleration direction
is the same as coil separation direction.
Hybrid Splitter (Receive Phase)

As explained in the transmit chain hybrid splitter section, the hybrid splitter circuit is used to split the RF
power into I and Q channels, which are 90° apart, during the transmit phase. Similarly in the receive phase,
the same hybrid splitter can function as a combiner to combine the signals from I and Q receive coils. It is
important to note that the I and Q signals from the coil are still in quadrature (90° apart). The hybrid splitter
provides additional phase shifts to the I and Q channels and combines them in-phase. If the RF coil is a
quadrature transmit-receive (T/R), then the same hybrid splitter circuit can be used for both splitting the RF
power during transmit and combining the MRI signal during receive. In quadrature phased-array coils,
however, various signal combination techniques can be used.44
The Preamplifier
The preamplifier(s) resides in the front of the MRI receive chain. Its primary purpose is to increase the MR
signal and its accompanying noise from the receive coil, to a level that dominates the effects of later receive
chain noise sources. The noise figure(NF) is defined as the loss in SNR as the information passes through a
stage of the system. A modern MR preamplifier will provide an NF of <0.5 dB. The receive chain stages that
follow can have higher noise figures since their effects are reduced by the gain of the stages that preceded
them. For a complete receive chain with N stages, the total NF will be:
where: F = 10NF/10
It is evident that the system noise figure will approach that of the preamplifier (STAGE1 in the total noise
figure expression), if the gain of the preamplifier is sufficiently high. Therefore, a low noise figure, high gain
preamplifier is essential for preserving the SNR of the received signal.
15 of 16 11-04-2010 23:54
Additionally, the preamplifier may be designed to exhibit very low input impedance. This low impedance (<5
ohms) is utilized in phased-array receive antennas to reduce current in adjacent coil elements, limiting the
coupling between them.
Receiver
Once the MRI signal is amplified and its noise figure is established, by the preamplifier, the amplitude of the
MRI signal is scaled to maximize the receive chain dynamic range. This is usually done by adding further gain
blocks to the receive chain followed by a programmable gain block (see Fig. 4-32).
The MRI receiver performs the down conversion and digitization of the MRI signal required for computer
processing and image reconstruction. The spectrum is down-converted to remove the RF central frequency
before digitization, since the relative frequencies of the various spectral components contain the encoded
spatial information. The receiver must preserve the amplitude, phase and relative frequency information
across a band of frequencies determined by the readout gradient.
The quadrature receiver23,26 is an efficient down-converter that minimizes filtering and A/D sampling speed
requirements. In-phase and quadrature (I, Q) radiofrequency local oscillators multiply the incoming MRI signal
in a pair of mixers (also known as frequency multipliers). If the MRI signal of bandwidth B Hz is
down-converted at the central frequency, it results in two base band signals of 0 to B/2 Hz each, containing
information from above and below the central frequency. The MRI signal is now represented as an in-phase
and quadrature complex pair (I, Q pair). After filtering, these two signals are digitized at their lowest
frequencies and bandwidths, allowing high-resolution analog to digital converters to be employed. See Figure
4-36 for a functional block diagram of the quadrature receiver. No loss of relative frequency information
occurs by this "folding" of the MRI signal spectrum about the central frequency. A complex fast fourier
transform (FFT) performed on the complex data set can discern components from above or below the central
frequency.
page 132
page 133
Add to lightbox
Figure 4-36 The basic blocks of the receive demodulation components. The signal is split and mixed with two
reference signals 90° apart to maintain information about the frequency spectrum of the signal.
The device shown in Figure 4-36 has several practical limitations. The two channels must exhibit very close
amplitude and phase tracking, or image ghosting may be present due to the inability to discern whether
information is coming from above or below the central frequency. Additionally, since the center of the
displayed image is at DC (zero Hertz), other common low-frequency disturbances, such as power line and
elevated semiconductor noise, may produce a central image artifact.
As analog to digital converter performance has improved, digitizing at intermediate or directly at radio
frequencies is becoming more common. The digitized signal can now be down-converted by a digital
multiplication and filtered in identical digital filters, avoiding the inaccuracies at DC and analog filter tracking
requirements of the conventional quadrature receiver (see Fig. 4-37).
© 2010 Elsevier
16 of 16 11-04-2010 23:54
PULSE SEQUENCE GENERATION

Overview
Magnetic resonance imaging requires the controlled application of magnetic field gradients, the
transmission of RF energy, and the collection and processing of RF response signals (MRI signal). The
pulse sequence generator is responsible for the control and timing of these stimuli and the detected
signal (FID or ECHO).
The pulse sequence generator translates requested scan parameters into gradient waveforms, RF
envelope waveforms, and control sequences that steer the response of the spin system through a
desired order in k-space. The waveforms and timing of these sequences depend on many factors,
e.g., the pulse sequence type, pulse sequence parameters, imaging options, and the system hardware
limitations such as gradient slew rate and peak gradient amplitude.
Add to lightbox
Figure 4-37 A basic block diagram of a digital quadrature receiver. Here the quadrature detection is
accomplished using digital multipliers. This technique has significant advantages over the analog
detection method shown in Figure 4-36.
Performance characteristics of the pulse sequence generator have a direct impact on the image
quality. Some of these characteristics include waveform synchronization, waveform resolution, and
waveform update rates. First, the pulse sequence generator must maintain tight time synchronization
among individual waveforms and waveform points to provide adequate control of the spin system, and
allow proper reception of the system response. Skew or a lack of synchronization can manifest itself in
phase artifacts, while jitter between waveform points tends to affect SNR. The dynamic range,
resolution, and update rates of individual waveforms also drive image quality limits. The degree to
which these parameters affect image quality highly depends on the MR imaging technique employed.
Gradient Waveform
Gradient waveform generation is restricted in consideration of patient comfort and safety limits as well
as gradient subsystem performance limitations. Some restrictions are placed on gradient waveform
generation to prevent peripheral nerve stimulation (PNS) in patients. Other restrictions are placed on
waveforms due to the slew rate limits, peak amplitude, and duty cycle limits of gradient coils and
amplifiers.
page 133
page 134
Gradient waveform generation may include a number of realtime computations and corrections to
optimize MRI system performance. Some of these waveform manipulations include gradient shim
offsets, eddy current compensation, and rotation of gradient axes. Gradient shim offsets are DC
offsets applied to compensate gradient waveforms based on measured inhomogeneity of the main
magnetic field B0. Eddy current compensation is applied to gradient waveforms to correct for currents
1 of 3 11-04-2010 23:54
that are induced in metal surfaces through application of the gradients, since the induced currents can
distort the expected gradient field. A calibration procedure is used to characterize and compensate the
induced eddy currents as explained in the system calibration section. Additional eddy current
compensation can be realized through modulation of the system center frequency synchronized with
realtime play out of the gradient waveforms. An additional optimization of system resources may be
employed through waveform rotation processing. Since the system gradient coils are oriented in three
orthogonal axes, a set of logical gradient waveforms can be defined for slice excitation in any
Cartesian plane. Oblique planes can be derived from these logical gradient waveforms by treating
each point in the gradient play out as a vector, and rotating that vector to the oblique plane. These
gradient waveform corrections and optimizations require realtime processing while maintaining
synchronization with the play out of RF waveforms and control sequences.
Radiofrequency Waveform
Radiofrequency waveform generation, like gradient waveform generation, observes restrictions to
maintain patient comfort and safety and to operate within limits of RF coils and amplifiers. Since RF
power is transmitted into patients to create resonance within a selected volume, RF waveform
amplitudes, durations, and duty cycles are restricted to prevent patients from exposure above specific
absorption rate (SAR) limits. The addition of power monitoring and system interlocks is typically
employed to eliminate unintended RF exposure.
The purpose of the RF transmit waveform is to create or suppress resonance in tissues with specific
MR imaging characteristics within a selected volume. This imaging volume is selected through the
combination of gradients and RF excitation during the stimulus portion of the MRI experiment. The RF
excitation pulse is provided through control of an RF modulator. Details of the RF modulator hardware
may be found in the RF chain section of this chapter. However, the key parameters in the control of the
RF excitation waveform are the frequency, the amplitude, and the phase of the RF pulse. The pulse
sequence generator may synchronously control some or all of these parameters during the RF
excitation pulse. Fat suppression pulses and selection of an off-center slice are examples where the
frequency and phase of the RF pulse needs to be controlled in realtime. Also, during the reception of
the detected MR signal, additional control of the RF modulator's phase and frequency may be needed.
Control Sequences
During an MR imaging experiment, a number of control signals are applied in synchronization with the
gradient and RF waveform play out. Some of these control signals protect sensitive electronics, others
are key to providing high MR image quality or enable more flexible pulse sequence control, while
others control the data flow through the image processing section of the MRI system.
Magnetic resonance imaging requires sensitive RF receive electronics to detect small MRI signals with
the highest possible SNR. Coil elements and sensitive receive electronics are commanded into a
protected mode when RF excitation is played out. During this RF excitation time, the high power RF
amplifier is commanded to provide output drive. This output is then disabled or blanked to allow for the
quietest possible RF environment during reception of the MRI signal response. In addition, some RF
coils provide transmit and receive capabilities. Synchronous control is required to command these coils
from the transmit state to the receive state in concert with sequence play out.
The control sequencer is responsible for resetting the phase of the RF modulator to maintain control of
the RF phase for the excitation and receive portions of the MRI experiment. The control sequencer
may also be used to indicate the timing of the receive data window for the filter and reconstruction
processors in the system. Other potential uses for sequence synchronous control signals include:
triggering external equipment based on scan timing, synchronizing realtime modifications to waveform
play out, realtime selection of coil interfaces, and synchronized control of RF transmit and receive gain.
Gating
Some MR imaging techniques require pulse sequence play out to be synchronized or gated with
external events. The pulse sequence generator may provide sequence-triggering inputs to enable these
2 of 3 11-04-2010 23:54
techniques. In the most common case, physiologic gating signals are used to control the effects of
cardiac and respiratory motion in imaging. In these cases, the challenge is to generate an accurate
and reliable gating signal. Physiologic sensors must provide good signals in the high magnetic field
environment. Since physiologic signals within the patient are continuously varying, predictive and
post-processing algorithms must often be applied to maintain acceptable gating performance.
© 2010 Elsevier
3 of 3 11-04-2010 23:54
SYSTEM CALIBRATION
Overview
Like other medical instrumentations, an MRI scanner needs to be calibrated before it is used for
diagnostic purposes. These calibration and tuning procedures can be divided into two categories: 1.
Calibrations that are performed after system installation and those performed regularly as a part of
plan maintenance; 2. The adjustments and tuning done before each patient scan.
Different manufacturers might have different ways of achieving this. In the past, MRI has been used as
a qualitative diagnostic imaging tool. Today a state-of-art MRI scanner can perform quantitative
measurements such as flow, diffusion, perfusion and pO 2 (oxygen pressure). These techniques may
require special calibration procedures.
page 134
page 135
System Installation and Regular System Calibrations

After an MRI system is installed, the following calibrations need to be performed at the top level: 1.
Gradient calibration: gradient coefficients are adjusted so that the slice thickness, and FOV are
correct; 2. Main field shimming: main field shimmed to meet the field homogeneity specifications using
passive shims, super-conducting shims, or both; 3. B1 RF calibrations: the RF amplifier is calibrated to
output predictable RF amplitude; 4. Eddy current calibrations: eddy current coefficients are calculated
using a special pulse sequence for compensation. The part of the gradient driver shaping the gradient
waveform to reduce eddy currents that induce phase and field shifts is called a pre-emphasis circuit.
Generally the calibration values for each subsystem are stored in separate calibration files that are
loaded during the system boot. The calibrations should be checked and adjusted if necessary at
regular time intervals as a part of preventative maintenance.
Calibrations and Tuning before each Patient Scan

MR image quality is sensitive to patient shape and weight, since each patient loads the system
differently. Thus the magnitude of the RF pulse, B0 field homogeneity, and the SNR depend on the
patient size and positioning. To account for these variations and maximize the image quality, most
manufacturers use a portion of the scan time (auto-prescan) to optimize parameters such as the
transmit power, receive attenuation, and field homogeneity (shimming) within the specified FOV. This
algorithm might be different for various pulse sequences and RF coils.
First the required transmit power is determined by varying the transmit gain (TG). Basically the
transmit power is increased in steps to obtain the maximum signal (FID magnitude). If sufficiently large
TR (repeat time) is used, this ensures that 90° flip angle was attained. Given the shape of the 180° RF
pulse, the required amplitude for 180° pulse can then easily be computed assuming good RF amplifier
linearity. Similarly, in the case of small flip angles, the amplitude is calculated using the 90° flip angle
amplitude as a reference.
Once TG has been determined, the MRI signal must be normalized to utilize the entire dynamic range
of the analog-to-digital converter (ADC) without over ranging the input of the ADC. To achieve this,
receiver gain or attenuation is automatically adjusted.
The field homogeneity within the prescribed FOV is shimmed, by applying small currents to the
gradient coils. This is also known as gradient shim. Some applications using fat suppression, and
others such as spectroscopy, spiral scanning, and functional imaging may require high-order shim. This
is achieved by adding extra shim coils and a shim power supply controlled by the scanner during the
calibration. All these calibrations are performed with the software without any user interaction prior to
each scan to ensure good image quality. Some manufacturers also incorporate, depending on the type
of the RF coil and the field strength, an auto-tuning of the RF coil into the algorithm.
1 of 3 11-04-2010 23:55
REFERENCES
1. Bloch F: The nuclear induction. Phys Rev 70:460-474, 1946.
2. Bloch F, Hansen WW, Packard M: The nuclear induction experiment. Phys Rev 70:474-485, 1946.
3. Bloch F, Hansen WW, Packard M: The nuclear induction. Phys Rev 69:127, 1946.
4. Purcell EM, Torrey HC, Pound RV: Resonance absorption by nuclear magnetic moments in a solid. Phys Rev 69:37-38,
1946.
5. Damadian R: Apparatus and method for detecting cancer in tissue, US Patent# 3789823 filed Mar 17, 1972.
242:190-191, 1973.
7. Mansfield P, Maudsley AA: Line scan proton spin imaging in biological structures by NMR. Phys Med Biol 23:847, 1976.
8. Mansfield P, Pykett IL, Morris PG, et al: Whole body line scan imaging by NMR. Br J Radiol 52:242, 1979. Medline
Similar articles
9. Mansfield P, Pykett IL: Biological and medical imaging by NMR. J Magn Reson 29:355-373, 1978.
10. Mansfield P, Maudsley AA: Medical imaging by NMR. Br J Radiol 50:188-194, 1977. Medline Similar articles
11. Weigner M, Pruessmann KP, Leussler C, et al: Specific coil design for SENSE: A six-Element Cardiac Array. Magn Reson
12. 2002 ISMRM Tenth Scientific Meeting and Exhibition, paper #834.
13. Sagawa M, Fujimura S, Togawa H, et al: New material for permanent magnets on a base of Nd and Fe. J Appl Phys
55:2083-2087, 1984.
14. Tilley DR, Tilley J: Superfluidity and superconductivity, 3rd ed. Bristol, UK: IOP Publishing, 1990.
15. Dresner L: Stability of Superconductors. New York: Plenum Press, 1995.
16. Gollay MW: Field Homogenizing Coils for NMR Instrumentation. Rev Sci Instrum 29:313, 1938.
17. Schenck JF, Hussain MA, Edelstein WE: Transverse gradient field coils for nuclear magnetic resonance. US Patent
4,646,024.
18. Roemer PB, Hickey JS: Self-shielded gradient coils for nuclear magnetic resonance imaging. US Patent 4,737,716.
19. Katzenelson E, Harvey PR: Modular MRI gradient coils. UK Patent 22950020 A.
20. Liu Q: Gradient coil capable of producing a variable field of view. US Patent 6,630,829B1.
21. Yoda K: Analytical design methods of self shielded planar coils. J Appl Phys 67:4349-53, 1990.
22. Silva AC, Merkle HE: Hardware considerations for functional magnetic resonance imaging. Concepts Magn Res
16A:35-49, 2003.
23. Hoult DI: The NMR receiver: A description and analysis of design. Progress in NMR spectroscopy 12:41, 1978.
24. Bleaney BI, Bleaney B: Electricity and magnetism, 2nd ed. Oxford: Clarendon Press, 1965.
25. Vizmuller P: RF design guide. Boston: Artech House, 1995.
26. Krauss HL, Bostian CW, Raab FH: Solid state radio engineering. New York: J. Wiley, 1980.
27. Stremler FG: Introduction to communication systems. Prentice Hall, New Jersey, 3rd ed, 1990, Addison Wesley, 197.
28. Gonzalez G: Microwave transistor amplifiers. London: Prentice Hall International, 1997.
29. Bowick C: RF circuit design. Boston: Newnes, 1982.
30. Chen C-N, Hoult DI, Sank VJ: A parallel algorithm for rotating frame zeugmatography. Magn Reson Med 1:354, 1984.
31. Glover GH, Hayes CE, Pelc NJ, et al: Comparison of linear and circular polarization for magnetic resonance imaging. J
Magn Res 64:255, 1985.
32. Jin J: Electromagnetic analysis and design in MRI. New York: CRC Press, 1999.
33. Hayes CE, Edelstein WA, Schenck JF, et al: An efficient, highly homogenous radiofrequency coil for whole-body NMR
imaging at 1.5 T. J Magn Res 63:622, 1985.
34. Guclu, C: Effect of body coil electric field distribution on receive-only surface coil heating. J Magn Res Imaging 14(4):484,
2001.
35. Boskamp EB, Schenck JF, Schaefer DJ, et al: ISMRM book of abstracts, 736, 1999.
36. Beresten K, Wright SM, Boskamp EB: ISMRM book of abstracts, 1095, 1994.
37. Caron, WN: Antenna impedance matching. Newington, CT: American Radio Relay League, 1993.
page 135
page 136
38. Gunston MAR: Microwave transmission line impedance data. Atlanta: Noble, 1997.
39. Sevick J: Transmission line transformers. Newington, CT: American Radio Relay League, 1987.
2 of 3 11-04-2010 23:55
40. Roemer PB, Edelstein WA, Hayes CE, et al: The NMR phased array. Magn Res Med 16:192, 1990.
41. Hayes CE, Roemer PB: Noise correlations in data simultaneously acquired from multiple surface coil arrays. Magn Res
Med 16:181, 1990.
42. Pruessmann K, Weiger M, Scheidegger M, et al: SENSE: Sensitivity encoding for fast MRI. Magn Res Med 42:952, 1999.
43. Weiger M, Pruessmann K, Leussler C, et al: Specific coil design for sense: a six element cardiac array. Magn Res Med
45:495, 2001.
44. Guclu CC, Boskamp E, Zheng T, et al: A method for preamplifier-decoupling improvement in quadrature phased-array
coils. Magn Res Imaging 19:255-258, 2004.
© 2010 Elsevier
3 of 3 11-04-2010 23:55
ULSE EQUENCE ESIGN

J. Paul Finn
Vibhas S. Deshpande
Orlando P. Simonetti
INTRODUCTION
Since the second edition of this book, continued advances in hardware, computer technology, image
reconstruction methods, and the applications of contrast agents have fueled explosive development in
magnetic resonance imaging (MRI). As most advances are based on sophisticated engineering or
physical principles, the gap in understanding between developers and users is greater than ever, and is
widening. This would not be a problem were it not for the fact that the choice and interpretation of MRI
techniques is best when based on a clear understanding of how these techniques work. This chapter is
written with the user in mind. Several excellent recent texts are available for the interested reader with
1,2
a strong technical background who is looking for mathematical rigor. What we aim to do is to
introduce the novice to some aspects of pulse sequence design which will give him or her insight into
how best to apply MRI in various clinical circumstances. We include, however, more detail on relevant
data acquisition, signal processing, and image reconstruction methods than do typical clinical texts.
The same basic physical and image reconstruction principles apply over a wide range of techniques,
albeit sometimes disguised in a different format.
MRI is a multiparametric technique, and many parameters can be associated in a more or less
arbitrary manner. A confounding array of different pulse sequences is therefore possible. The process
of applications development with MRI has features in common with drug discovery: from a huge pool of
potential candidates, relatively few find their way into routine clinical practice. There are several
reasons for this: in some cases there is no clear clinical application for a pulse sequence; in some
cases there is a substantial time lag between initial concept and availability as a commercial product;
and in some cases a technique proves unwieldy or unreliable in a clinical context. Such, however, is the
nature of research, and there is no progress without trial and error.
Pulse sequence design is a huge topic, comprising elements from many disciplines in the physical and
clinical sciences, and a comprehensive treatment of the subject is beyond the scope of any book
chapter. However, it is our hope that once the reader has grasped the relevant basic principles, he or
she will see how the same tools make subsecond imaging of the heart and spin-echo imaging of the
brain equally understandable. In this chapter, we confine the scope of the discussion to a description of
the relevant physics, data acquisition, signal processing, and image reconstruction as they relate to
conventional and ultrafast MRI techniques. We do not deal with hardware issues or parallel imaging
methods, which are covered in detail in Chapters 4 and 8 respectively. We also deal only with those
clinical issues relevant to illustrating the applications of specific families of pulse sequences. Some
mathematical detail is provided for those who are interested, but those who are not should not be
discouraged and should read through. Practical, plausibility comments usually accompany the
mathematical statements.
We take as our working definition of an MRI pulse sequence a series of radiofrequency (RF) events
(pulses) and magnetic field gradient events (pulses) generating nuclear magnetic resonance signals
that, when appropriately processed, result in the formation of an image. The details of how the RF and
gradient pulses are played out determine the form, speed, and information content of the image. We
concern ourselves with these processes.
© 2010 Elsevier
1 of 1 11-04-2010 23:56
INTRODUCTION
Since the second edition of this book, continued advances in hardware, computer technology, image
reconstruction methods, and the applications of contrast agents have fueled explosive development in
magnetic resonance imaging (MRI). As most advances are based on sophisticated engineering or
physical principles, the gap in understanding between developers and users is greater than ever, and is
widening. This would not be a problem were it not for the fact that the choice and interpretation of MRI
techniques is best when based on a clear understanding of how these techniques work. This chapter is
written with the user in mind. Several excellent recent texts are available for the interested reader with
a strong technical background who is looking for mathematical rigor.1,2 What we aim to do is to
introduce the novice to some aspects of pulse sequence design which will give him or her insight into
how best to apply MRI in various clinical circumstances. We include, however, more detail on relevant
data acquisition, signal processing, and image reconstruction methods than do typical clinical texts.
The same basic physical and image reconstruction principles apply over a wide range of techniques,
albeit sometimes disguised in a different format.
MRI is a multiparametric technique, and many parameters can be associated in a more or less
arbitrary manner. A confounding array of different pulse sequences is therefore possible. The process
of applications development with MRI has features in common with drug discovery: from a huge pool of
potential candidates, relatively few find their way into routine clinical practice. There are several
reasons for this: in some cases there is no clear clinical application for a pulse sequence; in some
cases there is a substantial time lag between initial concept and availability as a commercial product;
and in some cases a technique proves unwieldy or unreliable in a clinical context. Such, however, is the
nature of research, and there is no progress without trial and error.
Pulse sequence design is a huge topic, comprising elements from many disciplines in the physical and
clinical sciences, and a comprehensive treatment of the subject is beyond the scope of any book
chapter. However, it is our hope that once the reader has grasped the relevant basic principles, he or
she will see how the same tools make subsecond imaging of the heart and spin-echo imaging of the
brain equally understandable. In this chapter, we confine the scope of the discussion to a description of
the relevant physics, data acquisition, signal processing, and image reconstruction as they relate to
conventional and ultrafast MRI techniques. We do not deal with hardware issues or parallel imaging
methods, which are covered in detail in Chapters 4 and 8 respectively. We also deal only with those
clinical issues relevant to illustrating the applications of specific families of pulse sequences. Some
mathematical detail is provided for those who are interested, but those who are not should not be
discouraged and should read through. Practical, plausibility comments usually accompany the
mathematical statements.
We take as our working definition of an MRI pulse sequence a series of radiofrequency (RF) events
(pulses) and magnetic field gradient events (pulses) generating nuclear magnetic resonance signals
that, when appropriately processed, result in the formation of an image. The details of how the RF and
gradient pulses are played out determine the form, speed, and information content of the image. We
concern ourselves with these processes.
© 2010 Elsevier
1 of 1 11-04-2010 23:57
SUMMARY OF BACKGROUND PHYSICS

Magnetization
Add to lightbox
Add to lightbox
Figure 5-1 Net magnetization vector and spin precession around B0. As shown in A, the torque τ
acting on a typical spin is always perpendicular both to the spin and to the main field. This results in a
circular motion of the tip of the spin vector around the axis of the main field. As shown in B, there is a
slight excess of vectors pointing with the field, giving rise to a net magnetization, M. Because the spins
are randomly orientated in the x-y plane, there is no net x-y magnetization.
From a practical point of view, we need only concern ourselves with the "net magnetization vector," M
(Fig. 5-1), which is the vector sum of all the individual spins (small arrows) aligned parallel to the main
magnetic field, B0. At equilibrium, M is steady along the axis of B0 (conventionally the "z" axis) and
experiences no tendency to rotate off axis; i.e., there is no twisting effect or "torque." This is because
the main field and the magnetization vector are parallel; in order for there to be a torque, the
magnetization must be at least partially perpendicular to the main field. So, when aligned along z, the
magnetization is essentially static and does not generate a detectable signal in the presence of the
static external field. In order to be detectable, the magnetization must have a net component which lies
within the transverse (x-y) plane. If rotated away from z, the magnetization will experience a torque,
proportional to its component in the x-y plane, and will precess coherently about B0 and be detectable
in the x-y plane by a suitable probe. The Larmor frequency is the resonant frequency with which the
spins rotate in the presence of a magnetic field; it is specific for each nuclear species and, for any
given nucleus, it increases proportionately with the strength of the applied external magnetic field. At
1.5 Tesla, the Larmor frequency for hydrogen nuclei (protons) is approximately 63.75 megahertz
(MHz); at 3.0 Tesla, it is approximately 127.8 MHz. Recall that a changing magnetic field always
implies a changing electric field, and vice versa. The changing electric field constitutes a sinusoidally
alternating voltage in the transverse plane, and this is detectable by a probe. How might M be rotated
partially or fully into the transverse plane? The rotation of the magnetization vector, M, is accomplished
by the magnetic field of an RF pulse, generally designated the B1 field. Recall that radio waves are
part of the spectrum of electromagnetic radiation, and that all electromagnetic waves are composed of
electric and magnetic fields perpendicular to each other, and oscillating sinusoidally at a specific
1 of 4 11-04-2010 23:58
frequency. The magnetic field of the RF wave, B1, can be thought of as adding to the main field, B0,
but in a perpendicular direction. Now we have M and B1 perpendicular to each other, so B1 exerts a
torque on M. The direction of the torque is at all times perpendicular both to B1 and M, and is defined
by a "vector product" as detailed below (don't worry if you're not familiar with the vector product).
So, in order to rotate M away from the z direction:
The magnetic field of the RF pulse must lie somewhere in the x-y plane (this is the plane
perpenicular to M)
The B1 field must stay synchronous with the magnetization by rotating at the same (Larmor)
3
frequency as M (Fig. 5-2A). Otherwise, the effect of the B1 field would average out and
produce no net rotation.
In this case, as B1 tips M off axis, the motion of M relative to B1 is characterized by a simple rotation
around the B1 axis (Fig. 5-2B). At the Larmor frequency, the motion of M describes a cone-shaped
"precession" around B1 so that M lies within the y-z plane. The precessional frequency of M around B1
is γB1 and the angle α that M rotates through is given by
where t is the duration for which the RF field of strength B1 is applied (see Fig. 5-2B). α is generally
referred to as the "flip angle."
Add to lightbox
Add to lightbox
page 138
page 139
Figure 5-2 A, The B1 field is synchronous with the magnetization M in the x-y plane to produce a
nutation of M. B, The motion of M is due to the torque M × B1, which nutates M through α toward y' at a
rate γB1.
Relaxation
As we have seen, perturbation of the magnetization by application of a short RF pulse tips it away
from the longitudinal axis and generates a transverse component. If this magnetization is allowed to
precess freely, there is a regrowth of the longitudinal magnetization called longitudinal relaxation, and
2 of 4 11-04-2010 23:58
destruction of the transverse magnetization called transverse relaxation. Relaxation is described by

exponential time constants: T1 for longitudinal and T2 for transverse relaxation. Exponential processes
are those whose rate of change depends on how far they have left to go; the closer they get to their
final value, the more slowly they approach it. T1 is defined as the time taken for the longitudinal
magnetization to relax from 0 to 63% of the equilibrium magnitude. T2, or the transverse relaxation time
constant, is the measure of the time that the transverse magnetization takes to relax (decay) to 37% of
its initial magnitude. T2 decay occurs because individual spins (usually referred to as isochromats)
rotate at slightly different rates due to their chemical environment, and eventually they get out of sync
and begin to point in random directions and cancel each other. This process is referred to as
"dephasing," because the spins acquire different phases in the range 0 to 360°. In practice, field
inhomogeneities can be another source of transverse relaxation, other than the chemical environment.
The dephasing due to this process is accounted for by another time constant, T2'. The total transverse
relaxation time constant (T2*) then is the reciprocal sum of T2 and T2' and is given by:
Add to lightbox
Add to lightbox
Figure 5-3 Typical longitudinal and transverse relaxation curves plotted for T1 and T2 of 500 ms.
Both the longitudinal and transverse relaxations are modeled by exponential functions (Fig. 5-3) given
by the Bloch equations, which are stated without further justification:
Mz
longitudinal magnetization
3 of 4 11-04-2010 23:58
Mz0
longitudinal magnetization at t = 0+, i.e., immediately after the RF pulse
M0
equilibrium magnetization
Mxy
transverse magnetization.
Different tissues have different relaxation time constants, and these form the basis of some of the
contrast mechanisms in MRI.
© 2010 Elsevier
4 of 4 11-04-2010 23:58
PULSE SEQUENCE DESIGN

As mentioned earlier, a pulse sequence is defined as a sequence of radiofrequency and gradient
events applied as a function of time. Many different components are involved in sequence design. Each
individual component is first briefly introduced here to elucidate the underlying principles of sequence
design. This is followed by a summary of the commonly used existing pulse sequences.
Gradients
Gradient fields are spatially varying magnetic fields. Linear magnetic field gradients are used in MRI to
encode spatial information. The Larmor equation states that the precessional frequency (ω) of a spin
is directly proportional to the applied magnetic field strength (B):
where G (T/m) is the gradient field, r is the position of the spin along the gradient axis, and γ is a
proportionality constant called the gyromagnetic ratio. γ is constant for a given nuclear species,
relating the Larmor frequency to the applied field strength. It can be thought of as a scaling factor for
field strength.
page 139
page 140
Thus, the mapping of position to frequency by a magnetic field gradient follows directly from the
Larmor equation. A spatial variation in the magnetic field causes a corresponding spatial variation in
precessional frequency. The gradient field is produced by pulsing current through some combination of
coils to generate three spatially orthogonal, linear variations in the main field (Gx, Gy, and Gz).
Whereas the main magnetic field, B0, is generally fixed in direction and magnitude, the gradient field G
can assume a range of values, and any arbitrary direction achieved by a combination of Gx, Gy, and
Gz .
Note at this point that the field generated by the gradient coils is in the direction of the main field (z), so
that the magnitude of the Bz field is made to be a function of position along the direction of the
gradient. For example, an x-gradient augments the B0 field such that the magnetic field Bz varies as a
function of position in the x direction. Thus, "gradient" refers to the rate of change of the Bz field in the
specified direction. When the x gradient Gx is active, the net field Bnet has components from the main
field, B0, and Gx such that
and combining with Equation 5-6 yields
Equation 5-8 expresses resonance frequency as a linear function of position along the gradient
direction, where x is the offset from the magnet isocenter. At the isocenter, the net field is B0 at all
times, because x (and therefore xGx) is zero. The actual field at any position within the magnet bore is
the sum of contributions from all sources including the main field, applied gradients, and intrinsic
susceptibility gradients.
As Equation 5-8 shows, in the presence of gradients, each isochromat acquires a unique frequency of
precession. Thus, when the signal is read out, the position of each isochromat can then be decoded
(demodulated) depending on its frequency of precession. This property of "encoding" the position of a
spin is among the most important uses of gradients in pulse sequences.
For clinical machines, the maximal gradient strength typically lies in the range 20 to 50 mT/m. In typical
imaging systems, the gradient field is approximately two orders of magnitude smaller than the main
field. For example, if B0 is 1.5 T, the maximal field caused by the gradients is approximately ±12.5 mT,
assuming maximal gradient amplitudes of 50 mT/m over a linear range of ±25 cm (Fig. 5-4). The
6
gradient in frequency corresponding to the field gradient is given by γG, where γ = 42 × 10 Hz/T is the
6
gyromagnetic ratio. For G = 50 mT/m = 0.05 T/m, the spatial frequency gradient is therefore 2.1 × 10
6
Hz/m. For a field of view = 50 cm = ±25 cm, the spatial frequency bandwidth = 2.1 × 10 Hz/m × 0.5 m
1 of 29 11-04-2010 23:59
= 1.05 × 106 Hz or ±525 kHz. Sometimes the bandwidth is normalized to the imaging matrix and is
expressed in Hz per pixel. For a 256-image matrix, a bandwidth of 1.05 megahertz corresponds to 4.1
kHz/pixel.
Radiofrequency Events and Slice Selection

In MRI, RF energy is typically applied as a pulse lasting from tens of microseconds to tens of
milliseconds, depending on the application. When the RF is applied on resonance, that is, at the
Larmor frequency, the magnetization is rotated α° about B1, as discussed earlier. The degree of
rotation of the magnetization depends on the strength of the applied RF field and the duration of the
applied field. When α is 90°, all magnetization is rotated into the x-y plane, and the potential signal is
maximal. As discussed later, however, for many applications flip angles other than 90° are used. For
example, 180° pulses are used for inversion and refocusing, and pulses substantially less than 90° may
be appropriate for short repetition time (TR) imaging. Often, the direction of the B1 field in the
transverse plane, or what is referred to as its phase, is also important in sequences such as spoiled
gradient-echo sequences or steady-state free precession (SSFP) sequences.
Many different types of RF pulse are used in MRI. They are broadly classified into two types, "hard"
and "soft" pulses, or pulses that are spatially nonselective and selective, respectively. RF pulses of
short duration and high amplitude are described as hard pulses. These pulses, on account of their
short duration, excite a broad range of frequencies and are therefore referred to as nonselective
pulses. In general, the shorter the RF pulse, the larger the range of frequencies that it excites, or what
is commonly referred to as a large "bandwidth." Pulses of longer duration allow excitation of only a
narrow band of frequencies and are termed "soft" pulses. In the presence of slice selection gradients,
these RF pulses can be used for localized imaging. Other, specialized pulses include hyperbolic secant
pulses that are used for inversion, and fat suppression pulses that are narrow bandwidth frequency
selective pulses. These specialized pulses and their applications will be described in the later sections.
Slice Selection
Add to lightbox
page 140
page 141
Figure 5-4 Gradient field. Magnetic field gradient causes the frequency at position x to change by
x(δG/δx). The field experienced by spins far off center in the positive direction is therefore greater than
that seen by spins close to the magnet isocenter. As illustrated, the gradient strength is 25 mT/m.
One of the advantages of MRI over other imaging modalities is the ability to image planes or slices of
tissue with effectively any arbitrary orientation, thickness, and position. The process of slice selection
is carried out by applying band-limited RF energy, or soft RF pulses, simultaneously with a magnetic
field gradient. A schematic of the slice selection process is shown in Figure 5-5. Recall that a linear
gradient makes the precessional frequency of the spins vary with position. Therefore, to excite a slice
plane, the resonance frequency of the spin system perpendicular to the slice plane direction (ωs) is
made to vary linearly with position by a slice-selection field gradient Gs superimposed on the main field
B0:
2 of 29 11-04-2010 23:59
A slice of a given thickness therefore contains spins precessing over a range of frequencies. By
applying RF only in this frequency range, the bandwidth ∆f, the slice of interest can be selectively
excited. (Note that, in the general case, the bandwidth of the RF pulse bears no necessary relationship
to the receiver bandwidth during readout; these are entirely independent events.) The slice thickness
(T) is related to the RF bandwidth by
Add to lightbox
Figure 5-5 Slice selection using a gradient and a band-limited RF pulse. Given the strength of the slice
selection gradient (Gs), the thickness of the slice (T) will depend on the RF bandwidth, ∆f. The slice
position depends on the center frequency of the RF pulse, f. If the center frequency is changed (with
the bandwidth remaining unchanged), the position of the slice translates along the gradient axis, with
the thickness remaining constant.
Inspection of Equation 5-10 reveals that for a given RF bandwidth, stronger gradients produce thinner
slices and, for a given gradient amplitude, narrower bandwidths (longer pulses) produce thinner slices.
For example, if we fix the gradient at 10 mT/m (426 Hz/mm), a 5-mm slice spans a 2.1-kHz range of
frequencies. If the gradient is increased to 25 mT/m (1064.5 Hz/mm), a 2.1-kHz band covers only 2
mm, and the same pulse and bandwidth now excite a slice only 2.0 mm thick. To excite a 2.0-mm slice
using a 10 mT/m gradient, the RF pulse must have a bandwidth of 852 Hz.
As shown in Figure 5-5, the position of the slice in space can be controlled by the center frequency of
the RF pulse. Altering the center frequency of the RF pulse changes the slice position along the applied
gradient axis. However, the slice thickness remains the same as long as the bandwidth of the RF pulse
remains constant.
Therefore, a slice is defined by its
orientation, determined by the gradient direction

thickness, determined by the interplay between gradient amplitude and RF bandwidth
offset (position along the slice-selection direction), determined by RF center frequency (f).
The task is now reduced to limiting the excitation to a frequency band covering no more than the
desired slice thickness.
Slice Profile
3 of 29 11-04-2010 23:59
Ideally, one would fully excite all magnetization, M, within the slice of interest and leave everything
outside the slice unaffected. The formal representation of this arrangement is a square wave, where M
has its full value for all frequencies within the specified bandwidth, ±f0, and is zero for f greater than
|f0|. We must now prescribe an RF pulse waveform, played out over time, having this well-defined
frequency bandwidth. The frequency content of any time waveform is described by its Fourier
transform. For in-depth background on Fourier transforms and their applications, the interested reader
is referred to two excellent texts.4,5 It must be noted that the Fourier transform relationship between
the RF pulse envelope and the resultant magnetization holds only for flip angles less than
approximately 30°. The nonlinearity of the Bloch equations must be accounted for at higher flip angles.6
Nevertheless, the Fourier transform offers a good first approximation to the response of the spin
system to RF excitation in the presence of a linear field gradient. For a given desired slice profile and
the initial condition of the magnetization, the shape of slice selective RF pulse can be designed by
calculating the inverse Fourier transformation of the slice profile.
From Fourier transform theory, if we require a square frequency domain waveform, we need to apply
a time domain RF pulse as a sinc function of infinite duration, illustrated in Figure 5-6. The sinc function
is formally defined as sin(πx)/x. It is a ubiquitous function in Fourier analysis and signal processing
theory, and it has several interesting features. Nodes or zero crossings occur when the argument of
the sinc function is an integral multiple of π. The bandwidth of the RF pulse is ∆f = ±f0, and this in turn
is defined by the (reciprocal) time to the first zero crossing in the sinc function.
page 141
page 142
Add to lightbox
Add to lightbox
Figure 5-6 Rectangular frequency band (A) and its Fourier transform (B). A sinc RF profile over time
produces a well-defined spectrum in the frequency domain. The vertical lines in B denote the times of
the first zero crossings, which determine the bandwidth, ±(f0/2). The nodes of the sinc function in this
example occur at integer multiples of 1/f0. The example illustrated assumes the sinc function is infinite,
not truncated.
The broader the sinc function, the narrower the bandwidth, and vice versa. For example, in the above
example with a 10-mT/m gradient, the sinc function that produces a 5-mm thick slice has a bandwidth
of 2.1 kHz and has its first zero crossing 470 μs after maximal amplitude. To excite a 2.0-mm slice
using a 10-mT/m gradient, the RF pulse has a bandwidth of 852 Hz and its first zero crossing 1175 μs
after the maximal amplitude.
4 of 29 11-04-2010 23:59
Add to lightbox
Figure 5-7 Slice profile resulting from truncated RF pulse. Because the RF pulse is truncated in time,
the frequency convolution theorem implies that the attenuated sinc RF pulse transforms not to the
square wave P(f) but to the convolution of P(f) with the transform A(f) of the sampling envelope a(t).
This effect introduces ripples into the edges of the slice, most pronounced for extremely short pulses.
However, the definition of the Fourier transform considered here assumes that the RF pulse is played
out over infinite time! This is, of course, impractical, and for many applications severe constraints are
placed on the time available for RF excitation. A typical pulse lasts several milliseconds and may last
as little as a few hundred microseconds. The result of truncating (shortening) the RF pulse is a
distortion or "rippling" of the slice profile. The effects of truncation are summarized without
mathematical proof in Figure 5-7. The slice profile becomes distorted at the edges due to truncation of
the RF pulse. Furthermore, the Fourier transform procedure for RF pulse design does not work for
higher flip angles due to the non-linearity of the Bloch equations. In such cases, numerical solutions
may be used. The Shinnar-Le Roux algorithm7,8,9 is one such widely used tool to design slice-selective
RF pulses. This algorithm enables the quick computation of an RF pulse envelope given input
characteristics such as flip angle, RF bandwidth, duration of the pulse, and percent ripples in the
stopband and passband.
In practice, one cannot spend too long playing out the RF pulse, and the slice profile may be improved
by filtering or smoothing in the time domain. Most commonly, the amplitude of the RF time envelope is
weighted with a gaussian or Hanning filter function, smoothing out the ripples at the expense of
widening the main central lobe of the profile.
page 142
page 143
Recall that another factor of practical importance in improving the slice profile is the strength of the
slice-selection gradient. Because of the Larmor relationship between magnetic field strength and
precessional frequency, a field gradient can be expressed in terms of hertz per millimeter, as
discussed earlier. Thus, if the gradient is stronger, for a given slice thickness the bandwidth in the RF
pulse is greater. This means more zero crossings can be used for the same pulse duration, decreasing
truncation effects. So stronger gradients can be used to obtain thinner slices for a given pulse duration
and bandwidth, better slice profiles for a given pulse duration and slice thickness, or shorter RF pulses
for a given slice profile and thickness. The price to be paid for this, however, is an increase in the
5 of 29 11-04-2010 23:59
maximum transmitter amplitude and the maximum rate of power deposition. Specific absorption rate
(SAR) dose considerations may prove a limiting factor in certain circumstances.
Digitization Constraints
The arguments put forward so far are valid for a mathematical pulse envelope with a continuous range
of values over the truncation interval. However, because these RF waveforms are generated by
computer, the pulse envelopes are specified digitally by a finite number of support points as digital-
to-analog converter values, over the duration of the RF pulse. As a result, synthesized pulse
waveforms are subject to the limitations of discrete sampling-specifically, aliasing-and they must be
designed with this in mind.
According to the Nyquist theorem, frequency ambiguity occurs for frequencies greater than half the
sampling rate. This is the same phenomenon as occurs with pulsed Doppler or similar sampling
schemes.
Due to digitization, the waveform is reproduced periodically in the positive and negative frequency
domain. In effect, discrete sampling generates an infinite number of copies of the measured signal,
each copy separated by the sampling frequency-the greater the sampling frequency, the greater the
separation between adjacent copies. If the Nyquist condition is met, there is no overlap between
adjacent copies and the high-frequency copies of the pulse waveform are filtered out by low-pass
filters-hardware crystals with a well-defined frequency response. This leaves only the desired signal.
For example, if the pulse duration is 5120 μs and there are 512 sample points, the sampling frequency
-5
is 10 /s = 100 kHz, and no aliasing occurs below 50 kHz. The low-pass filter cutoff frequency is
therefore set at 50 kHz.
As pointed out earlier, for larger flip angles (>30°), the response of the spin system deviates from
linearity and the slice profile can no longer be predicted simply by Fourier transformation of the RF
envelope. This is particularly true for 180° pulses. In practice, optimization by numerical methods is
often utilized in RF pulse design.
Modulation
In addition to defining the thickness of the slice, its offset or displacement along the gradient axis
needs to be specified (see Fig. 5-5). To shift the center frequency of the RF pulse by frequency f m, the
RF waveform is modulated with a sinusoid of the appropriate carrier frequency fm. Multiplication by a
sinusoid in the time domain shifts the frequency envelope of the pulse without distorting its shape. The
bandwidth (and slice thickness) is the same, but it is now centered at f = f m, not f = 0. By making fm
continuously variable, any slice shift can be defined. The frequency offset, f m, that corresponds to a
slice position x is given by
As an example, if G(x) is 10 mT/m and x is 20 cm, fm is 85.2 kHz. Because it is necessary to be able
to shift the slice several times its thickness, the modulation frequency (slice shift in hertz) is usually
significantly greater than the bandwidth (slice thickness in hertz). In practice, modulation may be
performed by applying an incremented phase change over time to the RF pulse envelope. For
example, assume that we have a 5.12-ms pulse defined by 512 discrete points at 10-μs intervals. The
phase at each point is incremented by
where ϕ is in radians, τ is the duration of the RF pulse in seconds, and fm is the modulation frequency
in hertz.
Gradient Refocusing
A conventional approximation used by developers in slice-selection gradient waveform design is that
the transverse magnetization is generated instantaneously halfway through the RF pulse. When an RF
pulse is applied in the presence of a field gradient, the spins within the slice have a distribution of
frequencies determined by their position along the gradient axis. The phases of the transverse
6 of 29 11-04-2010 23:59
magnetization therefore fan out across the slice, but they can be brought back into alignment by a
reverse gradient. When the RF pulse is complete and the gradient is turned off, a representative spin
at any position has a phase according to its position on the gradient axis and the duration of the RF
pulse. This phase can be "unwound" by a gradient of equal and opposite area as shown on the slice-
select axis in Figure 5-8.
Practical Considerations in Radiofrequency Pulse Design

page 143
page 144
Add to lightbox
Figure 5-8 Gradient refocusing. ADC, analog-to-digital converter; Gp, phase-encoding gradient; Gr,
read gradient; Gs, slice-select gradient. During slice selection, the phases of the spins are dispersed
over the width of the slice. For stationary spins, these phases are unwound by a rephasing lobe equal
to half the area of the primary lobe and of opposite sign. In the read direction, the phase dispersion
caused by the first lobe is reversed by the rephasing lobe at the gradient-echo. As long as the integral
of the second lobe exactly counterbalances the first lobe, the relative durations of the gradient lobes
may be different. The duration of the readout lobe can therefore be expanded or contracted and its
amplitude changed accordingly. In this way the bandwidth of the detected signal may be modified. In
the phase-encode direction, a unipolar pulse imparts a constant phase to spins, proportional to
position along the gradient axis.
As summarized in Equation 5-2, the integral of the B1 field over time determines the flip angle, and that
required for a π pulse is twice that needed for a π/2 pulse. In practice, one typically fixes the duration
of the RF pulse and calibrates the B1 intensity. It follows that a flip angle α can be reached with a short
pulse of intense B1 or a longer pulse of less intense B1. In most situations, it is desirable to achieve α
as quickly as possible. However, a number of factors considered later may limit how quickly this can
be achieved:
The peak voltage of the RF transmitter sets a limit on the B1 field intensity, such that the flip
angle required may not be reached in the allotted time
For spatially selective pulses, the slice profile becomes degraded with truncation. The
availability of strong slice-selection gradients helps to offset this limitation, but it must still be
considered
For clinical imaging, the specific absorption rate for RF energy must remain within allowed
limits. Shorter RF pulses require higher power, and therefore higher specific absorption rate
values, to attain the same flip angle as longer pulses.
It is clear from Equation 5-2 that if the B1 field is not homogeneous in the volume of interest, α is not a
7 of 29 11-04-2010 23:59
single flip angle but is a distribution of flip angles. This may have unwanted effects on image contrast.
The discussion so far has assumed that the RF is applied uniformly throughout the relevant field of
view (FOV), in other words, that α is invariant over space. In practice, this is not the case, and the
spatial variation in the B1 field is strongly influenced by the design of the transmitter and how the
subject loads the coil. Generally, large transmitter coils have spatial B1 profiles that vary only slowly,
whereas the RF field of small transmitter coils may fall off dramatically toward the periphery. It may be
important to take these effects into account and, in the case of a surface receiver coil, to determine
whether it also serves as transmitter. B1 nonuniformity is more problematic at higher field strengths,
i.e., 3.0 T and above.
Add to lightbox
Figure 5-9 Changing effective field. As Beff changes its magnitude and rotates around y', the new
effective field is B'eff = Beff-(δθ/δτ)/γ. When the frequency sweep is slow, such that δθ/δt is small, M
follows Beff in the Beff-B1 plane.
When the B1 profile of a coil is known, it is possible to generate a correction algorithm for
post-processing and normalization in the same way as for a surface receiver coil. Although useful for
display purposes, such a scheme does not address SNR or contrast deficiencies that arise as a
consequence of the RF field. B1 nonuniformity is most apparent and most damaging when large flip
angles are used, as in inversion recovery (IR) imaging. Because the contrast in IR images is largely
determined by the inversion pulse, IR images are sensitive to the B1 profile during inversion. Certain
pulse envelopes show low sensitivity to B1 nonuniformity. Examples include adiabatic pulses, which are
used widely for inversion.
Adiabatic Passage
Adiabatic pulses do not obey the conventional relationship between the B1 amplitude and flip angle
stated earlier in Equation 5-2, as non-adiabatic pulses do. They follow the adiabatic passage principle,
which states that the magnetization vector follows the direction of the effective magnetic field provided
that the effective field does not change significantly during the course of one precession of the
magnetization about the effective field. The flip angle of an adiabatic pulse is then controlled by the B1
amplitude variation and the frequency modulation during its pulse duration. Even spins experiencing
different B1 fields can be excited with the same flip angle if the conditions for adiabatic passage are
met.
page 144
page 145
Consider the motion of M in the rotating frame as the RF field is swept through resonance. When the
frequency offset ∆ω = (ω0 - Ω) is large, there is a large residual Bz field B0 + Ω/γ (Fig. 5-9). Beff is
8 of 29 11-04-2010 23:59
therefore coincident with B0 and M precesses about the main field. As the frequency offset is slowly
decreased, Beff approaches B1 but at all times remains the precession axis for M. In other words, M
follows Beff as it changes its orientation from along z through B1 to -z. The magnitude of Beff decreases
to a minimum when aligned along B1 and then grows again. For M to follow Beff, the rate of precession
of M around Beff must be large compared with the velocity of rotation of Beff. This is the adiabatic
condition described earlier. Also, the sweep must be fast enough that no significant relaxation occurs
during the RF irradiation. These boundary conditions can be summarized as
When the offset frequency is incremented linearly over time, the RF power requirements can be
significant. Noting that Beff is much larger at the beginning and end of the passage than at the center
(when Beff = B1), Hardy and colleagues10 suggested a tangential frequency sweep of the form ωt - ω0
= ω1 tan(αω1t), where ωt - ω0 is the frequency offset at time t during the sweep, ω1 is the B1
frequency, and α is a scaling factor given by α = (2/ω1T0) arctan(θ/ω1), where T0 is the sweep time,
2θ is the range of the angular frequency sweep, and 0 < α < 1. This scheme provides an order of
magnitude increased efficiency compared with a linear sweep and has good inversion properties.
Because the B1 frequency is offset from the spin (Larmor) frequency, the B1 field is said to be
10
frequency modulated. In the scheme outlined by Hardy and coworkers, the amplitude of the B1 field
is constant. Other workers have described adiabatic pulses that are amplitude and frequency
modulated.11,12 These pulses provide insensitivity to inhomogeneity in the B1 spatial profile and can
therefore be of value when the RF transmitter field is nonuniform. B1 insensitivity is especially important
for inversion pulses as described in the following.
For spins lying at arbitrary orientations (not along z) at the start of the pulse, the trajectory in the
spatially varying B1 field is not defined. Classic adiabatic pulses can therefore be used for inversion
and excitation, but because they cannot induce plane rotations about an axis of the rotating frame, they
are inappropriate for refocusing. For a detailed analysis of the behavior of several classes of adiabatic
13
pulses, the interested reader is referred to the work of Ugurbil and colleagues. Following their
notation, the functions for amplitude and frequency modulation, respectively, of adiabatic pulses have
the form
where v is the ratio of the peak RF amplitude to the frequency modulation amplitude A, and fB and fω
10 11 12
are time waveforms. Function pairs for fB/fω include constant/tan, sech/tanh, and sin/cos.
By reversing the polarity of the B1 field at the half-passage point, Ugurbil and colleagues14 produced
refocusing adiabatic pulses. The effect of this pulse scheme is to invert magnetization oriented along z'
at the beginning of the pulse as for standard adiabatic passage. Spin vectors orthogonal to Beff at the
beginning of the pulse remain orthogonal to Beff throughout, but the phase "scrambling" that occurs
during the first half of the rotation is reversed during the second half by reversing the polarity of Beff.
Spins within the entire plane are therefore refocused (or inverted). However, this pulse breaks down
for spins far off-resonance.
11
Among the adiabatic pulses used, the hyperbolic secant pulse described by Silver and coworkers, is
the most widely used pulse. In addition to being insensitive to RF nonuniformity, the hyperbolic secant
pulse has excellent slice profile properties for inversion. Above a threshold B1 intensity, the
magnetization remains inverted, and outside the slice the spins are returned to their equilibrium position
along the z axis. Connolly and colleagues15 described a slice-selective, self-refocusing pulse that
leaves no phase variation across the slice and maintains insensitivity to B1 nonuniformity. Their pulse is
a composite one of 2π and π pulses; the 2π pulse performs no rotation, just compensates for the
phase scrambling caused by the π pulse.
9 of 29 11-04-2010 23:59
Adiabatic pulses have many advantages for certain imaging applications (such as IR sequences and
blood-tagging studies16 but they are relatively long, and significant relaxation can take place in some
tissues during irradiation. Tissue components with short relaxation times can lose signal during the
pulse, insensitivity to B1 variations may be compromised, and spurious high signal intensity may
17
originate from outside the slice.
In-Plane Spatial Encoding

An MR image is a map of the spatial distribution of spins. Conventionally, the imaging volume is
mapped onto a 3D Cartesian grid, the axes of which are labeled logically read, phase-encode, and
slice selection. The basis of spatial encoding with MRI is that the Larmor frequency depends on the
magnetic field experienced by the spins, and that this field in turn can be controlled. When this field is
made to vary with position by a gradient, the local frequency is also made to vary in a known way. By
means of the Fourier transform, it is possible to unravel the spectrum of frequencies generated by the
encoding gradient and derive from them a map of signal amplitude versus position.
We assume for the moment that the process of slice selection has reduced the task to encoding the
remaining two dimensions of the slice plane. In this case, the resolution in the slice direction is
reflected directly by the slice thickness in millimeters (ignoring complications due to RF truncation,
chemical shift and flow).
Frequency Encoding and k-Space

page 145
page 146
As discussed earlier, the application of a field gradient makes the precessional frequency a function of
position along the gradient axis. Analogous to slice selection, when a linear gradient G(r) is applied in
the frequency encoding ("read") direction, the precessional frequency ω(r) of a spin at position r along
this direction in the rotating frame is:
where ω(r) is in radians per second and f(r) is in hertz. For a read gradient G(r), the precessional
frequency in a plane perpendicular to G(r) at a displacement r from the isocenter is uniquely defined.
The signal amplitude at a given frequency is proportional to the number of spins at that frequency, as
determined by the Fourier transform of the time domain signal.
Mathematical analysis shows that immediately after an excitation pulse and ignoring relaxation and flow
effects, the signal S sampled at time t after the application of a linear gradient G(r) is the Fourier
transform of the spin distribution.
18
Here we introduce the concept of k-space or frequency space as the representation of the gradient-
encoded image before Fourier transformation. The raw time signal contains all the information about
the frequency distribution of the spins in the direction of the read gradient. Each value of k is
proportional to the zeroth time moment (or area), which is determined by the strength and duration of
the gradient waveform in that direction. But it must be inverted from a map of signal versus spatial
frequency into a readable image as signal versus position. A list of such values (spatial frequencies) is
collected in the gradient direction to be input to the Fourier transform for this purpose.
10 of 29 11-04-2010 23:59
Add to lightbox
page 146
page 147
Figure 5-10 k-space trajectory. Prephasing lobes in the phase-encode (ky) and read (kx) directions
assign the spins to the top left corner of the 2D k-space matrix. The read component is unwound by the
rephasing read gradient during signal sampling, causing the k trajectory to run from left to right across
the matrix. For the next phase-encode step, the amplitude of the phase-encoding gradient is
decreased, so that the ky offset is diminished at the start of the readout period. A table of such
incremental offsets spans the ky axis from maximal negative through zero to maximal positive values.
At each value of ky, an otherwise identical set of points is acquired along kx. The trajectory is therefore
a horizontal raster in the read direction, incremented in the vertical direction by each phase-encoding
iteration, analogous to scanning lines of print. The gradient-echo occurs halfway along kx, and the
maximal signal is generated when ky is zero, corresponding to zero amplitude of the phase-encoding
gradient. The ky values close to zero are low spatial frequencies, generally regarded as the primary
determinants of image contrast.
Defining
k(r) has the dimension of inverse distance or spatial frequency

When k(r) is multiplied by a position coordinate r, the result is a phase determined by the
magnitudes of k(r) and r (analogous to the product of temporal frequency and time)
The temporal sequence in which values for k(r) are encoded is termed the k-space trajectory
(Fig. 5-10). However, the same value of k(r) can be arrived at in a variety of ways, with the
trajectory controlled by gradient waveforms. This may have an indirect effect on the image in
several ways (e.g., contrast and motion sensitivity)
The FOV is the longest finite spatial wavelength (reciprocal of spatial frequency) that can be
encoded with discrete sampling. This corresponds to the minimum nonzero value of k. If N
discretely sampled data points are acquired equally spaced at intervals ∆t during a constant-
amplitude read gradient, FOV in the frequency-encoded direction can be defined as
11 of 29 11-04-2010 23:59
In the spatial frequency domain, small values of k (low spatial frequencies) are linked to large
objects in the image space. For a gradient to rotate the phase through 2π across an object, the
object must have a linear dimension λ = 2π/k along the direction of the gradient. The maximum
value of k defines the smallest dimension encoded; therefore, kmax defines the spatial resolution
(∆x). Again, assuming equidistant discrete sampling and equal coverage of negative and positive
spatial frequencies, the resolution in the read direction can be defined as
Equations 5-18 and 5-19 define the basic relationships between FOV and resolution, as well as
the read gradient strength (G), duration (N∆t), and sampling bandwidth (1/∆t). Equation 5-19
shows that the pixel size is twice the size of 1/(N∆k), because it is only the absolute distance
from k = 0 which contributes to spatial resolution. In other words, k = (±) N∆k/2 make the same
contribution to spatial resolution; the fact that they are each present in positive and negative
forms helps to increase signal and lessen artifact.
Sampling the Signal

As described in the previous section, if k-space has to be traversed in the read direction, the data are
sampled during the readout gradient. Since k is the gradient-time integral, the signal is sampled over a
range of k values during the gradient. If the read gradient is turned on and the signal is sampled
immediately after excitation, a free induction decay is acquired. To acquire positive and negative
spatial frequencies, usually, a dephasing read gradient is applied before the rephasing read gradient,
and an echo rather than a free induction decay is acquired. An echo is said to occur when
magnetization becomes coherent in the transverse plane. A gradient-echo occurs when the dephasing
effect of gradient activity is exactly balanced by the rephasing effect of later gradient activity. For this
to occur, two counterbalancing gradient pulses must subtend equal areas of opposite sign. In the
simplest gradient-echo measurement, a read gradient preparatory lobe is applied immediately after RF
excitation, as shown in Figure 5-8. This causes the spins to evolve a phase proportional to the gradient
area or k-space value as described earlier. If a gradient of opposite polarity is now applied, the spins
rephase when the area under the second lobe is the same size as that of the first. As the area of the
second lobe exceeds that of the first, the process of spin dephasing recurs, illustrated in Figure 5-8.
By acquiring data from the start of the rephasing gradient lobe, the build-up and decay of the
magnetization can be sampled. During the first half of the rephasing lobe, the free induction decay is
rebuilt to (potentially) its original amplitude, and during the second half it becomes dephased again.
In the laboratory frame, the raw time signal is in megahertz. This is down modulated to frequencies in
the audio range by a sinusoid at the Larmor frequency. The bandwidth in the signal is then determined
by the frequency distribution around the Larmor frequency associated with the read gradient. Although
a continuous range of frequencies is contained in the analog signal, the signal is sampled discretely
and is digitized by an analog-to-digital converter at a rate determined by the signal bandwidth. The
Nyquist condition must not be violated, in order to avoid aliasing.
To prevent aliasing, the sampling rate must be at least twice the bandwidth. For example, 256 complex
points are typically sampled during the readout period (analog-to-digital converter ON time) by a
quadrature detector (see Chapter 7). If the readout time is 5.12 ms and 256 complex points are
sampled (256 real and 256 imaginary), the sampling interval is 0.00001 s, the sampling rate is 100
kHz, and the resolvable bandwidth is 50 kHz. Low-pass filters are set to exclude signal (and noise)
outside the sampled frequency bandwidth.
Occasionally, the sampling rate is inadequate to prevent aliasing of parts outside the region of interest.
The low-pass filters are not effective, because the signal is wrapped around into the low-frequency
range and appears as a phantom limb or other body part. The only way to prevent this phenomenon is
to sample rapidly enough so that the unwanted signal is correctly localized outside the imaging
bandwidth. If the sampling frequency is doubled, an FOV twice as large can be correctly localized.
Overfolding is therefore prevented by oversampling the signal, which effectively increases the FOV but
12 of 29 11-04-2010 23:59
does not change the resolution, SNR, or gradient strength. Resolution is determined only by the kmax
value as described earlier. Oversampling increases both the number of pixels and FOV
proportionately.
It is convenient to normalize the sampling bandwidth to the number of pixels in the read direction, so
that a 50-kHz bandwidth corresponds to 50,000/256 = 195 Hz/pixel in a 5.12-ms readout time. If the
readout time is 10.24 ms, the bandwidth is 97.5 Hz/pixel and so on. This nomenclature is desirable
because
the bandwidth in hertz per pixel is not influenced by oversampling or variations in matrix size
the fat-water displacement caused by chemical shift can be conveniently expressed in pixels.
Phase-Encoding
Phase-encoding is a process whereby a spatially dependent phase shift in a direction orthogonal to
the read gradient is introduced by an additional gradient pulse. A series of such phase-modulated
echoes generated by a table of sequential pulses forms the basis for spatial encoding in this second
dimension. Each phase-modulated echo translates to one spatial frequency in the phase-encoding
direction, so the process must be repeated N times to encode N spatial frequencies (see Fig. 5-10).
Again, the FOV is defined by the smallest, non-zero value of k in this dimension. This differs from the
read direction, where N values of k are encoded during a single pulse of the read gradient (a single
readout period). We now consider this in more detail.
We have seen that a given value of k for a spin at position r results in a predictable phase shift.
However, as pointed out earlier, the definition of k as used in Equation 5-17 makes no assumptions
about how k was generated. Moreover, the expression is a generic one, equally applicable to any
spatial orientation. Recall that k is proportional to the gradient-time integral, so
k may be made a function of time if the gradient is constant and time is incremented, as is the
case during readout, or
k may be made a function of gradient strength, G, if the time is constant and the gradient
amplitude is incremented.
Now suppose that k(p) defines a gradient event in the phase-encoding direction y orthogonal to read;
then the phase of the echo is modulated as
page 147
page 148
Application of one such gradient event simply shifts the phase of the echo from each pixel in the read
direction by an amount determined by the mean contribution from all y offsets, and this provides no
useful information. However, if k(p) is successively incremented in value by ∆k(p) during a series of
measurements, the phase at position y undergoes periodic change as a function of k(p). This is
analogous to the periodic phase change at position r in the read direction as a function of k(r).
Therefore, when k(p) steps through a table of 256 values from -k(p)max to +k(p)max in 256 equal
increments ∆k(p) and k(r) steps through the same set of values under the read gradient, then
successive two-dimensional (2D) Fourier transformation along the read and phase-encoding directions
yields an image with isotropic spatial resolution.
Although the processes of spatial encoding in the read and phase-encoding directions are
mathematically similar, physically they are different and prone to different artifacts. In conventional
spin-warp imaging, adjacent values of k along the read direction are typically separated in time by
about 20 μs, and this is not long enough for significant changes to occur in the state of the
magnetization. In the phase-encoding direction, however, the interval between successive k values is
typically the TR of the sequence. Depending on the sequence structure, TR ranges from a few
milliseconds to several seconds, and during this interval substantial changes may occur to the
amplitude and phase of the MR signal. The sampled function may therefore be modulated by a variety
13 of 29 11-04-2010 23:59
of mechanisms, and the function modulating the signal is convolved in the reconstructed image.
Depending on the nature of the changes (e.g., motion, flow, nonuniform relaxation effects), the effects
may be manifest in the image as a loss of resolution, altered contrast, or, in the case of periodic
fluctuations, ghost artifact projected along the phase-encoding direction. In practice, phase-encoding is
carried out by applying a gradient pulse for a time tp after excitation and before readout (see Fig. 5-8).
19
If tp is constant and Gp is varied, the process is termed spin-warp imaging. A 2D image is calculated
by carrying out a 2D Fourier transform on the time domain data-first line by line, then column by
column.
Three-Dimensional Encoding
Phase-encoding can also be applied to the slice-selection direction to add a third dimension of spatial
encoding. If a thick slab rather than a thin slice is selected, the slab can be resolved into thin slices
(partitions) by incrementing a phase-encoding table in this direction, analogous to the 2D case. The
number of iterations (excitations) required to encode a 3D image is NyNz, where Ny and Nz refer to the
numbers of iterations in the phase-encoding and slice-selection directions, respectively. With
conventional acquisition, the minimal time required for this process is TR × Ny × Nz. Because
acquisition times with long TR sequences would be prohibitively long, the applications of 3D imaging
techniques were formerly limited to fast gradient-echo techniques.
Phase Accumulation
In most imaging pulse sequences, phase-encoding of each echo is accomplished with a single discrete
gradient pulse. The amplitude of the pulse is varied from echo to echo to encode a unique line of
k-space in each readout. The phase generated with each pulse is typically "rewound" with a pulse of
equal and opposite area, or the phase-encoded transverse magnetization is dephased or "spoiled"
after each readout. In this manner, no phase is accumulated from one echo to the next.
However, phase accumulation is an alternative means of phase-encoding. This strategy is used in

echo-planar imaging (EPI) in two modes: "blipped" phase-encoding, in which a series of constant-
amplitude gradient pulses are used for phase-encoding; and constant phase-encoding, in which a
constant-amplitude gradient is applied throughout the entire acquisition. In both cases, phase
accumulates from one echo to the next as the total area of the phase-encoding gradient increases. In
EPI with constant phase-encoding, it becomes difficult to distinguish the readout gradient from the
phase-encoding gradient in the traditional sense. Both gradient axes are active during data acquisition.
Speeding Up the Phase-Encoding Process

A practical point to note about phase-encoding is that it is time-consuming. Whereas in the read
direction all spatial frequencies are encoded in a matter of milliseconds, in the phase-encoding
direction the process may take several minutes. It is the iterative nature of phase-encoding and the
influence of longitudinal relaxation that necessitate the use of a TR in pulse sequences. Total
acquisition time is determined by TR × N × Acq, where N is the number of phase-encode iterations and
Acq is the number of acquisitions or averages (Acq is more than 1 if the SNR needs to be boosted or
motion artifacts need to be averaged out). Depending on the required contrast in the image, spins may
need to relax for a longer or shorter time between phase-encode steps. For pure T2-weighted images,
the TR should be several times the longest T1 in the region of interest. This typically includes water, so
TR should be several seconds. Conventional T2-weighted images are therefore time intensive. In the
case of IR imaging, the requirements for a long TR are even more stringent. T1-weighted spin-echo
measurements are designed to suppress partially the signal from long-T1 tissues, so these are done
with a relatively short TR-usually 300 to 600 ms.
Acquiring Multiple Lines Per Excitation

page 148
page 149
In conventional gradient-echo and spin-echo imaging, one line of k-space is acquired per α or 90°
pulse, and the number of iterations required is equal to the number of phase-encoding steps. If
14 of 29 11-04-2010 23:59
transverse magnetization persists after the signal is read, in principle this can be used to encode
further echoes. By reapplying the read gradients, interleaved with additional phase-encode (PE)
gradients, several Fourier lines can be acquired in much less time than with conventional encoding.
Two extreme forms of this process include the single-shot techniques of EPI 20,21 and RARE (rapid
acquisition with relaxation enhancement). Intermediate between conventional and single-shot
techniques are the hybrid fast imaging methods of turbo or fast spin-echo and segmented EPI. All of
these methods are considered in more detail later.
Rectangular Field-of-View
Because the high-cost item in terms of acquisition time is the number of iterations of the excitation
process (the number of repetitions of TR), time can be saved by decreasing this number. Reducing the
FOV in the phase-encoding direction by phase-encoding along the short axis of the subject (rectangular
FOV) yields equivalent spatial resolution in a shorter imaging time. For example, if the FOV in the read
direction is 350 mm with 256 points in read and the FOV in the PE direction is 175 mm, then 128 PE
steps yield an image with square pixels; the imaging time is halved. Because the SNR is proportional to
the square root of the number of PE steps, there is a penalty in the SNR with this method, depending
on the aspect ratio. Where the SNR is not limiting, the technique is successful. An important difference
to note between frequency encoding and phase-encoding is that hardware crystals cannot be used in
the phase-encoding direction to filter out unwanted frequencies. If the FOV is smaller than the object
size, overfolding will occur.
Note: Another way to speed image acquisition and/or prevent overfolding is by using parallel acquisition
techniques. Here, the sensitivity profiles of multiple receiver coil elements are used as a
complementary method for spatial encoding. With parallel acquisition, the FOV may be decreased
several-fold, effectively skipping points in k-space and using the coil sensitivity profiles to fill in the
missing data. Parallel imaging is dealt with in detail in Chapter 8 and is not discussed here further,
other than to say it can be used to enhance the performance of a wide variety of pulse sequences
(without necessarily changing their design).
Phase-Encode Ordering
In conventional gradient-echo and spin-echo imaging, the magnetization is sampled in a steady state
and no significant variations in the magnetization occur from one phase-encode step to the next. In this
case, the order in which the phase-encoding table is stepped through is unimportant. However, in many
fast imaging techniques, such as turbo fast low-angle shot (turboFLASH), segmented RARE, and EPI,
the magnetization is not in a steady state during acquisition and relaxation causes a modulation of
22-24
signal amplitude in the phase-encoded direction. Recall that the image contrast is determined
mainly by the low spatial frequencies (low phase-encode pulse amplitudes) and resolution by high
spatial frequencies (high phase-encode pulse amplitudes). Thus, signal modulation can have a
profound effect on the image contrast and effective resolution and can be manipulated by reordering
the phase-encoding table.
There are various different reordering strategies that have important implications for the image in terms
of signal weighting, contrast, effective resolution, or flow and motion properties. Some of the most
commonly used reordering schemes are described here.
Linear Reordering
Linear or sequential reordering is the most commonly used form of encoding. In linear reordering,
k-space lines are acquired sequentially, starting at one end of k-space and progressively acquiring
adjacent lines by incrementing the phase-encoding gradient in each step (Fig. 5-11A). It is used in
many conventional gradient-echo and spin-echo techniques where the magnetization does not vary
much from cycle to cycle. Linear reordering is also ideal for sequences where a steady state must be
achieved before data acquisition. A large number of RF pulses before the center of k-space minimize
possible signal modulations in the center of k-space. However, this strategy may not be suited for
magnetization-prepared (MP) acquisitions for the same reason, because the data acquisition RF
15 of 29 11-04-2010 23:59
excitations modify the MP signal weighting before the central k-space lines are acquired.
Centric Reordering
As the name suggests, phase-encoding begins at the center of k-space with the k = 0 line and then
progressively acquires the higher lines in k-space, alternating between the positive and negative k lines
(Fig. 5-11B). Such a strategy is typically used when a MP is used before the sequence readout. Since
the signal weighting from the magnetization preparation is of paramount importance, the central
k-space lines must be acquired immediately after the MP. As the data acquisition progresses to the
outer region of k-space, the signal weighting is altered by the application of data acquisition RF
excitations.
Reverse Centric Reordering

In reverse centric reordering the phase-encoding lines are collected starting from the outermost lines,
alternating between the positive and negative k lines, and moving inward until the k = 0 line is acquired
(Fig. 5-11C). Since the high spatial frequencies are enhanced in this reordering scheme, there is a
high-pass filtering effect on the image.
The reordering strategies mentioned above fall under the category of Fourier encoding, where the
k-space points are sampled on a rectangular grid, which is a requirement for direct Fourier
transformation. There are other specialized methods of encoding that deviate from the requirement of
sampling points on a rectangular grid. In these methods, the data are either reformatted to fit on a
rectangular grid after acquisition for direct Fourier transformation, or special image reconstruction
algorithms are used.
Radial Reordering
page 149
page 150
Add to lightbox
16 of 29 11-04-2010 23:59
Add to lightbox
Add to lightbox
Figure 5-11 Reordering schemes. A, Linear reordering. In linear reordering, the ky lines are acquired
sequentially starting at one end of k-space and incrementing the phase-encoding gradient in each
successive RF excitation. B, Centric reordering. In centric reordering, phase-encoding begins in the
center of k-space at the ky = 0 line, and progresses outwards by acquiring lines successively in
positive and negative k-space. C, Reverse centric reordering. In reverse centric encoding, data
acquisition begins at kymax and progresses inwards, acquiring lines successively in positive and
negative k-space.
The radial reordering method is commonly known as projection reconstruction, and is derived from
X-ray tomographic imaging. The archetypal Cartesian sampling encodes kx by varying the readout
gradient Gx during data sampling and encodes ky by stepping through the phase-encoding gradient Gy
in each step. In radial sampling, both the Gx and Gy gradients are switched on simultaneously during
sampling. Therefore, there are no specific readout and phase-encoding gradients in radial sampling.
The data are acquired as k-space points along the radius of a circle. A complete coverage of k-space
is achieved by changing the magnitudes of the Gx and Gy gradients such that the angle of the
projection θ is changed by an amount ∆kθ in each step. A schematic of radial k-space coverage is
shown in Figure 5-12.
17 of 29 11-04-2010 23:59
SPATIAL RESOLUTION IN RADIAL SAMPLING.

Spatial resolution in Cartesian sampling methods depends on the number of points sampled in the
readout direction and the number of lines in the phase-encoding direction. However, due to the polar
nature of radial sampling, encoding is now in terms of the radial direction, kr, and the angular direction,
kθ. The number of points in kr is akin to the number of readout points in Cartesian sampling and
determines the resolution in the radial direction, as the distance to the circumference of the k-space
circle. However, because the entire circumference is sampled by rays in all directions, spatial
resolution in the angular direction does not depend on θ, but is more related to the length of the
circumference of the k-space circle, which is also defined by kr. What varies, then, as the interval
between points kθ varies, is the FOV in the angular direction. Undersampling in kθ then does not lead
to a loss of resolution but streaking artifacts, which are the equivalent of aliasing in Cartesian sampling.
As can readily be appreciated from examination of Figure 5-12, the angular FOV decreases with
increasing distance from the center of k-space; i.e., higher spatial frequencies are associated with
smaller angular FOVs.
NYQUIST LIMITS IN RADIAL SAMPLING.

page 150
page 151
Add to lightbox
Figure 5-12 Radial scanning. Data are acquired along the radius of a circle. The distance between the
samples along the radius defines ∆kr. Multiple projections are acquired as different radii, with the
angle between the projections defining ∆kθ.
The Nyquist limits on the step sizes in k-space need to be adapted to polar coordinates for radial
sampling. The step sizes in the radial direction ∆kr and the angular direction ∆kθ are illustrated in
Figure 5-12. For angular sampling, the maximum angular step that must follow the Nyquist criterion is
1
given by :
where L = field of view. Therefore the minimum number of angular views necessary to prevent aliasing
is
Similarly, the radial step must satisfy
18 of 29 11-04-2010 23:59
and the number of total samples in the radial direction must be
This is the same Nyquist condition as that for the readout direction in Cartesian sampling.
Again, note from Figure 5-12 that the sampling density varies inversely as the radial distance from the
center of k-space. In Cartesian sampling, the sampling density is uniform throughout. The varying
density in k-space in radial sampling results in aliasing or streak artifacts in the images. Conversely,
the dense sampling in the center of k-space results in relatively higher SNR at the large distance
scales as compared to Cartesian sampling.
"ALIASING" ARTIFACTS IN RADIAL SAMPLING.

Aliasing artifacts appear in images due to inadequate sampling in k-space. Undersampling in the
angular direction by acquiring fewer projections than dictated by the Nyquist criterion and the varying
density of data sampling in k-space leads to aliasing artifacts in radially sampled images. Angular
undersampling in radially acquired data results in radial streak artifacts and decreased SNR as
opposed to Cartesian Nyquist violations that result in image wrap-around artifacts. For equal FOV in
the radial and angular directions, the number of projections acquired must be π/2 (approximately 1.5)
times the corresponding number of phase-encode steps in Cartesian sampling. In principle, radial
acquisitions allow a tradeoff between SNR and number of projections acquired rather than the tradeoff
between resolution and FOV in Cartesian sampling.
Spiral Reordering
Spiral scanning has also shown considerable promise for cardiac imaging, 25,26 functional brain
27 28
imaging, and flow imaging. Spiral scanning takes the blending of frequency and phase-encoding in
radial sampling one step further; identical gradient waveforms (with a phase shift between the two) are
applied to the two in-plane gradient axes, as shown in Figure 5-13. There are some advantages to
spiral scanning, such as the requirement of fewer excitations than in conventional Cartesian encoding,
and lower sensitivity to motion and flow artifacts. The main problem is the relatively long readout time
as compared to Cartesian sampling, which leads to blurring in the presence of off-resonance, chemical
shift, or susceptibility. The blurring is caused by the phase accumulated by off-resonant spins during
the entire readout. To reduce the readout time in a single shot, interleaved spirals are often used (see
Fig. 5-13).
Segmentation and Interleaving

In segmented k-space coverage, groups of lines are acquired in separate segments, and the
segments are merged together for a full coverage of k-space, e.g., we acquire lines 1, 3, 5, 7 … in
segment 1 and lines 2, 4, 6, 8 … in segment 2. The merger of these two segments then gives a
complete data set in k-space. The order in which the segments and lines are acquired may vary
depending on what reordering scheme is used, as explained above.
Segmented acquisitions form the basis of many multiecho techniques, such as turbo- or fast-spin-echo.
They can also be used for gradient-echo methods, when physiologic constraints, e.g., respiratory
motion, place constraints on the data acquisition time, and all the lines in an image cannot be acquired
continuously. It is then necessary to split the data acquisition into multiple segments, and the segments
are combined for image reconstruction, e.g., for a total matrix size of 75 lines, 25 lines are acquired in
each segment. A total of 75 lines are acquired in 3 segments and combined to form an image.
page 151
page 152
19 of 29 11-04-2010 23:59
Add to lightbox
Figure 5-13 Spiral scanning. Readout occurs continuously during simultaneous application of
sinusoidal read and phase-encoding gradients. A phase offset of π/2 between the gradients causes
the k trajectory to describe a spiral, starting at kx = ky = 0. As illustrated, the data are acquired in an
interleaved mode of four shots. Because the spins pass repeatedly through zero along kx and ky,
phase accumulation caused by flow is reversed, and the resulting images are intrinsically flow
compensated. Also, the low spatial frequencies are acquired early, so the minimal TE is small.
However, if the lines within each segment are acquired sequentially in k-space, a sharp signal
discontinuity will exist at the boundaries where the segments are merged. In the above example, there
will be two signal discontinuities, at line 26 and line 51. Such abrupt signal variations in k-space lead to
ghosting artifacts. These discontinuities are secondary mainly to the signal modulation by application of
successive excitations. To overcome this constraint, data are usually acquired in an interleaved
fashion. Interleaving minimizes the signal discontinuities by acquiring adjacent lines from similar echoes
within different segments. By appropriately stepping through k-space, the discontinuities in k-space can
be minimized and the signal envelope can approach a smooth filter; e.g., a schematic of a linearly
segmented interleaved data acquisition is shown in Figure 5-14. The total matrix size of 75 lines is
acquired in 3 segments with 25 lines acquired in each segment. As the diagram illustrates, the lines
adjacent to any line in k-space are acquired from similar echoes in different segments. This minimizes
the periodic signal variations in k-space and consequently reduces image artifacts.
Contrast Mechanisms in MRI

Basic Contrast Mechanisms
20 of 29 11-04-2010 23:59
Add to lightbox
page 152
page 153
Figure 5-14 Segmented and interleaved data acquisition. Schematic of a linear segmented and
interleaved data acquisition with a total matrix size of 75 lines and 25 lines per segment. Each set
therefore consists of 3 lines in k-space. In the schematic, each line style signifies lines acquired in one
segment, e.g., the solid line represents the first segment, the dotted line the second segment, and the
dashed line the third segment. The acquisition is interleaved such that the first echoes acquired in
each segment are adjacent to each other, i.e., set 1 consists of echo 1 acquired in 1st, 2nd, and 3rd
segments, set 2 contains the second echo from each segment, etc. In this way, the signal
discontinuities in k-space are reduced, which in turn reduces the image artifacts.
MRI has the advantage of generating different soft-tissue contrasts. Ignoring flow, chemical shift, and
many other subtle mechanisms, there are three fundamental parameters that determine signal and
contrast: proton density, T1, and T2 or T2*. In each case, the name suggests the difference in tissue
properties that is exploited. To gain an intuitive understanding of how each method of contrast can be
achieved, we will consider a simple gradient-echo sequence example. The contrast between two
tissues in a gradient-echo sequence is given by:
For a proton-density-weighted image, the

above equation must be reduced to:
Such a contrast will be achieved if TR>>T1 (e-TR/T1 ~0) and TE<<T2 (e-TE/T2 ~1).
-TE/T2
For T1-weighting, TE<<T2 (e ~1). The TR has to be chosen based on the T1 of the species to
be imaged such that the SNR is adequate; e.g., for very long T1s, a very short TR may not give
adequate signal.
-TR/T1
Similarly, for T2 or T2* weighting, TR>>T1 (e ~0), and TE~T2a or b.
It must be noted that T1 and T2 weighting cannot be isolated from the spin density ρ. There is always
an inherent spin density weighting in the signal. The T1 or T2 weighting can only be increased or
decreased to alter the signal. Since the difference in spin densities between various tissues is not very
large in practice, this is not a serious limitation.
21 of 29 11-04-2010 23:59
Magnetization Preparations
Magnetization preparations (MP) can be used to achieve a desired signal weighting using a
combination of RF pulses and gradients to modify the longitudinal magnetization before data readout.
Magnetization preparations allow flexibility in using different sequences for data acquisition because
the contrast is largely determined by the MP. There are numerous MPs in the literature, but here we
focus only on some of the most widely used.
Conventional Inversion Recovery

Preinversion may be used to generate T1-based contrast, as in IR imaging. As originally implemented
in spectroscopy, IR preparation was used to eliminate spectra from unwanted chemical species,
based on their T1.
Immediately after inversion, all magnetization lies along the -z axis and spins begin to relax
exponentially toward +z (Fig. 5-15). At the halfway stage, the z magnetization is zero, and no
longitudinal magnetization is available to tip into the transverse plane. The time between inversion and
excitation pulse or TI appropriate to null a species is related to its T1 as TI = 0.693T1, for TR much
greater than T1. A long TI is required to null long-T1 components, whereas short-T1 components are
nulled in a short TI.
Spins with long T1 (e.g., in fluids such as cerebrospinal fluid) may still be aligned along -z and have
negative phase at the time of readout. With phase-sensitive reconstruction (as opposed to magnitude
29
reconstruction), the phase as well as the amplitude of the signal is represented in the image. To
spins with negative phase, a pixel intensity less than background is ascribed. Spins with positive phase
have relaxed past zero (have short T1), and to these a pixel intensity greater than background is
ascribed. So the order of pixel intensity for a long-TI, IR brain image (assuming phase-sensitive
reconstruction) is white matter greater than gray matter greater than background greater than
cerebrospinal fluid.
Because long-TI IR imaging has poor slice efficiency and requires long acquisition times, it has come
into clinical use only with the advent of segmented RARE techniques. However, short-TI (tau) IR
30-32
(STIR) imaging has found many applications. In vivo, because spins with large T1 values also tend
to have large T2 values, contrast between long-T1, long-T2 and short-T1, short-T2 tissues is increased
by superimposing T2 effects on a short-TI preparation. In practice, this is done by using a moderately
long TE so that long-T2 components retain signal and short-T2 components do not. STIR imaging is
generally done with magnitude reconstruction, because the tissues with shortest T1 are near the null
point and all other tissues have negative phase; the range in signal is therefore -Mz to zero. Long-T1
and long-T2 tissues appear hyperintense compared with tissues with short T1 and short T2.
STIR is typically implemented with a TI value appropriate to null fat, so a narrow-bandwidth readout
can be used to boost the SNR without producing a chemical-shift artifact. The longer TE that
accompanies long readout times has a beneficial rather than a deleterious effect on contrast with
STIR, because the T1 and T2 effects are additive.
22 of 29 11-04-2010 23:59
Add to lightbox
page 153
page 154
Figure 5-15 T1 recovery curves. After inversion, spins recover longitudinal magnetization at a rate
dependent on their T1. Short-T1 components (such as fat) reach the null point quickly, whereas long-T1
components (such as water) recover slowly. The TI appropriate to null a given component then varies
with its T1 and with the TR of the sequence. Assuming TR is much greater than T1, spins with T1 = 300
ms are nulled when TI = 200 ms and spins with T1 = 1000 ms are nulled when TI = 700 ms, as
illustrated.
Add to lightbox
Figure 5-16 Double-inversion black blood RF preparation. After non-spatially selective inversion, a
spatially selective "reversion" pulse is immediately applied to spins within the slice plane. In-plane
spins (in this case in myocardium) effectively see no inversion pulse. Spins outside the slice plane (of
relevance to flowing blood) remain inverted and begin to relax. If the acquisition is timed so that the low
spatial frequencies are acquired during the null point of blood relaxation, no signal is seen from blood
that enters the slice between inversion and readout. ECG, electrocardiography. (From Simonetti OP,
Finn JP, White RD, Laub G, Henry DA: "Black blood" T2-weighted inversion-recovery MR imaging
of the heart. Radiology 199:49-57, 1996.)
23 of 29 11-04-2010 23:59
STIR has been shown to be an effective and sensitive technique for detection of a variety of disorders
in the liver32,33 and musculoskeletal system.34 As a fat suppression technique, STIR is generally more
robust than chemical-shift methods because it does not have stringent requirements for B0
homogeneity and is not vulnerable to susceptibility gradients. It can, however, fail with poor B1
homogeneity, and one must be careful in the design of inversion pulses. In general, one should use
pulses that are B1 insensitive, particularly with local transmit-receive coils. Specialized adiabatic RF
pulses are generally used for inversion because they are insensitive to B1 errors.
Saturation Recovery
A saturation recovery preparation consists of a π/2 RF pulse that converts all the longitudinal
magnetization into transverse magnetization, after which a gradient spoiler dephases it. The
magnetization of the tissues then recovers depending on their respective T1 values. Therefore, if data
are sampled after an appropriate time delay, T1-weighted images can be obtained. Although this
preparation is not widely used, one of its major applications is in cardiac perfusion imaging.35
Spatial Presaturation
Unwanted signal from within or outside the slice plane may be suppressed by means of spatial
36
presaturation pulses. This is typically implemented by selectively exciting the region of interest with a
π/2 pulse and then spoiling coherent magnetization with a crusher gradient. The effect persists until
longitudinal magnetization is restored and so is T1 limited. The most common application for spatial
presaturation is to eliminate artifact from flowing blood or to perform selective MR angiography (see
Chapter 27). It can also be used to decrease signals from organs or soft-tissues that are the source of
motion artifact, such as the larynx or anterior abdominal fat.
Black Blood Preparation

As discussed earlier, blood can be nulled effectively with a nonselective inversion pulse, based on its
T1 and independent of flow. A limitation of this process is that all spins in the slice plane are also
inverted and have their contrast behavior so determined. Magnetization within the slice plane can be
37
"reset" after inversion by a selective 180° pulse, as described by Edelman and colleagues and shown
in Figure 5-16. The effect of the combined pulses on spins within the slice is as if they had seen no RF,
while all spins outside the slice (most relevantly in blood) are inverted. As originally described, a fast
gradient-echo readout of the slice is timed to coincide with the null point of blood, so that blood that
has entered (and exited) the slice during TI gives no signal but the image is otherwise proton density
weighted. Excellent contrast between stationary tissue and the blood pool has been shown with this
technique, and it has been used to distinguish between blood and thrombus in abdominal aneurysms.
In studies of the heart, blood nulling has been used as a preparatory event in vessel wall imaging,38
breath-hold T1- and T2-weighted TSE imaging, and breath-hold STIR imaging by Simonetti and
co-workers39 with promising results.
Magnetization Transfer Contrast

page 154
page 155
40
Wolff and Balaban were the first to establish the use of magnetization transfer contrast in MRI. This
method derives its contrast mechanism by exploiting the difference between the two separate groups
of protons present in tissues-the mobile or free protons, and the immobile or restricted protons. The
immobile protons are those that are bound to large protein, lipid, carbohydrate, or nucleic acid
molecules (macromolecules, e.g., myocardium). Spins that are bound to macromolecules have a short
T2 and a broad spectral peak that is distributed symmetrically about the comparatively sharp peak of
the free protons at resonant frequency. Unbound spins, on the other hand, have a narrow-frequency
bandwidth and must be excited at or near the Larmor frequency. By applying a long RF pulse, typically
centered 1.5 kHz off-resonance at 1.5 T, the pool of bound spins can be saturated before exciting the
unbound pool by a pulse at the Larmor frequency. When imaging is performed following such
24 of 29 11-04-2010 23:59
excitations, the resulting image will show suppression of the signal from the bound group of protons.
The effect of magnetization transfer preparation is to suppress partially the signal from certain "solid"
tissues such as muscle and brain, depending on the relative populations of bound and unbound spins.
However, the free protons are coupled to the macromolecular protons by means of chemical exchange
or dipolar interactions. The result is that there is an exchange of protons between the free and bound
proton pools. Saturation of one pool therefore has the effect of reducing the signal of the other group
of protons. Applications of magnetization transfer preparation include background suppression in MR
angiography,41,42 T2-type contrast in the heart at shorter TE values,43-45 and gadolinium-enhanced
46
T1-weighted imaging of the brain.
Fat Suppression
Hydrogen nuclei in different molecules are surrounded by varying electron clouds. These orbital
electrons are constantly in motion within the cloud and, when subjected to an external magnetic field,
according to Lenz's law adjust their trajectories so as to induce an opposing field. The induced fields
shield the nuclei from the applied field B0 to different degrees, expressed by the shielding constant σ,
such that the net field seen by the nuclei is B0 (1 - σ). This phenomenon forms the basis for NMR
spectroscopy.3 Protons in lipid molecules are associated with denser electron cloud cover than those
in water molecules, and so, when placed in a magnetic field, lipid protons precess at a lower
frequency. The difference in precessional frequencies between nuclei in various chemical environments
is expressed by the chemical shift. The units used to quantify chemical shift are either hertz or parts
per million, depending on whether normalization for field strength has been performed. For example, at
1.5 T, hydrogen nuclei in fat molecules precess at approximately 210 Hz below hydrogen nuclei in
water. The Larmor frequency for water protons at 1.5 T is 63.75 MHz, and the chemical shift for fat
protons is (-210 Hz/63.75 × 106 Hz) × 106 = -3.3 ppm. Conversion from parts per million to shift in
hertz is performed simply by multiplying by the frequency of RF irradiation in megahertz.
For many clinical applications, it is desirable to eliminate the signal from protons in one or other
chemical environment. This most commonly involves suppression of the fat signal either by exciting and
47-49 50,51
then crushing the fat or by exciting the water selectively. Chemical-shift-selective (CHESS) fat
saturation is accomplished by applying a frequency-selective, spatially nonselective excitation pulse to
the fat protons and then spoiling the transverse magnetization with a strong gradient pulse. The water
spins are unaffected and can then be excited and imaged without interference from the fat. The
technical challenge in chemical-shift-selective saturation is selecting only the fat or the water without
selecting the other species. This is not a problem when there is no overlap between fat and water
resonances. For example, at 1.5 T an RF pulse could be centered on the fat frequency with a
bandwidth of, say, ±100 Hz. Such a pulse would leave water spins relatively untouched. However, in
practice there are nonuniformities ∆B0 in the main field, caused mainly by patient-induced susceptibility
gradients that may exceed the fat-water chemical-shift separation. Water protons precessing at
(γ/2π)(B0 + ∆B0) may well lie within the bandwidth of the fat saturation pulse. The result is that in some
portions of the image fat is suppressed, whereas in other portions water is suppressed. To minimize
this effect, B0 homogeneity must be optimized by shimming before all spectral selective
measurements.
Fat suppression is most often carried out by using either a gaussian pulse centered on the fat
frequency or a binomial excitation scheme with short, wide-bandwidth ("hard") pulses. Because the
frequency separation of fat and water is small, the RF saturation bandwidth must be narrow (typically
less than 200 Hz) to avoid crosstalk.
Binomial excitation is performed by applying a series of hard RF pulses.49 The series describes a
binomial distribution when the amplitude of the pulses is plotted over time, and the time between pulses
corresponds to a chosen phase evolution between the fat and water spins. For example, if the
interpulse interval is 2.38 ms, the fat-water phase evolution between pulses is 180° at 1.5 T. The
pulses may be designed to excite water or fat selectively by appropriate choice of the RF phase. 51 If
the phase is kept constant, the sense of nutation for fat resonances is reversed with alternate pulses,
25 of 29 11-04-2010 23:59
and because the binomial distribution is symmetric, the net flip angle is zero. For the water
resonances, the pulses are additive, such that the net flip angle is the sum of the individual pulses. For
example, with a 1:2:1 excitation scheme, if each unit contains enough energy for a flip of α, the
maximal achievable flip angle is 4α for spins on resonance (Fig. 5-17). To excite and then spoil the fat
resonances, the RF phase would be alternated (e.g., from x to -x) so that the phase of the RF always
coincides with the phase of the fat spins and the net flip angle for water is zero.
Better spectral selectivity is achieved with higher-order binomial schemes such as 1:3:3:1 or higher.
However, the time penalty can be costly, and in some cases the minimal TR of a gradient-echo
sequence may be more than doubled. This may have undesirable effects not only on acquisition time
but also on image contrast in sequences that depend on steady-state effects. The simplest binomial
series is a 1:1 "jump return," where one species sees θ/2 + θ/2 = θ and the other sees θ/2 - θ/2 = 0.
This approach has been extended by Thomasson and co-workers,52 who offset the phase of the
second RF pulse by π/2 or less, shortening the interpulse delay by a factor of 2 or more. They
achieved a substantial time saving in short-TR three-dimensional (3D) gradient-echo sequences and
extended the concept to spatial-spectral excitation.
page 155
page 156
Add to lightbox
Figure 5-17 A 1:2:1 binomial water-selective pulse sequence. Fat spins (f) and water spins (w) evolve
a π phase shift in a time τ. When a series of pulses are applied at intervals of τ, the effects on fat and
water spins are opposite, so the flip angle for water sums to 90° whereas that for fat is zero.
From a practical perspective, the major disadvantage to using CHESS fat suppression is the additional
time required to play out the narrow bandwidth RF pulses. When these pre-pulses are added to
short-TR sequences, the penalty in slices (or partitions with 3D) can be prohibitive. Considerable cost
savings are possible when the presaturation pulses are played out less frequently, while maintaining
sufficient fat suppression.53 Here, the CHESS pulses are shared among several slices or partitions, by
54
nesting the presaturation loop within the slices (or partitions) loop. When used in multislice 2D FLASH
26 of 29 11-04-2010 23:59
or 3D FLASH imaging,55 the cost in time or slices can be negligible.
Spin-Echoes and Gradient-Echoes

Based on the method of generation of echoes, sequences are broadly classified into two types:
spin-echo sequences and gradient-echo sequences. Before we describe the various different
sequences, the basics of spin-echo and gradient-echo are revisited. It should be noted that, without
several highly simplifying assumptions, the details of multiecho pulse experiments are extremely
involved, and in what follows we consider only the simplest examples. For a detailed description, the
interested reader is referred to several outstanding works by Hennig and Scheffler.56-58
Add to lightbox
Figure 5-18 Spin-echo. In A, a 90° pulse has been applied along x' and the magnetization lies along y'.
B, At time t = TE/2 later, the spins have dephased in the x-y plane through T2* decay. In C, a 180°
refocusing pulse applied along x' reflects the spin vectors through the x' axis. The spins continue to
precess as before the refocusing pulse, and in D, at time t = TE, a spin-echo occurs. With multiple
refocusing pulses, imperfections in the RF are propagated. When the refocusing pulse is applied
along y' as in C', the spin-echo occurs along y' without phase inversion as in D'. Flip angle errors are
corrected on alternate echoes and are not propagated. This is the Meiboom-Gill modification of the
Carr-Purcell multiecho pulse. It is standard in long spin-echo train sequences.
Spin-Echoes
When the magnetization is rotated into the x-y plane, initially all of it is available for detection and the
MR signal is maximal. However, through relaxation processes, the signal begins to decay
exponentially. The falling signal profile over time is known as free induction decay, and the speed with
which the signal falls off is determined by how efficient transverse relaxation processes are. Spins
dephase through a combination of T2 decay, main field nonuniformity, and local field distortions caused
by susceptibility gradients. The inhomogeneity in the main field ∆B0 causes spins to dephase at a rate
γ∆B0 in addition to the intrinsic spin-spin interaction 1/T2. The net rate of decay is therefore the sum
(1/T2 + γ∆B0). In the presence of main field nonuniformity, the effective decay time is expressed by
T2* = 1/(1/T2 + γ∆B0). The rate of decay of the transverse magnetization is therefore greater in the
presence of significant ∆B0.
page 156
27 of 29 11-04-2010 23:59
page 157
Add to lightbox
Figure 5-19 Stimulated-echo acquisition mode (STEAM). A stimulated echo is formed by a train of
three 90° pulses. In A, B1 is applied along x' and the magnetization is laid down along y'. B, After a
suitable time TE/2, the spins have dephased in the x-y plane as a result of T2* processes or applied
gradients. A 90° pulse is now applied along y', rotating the plane of magnetization to the y'-z' plane (C).
Spins aligned along z' undergo no further phase changes, whereas those along y' dephase completely
(D). We refer to this interval as the STEAM time. Another 90° pulse is now applied along y', rotating
the z magnetization along x' (E). After TE/2, these spins all have components along positive y' (F),
forming a stimulated echo. Note that only half the magnetization is recoverable in this way.
Hahn59 described a technique of RF refocusing to neutralize the effects of B0 inhomogeneity and local
susceptibility effects, which is termed spin-echo. After the 90° excitation, a 180° pulse is applied, and
an echo is formed at a time determined by twice the interval between the two pulses. As outlined in
Figure 5-18, the 180° pulse has the effect of refocusing spins that fanned out before its application.
The spin-echo refocuses dephasing caused by fixed differences such as ∆B0, chemical-shift effects,
and local susceptibility gradients. It does not refocus dephasing caused by random processes such as
spin-spin interaction1 ("true T2") and diffusion.60 The refocusing pulse can be reapplied multiple times
18
to generate a train of spin-echoes as described by Carr and Purcell. The T2 decay envelope can
then be measured and values of T2 computed.
In the modification of the Carr-Purcell spin-echo experiment described by Meiboom and Gill, 61 the
phase of the refocusing 180° pulses is offset by 90° relative to the excitatory 90° pulse (see Fig.
5-17C' and D'). Thus, if a π/2 pulse is applied along x, the refocusing π pulses are applied along y.
The effect of the π pulse is to form the complex conjugate M* of M by rotation about y. This serves to
prevent propagation of imperfections in the π pulses as follows. Consider a situation in which instead
of π refocusing pulses, 150° pulses are applied along x, the same axis as the excitatory 90° pulse. At
28 of 29 11-04-2010 23:59
the first spin-echo, the magnetization M lies 30° below the x-y plane, at the second spin-echo M lies
60° above the x-y plane, and so on for subsequent spin-echoes. If the refocusing RF pulses are
applied along y, then the undershoot on the first echo is corrected on the second, the undershoot on
the third is corrected on the fourth, and so on. Every second spin-echo now has the correct amplitude,
and the initial error is not propagated. Formerly, generation of multiple spin-echoes had as its main
application the measurement of T2 values. With the development of rapid acquisition with relaxation
57,62 63-65
enhancement (RARE) and related rapid spin-echo imaging techniques, the use of multiecho
sequences has increased significantly. This buffering effect of offsetting the phase of the π pulses may
be useful clinically when the application of multiple refocusing pulses results in RF energy deposition
outside recommended limits. The flip angle of the π pulses can be scaled back as described earlier
with little perceptible change in image quality but substantial reduction in specific absorption rate. This
is particularly relevant in the design of hybrid fast RF refocused sequences such as turbo (fast)
spin-echo.
Stimulated Echoes
page 157
page 158
Stimulated echoes occur when a train of three or more RF pulses is applied sequentially.56 Each
stimulated echo is formed by three (nominal) 90° pulses. The first pulse nutates the magnetization into
the x-y plane so that it initially lies, for example, along y' (Fig. 5-19). Subsequently, over an interval
TE/2, where TE is the echo time, the spins fan out in the x-y plane as a result of T2* decay. The
second pulse now rotates the plane of magnetization so that it has one axis parallel to z. The
components lying along z no longer precess, and those lying along y' undergo irreversible T2 decay.
The third pulse recalls the magnetization stored along ±z to ±x'. After another interval TE/2, the
magnetization has a component along +y'; this is the stimulated echo. As illustrated in Figure 5-19, the
delay between the first two pulses is the same as the delay between the last pulse and the stimulated
echo.
The following are noteworthy about stimulated echoes:
The amplitude of the echo is smaller than that of a spin-echo of comparable TE. This is because
only half the available magnetization is used in the stimulated echo
For the same reason, not all spins are exactly in phase along y' at the echo (only spins 3 and 7
in Fig. 5-19 are), as would occur with a spin-echo. Rather, there is a distribution of spin vectors
oriented in the positive y' direction; the largest are refocused along y' and others also have a
component along ±x'
The interval between the second and third pulses can be substantially larger than TE. This long
interval can be exploited to heighten sensitivity to diffusion and flow effects
T1 relaxation occurs between the second and third pulses. If this is significant, the amplitude of
the stimulated echo is decreased still further
Stimulated echoes occur in a variety of circumstances in which one might not expect them. For
example, in multiecho spin-echo imaging, three sequential RF pulses produce a stimulated echo
that may be superimposed on a spin-echo. It is important in fast or turbo spin-echo (TSE)
imaging to be aware of the effects stimulated echoes can have on image contrast 66-68 and to
engineer them appropriately. In certain forms of fast gradient-echo imaging, stimulated echoes
modulate the amplitude of subsequent gradient and RF echoes and can alter image contrast
and signal-to-noise ratio
Stimulated-echo imaging69 has found applications in several areas, including ultrafast imaging, 70
diffusion imaging,71,72 motion tracking,73 brain imaging,74 and cardiac imaging.75
© 2010 Elsevier
29 of 29 11-04-2010 23:59
SOME SPECIFIC PULSE SEQUENCE TYPES

The building blocks of all sequences in common use have now been described, and several features
specific to certain classes of pulse sequences have also been outlined. We now review the major
categories of pulse sequences in use or in development for clinical imaging, highlight similarities and
differences among them, and allude to their major applications. For detailed treatment of clinical
applications of these and other techniques, the reader is referred to the relevant chapters in this
edition.
Spin-Echo versus Gradient-Echo

A spin-echo occurs by the refocusing effect of a 180° RF pulse as outlined in Figure 5-18. The timing
of the spin-echo is determined by the time interval between the 90° and 180° pulses, independent of
gradient events. A gradient-echo occurs when the area subtended by a gradient pulse is exactly
neutralized by an equal and opposite area. The parameter that determines when the gradient-echo
occurs is the rate at which these areas are traced out, and this is proportional to the gradient
amplitude. In general, sequences are designed so that the gradient-echo is made to coincide with a
spin-echo (if one exists), but this is not always the case. The steps outlined earlier for spatial encoding
can be applied to data acquired under either a spin-echo or gradient-echo envelope. With spin-echo
refocusing, dephasing effects caused by magnetic field nonuniformities and chemical shift are
reversed. Irreversible dephasing persists because of T2 decay and diffusion, providing mechanisms for
tissue contrast. Gradient reversal does not undo dephasing effects resulting from chemical shift or
magnetic field inhomogeneity, whether caused by nonuniformity of the B0 field or magnetic
susceptibility gradients. The choice of TE in a gradient-echo sequence may profoundly influence image
contrast, depending on whether fat and water are in phase or opposed in phase.
Gradient-Echo Pulse Sequences

FLASH or Spoiled GRASS
Gradient-echo techniques have proved useful for fast MRI in a variety of circumstances. Yet another
contrast parameter is available with gradient-echoes: the flip angle. The simplest of all gradient-echo
imaging techniques is the FLASH76 or spoiled GRASS (gradient-recalled acquisition in the steady
state) sequence (see Fig. 5-20), widely used for fast T1-weighted imaging.
77
In short-TR gradient-echo imaging, phase coherences may build up in the transverse magnetization.
Depending on the desired contrast in the image, these coherences may be detrimental or beneficial.
To generate T1 contrast in a short-TR gradient-echo sequence, it is desirable to eliminate all
transverse coherence between phase-encoding steps, because such coherences accentuate signal
from long-T2* (and therefore long-T1) components. These coherences can be suppressed by applying
a crusher gradient78 after sampling the echo and before the next excitation; this process is termed
gradient spoiling. The aim is to cause dephasing of all coherent x-y magnetization. Another technique,
called RF spoiling,78,79 uses a table of pseudorandom values for the phase of the B1 field on
successive iterations, thus preventing the build-up of transverse coherences. The phase of the echo is
varied between phase-encoding steps; this phase is also locked into the receiver reference signal so
that image calculation is not affected. Unlike gradient spoiling, RF spoiling is effective even in the
center of the magnetic field. Attention must be paid to the details of the B1 phase series over time, to
avoid unexpected coherences.
page 158
page 159
1 of 20 12-04-2010 00:00
Add to lightbox
Figure 5-20 2D FLASH sequence structure. After slice selection, the rephasing slice-selection,
dephasing read, and phase-encoding gradients are applied simultaneously. This is followed by the
rephasing read gradient and signal sample window (ADC on time). To destroy coherent magnetization
between excitations, a gradient spoiling table is applied after data acquisition. RF spoiling is typically
also employed with this structure.
The spin-warp sequence structure for gradient-spoiled 2D FLASH as outlined in Figure 5-20 provides a
good template for illustration of conventional spatial encoding and signal sampling. After slice selection,
gradients in slice, phase-encode, and read are applied simultaneously to maximize time efficiency. The
rephasing lobe in slice selection is to undo the dephasing effects of the first lobe. Simultaneous
application of the dephasing read gradient and maximal-amplitude phase-encoding gradient sets the
magnetization to one corner of k-space in a diagonal trajectory, as shown in Figure 5-10. The
rephasing read gradient then unwinds the phase of the dephasing lobe and causes maximal rephasing
in read at the time of the echo. The k-space trajectory of this last step is linear in the read direction, at
an offset in the phase-encoding direction determined by the strength of the phase-encoding gradient
pulse (see Fig. 5-10). Because the phase-encoding gradient steps through a table of values from
-G(p)max to G(p)max, the k-space trajectory is a linear raster of values. Discrete signal sampling occurs
via the analog-to-digital converter during the build-up and decay of the gradient-echo. In Figure 5-20, a
gradient spoiling table is shown. When present, this increases the minimal TR of the sequence and
slightly decreases slice efficiency. The same basic FLASH structure (typically with a wide bandwidth)
is used as the readout module after the preparatory period in turboFLASH.
TurboFLASH
TurboFLASH is a gradient-echo sequence that uses an RF preparation followed by multiple, short TE
and TR FLASH acquisitions, as depicted in Figure 5-21. The Ernst angle is small because of the short
TR, and, without preparation of the magnetization, the images would be basically proton density
weighted and appear "flat." To introduce T1 contrast, a single inversion pulse is applied before
acquiring all of the image data (k-space lines). This differs substantially from the classic method of IR
imaging in which an inversion pulse precedes every line of data (every phase-encoding step).
Nonetheless, turboFLASH images have IR-type contrast that is typically greater than can be achieved
with non-IR spin-echo or gradient-echo methods. T1 relaxation is changing throughout the acquisition
period, and the data are modulated in the phase-encoded direction by the T1 recovery curve. Despite
this, image quality with turboFLASH is typically limited more by SNR than by image blurring.
The type of T1 contrast generated with turboFLASH depends on how much relaxation has progressed
by the time the central lines of k-space are collected. This in turn is influenced by the order in which the
phase-encoding steps are acquired. To clarify this point, consider that a nominal delay time TI follows
the inversion pulse and that the FLASH readout is acquired with centric phase-encoding. The center of
2 of 20 12-04-2010 00:00
k-space that dominates image contrast is therefore acquired at the nominal TI, and the contrast is
comparable to that of an IR image at that TI. If, however, the FLASH readout is acquired with linear
phase-encoding, then the central lines of k-space are acquired not at the nominal TI but at an effective
TI of TI + (TR × N/2), where N is the number of phase-encoding steps. Thus, if TI = 500 ms, TR = 12
ms, and N = 128, then TIeff = 500 + (12 × 64) = 1268 ms! Therefore, with linear phase-encoding, TIeff
scales with the number of phase-encoding steps. If TI = 0, then TIeff = 768 ms, a typical value for
T1-weighted abdominal imaging at 1.0 or 1.5 T. It should be noted, however, that with linear ordering
TIeff is related to the number of phase-encoding steps and that with reasonable spatial resolution the
minimal TIeff is large. TurboFLASH has been extended to a 3D acquisition by Mugler and Brookeman80
for rapid T1-weighted imaging of the brain.
The inversion pre-pulse for turboFLASH may be spatially selective or nonselective. A spatially
nonselective RF pulse is typically applied over a wide bandwidth in the absence of a field gradient. To
excite spins evenly over a wide range of frequencies, the envelope of the RF pulse approaches a spike
function in the time domain; its Fourier transform approaches a constant value in the frequency domain.
Such a pulse is often referred to as a "hard" pulse, and in practice it is applied as a short rectangular
pulse that excites a broad-frequency envelope. Because hard pulses are short, they require relatively
high peak transmitter voltage.
page 159
page 160
Add to lightbox
3 of 20 12-04-2010 00:00
Add to lightbox
Figure 5-21 TurboFLASH pulse sequence diagram. After a single magnetization preparation phase
(A), a fast FLASH sequence is played out. The time between the preparation event and the zero
phase-encoding step (here illustrated halfway through the FLASH sequence) determines the image
contrast. For example, in the case of inversion (B), the effective inversion time is the interval from the
inversion pulse to the central phase-encode step (arrow). With centric phase-encode reordering, this
step occurs at the beginning of the FLASH sequence, and the effective TI is the nominal TI.
In turboFLASH, a hard inversion pulse is combined with spatially selective α pulses for slice selectivity.
For single-slice imaging, the main difference between using selective and nonselective inversion pulses
is the effect on spins outside the slice. There are two major consequences of using a nonselective
inversion pulse:
Because the entire FOV of the transmitter is inverted, the blood signal may be uniformly nulled
without interference from flow-related effects and independent of the blood flow velocity. Flow
artifacts may therefore be completely eliminated by choice of the appropriate TIeff, for example,
in coronal imaging of the abdomen
Because the entire FOV of the transmitter is inverted, sufficient time must be allowed for
relaxation before the next slice is acquired. This decreases time efficiency in multislice
applications.
If selective inversion pulses are used, sequential slice acquisition may proceed without waiting for
adjacent slice positions to relax. The blood signal is less predictable with selective pulses and is often
bright because of inflow effects. Selective inversion pulses are best implemented with the hyperbolic
secant (sech) or other optimized pulses.
Segmented TurboFLASH
By shortening the acquisition window of the FLASH readout, both the T1 filtering effect and the minimal
TI limitation can be addressed. This has been accomplished by segmenting81 the acquisition into
several sequential windows, each containing a fraction of the data, as shown in Figure 5-22. For
example, 128 phase-encoding steps can be acquired as four segments, each containing 32 lines. An
inversion pulse precedes each segment and a delay time separates the segments from each other to
allow relaxation to occur. The intersegment delay should be several seconds, and one dummy segment
should precede the first segment acquisition to establish a steady state for the remaining ones. The
concept of segmentation of k-space in the phase-encoding direction has been applied to several other
82
techniques for fast imaging, including breath-hold cine MRI of the heart, breath-hold-triggered MR
83 84
angiography, TSE, and segmented EPI.
TurboSTEAM
page 160
page 161
4 of 20 12-04-2010 00:00
Add to lightbox
Figure 5-22 Segmented turboFLASH. The basic structure is similar to that of turboFLASH, but the
acquisition is broken into a number of segments, each preceded by magnetization preparation. The
k-space is traversed in a number of interleaved shots, and the acquisition window of the FLASH
sequence is correspondingly narrowed compared with the single-shot case. Less relaxation time
filtering of the data occurs, and image contrast can be more finely controlled.
Stimulated-echo acquisition mode (STEAM) imaging can be carried out, like turboFLASH, as a
single-shot technique. In the method described by Frahm and colleagues, 85 two 90° pulses are applied
TE/2 apart. A series of fast, low-flip-angle readouts is used to capture the magnetization stored along
z, each separately phase-encoded as a stimulated echo (Fig. 5-23). The method can be used for fast
diffusion imaging or black blood imaging of the heart.
Fast Imaging with Steady-State Precession
Add to lightbox
page 161
page 162
5 of 20 12-04-2010 00:00
Figure 5-23 TurboSTEAM sequence. Slice-selective π/2 pulses are separated in time by TE/2. Slice-
selection and read gradient lobes ensure full dephasing of transverse magnetization, which is aligned
along the z-axis by the second π/2 pulse. A series of fast, low-flip-angle gradient-echo readouts
follows, each separately phase-encoded. Within STEAM time (TM), the slice-selection lobe balances
the initial dephasing lobe and the read gradient balances the earlier dephasing read gradient lobe.
The gradient-echo occurs at TE/2 after the α pulse, coinciding with the stimulated echo (see Fig.
5-18). Because α is small, only a portion of the stored z magnetization is read with each phase-
encode step, so the process can be repeated many times.
Add to lightbox
Figure 5-24 2D FISP sequence structure. Compared with the structure in FLASH, the phase of the RF
is alternated from one line to the next, and the gradient spoiling table in the FLASH sequence is
replaced by a phase-encoding rewinding table. Coherent magnetization that persists after signal
sampling (for example, for fluid) is realigned along z and is made available for re-excitation with the
subsequent α pulse. Long-T2* components are therefore accentuated.
Fast imaging with steady-state precession (FISP)86 is a fast gradient-echo technique used to enhance
the signal from long-T2* components (such as fluid). When TR is short compared with the rate of
decay of transverse magnetization, it is possible to recycle a large part of the magnetization by
alternating the phase of the RF pulse on successive excitations; that is, the direction of the B1 field is
alternated. The effect is to realign along z the component of magnetization that lies along y at a time
TR after the previous excitation, while at the same time tilting longitudinal magnetization from z to -y. It
follows that the phase of Mxy is alternated from positive to negative with successive phase-encoding
steps; this is accounted for by matching the transmitter and receiver phase. Also, for this scheme to be
effective in restoring Mxy along z, Mxy must be made to lie coherently along y, because B1x has no
effect on spins aligned along ±x. In general, at the end of a readout period, the spins are dephased in
x-y because of a combination of gradient moments from the readout and phase-encoding directions.
To rephase Mxy along y, these gradient moments must be reversed before the next excitation (Fig.
5-24). Balancing the gradients rephases spins that have not moved over the duration of the interval.
However, flowing spins have acquired a velocity-dependent phase shift that is not unwound with
zero-order gradient balancing. As illustrated in Figure 5-24, the gradients in FISP are not fully balanced
in the slice-selection and read directions.
Time-reversed FISP (PSIF)

Unlike FLASH and FISP, which both acquire the primary gradient-echo, the time-reversed FISP (PSIF)
or contrast-enhanced, Fourier-acquired, steady-state (CE-FAST)87,88 acquisition utilizes the RF echo
formed by the repeated α pulses. The signal intensity of the RF-refocused echo is weighted by T2 over
the interval from one α pulse to the readout after the next α pulse, as depicted in Figure 5-25.
Stimulated echoes as well as spin-echoes contribute to the PSIF signal, making the contrast highly
dependent on flip angle. PSIF is useful in rapid acquisition of fluid-sensitive images.
Double Echo in the Steady State (DESS)
6 of 20 12-04-2010 00:00
The double echo in the steady state (DESS)89 or fast acquisition double echo (FADE)90 sequence (Fig.
5-26) utilizes both the free induction decay gradient-echo used in the FISP sequence, and the RF echo
of PSIF. The gradient-echo is offset from the RF echo, and they may be used to generate separate
images that can be combined or displayed separately. The RF echo tends to accentuate long-T2
components such as fluid. This may be superimposed on a more T1-weighted anatomic image related
to the FISP echo.
Spin-Echo Pulse Sequences
Add to lightbox
page 162
page 163
Figure 5-25 PSIF pulse sequence diagram. In this technique, the RF echo is of prime importance in
generating image contrast. It is mainly the RF portion of the steady-state signal that is acquired. The
gradient structure is balanced within the TR scope, and the RF echo occurs during readout in the
following cycle.
Add to lightbox
Figure 5-26 DESS pulse sequence diagram. The gradients are balanced along all three axes and the
gradient and RF echoes are offset from each other. They may be used to generate separate images,
which can be combined or displayed separately.
7 of 20 12-04-2010 00:00
Conventional spin-echo pulse sequences have formed the basis for clinical MRI since the modality was
introduced. Various derivative techniques have been proposed that dramatically speed up image
acquisition and make possible applications that would otherwise not be feasible. Strictly speaking, any
technique that uses a 180° refocusing RF pulse and acquires data under the RF echo is a spin-echo
pulse sequence. With the explosive pace of development in fast imaging, a rapidly increasing number
of techniques now fall within the scope of this definition.
Conventional Spin-Echo Sequences
Add to lightbox
Figure 5-27 Spin-echo pulse sequence diagram. After a π/2 excitation pulse, a π refocusing pulse is
applied at TE/2. A spin-echo occurs at TE as outlined in Figure 5-18. By comparison with Figure 5-18,
note that the polarity of the dephasing read gradient is the same as that of the rephasing read
gradient. This is a consequence of the π pulse, which inverts the phase of the echo.
The design of a conventional spin-echo pulse sequence is straightforward and is illustrated in Figure
5-27. The initial gradient and RF events are similar to the 2D FLASH structure of Figure 5-20, but a
refocusing π pulse is applied at time TE/2, immediately after the phase-encoding, slice-rephasing, and
read-dephasing lobes. Note that the polarity of the dephasing and rephasing read gradients is the
same, because the π pulse has inverted all phase accumulations. Note also that the π pulse is slice
selective but does not have a rephasing lobe; π pulses are self-refocusing (for stationary spins). The
spin-echo occurs at TE, and the gradient-echo is made to coincide with this. In the sequence diagram
in Figure 5-27, the shortest possible TE is used, so this corresponds to the structure for a
T1-weighted, short-TR, short-TE sequence. For a T2-weighted spin-echo sequence, a delay time is
added on either side of the π pulse. In Figure 5-27, the area of the gradient pulses from the 90° pulse
to the center of the echo is zero when the phase-reversal effect of the 180° pulse is taken into
account. First-order or higher-order gradient moment nulling may be employed to diminish sensitivity to
flow and motion artifact on spin-echo as well as gradient-echo sequences.
Turbo (Fast) Spin-Echo Sequences

page 163
page 164
The concept of independently phase-encoding sequential spin-echoes of a multiecho train and using all
the echoes to generate one image was first expounded by Hennig and colleagues. 62 Their original
implementation of a single-shot technique, which they called RARE, acquired all the data for an image
with just one 90° pulse and multiple 180° pulses. The phase-encoding gradient was incremented with
each spin-echo, such that all of k-space was traversed in a single excitation. This was an extension of
an earlier concept of Mansfield for single-shot imaging, which he called EPI, using a rapidly oscillating
read gradient and a constant, low-amplitude phase-encoding gradient. Whereas EPI requires fast
8 of 20 12-04-2010 00:00
gradients and associated hardware, RARE techniques are feasible on conventional MRI machines.
Both RARE and EPI are subject to significant artifacts. With RARE, T2 decay progresses between
points in k-space, whereas with EPI T2* decay progresses between points in k-space. The net effect
of these dependences is that RARE images are highly T2 filtered, causing severe blurring of all but
long T2 components, and EPI images are highly susceptibility weighted. Both of these effects are
minimized by the use of shorter acquisition windows, either by enhanced gradient performance or by
segmenting the acquisition into several narrower windows. This latter technique is employed in TSE
and segmented EPI. Whatever technique is used, relaxation time filtering is diminished when the
interval between successive echoes is minimized. This requires short gradient rise times and relatively
short readout periods (wide bandwidths).
The method of TSE or segmented RARE employs a variable number of separately phase-encoded
echoes per 90° excitation pulse, as shown in Figure 5-28. The number of spin-echoes per excitation is
called the echo train length or turbo factor. If, for example, the turbo factor is 15, then for each pass of
the TR, 15 phase-encoding steps rather than 1 are acquired, speeding acquisition of a single slice by a
factor of 15. It should be noted, however, that with multislice acquisition, the minimal TR per slice is
increased, so that overall multislice acquisition time is decreased by a factor less than the turbo factor.
Add to lightbox
Figure 5-28 Turbo spin-echo. TSE sequence diagram showing T2 decay that occurs from one echo to
the next. The exact modulation function that this decay imposes on the image data is determined by
the phase-encode ordering.
The T2 contrast in TSE is determined by the TE of the echoes used to encode the center of k-space.23
If, for example, the low spatial frequencies correspond to TE values of about 100 ms, the T2 weighting
is similar to that for a conventional spin-echo image with TE = 100 ms. The order in which the echoes
are phase-encoded is important in TSE and variants. Sharp discontinuities and periodicities in k-space
must be avoided to minimize ring and ghost artifacts, and this generally implies that the acquisition
order must be interleaved from segment to segment. Blurring is more of a problem with short TE
values because short-T2 components contribute more to the signal, and these are the source of the
worst T2 filtering. One feature that distinguishes TSE T2-weighted images from conventional
T2-weighted images is the intensity of the fat signal. With conventional T2, fat is subdued, whereas
with TSE, fat is intense91; this can be a source of confusion. For this reason, TSE images may
sometimes give the false impression of being T1 weighted. Another feature of TSE that has caused
some concern in brain imaging is its relative insensitivity to the susceptibility effects of blood products.
Repeated and rapid application of refocusing RF pulses prevents the dephasing by diffusion through
9 of 20 12-04-2010 00:00
local fixed field nonuniformities that is responsible for the signal loss in hematomas and hemosiderin.
A half-Fourier version of single-shot RARE, termed half-Fourier single-shot TSE (HASTE), can be used
to acquire a single image with standard spatial resolution in a few hundred milliseconds. An inversion
pre-pulse can be appended to any of the TSE group of sequences to give IR-type contrast. Also,
because of the increased time efficiency of TSE, it is feasible to run a 3D version for isotropic
coverage of the brain or spine in less than 10 minutes.
Turbo Inversion Recovery

Conventional STIR imaging is time inefficient, and this has limited its widespread clinical use. With the
advent of TSE techniques, turboSTIR imaging (Fig. 5-29) became an obvious and immediate
extension.30 In fact, some of the less desirable features of TSE imaging may be offset by using a STIR
preparation. The bright fat signal, distracting on many TSE images, is nulled on turboSTIR, and the T2
filtering effect, caused largely by short-T2 components, is less of a problem. For these reasons,
turboSTIR may have higher contrast and less blurring than its TSE counterpart, and turboSTIR is the
only TSE technique in widespread clinical use for musculoskeletal imaging.
page 164
page 165
Add to lightbox
Figure 5-29 TurboIR pulse sequence diagram. The structure is that of a TSE sequence after an
inversion pulse. After a suitable TI, a 90° excitation pulse is followed by a train of refocusing RF pulses,
each giving rise to a spin-echo. Each echo is separately phase-encoded, and the effective TE is
determined by the low-spatial-frequency echoes.
When used in imaging of the brain, STIR produces bright cerebrospinal fluid, which may render
periventricular white matter lesions difficult to see. It has been shown that by using a long TI to null the
cerebrospinal fluid, together with a long TE for T2 weighting, periventricular white matter lesions are
92
made conspicuous in the brain. This technique, called fluid-attenuated IR (FLAIR), is time-consuming
but is much more efficient when used as a turboIR implementation. Slice looping structures for IR
imaging can be optimized at the extremes of TI:
For short TI, the most efficient loop structure has the inversion and readout completed on one
slice before inverting the next
For long TI, an interleaved structure is used, whereby slices 1 to N are first inverted and then
slices 1 to N are excited and read.
With intermediate values of TI, either limitations are imposed on the number of slices available or
acquisition time is prolonged.
Gradient and Spin-Echo (GRASE)
10 of 20 12-04-2010 00:00
One of the limitations of TSE is its high specific absorption rate in multislice applications with large
turbo factors. A solution to this problem has been suggested by Feinberg and Oshio93 in a technique
that they called gradient and spin-echo (GRASE). By combining data from spin-echo refocusings with
gradient-echo refocusings, images can be acquired with fewer RF pulses and more quickly than with
TSE (Fig. 5-30). In principle, sensitivity to blood products, which is diminished in TSE, can be
recovered in GRASE through the gradient-echo refocusings. Because the polarity of the read gradient
in GRASE is alternated, additional steps must be taken to prevent N/2 ghosting. GRASE can also be
used as a single-shot technique.
Hyperechoes
Spin-echo-based sequences are limited at higher field strengths such as 3T by the specific absorption
rate (SAR) limits. Hennig and Scheffler developed a new spin-refocusing strategy using echoes of
94
echoes, called hyperechoes. Hyperechoes can alleviate the SAR limitations of TSE imaging by
reducing the flip angles for data acquisition without compromising on the SE-like signal behavior.
In a CPMG (Carr-Purcell-Meiboom-Gill) experiment, where a 90° pulse is followed by multiple 180°

pulses, images are acquired within a reasonable time while maintaining the spin-echo properties of the
signal. As a result of the altered phases of the RF pulses in a CPMG sequence, the sequence
develops high signal intensity even for low refocusing flip angles. However, with large turbo factors at
low flip angles, the efficiency of signal usage may be reduced due to destructive interference between
the spin configurations. Hennig and Scheffler94 showed in their study that the magnetization scrambled
by any arbitrary multipulse sequence can be unwound by a 180° pulse and a spin-echo with full
intensity can be formed.
Add to lightbox
page 165
page 166
Figure 5-30 GRASE pulse sequence diagram. The structure differs from that of TSE in that additional
gradient refocusings surround the spin-echoes. These additional gradient-echoes are separately
phase-encoded, increasing time efficiency and lowering RF power deposition. Note that the polarity of
the additional read gradient refocusings is opposite to that of the central lobe. Time reversal of
alternate echoes can give rise to artifacts analogous to those in segmented EPI. Because of the
frequent spin-echoes, however, sensitivity to susceptibility gradients is less than with EPI. In general,
the low spatial frequencies coincide with the spin-echoes. The k-space offset between adjacent
gradient-echo and spin-echo refocusings is constant, and the gradient tables are incremented
between adjacent gradient-echo refocusings.
11 of 20 12-04-2010 00:00
Add to lightbox
Figure 5-31 Illustration of a hyperecho sequence. Any sequence of pulses that show the above
symmetry and are time reversed centered around a π pulse will act as a single π pulse, as long as
relaxation effects are neglected (α = flip angle, Φ = RF phase). The π pulse is applied with a 0 phase.
If a π/2 excitation precedes these pulses, a hyperecho will be generated as shown.
A schematic of a hyperecho sequence is shown in Figure 5-31. Low flip angles are used for the
refocusing pulses instead of 180° as in a CPMG sequence. If the phase-encoding is such that the k =
0 line is sampled at the (2n + 1) echo, the central 180° pulse is placed before the (n + 1) refocusing
period. The hyperecho mechanism time-reverses all the echoes after the 180° pulse and a full-intensity
spin-echo is formed at the (2n + 1) echo. Since the contrast behavior of TSE sequences has been
shown to be determined by the echoes with no phase-encoding, the spin-echo characteristics of the
image are retained, but with a much reduced SAR.
Steady-State Free Precession (SSFP, TrueFISP, Balanced FFE, FIESTA)
Add to lightbox
Figure 5-32 SSFP pulse sequence diagram. The gradients are balanced along all three axes so that
steady-state effects related to long-T2* species are emphasized. The gradient-echo and the RF echo
are superimposed at TE, and the gradient structure is motion insensitive.
With improved gradient capabilities in recent years that permit the use of ultrashort TRs, SSFP has
gained wide popularity for various applications in cardiac imaging,35,82,95 flow imaging,96 T1/T2
quantification,97,98 whole body imaging,99 etc. SSFP is a unique sequence in that it belongs to both the
spin-echo and gradient-echo families. In an SSFP sequence, a gradient-echo is acquired similar to a
FLASH or FISP sequence. However, the gradient structure is symmetrically balanced in the slice,
phase-encoding, and read directions (Fig. 5-32) and no RF spoiling is implemented. Coherences are
maintained in successive RF cycles so that the transverse magnetization from one TR contributes to
the signal in the succeeding TRs. The RF echoes generated by the train of α pulses in the steady state
are then superimposed on the gradient-echo (assuming a uniform main field). Signal is therefore
highlighted from long-T2* and long-T2 components. Since there is no magnetization spoiling, there are
no saturation effects as in spoiled techniques, and high flip angles can be used in SSFP. The balanced
and symmetric gradient structure also generates motion insensitivity because of first-order gradient
moment nulling.
12 of 20 12-04-2010 00:00
While the TR remains shorter than T2 and there is no RF spoiling, the magnetization develops a
coherent steady state, which is a combination of the longitudinal and transverse components. Starting
from equilibrium magnetization, a finite time is necessary for the magnetization to reach steady state,
during which RF excitations are continuously applied. In steady state, the SSFP signal is T2/T1
weighted, which is a unique form of contrast and is seeing increasing applications for in vivo imaging.
The SSFP signal is a complicated function of parameters such as TR, T1, T2, flip angle, and
off-resonance angle β (β = γ×∆B0 ×TR). Despite its many advantages, such as high SNR compared to
spoiled gradient-echo sequences and high imaging efficiency, signal sensitivity to off-resonance is one
of the major limitations of SSFP.
page 166
page 167
Add to lightbox
Figure 5-33 Spectral response of the SSFP signal. Observe that the signal is relatively uniform when
the off-resonance angle is around π radians. (TR = 5 ms, T1 = 1200 ms, T2 = 250 ms, flip angle = 50°.)
Off-resonance
Off-resonance artifacts often manifest as banding patterns in SSFP images. Since the gradients are
balanced in each TR, the resonance frequency offset dominates the phase behavior of the spins in
SSFP. One of the main sources of off-resonance is B0 field inhomogeneity. For high flip angles, the
signal dependence in SSFP on the off-resonance angle β is shown in Figure 5-33. If β is in the vicinity
of 0 or 2π radians, there is a strong signal dependency on the resonance offset that is undesirable. On
the other hand, the signal is relatively uniform for a range of β centered around π.
The static B0 field varies within an object depending on position. This leads to extraneous signal
variations in an image. If there is a changing β across the object, the signal sensitivity to off-resonance
will degrade the uniformity in the reconstructed image. However, if β remains close to π, the above
problem can be avoided as long as the variation in β across the imaged object is small. The variation in
β can be kept small by careful shimming and by keeping the TR as short as possible. To uniformly bias
β around π in the entire image, the phase of the RF pulses is incremented linearly by π radians with
each excitation. This scheme of incrementing RF phases by π in each cycle is often referred to as
alternating RF pulses. It must be kept in mind that for lower flip angles (~10°), the signal is lower for β
= π than for β = 0, and the contrast in an image is reversed as compared to that at higher flip angles.
However, the signal uniformity is preserved at β = π even for lower flip angles.
Approach to Steady State

When the magnetization starting from its equilibrium value M0 experiences repeated RF excitations, it
takes time on the order of 3T1 before it reaches steady state.79 During the transient period, the
magnetization oscillates as it approaches steady state. The oscillations decay as the number of
applied RF excitations increases. Data cannot be collected during this transient period because the
13 of 20 12-04-2010 00:00
signal oscillations in k-space give rise to image artifacts. Due to a prohibitively long preparation time
that may be required for tissues with long T1, a few different strategies have been proposed to
address the problem of a quick approach to steady state.
When alternating RF pulses are applied, the magnetization reaches a steady state where the
magnetization vector oscillates between α/2 and -α/2 (α = flip angle) about the z-axis with successive
RF pulses. Therefore, if the phase-alternating α pulse train is preceded by an RF pulse with a flip
angle of α/2, the magnetization vector will be forced into steady state immediately if the frequency is
on-resonance.100 For off-resonant spins, the signal oscillations persist despite the α/2 pre-pulse.
Another method that improves over the off-resonant response of the α/2 preparation is the use of a
101,102
linear flip angle preparation. Both these magnetization preparations substantially reduce the
magnetization fluctuations in the approach to steady state and permit data acquisition during signal
transience to steady state.
Echo-Planar Imaging
EPI is the fastest MRI method to have found potential clinical applications. 20,21,103 All of the spatial
encoding is carried out after a single excitation under a spin-echo or gradient-echo envelope. The read
gradient is oscillated, generating a series of rapid gradient-echo refocusings, during each of which one
Fourier line is read. Phase-encoding is carried out with either a constant, low-amplitude pulse or a
series of blipped pulses between readouts, which progressively increase the gradient moment in
discrete steps (see Figs. 5-34 and 5-35). The polarity of the read gradient is alternated plus and
minus, resulting in a rectilinear zigzag raster in k-space. With blipped phase-encoding, the k-space
trajectory is mapped directly to a Cartesian grid, so no interpolation is necessary before image
reconstruction. With constant phase-encoding, the k-space trajectory is an oblique zigzag, so the k
values do not fit exactly on a Cartesian grid and must be interpolated before Fourier transformation.
page 167
page 168
Add to lightbox
14 of 20 12-04-2010 00:00
Add to lightbox
Figure 5-34 Blipped phase-encoding EPI pulse sequence (A). The k-space is traversed by monotonic
progression of phase from a series of small, discrete gradient pulses (B). Of particular importance in
EPI is the time reversal of the echoes from line to line. The k trajectory in read therefore alternates
between left to right and right to left, and the blipped phase-encoding gradient increments the offset in
ky between lines. Without proper correction for phase drift between odd and even lines, N/2 ghosts
may be severe.
Signal sampling with EPI depends on the read gradient waveform. When "resonant" gradients are
used, the fastest waveform is a sinusoidal pattern at the resonant frequency of the gradient coil (see
Chapter 7). This generates a smooth transition without sharp discontinuities at plateaus. It also means
that the signal can be sampled through the sinusoidal half-period, including the ramps. In this case,
correction must be made for the unequal rate of k-space encoding with linear sampling, again with
interpolation. An alternative scheme samples nonlinearly in time but linearly in k-space, and this
requires matching the sampling window to the gradient waveform with high precision.104
With EPI, a variety of contrast-altering preparatory schemes may be used. These include preinversion
and diffusion encoding with either a spin-echo or a stimulated-echo readout.
Add to lightbox
15 of 20 12-04-2010 00:00
Add to lightbox
Figure 5-35 Constant phase-encoding EPI pulse sequence (A). The k-space is traversed by
accumulation of phase throughout a constant, low-amplitude gradient (B). Because ky is incremented
continuously and not as a series of discrete steps as with blipped EPI, the k trajectory now oscillates
obliquely rather than horizontally. This must be interpolated onto a Cartesian k-space grid before
Fourier transformation.
Motion and Flow

Because conventional MR image acquisition typically takes several minutes, significant motion and flow
can take place within the body, giving rise to a variety of effects and artifacts.105,106 These can be
separated into two classes: time-of-flight effects, which are based on the macroscopic motion of fluid
and tissue both through and within the image plane; and phase effects, which are due to the motion of
spins along the applied imaging field gradients. Both classes of motion effects have been exploited in
MR angiographic techniques107,108 and quantitative flow measurement.109,110
Time-of-Flight Effects
page 168
page 169
In conventional spin-echo imaging, washout and signal loss occur when spins initially excited in the slice
plane move out of the slice in the time interval between the 90° and 180° pulses. This effect can be
useful in generating black blood contrast and eliminating motion artifact by suppressing the signal from
flowing blood. Washout of flowing blood occurs by the same mechanism in TSE, in which each 90°
excitation is followed by a train of 180° refocusing pulses, and in spin-echo EPI, in which the 90° to
180° pulse pair is followed by the readout of multiple echoes.
Black blood contrast is deliberately enhanced through the use of spatial presaturation pulses and
out-of-slice inversion pulses to reduce selectively the signal from blood flowing into the image plane
from either or both sides. Ordering the slice acquisition in the direction of flow also serves to saturate
the signal from blood before it enters the slice plane. Alternatively, ordering the slice acquisition
opposite to the flow direction emphasizes in-flow enhancement.
When TR is short, the signal from stationary tissue saturates, while the blood signal is constantly
refreshed by inflow. Inflow enhancement occurs when the TR between RF excitations is sufficient to
allow partially saturated spins to be replaced by fully magnetized spins from outside the imaged slice.
Inflow enhancement is the mechanism of contrast in time-of-flight angiography techniques. This is
described in detail in Chapter 27.
16 of 20 12-04-2010 00:00
For the case of constant flow velocity, bulk motion generates the signal suppression or enhancement
patterns just described but no discernible image artifacts. On the other hand, if flow or motion is time
varying throughout the image acquisition, the washout and inflow effects are time varying as well and
cause a modulation of the amplitude of the MR image data. Any modulation of the signal amplitude not
generated intentionally by spatial encoding gives rise to image artifacts. Most physiologic motion is
approximately periodic with cardiac and respiratory cycles and generates discrete "ghost" artifacts in
the image. Cardiac and respiratory triggering or gating is commonly employed to eliminate this signal
modulation by acquiring data at the same point in the cardiac or respiratory cycles. Fast image
acquisition techniques can be used to reduce the data acquisition to a reasonable breath-hold period
(<20 seconds); this is an effective means of eliminating respiratory motion artifact. A similar approach
to avoiding motion can be used in cardiac imaging by gating image acquisition to diastole, when the
heart is relatively stationary.
Another time-of-flight effect is the misregistration artifact, which is due to motion with components
along at least two of the imaging axes. This is often referred to as the oblique flow artifact.111 This
artifact occurs because in conventional 2D Fourier transform imaging, the spatial positions along the
imaging axes are encoded at different distinct points in time: at the center of the RF pulse on the slice-
select axis, at the center of a unipolar phase-encoding pulse on the phase-encoded axis, and at the
echo on the read axis. Motion with directional components along two or three axes causes a
misregistration of spins to appear in the image at a location they never actually occupy. This effect is
most obvious when blood vessels are lying in the image plane and oriented oblique to the readout and
phase-encoded axes. It is possible to design the phase-encoding waveform to encode position at an
arbitrary point in time, eliminating misregistration between two of the three axes. By extending this to a
3D Fourier transform pulse sequence, where two of the three axes are phase-encoded, oblique flow
misregistration can be eliminated entirely.
Phase Effects
In addition to time-of-flight effects, motion in the presence of magnetic field gradients contributes to the
phase of the MR signal. The phase of the magnetization at a specific point in space and time is defined
in terms of the gradient waveform, G(t), and the position, r, of a spin by the equation
This equation is central to the design of gradient waveforms for imaging but is valid only under the
assumption that the positions of spins remain unchanged over the interval of integration, that is, during
the application of the gradient. When human subjects are imaged, this assumption is frequently
violated. If a spin is moving along the direction of a time-varying magnetic field gradient G(τ), Equation
5-26 becomes
where τ is the variable of integration and r(τ) is the time-varying position of a moving spin in the
direction defined by the gradient. Thus, the phase is a function of not only the applied gradient but also
the motion itself. It is important to understand the effects that motion may have on the phase of the MR
signal when designing a pulse sequence for clinical application. At times it may be desirable to quantify
physiologic motion, to utilize motion as a contrast mechanism, or to minimize the effects of motion in
the final image. Gradient waveforms can be defined that exploit the motion dependence of the phase
of the NMR signal to achieve these goals.
To facilitate an understanding of the effects of motion on the MR signal phase, it is helpful to expand
the instantaneous position as a Taylor series, which expresses a function in terms of its value and its
instantaneous derivatives. For example, substituting
where
x0
position at time t0
x'0
v0 = velocity at time t0
17 of 20 12-04-2010 00:00
x'0
a0 = acceleration at time t0
back into Equation 5-27 and rearranging terms yield
page 169
page 170
Add to lightbox
Figure 5-36 FLASH gradient waveforms with first-order moment nulling (velocity compensation). The
waveform is designed so that the zeroth- and first-order gradient moments are nulled. Position-
dependent phase shifts and velocity-dependent phase shifts (independent of position) are nulled. Note
that the echo occurs asymmetrically during the readout window. This is designed to shorten the TE and
minimize sensitivity to higher-order motion.
Thus, the phase can be separated into terms proportional to specific instantaneous derivatives of
position. The gradient waveform moment integrals define the sensitivity of a gradient waveform to each
order of motion. This forms the basis for motion artifact suppression by gradient moment nulling,
phase-contrast angiography, and flow velocity quantification.
Gradient Moment Nulling

The relationship between data space (k-space) and image space can be corrupted by motion-induced
phase, resulting in image artifacts by several mechanisms. Dispersion of phase by spatial variation of
motion within a voxel causes a loss of signal intensity. If this phase dispersion is time varying as well, it
may cause ghosting and blurring. If all of the spins within a voxel are moving in the same manner, they
all acquire the same phase shift as a result of motion. This can result in a misregistration artifact if
there is a motion component in more than one imaging axis and in ghosting and blurring if it varies from
view to view. All of these deleterious effects of motion can be suppressed by the technique of gradient
moment nulling.112,113
Examining Equation 5-29, one can see that phase related to velocity and higher derivatives of position
can be nullified by designing the gradient waveform G(t) so that its first-order moment is zero. This can
be extended to higher-order derivatives as well, although for the relatively short time intervals defined
by the typical TE values used in MRI, most physiologic motion may be represented by a finite Taylor
series expansion of two or three terms. A gradient waveform consisting of n + 1 pulses is required to
null n moments. For example, a three-lobe gradient waveform is required to achieve zero phase
related to position and velocity, as shown in Figure 5-36. Gradient waveforms with any order of motion
compensation are easily designed by using linear algebra techniques.112
18 of 20 12-04-2010 00:00
Equation 5-27 is usually evaluated at the echo (t = TE), where the result indicates the phase shift of
the dominant kx = 0 spatial frequency component of the image. Phase at other spatial frequencies is
measured relative to the linear phase generated by the spatial encoding of stationary spins and cannot
be fully compensated by standard moment-nulling techniques. Spins moving along a constant linear
field gradient exhibit a nonlinearly varying phase as a function of time, and this causes a blurring in the
image.
Gradient moment nulling can also be used to compensate for the oblique flow misregistration artifact
described in the previous section. Defining the TE as the time origin, or center of the moment
expansion, and designing the phase-encoding waveform with the first moment nulled effectively encode
the position at TE of spins with a constant velocity component along the phase-encoded axis. Thus, the
phase-encoded and readout coordinates of a spin are defined simultaneously, eliminating
misregistration. This can be extended to the slice-select direction in a 3D sequence as well.
TSE pulse sequences differ from conventional spin-echo imaging in that stimulated echo components
are included in the MR signal. Moment nulling for the stimulated echoes must consider only the time
that the signal spends in the transverse plane, as this is the only time that phase caused by motion can
accumulate. Hinks and Constable114 described a technique for moment nulling the spin-echo and
stimulated echoes in TSE; resultant waveforms are shown in Figure 5-37.
Despite its speed, EPI can be highly sensitive to flow and motion effects, 115,116 particularly when
in-plane velocities are high. The time-of-flight and phase effects found in conventional MRI have
analogues in EPI. The washout effect described for spin-echo imaging occurs in spin-echo EPI and can
be used to suppress the signal from blood. Inflow enhancement does not normally occur in single-shot
EPI because the TR is effectively infinite. However, intentional presaturation of stationary spins can be
used to generate inflow enhancement if desired. Misregistration artifacts caused by directional
components of motion along multiple axes can occur in EPI, particularly in single-shot techniques, in
which the TE values are typically long and the time difference between encodings along the slice-select
and phase-encoding axes is significant.
page 170
page 171
Add to lightbox
Figure 5-37 Velocity-compensated TSE pulse sequence. First-order moment nulling is applied in
slice-selection and read directions. Both the primary spin-echoes and stimulated echoes are flow
compensated.
Phase effects of motion in EPI are somewhat different from those in conventional MRI. Whereas the
phase-encoded axis in EPI is directly analogous to the conventional readout axis in terms of motion
sensitivity, significant differences exist in the EPI slice-selection and readout axes. Single-shot EPI is
19 of 20 12-04-2010 00:00
insensitive to motion along the slice direction for two reasons: the high-powered gradient system
required for EPI can be used to keep the duration (and motional dephasing) along the slice-select axis
to a minimum, and there is no line-to-line variation in effects caused by through-plane motion. Slice
selection occurs only once per image, effectively freezing motion and eliminating any signal modulation.
The EPI phase-encoded axis, either blipped or constant, can be treated exactly like a conventional
readout axis in terms of moment nulling. It should be noted that the duration of the EPI readout is
typically on the order of 20 to 100 ms, and significant dephasing can occur. The oscillating EPI readout
gradient has no analogue in conventional MRI. The first moment of this waveform oscillates between
two values and causes a modulation of signal phase even with constant-velocity motion. Fortunately,
these oscillating moments are typically quite small, and high velocities along the read direction are
required for this effect to be significant. The line-to-line oscillation results in N/2 ghosts. In practice, the
same strategies of keeping TE short, gradient moment nulling, and blood signal suppression are
effective in suppressing motion artifacts in EPI.
© 2010 Elsevier
20 of 20 12-04-2010 00:00
CONCLUSION
We have seen that, although the details of how a specific pulse sequence is designed may vary greatly
from one technique to another, all pulse sequences share a set of common elements. Spins must be
excited and the resulting signal spatially encoded, detected, and digitally processed to produce an
image. Within this general framework, a variety of RF pre-pulses and gradient pulse modifications may
be used for specific purposes, as described. The resulting images may contain information that
focuses on detailed anatomic structure or on correlates of organ function such as flow, perfusion,
diffusion, oxygenation, muscle mechanics, or kinematics. The power of MRI lies in its flexibility, and its
clinical potential will continue to grow with developments in its underlying technology. Pulse sequence
design will remain a cornerstone for future development.
REFERENCES
1. Haacke EM, Brown RW, Thompson MR, Venkatesan R: Magnetic Resonance Imaging: Physical Principles and Sequence
Design. New York: John Wiley & Sons, 1999.
2. Liang Z, Lauterbur PC: Principles of Magnetic Resonance Imaging: A Signal Processing Perspective. Piscataway, NJ:
Wiley-IEEE Computer Society Press, 1999.
3. Emsley J, Feeney J, Sutcliffe LH: High Resolution NMR Spectroscopy. Oxford: Pergamon Press, 1965.
4. Bracewell R: The Fourier Transform and Its Applications, 2nd ed. New York: McGraw-Hill, 1986.
5. Brigham EO: The Fast Fourier Transform and its Applications. Englewood Cliffs, NJ: Prentice-Hall, 1988.
6. Hoult DI: Solution of the Bloch equations in the presence of a varying B1 field-Approach to selective pulse analysis. J Magn
Reson 35:69-86, 1979.
7. Le Roux P: French Patent 8610179, 1986.
8. Shinnar M, Bolinger L, Leigh JS: The synthesis of soft pulses with a specified frequency response. Magn Reson Med
9. Pauly JM, Le Roux P, Nishimura DG, Macovski A: Parameter Relations for the Shinnar-Le Roux Selective Excitation Pulse
Design Algorithm. IEEE Trans Med Imag 10(1):53-65, Mar 1991.
10. Hardy CJ, Edelstein WA, Vatis D: Efficient adiabatic fast passage for NMR population inversion in the presence of
radiofrequency field inhomogeneity and frequency offsets. J Magn Reson 66:470-482, 1986.
11. Silver MS, Joseph RI, Hoult DI: Highly selective π/2 and π pulse generation. J Magn Reson 59:347-351, 1984.
12. Bendall MR, Garwood M, Ugurbil K, Pegg DT: Adiabatic refocusing pulse which compensates for variable RF power and
off-resonance effects. Magn Reson Med 4:493-499, 1987. Medline Similar articles
13. Ugurbil K, Garwood M, Rath AR: Optimization of modulation functions to improve insensitivity of adiabatic pulses to
variations in B1 magnitude. J Magn Reson 80:448-469, 1988.
14. Ugurbil K, Garwood M, Bendall MR: Amplitude- and frequency-modulated pulses to achieve 90° plane rotations with
inhomogeneous B1 fields. J Magn Reson 72:177-185, 1987.
15. Connolly S, Nishimura D, Macovski A: A selective adiabatic spin-echo pulse. J Magn Reson 83:324-344, 1989.
16. Edelman RR, Siewert B, Darby DG, et al: Qualitative mapping of cerebral blood flow and functional localization with
echo-planar MR imaging and signal targeting with alternating radiofrequency. Radiology 192:513-520, 1994. Medline
Similar articles
page 171
page 172
17. Norris DG, Ludemann H, Leibfritz D: An analysis of the effects of short T2 values on the hyperbolic-secant pulse. J Magn
Reson 92:94-101, 1991.
18. Carr HY, Purcell EM: Effects of diffusion on free precession in nuclear magnetic resonance. Physical Review 94:630-638,
1954.
19. Edelstein WA, Hutchison JM, Johnson G, Redpath T: Spin warp NMR imaging and applications to human whole-body
imaging. Phys Med Biol 25:751-756, 1980. Medline Similar articles
20. Mansfield P, Pykett IL: Biological and medical imaging by NMR. J Magn Reson 29:355-373. 1978.
21. Stehling MK, Turner R, Mansfield P: Echo-planar imaging: magnetic resonance imaging in a fraction of a second. Science
22. Chien D, Atkinson DJ, Edelman RR: Strategies to improve contrast in turboFLASH imaging: reordered phase encoding
and k-space segmentation. J Magn Reson Imaging 1:63-70, 1991. Medline Similar articles
23. Mulkern RV, Melki PS, Jakab P, et al: Phase-encode order and its effect on contrast and artifact in single-shot RARE
sequences. Med Phys 18:1032-1037, 1991. Medline Similar articles
24. Mulkern RV, Wong ST, Winalski C, Jolesz FA: Contrast manipulation and artifact assessment of 2D and 3D RARE
sequences. Magn Reson Imaging 8:557-566, 1990.
1 of 5 12-04-2010 00:01
25. Yang PC, Meyer CH, Terashima M, et al: Spiral magnetic resonance coronary angiography with rapid real-time
localization. J Am Coll Cardiol 41:1134-1141, 2003. Medline Similar articles
26. Spuentrup E, Schroeder J, Mahnken AH, et al: Quantitative assessment of left ventricular function with interactive real-time
spiral and radial MR imaging. Radiology 227:870-876, 2003. Medline Similar articles
27. Lai S, Glover GH: Three-dimensional spiral fMRI technique: a comparison with 2D spiral acquisition. Magn Reson Med
39:68-78, 1998.
28. Sommer G, Corrigan G, Fredrickson J, et al: Renal blood flow: measurement in vivo with rapid spiral MR imaging.
Radiology 208:729-734, 1998. Medline Similar articles
29. Kellman P, Arai AE, McVeigh ER, Aletras AH: Phase-sensitive inversion recovery for detecting myocardial infarction using
gadolinium-delayed hyperenhancement. Magn Reson Med 47:372-383, 2002. Medline Similar articles
30. Walker R, Kessar P, Blanchard R, et al: Turbo STIR magnetic resonance imaging as a whole-body screening tool for
metastases in patients with breast carcinoma: preliminary clinical experience. J Magn Reson Imaging 11:343-350, 2000.
31. Nakatsu M, Hatabu H, Itoh H, et al: Comparison of short inversion time inversion recovery (STIR) and fat-saturated
(chemsat) techniques for background fat intensity suppression in cervical and thoracic MR imaging. J Magn Reson Imaging
32. Bydder GM, Steiner RE, Blumgart LH, et al: MR imaging of the liver using short TI inversion recovery sequences. J Comput
Assist Tomogr 9:1084-1089, 1985. Medline Similar articles
33. Earls JP, Rofsky NM, DeCorato DR, et al: Echo-train STIR MRI of the liver: comparison of breath-hold and non-breath-hold
imaging strategies. J Magn Reson Imaging 9:87-92, 1999. Medline Similar articles
34. Dwyer AJ, Frank JA, Sank VJ, et al: Short-Ti inversion-recovery pulse sequence: analysis and initial experience in cancer
35. Fenchel M, Helber U, Simonetti OP, et al: Multislice first-pass myocardial perfusion imaging: Comparison of saturation
recovery (SR)-TrueFISP-two-dimensional (2D) and SR-TurboFLASH-2D pulse sequences. J Magn Reson Imaging
36. Felmlee JP, Ehman RL: Spatial presaturation: a method for suppressing flow artifacts and improving depiction of vascular
anatomy in MR imaging. Radiology 164:559-564, 1987. Medline Similar articles
37. Edelman RR, Chien D, Kim D: Fast selective black blood MR imaging. Radiology 181:655-660, 1991. Medline
Similar articles
38. Kim WY, Stuber M, Bornert P, et al: Three-dimensional black-blood cardiac magnetic resonance coronary vessel wall
imaging detects positive arterial remodeling in patients with nonsignificant coronary artery disease. Circulation 106:296-299,
39. Simonetti OP, Finn JP, White RD, et al: "Black blood" T2-weighted inversion-recovery MR imaging of the heart. Radiology
40. Wolff SD, Balaban RS: Magnetization transfer contrast (MTC) and tissue water proton relaxation in vivo. Magn Reson Med
41. Edelman RR, Ahn SS, Chien D, et al: Improved time-of-flight MR angiography of the brain with magnetization transfer
contrast. Radiology 184:395-399, 1992. Medline Similar articles
42. Pike GB, Hu BS, Glover GH, Enzmann DR: Magnetization transfer time-of-flight magnetic resonance angiography. Magn
43. Balaban RS, Chesnick S, Hedges K, et al: Magnetization transfer contrast in MR imaging of the heart. Radiology
44. Semple SI, Redpath TW, McKiddie FI, Waiter GD: Comparison of four magnetization preparation schemes to improve
blood-wall contrast in cine short-axis cardiac imaging. Magn Reson Med 39:291-299, 1998. Medline Similar articles
45. Wielopolski PA, van Geuns RJ, de Feyter PJ, Oudkerk M: Breath-hold coronary MR angiography with volume-targeted
46. Haba D, Pasco Papon A, Tanguy JY, et al: Use of half-dose gadolinium-enhanced MRI and magnetization transfer
saturation in brain tumors. Eur Radiol 11:117-122, 2001. Medline Similar articles
47. Frahm J, Haase A, Hanicke W, et al: Chemical shift selective MR imaging using a whole-body magnet. Radiology
48. Rosen BR, Wedeen VJ, Brady TJ: Selective saturation NMR imaging. J Comput Assist Tomogr 8:813-818, 1984.
49. Hore PJ: Solvent suppression in Fourier-Transform nuclear magnetic-resonance. J Magn Reson 55:283-300, 1983.
50. Meyer CH, Pauly JM, Macovski A, Nishimura DG: Simultaneous spatial and spectral selective excitation. Magn Reson Med
51. Flamig DP, Pierce WB, Harms SE, Griffey RH: Magnetization transfer contrast in fat-suppressed steady-state three-
dimensional MR images. Magn Reson Med 26:122-131, 1992. Medline Similar articles
52. Thomasson D, Purdy D, Finn JP: Phase-modulated binomial RF pulses for fast spectrally-selective musculoskeletal
2 of 5 12-04-2010 00:01
imaging. Magn Reson Med 35:563-568, 1996. Medline Similar articles

53. Finn JP: Rapid, fat-or water-suppressed multislice MR pulse sequence. US Patent # 5633586, 1997.
54. Finn JP, Thomasson DM, Moore JR: Minimum cost presaturation by slice segmentation in fast 2-D gradient-echo imaging.
New York: Proceedings of the ISMRM, 1996.
55. Rofsky NM, Lee VS, Laub G, et al: Abdominal MR imaging with a volumetric interpolated breath-hold examination.
56. Hennig J: Echoes-how to generate, recognize, use or avoid them in MR imaging sequences. I. Fundamental and not so
fundamental properties of spin echoes. Concepts Magn Reson 3:125-143, 1991.
57. Hennig J, Scheffler K: Easy improvement of signal-to-noise in RARE-sequences with low refocusing flip angles. Rapid
acquisition with relaxation enhancement. Magn Reson Med 44:983-985, 2000. Medline Similar articles
58. Hennig J, Weigel M, Scheffler K: Multiecho sequences with variable refocusing flip angles: optimization of signal behavior
using smooth transitions between pseudo steady states (TRAPS). Magn Reson Med 49:527-535, 2003. Medline
Similar articles
60. Stejskal EO, Tanner JE: Spin diffusion measurements: spin-echoes in the presence of a time-dependent gradient field. J
Chem Physiol 42:288-292, 1965.
61. Meiboom S, Gill D: Modified spin echo method for measuring nuclear relaxation rates. Rev Sci Instrum 29:688, 1958.
63. Melki PS, Jolesz FA, Mulkern RV: Partial RF echo planar imaging with the FAISE method. I. Experimental and theoretical
assessment of artifact. Magn Reson Med 26:328-341, 1992. Medline Similar articles
64. Melki PS, Jolesz FA, Mulkern RV: Partial RF echo-planar imaging with the FAISE method. II. Contrast equivalence with
spin-echo sequences. Magn Reson Med 26:342-354, 1992. Medline Similar articles
65. Oshio K, Feinberg DA: GRASE (Gradient- and spin-echo) imaging: a novel fast MRI technique. Magn Reson Med
66. Constable RT, Smith RC, Gore JC: Signal-to-noise and contrast in fast spin echo (FSE) and inversion recovery FSE
imaging. J Comput Assist Tomogr 16:41-47, 1992. Medline Similar articles
67. Constable RT, Anderson AW, Zhong J, Gore JC: Factors influencing contrast in fast spin-echo MR imaging. Magn Reson
68. Williams CF, Redpath TW, Smith FW: The influence of stimulated echoes on contrast in fast spin-echo imaging. Magn
69. Sattin W, Mareci TH, Scott KN: Exploiting the stimulated echo in nuclear magnetic-resonance imaging. 1. Method. J Magn
Reson 64:177-182, 1985.
70. Hennig J, Hoddap M: Burst imaging. MAGMA 1:39-48, 1993.
71. Ardelean I, Kimmich R: Diffusion measurements using the nonlinear stimulated echo. J Magn Reson 143:101-105, 2000.
72. Scheenen TW, Vergeldt FJ, Windt CW, et al: Microscopic imaging of slow flow and diffusion: a pulsed field gradient
stimulated echo sequence combined with turbo spin echo imaging. J Magn Reson 151:94-100, 2001. Medline
Similar articles
73. Wedeen VJ, Weisskoff RM, Reese TG, et al: Motionless movies of myocardial strain-rates using stimulated echoes. Magn
74. Heiland S, Dietrich O, Sartor K: Diffusion-weighted imaging of the brain: comparison of stimulated- and spin-echo
echo-planar sequences. Neuroradiology 43:442-447, 2001. Medline Similar articles
75. Aletras AH, Wen H: Mixed echo train acquisition displacement encoding with stimulated echoes: an optimized DENSE
method for in vivo functional imaging of the human heart. Magn Reson Med 46:523-534, 2001. Medline Similar articles
76. Haase A, Frahm J, Matthaei D, et al: Flash imaging-Rapid NMR imaging using low flip-angle pulses. J Magn Reson
67:258-266, 1986.
77. Frahm J, Hanicke W, Merboldt KD: Transverse coherence in rapid flash NMR imaging. J Magn Reson 72:307-314, 1987.
78. Crawley AP, Wood ML, Henkelman RM: Elimination of transverse coherences in FLASH MRI. Magn Reson Med
79. Zur Y, Wood ML, Neuringer LJ: Spoiling of transverse magnetization in steady-state sequences. Magn Reson Med
80. Mugler JP III, Brookeman JR: Three-dimensional magnetization-prepared rapid gradient-echo imaging (3D MP RAGE).
Magn Reson Med 15:152-157, 1990. Medline Similar articles
81. Edelman RR, Wallner B, Singer A, et al: Segmented turboFLASH: method for breath-hold MR imaging of the liver with
flexible contrast. Radiology 177:515-521, 1990. Medline Similar articles
82. Carr JC, Simonetti OP, Bundy JM, et al: Cine MR angiography of the heart with segmented true fast imaging with
3 of 5 12-04-2010 00:01
steady-state precession. Radiology 219:828-834, 2001. Medline Similar articles

page 172
page 173
83. Li D, Paschal CB, Haacke EM, Adler LP: Coronary arteries: three-dimensional MR imaging with fat saturation and
magnetization transfer contrast. Radiology 187:401-406, 1993. Medline Similar articles
84. Wielopolski PA, Manning WJ, Edelman RR: Single breath-hold volumetric imaging of the heart using magnetization-
prepared 3-dimensional segmented echo planar imaging. J Magn Reson Imaging 5:403-409, 1995.
85. Frahm J, Hanicke W, Bruhn H, et al: High-speed STEAM MRI of the human heart. Magn Reson Med 22:133-142, 1991.
86. Oppelt A, Graumann R, Barfuss H: FISP: a new fast MRI sequence. Electromedica (English ED) 3:15-18, 1986.
87. Gyngell ML: The steady-state signals in short-repetition-time sequences. J Magn Reson 81:474-483, 1989.
88. Gyngell ML: The application of steady-state free precession in rapid 2DFT NMR imaging: FAST and CE-FAST sequences.
Magn Reson Imaging 6:415-419, 1988.
89. Bruder H, Fischer H, Graumann R, Deimling M: A new steady-state imaging sequence for simultaneous acquisition of two
MR images with clearly different contrasts. Magn Reson Med 7:35-42, 1988. Medline Similar articles
90. Redpath TW, Jones RA: FADE-a new fast imaging sequence. Magn Reson Med 6:224-234, 1988. Medline
Similar articles
91. Listerud J, Einstein S, Outwater E, Kressel HY: First principles of fast spin echo. Magn Reson Q 8:199-244, 1992.
92. White SJ, Hajnal JV, Young IR, Bydder GM: Use of fluid-attenuated inversion-recovery pulse sequences for imaging the
spinal cord. Magn Reson Med 28:153-162, 1992. Medline Similar articles
93. Feinberg DA, Oshio K: Gradient-echo shifting in fast MRI techniques (GRASE imaging) for correction of field
inhomogeneity errors and chemical-shift. J Magn Reson 97:177-183, 1992.
94. Hennig J, Scheffler K: Hyperechoes. Magn Reson Med 46:1-12, 2001. Medline Similar articles
95. Deshpande VS, Shea SM, Laub G, et al: 3D magnetization-prepared true-FISP: A new technique for imaging coronary
arteries. Magn Reson Med 46:494-502, 2001. Medline Similar articles
96. Markl M, Alley MT, Pelc NJ: Balanced phase-contrast steady-state free precession (PC-SSFP): a novel technique for
velocity encoding by gradient inversion. Magn Reson Med 49:945-952, 2003. Medline Similar articles
97. Scheffler K, Hennig J: T(1) quantification with inversion recovery TrueFISP. Magn Reson Med 45:720-723, 2001.
98. Schmitt P, Griswold MA, Jakob PM, et al: Inversion recovery TrueFISP: quantification of T(1), T(2), and spin density. Magn
99. Fautz HP, Leupold J, Hennig J, Scheffler K: Signal behavior in continuously ramped 2D TrueFISP for whole-body imaging.
Magn Reson Med 48:1085-1090, 2002.
100. Deimling M, Heid O: True FISP Imaging with Inherent Fat Cancellation. Denver: Proceedings of the ISMRM, 1999, p
1500.
101. Nishimura DG, Vasanawala S: Analysis and Reduction of the Transient Response in SSFP Imaging. Denver:
Proceedings of the ISMRM, 2000, p 301.
102. Deshpande VS, Chung YC, Zhang Q, et al: Reduction of transient signal oscillations in true-FISP using a linear flip angle
magnetization preparation. Magn Reson Med 49:151-157, 2003. Medline Similar articles
103. Cohen MS, Weisskoff RM: Ultra-fast imaging. Magn Reson Imaging 9:1-37, 1991. Medline Similar articles
104. Bruder H, Fischer H, Reinfelder HE, Schmitt F: Image reconstruction for echo planar imaging with nonequidistant k-space
sampling. Magn Reson Med 23:311-323, 1992. Medline Similar articles
106. Mills CM, Brant-Zawadzki M, Crooks LE, et al: Nuclear magnetic resonance: principles of blood flow imaging. Am J
Roentgenol 142:165-170, 1984.
107. Laub GA, Kaiser WA: MR angiography with gradient motion refocusing. J Comput Assist Tomogr 12:377-382, 1988.
108. Dumoulin CL, Hart HR Jr: Magnetic resonance angiography. Radiology 161:717-720, 1986. Medline
Similar articles
109. Moran PR: A flow velocity zeugmatographic interlace for NMR imaging in humans. Magn Reson Imaging 1:197-203,
110. Edelman RR, Mattle HP, Kleefield J, Silver MS: Quantification of blood flow with dynamic MR imaging and presaturation
bolus tracking. Radiology 171:551-556, 1989. Medline Similar articles
111. Nishimura DG, Jackson JI, Pauly JM: On the nature and reduction of the displacement artifact in flow images. Magn
4 of 5 12-04-2010 00:01

Roentgenol 148:1251-1258, 1987.
114. Hinks RS, Constable RT: Gradient moment nulling in fast spin echo. Magn Reson Med 32:698-706, 1994. Medline
Similar articles
115. Butts K, Riederer SJ: Analysis of flow effects in echo-planar imaging. J Magn Reson Imaging 2:285-293, 1992.
116. Duerk JL, Simonetti OP: Theoretical aspects of motion sensitivity and compensation in echo-planar imaging. J Magn
© 2010 Elsevier
5 of 5 12-04-2010 00:01
IOCHEMICAL ASIS OF THE PPEARANCE OF

EREBRAL EMORRHAGE
Keith R. Thulborn
INTRODUCTION
The highly variable appearance of cerebral hemorrhage by magnetic resonance imaging (MRI)
provides a wealth of information about the underlying biochemical processes. This information is often
neglected in the MRI evaluation of hemorrhagic pathologic processes. A complete analysis using the
full information content of MR images requires an understanding of the biochemical processes
influencing the relaxation mechanisms controlling signal intensity.
Extrapolating from in vitro studies and clinical observations, researchers have proposed hypotheses to
explain the evolution of the features of cerebral hematomas on MR images.1-6 The explanations have
emphasized the role of iron from hemoglobin in determining relaxation mechanisms of these variable
patterns. This is based on the high concentration and changing magnetic properties of iron as the
biochemical form, oxidation state, and spatial distribution change with time. Other pathophysiologic
processes such as changes in integrity of the blood-brain barrier with alteration of the degree of
edema and protein concentration have explained other features of these images. In this chapter, the
physiochemical principles of the magnetic properties of biologic systems are reviewed, the biochemical
pathways of iron metabolism in resolving hematoma are discussed, and a non-mathematic scheme is
presented for relating these biochemical aspects to the relaxation phenomena that produce the
variable signal contrast observed in MR images of hemorrhage. Clinical aspects involved in using MRI
to evaluate hemorrhage are considered in Chapter 45.
© 2010 Elsevier
1 of 1 12-04-2010 00:02
BIOCHEMICAL EVOLUTION OF IRON IN CEREBRAL HEMATOMA

Hemorrhage is a complex process, biochemically and pathologically, evolving over many months. The
pattern of evolution of hemorrhage described by computed tomography is relatively simple, reflecting
the protein content of the tissues at the site of the lesion,7,8 as compared with the richness of
3-6
information content reflected in the variability reported for MR images of the same process. The rate
of the many biochemical changes depends on the location and size of the lesion and on the physiologic
status of the patient.9 Better vascularized areas would be expected to show more rapid repair.
Because of the integrity of the blood-brain barrier, repair and clearance processes in the parenchyma
differ from those occurring in subarachnoid and subdural hemorrhages. Such variability in hematoma-
induced repair response has been reported also from biochemical studies of hemorrhage at different
10
sites outside the central nervous system.
Iron Metabolism in Hemorrhage

page 174
page 175
Add to lightbox
Figure 6-1 Schematic depiction of iron metabolism in a hematoma showing simplified iron salvage
pathways in the red blood cell, phagocyte, and extracellular space. BBB, blood-brain barrier; deoxyHb,
deoxyhemoglobin; metHb, methemoglobin; oxyHb, oxyhemoglobin; PO2, partial pressure of oxygen.
Iron is a transition metal with an atomic number of 26 and an electronic configuration described as
The distribution of electrons in the outer 3d electronic orbital is dependent on the number of electrons.
2+
As the oxidation state increases, outer electrons are lost, leaving the ferrous ion (Fe ) with six 3d
3+
electrons and the ferric ion (Fe ) with five 3d electrons. Unpaired electrons within the outer 3d orbitals
of iron are of critical importance in determining the magnetic properties of this atom.
As the most abundant transition metal in the human body, iron is vital for oxygen transport by
hemoglobin in the erythrocyte of blood, oxygen storage by myoglobin in the tissues, and multiple
catalytic functions in enzyme systems throughout the body. In contrast to its functional chelated form,
free iron is toxic, as is evident from toxic ingestions and iron overload states.11-13 Toxicity is believed to
1 of 5 12-04-2010 00:04
be due to enhanced catalysis of free radical production by unchelated iron. 14 Hemorrhage can be
thought of as having a tightly controlled iron salvage pathway, in which iron from the extravasated
erythrocytes is mobilized from hemoglobin, detoxified by chelation to short-term iron transport proteins
for transfer back to the reticuloendothelial system, or converted to long-term storage proteins for local
12,15,16
deposition. The different forms of iron in this still incompletely understood pathway have different
magnetic properties that can influence the magnetic relaxation properties of the proton spin system
used to form the MR image.
The iron salvage pathway of arterial hemorrhage is now examined stepwise in terms of the magnetic
properties of the iron (Fig. 6-1). No time scale is given because this can only be misleading, given the
multiple factors that determine rate of repair as described previously.
Add to lightbox
Figure 6-2 Energy diagram of the 3d electronic orbitals of the ferrous form of iron with different ligands.
In the absence of ligands, all 3d orbitals have the same energy. In deoxyhemoglobin, the five ligands
produce little energy separation between the eg and t2g groups of 3d orbitals, allowing the six
electrons to distribute such that four electrons are unpaired (i.e., paramagnetic). In oxyhemoglobin, the
six ligands produce marked energy difference E between the eg and t2g groups of 3d orbitals, causing
electron pairing in the lowest energy state of the t2g orbitals (i.e., diamagnetic).
Arterial Blood
In arterial hemorrhage, the freshly extravasated erythrocytes contain fully oxygenated hemoglobin.
Hemoglobin is a tetramer of polypeptide chains, with each polypeptide chain having considerable
17
ordered secondary and tertiary structure. Each chain has a prosthetic heme group bound within a
hydrophobic cleft. The heme group is protoporphyrin IX with a centrally chelated ferrous ion. Initially, in
arterial blood, iron is bound in octahedral geometry with six ligands. The tetrapyrrole nitrogens of the
protoporphyrin constitute four ligands in a plane around the iron. The imidazole group of a histidine
from the polypeptide chain is the axial ligand below the protoporphyrin plane, whereas molecular
oxygen is the sixth exchangeable ligand above the plane. The interaction of the six ligands with the
metal center in oxyhemoglobin causes the six outer electrons of the ferrous ion to pair in the t2g group
of 3d orbitals of the lowest energy (Fig. 6-2). Because all the electrons of the iron and other atoms
that constitute hemoglobin are paired, oxygenated blood is diamagnetic (χ < 0).
Deoxygenation
page 175
page 176
2 of 5 12-04-2010 00:04
Add to lightbox
Figure 6-3 Schematic representation of the exclusion of water from the paramagnetic iron of
deoxyhemoglobin by the surrounding globin protein until the protein undergoes conformational
changes that lead to access of water to the heme cleft and oxidation of the ferrous iron to ferric iron in
methemoglobin. Water access explains relaxivity effects of methemoglobin and the absence of such
effects for deoxyhemoglobin, although both are paramagnetic.
In tissues undergoing aerobic respiration, the partial pressure of dissolved oxygen is lower than that
required to fully saturate the oxygen-binding sites of hemoglobin. The binding equilibrium is reversed,
and molecular oxygen is delivered to the tissues. A cerebral hematoma has a mass effect compressing
surrounding tissue, thereby reducing perfusion to these regions. A gradient in the partial pressure of
oxygen from the hematoma to the compromised surrounding tissue results in hemoglobin desaturation.
In addition, there is reduced washout of byproducts, such as carbon dioxide from remaining aerobic
respiration and lactate from anaerobic glycolysis, which results in a decline in pH as the buffering
capacity of the tissue is exceeded. This leads to further deoxygenation of the hemoglobin by
displacement of the hemoglobin oxygen-binding equilibrium toward dissociation (the Bohr effect). The
removal of molecular oxygen changes the coordinate geometry of the heme ferrous ion to a five-ligand
system of deoxyhemoglobin, which decreases the energy separation between the eg and t2g groups of
electronic orbitals. The six 3d electrons redistribute among the five 3d orbitals, leaving four unpaired
electrons of parallel spin (see Fig. 6-2). These unpaired electrons confer deoxyhemoglobin with its
paramagnetic properties (χ > 0). Because the paramagnetic ferrous ion is bound to the heme group in
the hydrophobic cleft of the globin protein, water molecules are unable to approach the iron center.
This restricts relaxivity effects (Fig. 6-3). As the deoxyhemoglobin is packaged in red blood cells, the
magnetic susceptibility of the interior of the cell is different from that of the suspending fluid, resulting in
susceptibility variations within the hematoma (Fig. 6-4). The importance of the integrity of the red blood
18,19
cell membrane has been demonstrated in vitro.
3 of 5 12-04-2010 00:04
Add to lightbox
Figure 6-4 Simplified schematic representation of the different effective magnetic fields Heff produced
by tissues of diamagnetic (χ1 < 0) and paramagnetic (χ2 > 0) magnetic susceptibilities in the applied
field H0. The signal frequency ω is shifted by Heff according to the Larmor equation (ω = γHeff). The
signal intensity of the voxels (4 and 9 of voxels 1 through 12) at the boundaries is reduced owing to
diffusion in the field gradients on spin-echo images and, in addition, due to intravoxel dephasing on
gradient-echo images.
The time course and distribution of deoxygenation can be used to explain some features of hyperacute
hemorrhage20 in which a peripheral rim of reduced signal intensity is observed around parenchymal
hematomas of less than 24 hours' duration on T2-weighted images. Histologic examination shows that
the hematoma-tissue interface is not smooth but rather shows periodic interdigitation of strands of
blood clot containing intact red blood cells with surrounding brain tissue. Due to mass effect of the clot,
this tissue can have compromised perfusion resulting in decreasing pH due to accumulation of
metabolic byproducts. This periodic geometry and rapid deoxygenation of the periphery of the blood
clot due to the Bohr effect from acidification results in magnetic susceptibility effects with concomitant
MR signal loss.
Red Blood Cell Lysis

page 176
page 177
The tissue damage elicits an inflammatory repair response within the surrounding tissue, with
phagocytes, such as macrophages, infiltrating the boundaries of the hematoma to clear extravasated
materials and damaged tissues. Glial cells also show phagocytic activity. Red blood cells may be
phagocytosed entirely or partially, or lysed by enzymes released into the region by the inflammatory
cells.8,9 Loss of red blood cell membrane integrity releases hemoglobin. In the absence of the
functional reductase enzymes of the red blood cell (NADH-cytochrome b5 reductase, NADPH-flavin
reductase),17 hemoglobin is rapidly converted to methemoglobin, in which the iron, still bound to the
heme moiety within the globin protein, is oxidized to the ferric state with five 3d electrons. Once in the
ferric oxidation state, the iron is paramagnetic (χ > 0). The protein undergoes a number of changes,
ultimately irreversible, in secondary, tertiary, and quaternary structures in which the ferric iron is no
21
longer protected from the surrounding solvent. The electronic configuration of the iron changes from
initially five unpaired electrons, one in each of the five 3d suborbitals, to one unpaired electron as the
weak sixth ligand of water is exchanged for a hydroxide and then another imidazole nitrogen of a
histidyl residue of the protein. These changes define the hemichromes as described by electron
paramagnetic resonance spectroscopy.22 The time course of these processes in vivo remains
unknown.
4 of 5 12-04-2010 00:04
Extracellular Iron-Binding Proteins

Extracellular protein is further degraded with release of the iron to localized extracellular binding
proteins such as lactoferrin and transferrin. Some extracellular ferritin is also present. These binding
proteins detoxify free iron for recycling to the reticuloendothelial system through the circulation and for
23-25
local storage by glial cells and macrophages. The chelated iron remains paramagnetic.
Intracellular Iron Processing

Red blood cells and hemoglobin, phagocytosed by macrophages and glial elements of the central
nervous system, are digested by the lysosomal system, with the iron being stored as ferric
oxyhydroxide in the hydrophobic center of the major iron storage protein called ferritin.15,26 Ferritin is a
water-soluble protein of approximately 450 kd with 24 polypeptide subunits surrounding a core of as
many as 4500 ferric ions. If the quantity of available iron exceeds the capacity of the cell to synthesize
apoferritin, excess iron is stored as hemosiderin.16 Hemosiderin is an insoluble larger aggregation of
ferric oxyhydroxide with less protein than ferritin and, as yet, a poorly characterized biochemical
structure. These storage forms with large aggregates of iron behave antiferromagnetically and
ferromagnetically, sometimes with superparamagnetic properties (χ > 0).27-29 These aggregates of
iron have reduced accessibility to surrounding water, thereby minimizing relaxivity effects. However,
magnetic susceptibility variations can be expected in tissues containing such materials. The iron
storage processes occur throughout resolution of hemorrhage but become significant later, presumably
reflecting concentration changes and the cessation of other relaxation processes. Much of the ferritin is
intracellular within both macrophages and astrocytes, whereas the hemosiderin is in macrophages.30
© 2010 Elsevier
5 of 5 12-04-2010 00:04
INTEGRITY OF THE BLOOD-BRAIN BARRIER

Edema
The loss of the blood-brain barrier around the site of hemorrhage causes vasogenic edema. The
damage to the tissue in and around the site from mass effect, reduced perfusion, and inflammation
worsens the edema. Although such changes do not affect the magnetic properties of the tissues that
remain diamagnetic (χ < 0), other magnetic relaxation phenomena occur to alter the MR image.
Coagulation
The extravasated blood initiates the coagulation cascade, leading to clot formation that limits further
bleeding. The protein network of the clot with trapped red blood cells is expected to undergo a number
of changes, including clot contraction with changes in the concentration and distribution of blood
products, which can change magnetic properties of the tissue and thus the MRI characteristics. This
has not been systematically studied in vivo.31,32
© 2010 Elsevier
1 of 1 12-04-2010 00:04
RELAXATION MECHANISMS
For totally diamagnetic tissues, the most important relaxation mechanism for both longitudinal and
transverse relaxation is attributed to dipole-dipole interactions. 33,34 Other mechanisms of scalar-spin
coupling, chemical shift anisotropy, and quadripolar and spin-rotational effects are usually less
33-38
important in proton MRI and are described elsewhere. Paramagnetic substances have several
effects of much greater magnitude than diamagnetic substances, as discussed in the Appendix at the
end of this chapter. These effects include: 1. relaxivity effects due to dipole-dipole interactions, which
produce T1 and T2 relaxation, generally with T1 effects dominating to produce increased signal
intensity; and 2. susceptibility effects, which produce only T2 relaxation and signal loss on MR images.
The dipole-dipole effects of paramagnetic substances are discussed in the Appendix, but consideration
must be given to other exchange processes not involving paramagnetic species.
Changes in the protein content within the hematoma are also seen as clot formation, clot contraction,
38
and necrosis occur. From in vitro studies, increasing protein concentration would be expected to
promote T1 and T2 relaxation rates, although rigorous in vivo studies have not been reported. It is
possible that the exchange of water between bulk and protein-bound phases may be a significant
relaxation process in some situations. Increasing edema has been suggested to allow increased
diffusion by removing diffusional barriers such as macromolecules and cell membranes, thereby
promoting T2 relaxation. In areas of necrosis, diffusional barriers may not be removed to the same
degree as in vasogenic edema in intact, albeit damaged, tissue. The relative influences of diffusion and
protein exchange on in vivo relaxation processes cannot be predicted as yet.
© 2010 Elsevier
1 of 1 12-04-2010 00:05
SENSITIVITIES OF MRI PULSE SEQUENCES

The basic principles of imaging pulse sequences have been presented in numerous excellent texts, 39-43
and only features pertaining to hemorrhage are elaborated on here.
page 177
page 178
Add to lightbox
Figure 6-5 Simplified schematic of timing diagrams of the spin-echo (SE) (top), asymmetric-echo
(ASE) (middle), and gradient-echo (GRE) (bottom) pulse sequences with slice selection by gradient
Gz shown only on the 90° RF pulse, frequency encoding using Gx, and phase-encoding using Gy. The
echoes are labeled. The pre-encode Gx gradient is positive for the SE sequence as it is before the
180° RF pulse but negative for the GRE sequence because there is no 180° RF pulse.
Spin-Echo Pulse Sequence

The spin-echo (SE) sequence, shown in Figure 6-5, uses a 180° radiofrequency (RF) pulse centered
between the initial 90° RF pulse and the center of the acquisition time period to refocus nuclear spins
of variable Larmor frequencies into an echo, producing the MR signal. This minimizes the effect of
static H0 inhomogeneities on transverse relaxation that would otherwise reduce the signal intensity. By
selecting appropriate timing parameters for the pulse sequence, the MR image can be made
selectively sensitive to relaxivity and magnetic susceptibility effects. If significant diffusion through field
inhomogeneities occurs during the echo time (TE), signal loss occurs, as discussed in the Appendix. As
TE is made longer relative to the diffusional correlation time, the pulse sequence becomes more
sensitive to these inhomogeneities. Because susceptibility differences are a source of field
nonuniformity, increasing TE increases sensitivity of SE images to processes such as hemorrhage that
generate such susceptibility variation. The effect is readily recognized as signal loss on T2-weighted
images, which is not present on T1-weighted images.
Faster spin-echo imaging has been achieved by expanding the single refocusing 180° RF pulse with a
train of refocusing RF pulses in which spatial phase-encoding is performed between each refocusing
pulse. Although these pulses can be 180° at lower magnetic fields, high power deposition arises from
long trains of such pulses at higher magnetic fields, as described elsewhere (Chapter 18). Typically,
1 of 3 12-04-2010 00:05
RF pulses of lower flip angles such as 120° are used on 3 tesla scanners to maintain the specific
absorption rate (SAR) within the Food and Drug Administration guidelines. The echo times between the
refocusing RF pulses are short so that little time is allowed for dephasing. Although this is an
advantage for maintaining signal-to-noise performance in the images, sensitivity to magnetic
susceptibility effects from different phases of the evolution of hemorrhage is lost. If these effects are
to be observed, it is advisable always to use GRE sequences in addition to the faster SE sequences
based on long RF pulse trains.
An alternative sequence that is very sensitive to magnetic susceptibility effects is echo-planar imaging
(EPI) in which a single RF pulse is used in either a gradient-recalled echo or a spin-echo format. The
echo-planar image is generated using a rectilinear trajectory through the entire k-space by rapid
switching of phase and frequency encoding gradients multiple times, as discussed elsewhere (Chapter
7). Because all of k-space is covered in a single RF pulse or pulse pair, phase errors induced by the
magnetic susceptibility effects of the various stages of the evolution of hemorrhage accumulate during
the acquisition. The phase cancellation results in reduced signal. As both diffusion-weighted and
non-diffusion-weighted spin-echo EPI can be rapidly acquired, the non-diffusion-weighted images can
be used for detection of hemorrhage. Similarly, dynamic susceptibility contrast perfusion imaging is
usually performed with gradient-echo EPI. The images prior to the arrival of the contrast in the tissue
are also sensitive to hemorrhage which produces signal loss.
Two other imaging sequences can be used to enhance detection of the susceptibility effects on T2
relaxation. These are the gradient- or field-echo (GRE) and the asymmetric spin-echo (ASE)
sequences that have been described in detail elsewhere44,45 (see Fig. 6-5).
Asymmetric Spin-Echo Pulse Sequence

The ASE sequence offsets the 180° refocusing pulse by a time interval that can be varied to alter the
amount of signal refocusing. This sequence is sensitive not only to the effects of diffusion through
magnetic field gradients as used by the SE sequence but also to variations of Larmor frequencies
within a single voxel due to nonuniform magnetic susceptibility (Fig. 6-6). This results in rapid loss of
phase coherence among the nuclear spins within the voxel and hence rapid signal loss. The amount of
signal loss, and thus the sensitivity to intravoxel susceptibility heterogeneity, can be altered by varying
the size of the offset. Comparison with ASE and SE images of the same TE allows calculation of T2*
and has been suggested as a means of quantifying the iron content of tissue.45
page 178
page 179
2 of 3 12-04-2010 00:05
Add to lightbox
Figure 6-6 Schematic representation of the effects of nondiamagnetic substances on the intravoxel
distribution of susceptibility χ, Heff, and resonance frequency ω. Variation of resonance frequencies
within the voxel produces signal loss on GRE images.
This sequence has been adapted to an echo-planar sequence for detection of magnetic susceptibility
effects arising from tissue oxygenation changes during neuronal activation in functional MRI.46
Gradient-Echo Pulse Sequence

The GRE sequence does not use a 180° refocusing pulse and is thus sensitive to both static magnetic
field inhomogeneities (magnet imperfections and tissue susceptibility heterogeneity) and the effects of
diffusion.44 Magnet imperfections over the volume of the imaging voxel have become less important
with improved magnet technology, allowing detection of tissue susceptibility variations by means of the
diffusional effects. Even without diffusion, signal cancellation occurs owing to the range of Larmor
frequencies caused by susceptibility variations within a voxel, as is also the case for the ASE
sequence. This makes the GRE sequence useful for enhancing detection of susceptibility effects that
may be less clearly identified on SE images. The artifacts from unwanted susceptibility effects, such
47
as from the paranasal sinuses on intracranial lesions close to the base of the skull, make the GRE
sequence less useful for initial clinical screening examinations.
© 2010 Elsevier
3 of 3 12-04-2010 00:05
FIELD STRENGTH DEPENDENCE

The development of clinical 3.0 T scanners since 1993 has increased sensitivity to magnetic
susceptibility effects. Just as the susceptibility effects of blood oxygenation level dependent (BOLD)
contrast used for functional MRI of the brain are improved at 3.0 T, the detection of hemorrhage and
its blood products are also improved. Although computed tomography has been used routinely for
detection of blood products in the brain and subarachnoid space, clinical experience indicates that MRI
at 3.0 T is more sensitive than CT. Whereas CT attenuation of blood is dependent on increased protein
content that is rapidly dispersed, MRI is sensitive to the form of the iron products within the tissues
which can persist for many years.20 The detection of blood in the subarachnoid space by MRI using
FLAIR sequences is not based on iron products of blood but rather on changes in the T1 relaxation
properties of CSF by the constituents of blood.
page 179
page 180
Table 6-1. The Influence of Iron Metabolism on the MRI Appearance of

Hemorrhage*
Relaxation MR Signal
Mechanism Intensity†
Biochemical Magnetic
Stage Form Location Property R χ T1 T2
Oxyhemoglobin II RBC Diamag - - Dark Bright
Fe oxyHb
Deoxygenation II RBC Paramag - + Dark Dark
Fe deoxyHb
RBC lysis + FeIII metHb, Extracellular Paramag + - Bright Bright
oxidation hemichromes
Extracellular iron FeIII transferrin, Extracellular Paramag + - Bright Bright
processing lactoferrin
Intracellular iron FeIII ferritin, Phagocytes Superpar - + Iso Dark
storage hemosiderin
*deoxyHb, deoxyhemoglobin; Diamag, diamagnetic; iso, isointense; oxyHb, oxyhemoglobin;

Paramag, paramagnetic; metHb, methemoglobin; RBC, red blood cell; R, relaxivity; χ, susceptibility;
Superpar, superparamagnetic;T1,T1 weighted;T2,T2 weighted.
†
Signal intensities are estimated relative to cerebral cortex.
The variation of 1/T1 and 1/T2 with magnetic field strength is termed nuclear magnetic relaxation
dispersion. Because the mathematic treatment of nuclear magnetic relaxation dispersion is beyond the
27
scope of this chapter, the interested reader is referred elsewhere for a more formal introduction.
Only the observations relevant to hemorrhage are discussed here. At the limit of zero field, 1/T1 =
1/T2. As field strength changes, the efficiency of relaxation is dependent on matching correlation times
of local fluctuating magnetic fields generated by diffusional processes to the Larmor frequency. This
occurs over a wide range (10-10 to 10-11 second) corresponding to low imaging field strengths, and the
effects on T1 and T2 relaxation rates are comparable. In contrast, at higher field strengths, 1/T1 tends
toward zero and 1/T2 tends to a nonzero value termed the secular contribution to T2 relaxation.
Therefore, higher magnetic field strengths emphasize susceptibility effects. In vitro studies of
deoxygenated red blood cells indicate a quadratic dependence of 1/T2 on magnetic field strength over
a range of 2 to 5 T.18 The susceptibility effects of ferritin as a function of field strength in vivo have
48
been measured, and less than a quadratic dependence was found. The distribution of the iron
aggregates in vivo is not known, and modeling is difficult. Although the theoretic treatment is
incomplete for susceptibility effects induced by ferritin and hemosiderin (speculated to be due to
antiferromagnetic and ferromagnetic properties of these iron aggregates), it is clear that higher
1 of 2 12-04-2010 00:06
imaging magnetic field strengths increase sensitivity of MR images to susceptibility-induced relaxation

mechanisms, regardless of the source of the susceptibility variation.27,28
© 2010 Elsevier
2 of 2 12-04-2010 00:06
EVALUATION OF HEMORRHAGE WITH MRI

Role of Iron Products
The magnetic properties of the iron products of resolving hematoma have been discussed earlier.
Although deoxygenated hemoglobin and all the products containing iron in the ferric oxidation state are
paramagnetic (methemoglobin, hemichromes, transferrin, lactoferrin, low-molecular-weight iron
chelates) or antiferromagnetic (ferritin, hemosiderin), the production of relaxivity effects is dependent
on the close approach of water protons to the iron and the production of the susceptibility effects is
dependent on the distribution of the iron. The relaxation properties and the imaging characteristics are
discussed and summarized in Table 6-1.
Oxyhemoglobin
The MR image of the center of an acute hematoma is essentially a collection of protein-rich
diamagnetic fluid. Relaxivity and susceptibility effects are not observed for the diamagnetic iron of
oxygenated hemoglobin. Image intensity is determined by other dipole-dipole mechanisms as operating
in normal surrounding tissue. This means a variably long T1 (isointense to dark on T1-weighted
images) and relatively long T2 (bright on T2-weighted images). The change in water content and
distribution is clearly evident as areas of long T1 (dark on T1-weighted SE images) and long T2 (bright
on T2-weighted SE images) in the areas bordering the hematoma.49 Changes in the protein content
within the hematoma also occur as clot formation, clot contraction, and necrosis occur. From in vitro
50
studies, increasing protein concentration would be expected to promote T1 and T2 relaxation rates,
although rigorous in vivo studies have not been reported. Exchange of water between bulk phase and
protein-bound phases may provide an explanation for small modulations in image intensity.
Deoxygenated Hemoglobin
The paramagnetic iron of deoxygenated hemoglobin is held within a hydrophobic cleft. The exclusion of
water from close approach to the paramagnetic center prevents relaxivity effects. T1 relaxation is not
affected so that the hematoma remains dark on T1-weighted images. The packaging of the
paramagnetic centers within the red blood cells produces susceptibility variations that produce
transverse relaxation. This explains the loss of signal on T2-weighted SE and GRE images (dark on
T2-weighted images). The deoxygenation extends inward over time from the periphery of the
hematoma where there is close interdigitation of clot and tissue.
Red Blood Cell Lysis

page 180
page 181
Loss of integrity of the red blood cells homogenizes the distribution of paramagnetic iron to minimize
the susceptibility variations and reduces transverse relaxation. As the enzyme systems employed
within the red blood cell to maintain the ferrous oxidation state of the iron become nonfunctional,
methemoglobin and other hemichromes are formed. These proteins allow water access to the
paramagnetic iron to induce the relaxivity effects that shorten T1 and to a lesser degree T2. Thus, a
hematoma at this stage shows increasing brightness on T1-weighted SE images. Although the
concomitant shortening of T2 from relaxivity effects may suggest further loss of brightness on
T2-weighted SE and GRE images, the loss of red blood cell integrity removes the paramagnetic
aggregation responsible for susceptibility-induced relaxation effects. Because this is the dominant T2
relaxation process, loss of this mechanism means that the effective T2 is still longer than with intact
red blood cells. Hence T2-weighted SE and GRE images appear brighter after cell lysis has occurred.
This effect extends from the periphery inward to the center of the hematoma.
Extracellular Iron-Binding Proteins

The specific relaxivity and susceptibility effects of ferric ions chelated to these proteins described in
vitro23-25 remain unknown for resolving in vivo hemorrhage. The concentration of these substances may
be too low to have a significant role in determining signal intensity within an MR image.
1 of 2 12-04-2010 00:06
Intracellular Iron Storage

The structure of iron storage proteins such as ferritin excludes water from close approach to most of
the paramagnetic ferric ions, minimizing relaxivity effects. The antiferromagnetic properties of these
crystalline aggregates of ferric oxyhydroxide induce susceptibility variations from surrounding tissue.
The implications for imaging of old resolving hematomas are that T1-weighted MR images show
isointensity with surrounding tissue owing to the absence of relaxivity effects and that T2-weighted and
GRE images show signal loss owing to susceptibility effects. If the T2 is significantly shortened below
the TE used for T1-weighted images, then the susceptibility effect is observed on these images. This is
not a relaxivity effect. Rather, the term T1 weighted is inappropriate under these conditions. This
assumes that cavitation has not occurred. If a cavity develops in the region of tissue loss, then the
signal characteristics of cerebrospinal fluid that fills the cavity dominate the image.
Images of Evolving Cerebral Hematoma

The best controlled longitudinal study of resolving intraparenchymal hematoma was reported for an
experimental model of hemorrhage in the monkey in which venous blood was injected into the right
cerebral hemisphere and followed by MRI for several months.49 Selected images from this study are
reproduced with permission for discussion.
At 2 hours after injection of blood (Fig. 6-7A and B), the acute hematoma has low signal intensity on
the T1-weighted and high signal intensity on the T2-weighted images relative to normal cortex. This is
consistent with the absence of significant relaxivity and susceptibility relaxation mechanisms in the
acute setting. Although the injection used venous blood, marked susceptibility effects are not observed,
suggesting that greater deoxygenation is required. The T1-weighted image shows greater signal loss
in the periphery consistent with edema in surrounding tissue having slightly different relaxation
phenomena from the center of the hematoma. During the next 2 days (Fig. 6-7C and D), the signal
intensity of the center of the hematoma on the T1-weighted image increases, presumably from the
relaxivity mechanism of methemoglobin, hemichromes, and other paramagnetic centers, allowing close
approach of water protons. The signal intensity from the surrounding edematous zone shows little
change. In contrast, the same area of hematoma on the T2-weighted image displays decreased signal
intensity, presumably from susceptibility-induced relaxation mechanisms as the intact red blood cells
become increasingly deoxygenated and from any intracellular methemoglobin that may form as the
energy status of the cells declines. It is not clear whether the methemoglobin and hemichromes form
intracellularly or that the hematoma is a mixture of intact deoxygenated cells suspended within a
solution of hemoglobin degradation products. Over 6 to 10 days (Fig. 6-7E to H), the T1-weighted
images show increasing signal intensity due to relaxivity effects from increasing concentrations of
hemoglobin degradation products. The T2-weighted images show increasing signal intensity when
susceptibility effects diminish as red blood cell integrity is lost within the hematoma. The edematous
periphery of the lesion shows minimal changes during this time interval. After 2 months (Fig. 6-7I and
J), the T1-weighted image shows little evidence of the lesion, whereas an area of decreased signal
intensity remains on the T2-weighted image. This can be attributed to the susceptibility-induced
relaxation mechanism of the iron storage products.
© 2010 Elsevier
2 of 2 12-04-2010 00:06
CONCLUSION
The basis of the highly variable MRI appearance of resolving intraparenchymal cerebral hematoma
reported clinically can be rationalized largely in terms of a model encompassing current concepts of
iron metabolism and integrity of the blood-brain barrier. The time scale is dependent on the size of the
lesion, its location with respect to both vascular supply and white and gray matter, and the physiologic
status of the patient. The model should be regarded as a working hypothesis. The delineation of the in
vivo biochemistry of iron metabolism and water balance remains to be performed. It is unlikely that
descriptive analysis of the MRI appearance of cerebral hemorrhage can be verified with biochemical
studies in patients, making animal models a necessary vehicle for further detailed studies.
page 181
page 182
Add to lightbox
Add to lightbox
1 of 4 12-04-2010 00:07
Add to lightbox
Add to lightbox
Add to lightbox
2 of 4 12-04-2010 00:07
Add to lightbox
page 182
page 183
Add to lightbox
Add to lightbox
3 of 4 12-04-2010 00:07
Add to lightbox
Add to lightbox
Figure 6-7 Longitudinal study, using inversion recovery (T1-weighted) SE (A, C, E, G, and I) and
T2-weighted SE coronal images (B, D, F, H, and J) of a monkey after injection of 3 mL of venous
blood into the right cerebral hemisphere. Images were selected at 2 hours (A and B), 2 days (C and
D), 6 days (E and F). 10 days (G and H), and 2 months (I) and (J) from a more complete study. (A to J
from Di Chiro G, Brooks RA, Girton ME, et al: Sequential MR studies of intracerebral hematomas in
monkeys. AJNR 7:193-199, 1986 © by American Society of Neuroradiology 1986.)
© 2010 Elsevier
4 of 4 12-04-2010 00:07
APPENDIX Magnetic Properties of Biologic Tissues

Origin of Magnetic Properties
A magnetic field is generated by a moving electric charge, that is, an electric charge with momentum.51
The strength of the magnetic field is determined by the magnitude of both the charge and momentum.
Such a charge with a magnetic field is termed a magnetic dipole. Electrons moving in orbitals about a
nucleus represent moving charges with both orbital angular momentum and spin angular momentum
and generate a magnetic field. The nucleus also has momentum and charge. However, because the
magnetic moment of the charged particle is inversely proportional to its mass, and the mass of the
nucleus is three orders of magnitude greater than that of the electron, the contribution of the nucleus to
the magnetic properties of the atom is much less than that of the electrons. Hence, although nuclear
magnetic interactions occur and nuclear magnetization is the source of the signal in the MR image, the
magnetic properties of tissue are determined predominantly by the electronic configuration of the
atoms and molecules. The dominant effects encountered on MR images are discussed below.
Diamagnetism
Most biologic materials consist of low-atomic-weight elements such as carbon and hydrogen in which
the electrons are paired in atomic and molecular suborbitals. When the electrons are paired, the spin
angular momentum is canceled and no magnetic dipole is observed. However, the paired electrons still
have orbital angular momentum that produces a magnetic field (termed a Lenz field) opposing the
applied magnetic field. The resultant field within such a material is less than that of the original applied
magnetic field. Such materials are termed diamagnetic.
page 183
page 184
Paramagnetism
Some biologic substances have atomic or molecular structures in which some of the electrons are
unpaired. Transition metal ions, such as iron with the ferrous and ferric oxidation states, are important
examples in which the number of unpaired electrons varies with the biochemical state of the metal ion.
An unpaired electron has a spin angular momentum and therefore a magnetic moment that is not
canceled as in the paired state. At physiologic temperatures, more electrons align parallel to the
applied field, resulting in an enhancement of that applied field. Materials that have no magnetic field in
the absence of an applied magnetic field but that respond to enhance an applied magnetic field are
termed paramagnetic. Examples of paramagnetic substances include gadolinium, used as an MRI
contrast agent, and ferrous and ferric iron. On T1-weighted images, uniform distributions of such
species produce increased signal intensity. However, nonuniform distributions of such species alter the
MR image by producing a range of effective magnetic fields within the sample and therefore a range of
resonance frequencies of the MR signal. Depending on the type of imaging pulse sequence used, the
frequency dispersion can be manipulated to decrease signal in the area of the paramagnetic species
(see Sensitivities of MRI Pulse Sequences).
Other Forms of Magnetism

There are other biologically important closely packed ensembles of atoms, such as crystalline
structures, in which unpaired electrons of neighboring atoms interact to minimize the magnetic forces
outside the material in the absence of an applied magnetic field. The magnetic forces producing
preferred patterns of spin alignments to reach magnetic equilibrium are termed exchange forces.
Antiferromagnetism (Individual Opposition)

If unpaired electrons of neighboring atoms interact to align with opposing spins, the magnetic forces
are minimized. In an applied magnetic field, spin pairing must be disrupted for realignment. Thus, the
response to the applied field is less than that of a paramagnetic substance, but the effective field is still
enhanced. Such materials are termed antiferromagnetic. The alignment pattern can be disrupted if the
thermal energy is increased, and initially the response to an applied field is enhanced as the
1 of 5 12-04-2010 00:08
temperature is increased. Above a critical temperature, known as the Neel temperature, adjacent spin
pairing is disrupted and the substance becomes paramagnetic. The effects of antiferromagnetic
substances on MR images are similar to those of paramagnetic substances although reduced in
magnitude and with different temperature dependence.
Ferromagnetism (Group Opposition)

If the unpaired electrons of a group of atoms can align in domains, each domain has a net magnetic
field. Adjacent domains can then interact by means of these magnetic fields to minimize, although not
completely cancel, the field outside the material. If that material is immersed in a high magnetic field,
domains respond to both the applied field and neighboring domain fields to markedly enhance the field.
Such materials show a magnetic field in the absence of an applied magnetic field and are termed
ferromagnetic. The effect of such substances on the MR image is of greater magnitude than that of
paramagnetic substances.
If a ferromagnetic crystal is reduced in size to that of a single domain, this single-domain particle has a
net magnetic dipole equivalent to that of a domain. If a collection of such particles is free to rotate in an
applied magnetic field on a time scale that is shorter than the observation time, the magnetic dipoles
behave as expected for paramagnetism discussed previously. However, the larger magnetic moments
of the particles produce a greater enhancement of the applied magnetic field. Such particles are
termed superparamagnetic.52,53
If the size of the particles, usually containing many domains, is reduced below the size of a single
domain, the aligning exchange forces and misaligning thermal forces become comparable. If the time
scale of the observation is longer than the switching rate of the equilibrium between the aligned and
disordered states, then the magnetic properties of the particles are dependent on the temperature-
volume relationships that determine the switching frequency. An aggregate of such particles behaves
paramagnetically but with a greater magnetic dipole than if no domain formed at all and thus is termed
51
supermagnetic. Thus, the effects of superparamagnetic and supermagnetic substances on the MR
image are similar to those of paramagnetic substances but of a magnitude between that of
paramagnetic and ferromagnetic species.
Magnetic Susceptibility
Because all biologic materials have at least one of the previously discussed magnetic properties, they
interact with a static magnetic field, H0, to produce a magnetization, m, that reduces (diamagnetism) or
enhances (paramagnetism, antiferromagnetism, ferromagnetism) the effective magnetic field, Heff,
established within the material. Note that m in this sense is usually referring to the effects of electronic
configurations, not nuclear magnetization M. This effect can be expressed in terms of the magnetic
susceptibility χ of the material in which
where χ = m/H0. Thus, χ < 0 for diamagnetic materials, χ > 0 for paramagnetic materials, and χ = 0
for a vacuum.
page 184
page 185
When placed in a static magnetic field of the imaging magnet, tissues of different magnetic
susceptibility establish different effective magnetic fields experienced by the nucleus under observation.
Thus, the response of the nuclear magnetization generated in the MRI study is altered with resultant
changes in the image. Susceptibility-induced field variations within a voxel broaden the Larmor
frequency range (see Fig. 6-6). SE imaging minimizes this effect by using a 180° RF pulse to refocus
the dispersion that occurs in the transverse magnetization Mxy, during the pulse sequence. If significant
molecular diffusion occurs during TE (time from initial 90° RF pulse to formation of the echo) through
regions of variable Heff, due to either magnet imperfections or susceptibility variations in the tissue,
incomplete refocusing results in loss of transverse magnetization and thus signal loss in the MR image.
The effect becomes more apparent as the TE exceeds the diffusional correlation time (time taken for a
proton to move from one position to the next position). Spins diffusing in the x-y plane experience these
2 of 5 12-04-2010 00:08
variations in Heff as a fluctuating magnetic field hz, with resultant transverse but not longitudinal
34-36
relaxation. This effect is well known in MR spectroscopy, in which a known magnetic field gradient
(G) can be applied to a sample to measure the diffusion coefficient (D) of protons through a solvent.
The signal intensity (S) for a Hahn spin-echo at TE is given as
Hence, differences in susceptibility at the boundary of a hematoma containing paramagnetic blood

products and surrounding normal diamagnetic tissue produce T2 relaxation if sufficient diffusion is
permitted to occur during the imaging sequence. Such effects cause signal loss at these boundaries on
T2-weighted MR images of sufficiently long TE without corresponding signal loss on T1-weighted
images.
Relaxivity
The relative rotational and translational motions of water molecules and paramagnetic entities in
biologic systems occur on a time scale that produces an apparent isotropically fluctuating magnetic
field in the range of the Larmor frequencies for protons at current imaging field strengths. If the water
molecules are able to approach the paramagnetic center, then magnetic interactions allow efficient
energy exchange to occur and the magnetically perturbed water proton spin system can relax to its
equilibrium state.54 The phenomenologic equation for intermolecular relaxation interactions of a
paramagnetic agent (P) in bulk solution is
where i = 1, 2; R is the relaxivity constant (mM/s); and [P] is the concentration (mM) of the
paramagnetic substance P. (1/T)d is the rate attributable to diamagnetic relaxation processes and
(1/T)p is the relaxation rate in the presence of P. In the presence of a suitable paramagnetic
substance, the paramagnetic term dominates over the diamagnetic term of Equation 6-3. The same
equation applies for both longitudinal and transverse relaxation. Because T1 is generally longer than
T2, 1/T1 is smaller than 1/T2, and so the constant term R[P] contributes a greater proportion to the
longitudinal relaxation rate (1/T1) than transverse relaxation rate (1/T2). The implication for MRI is that
the relaxivity effects of paramagnetic substances are detected with greater sensitivity on T1-weighted
images than on T2-weighted images. The paramagnetic relaxation rate can be further analyzed
mechanistically as inner sphere (ligand exchange in which water molecules are in the first coordination
sphere of P) and outer sphere (diffusion with close approach of water near to but without coordination
to P) contributions, but it is not the purpose of this review to describe the more mechanistic Solomon-
Bloembergen equations that are presented in detail elsewhere.54 Application of Equation 6-3 in biologic
systems as complex as cerebral hematomas can only be approximate because of the heterogeneity in
type and distribution of the various paramagnetic substances involved. When water is unable to
approach the paramagnetic center, no magnetic interaction and therefore no relaxation occurs by
relaxivity mechanisms (see Fig. 6-3). However, susceptibility effects can still be manifested so that,
whereas T1 is unaffected, T2 is shortened and the MR signal is decreased.
REFERENCES
1. Sipponen JT, Sepponen RE, Sivula A: Nuclear magnetic resonance (NMR) imaging of intracranial hemorrhage in the acute
and resolving phases. J Comput Assist Tomogr 7:954-959, 1983. Medline Similar articles
2. DeLaPaz RL, New PFJ, Buonanno FS, et al: NMR imaging of intracranial hemorrhage. J Comput Assist Tomogr 8:599-607,
3. Gomori JM, Grossman RI, Goldberg HI, et al: Intracranial hematomas: imaging by high-field MR. Radiology 157:87-93, 1985.
4. Gomori JM, Grossman RI: Head and neck hemorrhage. In Kressel HY (ed.): Magnetic Resonance Annual 1987. New York:
Raven Press, 1987, pp 71-112.
5. Gomori JM, Grossman RI, Hackney DB, et al: Variable appearances of subacute intracranial hematomas on high-field
spin-echo MR. Am J Neuroradiol 8:1019-1026, 1987.
6. Zimmerman RD, Heier LA, Snow RB, et al: Acute intracranial hemorrhage: intensity changes on sequential MR scans at 0.5
T. Am J Neuroradiol 9:47-57, 1988.
7. New PFJ, Scott WR: Blood. In New PFJ, Scott WR (eds): Computed Tomography of the Brain and Orbit (EMI Scanning).
Baltimore: Williams & Wilkins, 1975, pp 263-267.
3 of 5 12-04-2010 00:08
8. Enzmann DR, Britt RH, Lyons BE, et al: Natural history of experimental intracerebral hemorrhage: sonography, computed
tomography and neuropathology. Am J Neuroradiol 2:517-526, 1981. Medline Similar articles
9. Kreindler A, Marcovici G, Florescu I: Histoenzymologic and biochemical investigations into the nervous tissue round an
experimentally-induced cerebral hemorrhagic focus. Rev Roum Neurol 9:313-319, 1972. Medline Similar articles
10. Lalonde JMA, Ghadially FN, Massey KL: Ultrastructure of intramuscular haematomas and electron-probe X-ray analysis of
extracellular and intracellular iron deposits. J Pathol 125:17-23, 1978. Medline Similar articles
11. Finch CA, Huebers HA: Iron metabolism. Clin Physiol Biochem 4:5-10, 1986. Medline Similar articles
12. Trump BF, Valigorsky JM, Arstila AU, et al: The relationship of intracellular pathways of iron metabolism to cellular iron
overload and the iron storage diseases. Am J Pathol 72:295-324, 1973. Medline Similar articles
13. Weinberg ED: Iron, infection and neoplasia. Clin Physiol Biochem 4:50-60, 1986. Medline Similar articles
14. Southern PA, Powis G: Free radicals in medicine. I. Chemical nature and biologic reactions. II. Involvement in human
disease. Mayo Clin Proc 63:390-408, 1988. Medline Similar articles
15. Munro HN, Linder MC: Ferritin: structure, biosynthesis and role in iron metabolism. Physiol Rev 58:317-396, 1978.
16. Wixon RL, Prutkin L, Munro HN: Hemosiderin: nature, formation and significance. Int Rev Exp Pathol 22:193-225, 1980.
17. Bunn HF, Forget BG (eds): Hemoglobin: Molecular, Genetic and Clinical Aspects. Philadelphia: WB Saunders, 1986, pp
13-35, 634-662.
18. Thulborn KR, Wateron JC, Matthews PM, Radda GK: Oxygenation dependence of the transverse relaxation time of water
protons in whole blood at high field. Biochim Biophys Acta 714:265-270, 1982. Medline Similar articles
19. Brindle KM, Brown FF, Campbell ID, et al: Application of spin-echo nuclear magnetic resonance to whole cell systems:
membrane transport. Biochem J 180:37-44, 1979. Medline Similar articles
page 185
page 186
20. Atlas SW, Thulborn KR: MRI Detection of hyperacute parenchymal hemorrhage of the brain. Am J Neuroradiol
21. Koenig SH, Brown RD, Lindstrom TR: Interactions of solvent with the heme region of methemoglobin and fluoro-
methemoglobin. Biophys J 34:397-408, 1981. Medline Similar articles
22. Blumberg WE: The study of hemoglobin by electron paramagnetic resonance spectroscopy. Methods Enzymol
23. Windle JJ, Weirsema AK, Clarke JR, Feeney RE: Investigation of the iron and copper complexes of avian conalbumins and
human transferrins by electron paramagnetic resonance. Biochemistry 2:1341-1345, 1963. Medline Similar articles
24. Aasa R, Aisen P: An electron paramagnetic resonance study of the iron and copper complexes of transferrin. J Biol Chem
25. Koenig SH, Schillinger WE: Nuclear magnetic relaxation dispersion in protein solutions. II. Transferrin. J Biol Chem
26. Harrison PM, Fischbach FA, Hoy TG, Haggis GH: Ferric oxyhydroxide core of ferritin. Nature 216:1188-1190, 1967.
27. Koenig SH, Brown RD: Relaxometry of magnetic resonance imaging contrast agents. In Kressel HY (ed.): Magnetic
Resonance Annual 1987. New York: Raven Press, 1987, pp 263-286.
28. Gillis P, Koenig SH: Transverse relaxation of solvent protons induced by magnetized spheres: application to ferritin,
erythrocytes and magnetite. Magn Reson Med 5:323-345, 1987. Medline Similar articles
29. Weir MP, Peters TJ, Gibson JF: Electron spin resonance studies of splenic ferritin and haemosiderin. Biochim Biophys
Acta 828:298-305, 1985. Medline Similar articles
30. Boas JF, Troup GJ: Electron spin resonance and Mossbauer effect studies of ferritin. Biochim Biophys Acta 229:68-74,
31. Darrow YC, Alvord EC, Hodson WA: Histological evolution of the reactions to hemorrhage in the premature human infant's
brain: a combined autopsy study and a comparison with the reaction in adults. Am J Pathol 130:44-58, 1988. Medline
Similar articles
32. Cohen MD, McGuire W, Cory DA, Smith JA: MR appearance of blood and blood products: an in vitro study. Am J
Roentgenol 146:1293-1297, 1986.
33. Hayman LA, Ford JJ, Taber KH, et al: T2 effect of hemoglobin concentration: assessment with in vitro spectroscopy.
34. Abragam A: Principles of Nuclear Magnetism. Oxford: Clarendon Press, 1985.
35. Becker ED: High Resolution NMR. Theory and Chemical Applications. New York: Academic Press, 1980.
36. Farrar TC, Becker ED: Pulse and Fourier Transform NMR. New York: Academic Press, 1971, pp 2-15.
37. Fukushima E, Roeder SBW: Experimental Pulse NMR. A Nuts and Bolts Approach. London: Addison-Wesley Publishing,
1981, pp 1-125.
4 of 5 12-04-2010 00:08
38. Yoder CH, Schaeffer CD Jr: Introduction to Multinuclear NMR. Menlo Park, CA: Benjamin/Cummings Publishing, 1987.
39. Fullerton GD: Basic concepts for nuclear magnetic resonance imaging. Magn Reson Imaging 1:39-53, 1982. Medline
Similar articles
40. Bottomley PA: NMR imaging techniques and applications. Rev Sci Instrum 53:1319-1337, 1982. Medline
Similar articles
41. Brant-Zawadzki M, Norman D: Magnetic Resonance Imaging of the Central Nervous System. New York: Raven Press,
1987.
42. Pykett IL, Newhouse JH, Buonanno FS, et al: Principles of nuclear magnetic resonance imaging. Radiology 143:157-168,
43. Wehrli FW: Principles of magnetic resonance. In Stark DD, Bradley WG (eds): Magnetic Resonance Imaging. St. Louis:
CV Mosby, 1988, pp 1-23.
44. Edelman RR, Buxton RB, Brady TJ: Rapid MR imaging. In Kressel HY (ed.): Magnetic Resonance Annual 1988. New York:
Raven Press, 1988, pp 189-216.
45. Wismer GL, Buxton RB, Rosen BR, et al: Susceptibility induced magnetic resonance line broadening: applications to brain
iron mapping. J Comput Assist Tomogr 12:259-265, 1988. Medline Similar articles
46. Hoppel BE, Weisskoff RM, Thulborn KR, et al: Measurement of regional blood oxygenation and cerebral hemodynamics.
47. Ludeke KM, Roschmann P, Tischler R: Susceptibility artefacts in NMR imaging. Magn Reson Imaging 3:329-343, 1985.
48. Schenck JF, Mueller OM, Souza SP, Dumoulin CL: Magnetic resonance imaging of brain iron using a 4 Tesla whole-body
scanner. In Frankel RB, Blakemore RP (eds): Iron Biominerals. New York: Plenum Publishing, 1990, pp 373-385.
49. Di Chiro G, Brooks RA, Girton ME, et al: Sequential MR studies of intracerebral hematomas in monkeys. Am J Neuroradiol
50. Kamman RL, Go KG, Brouwer W, Berendsen HJC: Nuclear magnetic resonance relaxation in experimental brain edema:
effects of water concentration, protein concentration and temperature. Magn Reson Med 6:265-274, 1988. Medline
Similar articles
51. Burke HE: Handbook of Magnetic Phenomena. New York: Van Nostrand Reinhold, 1986, pp 9-57.
52. Bean CP, Livingston JD: Superparamagnetism. J Appl Phys 30:120S-129S, 1959.
53. Bean CP: Hysteresis loops of mixtures of ferromagnetic micropowders. J Appl Phys 26:1381-1383, 1955.
54. Lauffer RB: Paramagnetic metal complexes as water proton relaxation agents for NMR imaging: theory and design. Chem
Rev 87:901-927, 1987.
© 2010 Elsevier
5 of 5 12-04-2010 00:08
DVANCED MAGING ECHNIQUES NCLUDING AST

MAGING
Timothy J. Carroll
Ken E. Sakaie
Piotr A. Wielopolski
Robert R. Edelman
In the continuing efforts to improve the diagnostic capability of MRI, a number of fast imaging, novel
magnetization preparation, and data acquisition strategies have been introduced. Each of the methods
discussed here provides new information in terms of image contrast, quality, or time resolution.
Because of their clinical relevance, fast imaging techniques will be considered first (exclusive of parallel
imaging methods that are reviewed in Chapter 8, Parallel Imaging Techniques). The discussion of
magnetization preparation begins with fat-suppression methods, which reduce undesirably bright
adipose signal that can exacerbate artifact and reduce contrast. Magnetization transfer, spin lock
imaging, and zero quantum coherence imaging provide types of contrast distinct from the standard
forms; the methods may prove particularly valuable in differentiating between tissue in which the
concentration of proteins and other macromolecules correlates with function. Electric current imaging
uses applied or intrinsic currents to provide contrast and may provide insights into the function and
viability of certain tissue. Imaging of intrinsic currents among clusters of neurons may provide a new
way to detect functional activation of the brain.
Changing the manner of data acquisition can have a large impact on image quality. Non-Cartesian
trajectories in k-space-namely projection reconstruction and spiral imaging-provide certain advantages,
particularly in fast imaging. The oversampling of the center of k-space inherent to the non-Cartesian
acquisitions makes these approaches less susceptible to artifact in the presence of motion even for
high resolution scans. Time-resolved imaging takes fast imaging a step further, allowing time resolution
of dynamic changes and has proven particularly useful in MR angiography. Motion correction
techniques, on the other hand, aim to eliminate the effects of dynamic changes on the image, resulting
in higher image quality.
© 2010 Elsevier
1 of 1 12-04-2010 00:08
PULSE SEQUENCES FOR FAST IMAGING

page 187
page 188
Add to lightbox
Figure 7-1 Time line showing the different time frames of certain body functions.
Along with the expansion of clinical applications has come a heightened appreciation of the problems posed by
respiratory and cardiac motion and the need for fast imaging to overcome these artifacts and to provide the speed
needed for functional imaging studies (Fig. 7-1). Although standard spin-echo (SE) pulse sequences are well
validated, they lack the flexibility to address these problems. Fortunately, solutions now exist in the area of fast MRI
techniques. It is now possible, for instance, to eliminate artifacts from respiratory motion by use of subsecond
imaging methods such as turbo fast low-angle shot (turbo FLASH); cardiac motion can be tamed by use of
balanced steady-state free precession (SSFP) sequences such as true FISP or FIESTA in conjunction with cardiac
gating. In addition, fast MRI methods can produce images with useful types of tissue contrast and improved spatial
resolution, completely aside from the ability to reduce scan times.
Practical aspects of imaging techniques, including a basic explanation of fast imaging methods, are presented in
Chapter 3, Practical Considerations and Image Optimization. In this section, fast MRI methods are reviewed in
greater depth with consideration of technical issues as well as illustration of typical clinical applications. These
methods, summarized in Table 7-1, can greatly improve the image quality and breadth of applications of MRI.
Fast Spin-Echo Imaging Methods

Though largely supplanted by fast SE, standard SE pulse sequences are still commonly used, particularly for
T1-weighted acquisitions (e.g., brain, spine). The single-echo SE sequence consists of a pair of RF pulses and can
be represented as:
where the echo time (TE) defines the center of the SE signal. For each excitation, the number of times the
sequence is repeated is determined by the spatial resolution (specifically the number of pixels) along the phase-
encoding axis and is equal to the number of phase-encoding steps. A high level of spatial resolution requires a
correspondingly large number of phase-encoding steps.
The first RF pulse in the SE sequence, used to excite the spins, is set to 90° to tip the existing longitudinal
magnetization completely into the transverse plane. This maximizes the amount of signal that can be obtained per
measurement. However, immediately after the 90° pulse, the remaining longitudinal magnetization is zero. If the
spins were again excited at this time, no signal would be produced. The next 90° pulse produces a signal
proportional to the accumulated magnetization, as expressed by:
where S0 is the signal that would be produced by a 90° pulse when the spins are fully aligned with the magnetic
field. The signal (S) depends on the amount of T1 relaxation that occurs during the interval TR. To maintain a
sufficient degree of tissue contrast and SNR, the TR cannot be made much shorter than approximately 300 to 400
msec. This limitation essentially precludes a conventional SE sequence from being used for fast imaging, the reason
being that scan time is proportional to TR, as given by the following equation:
where Ny is the number of phase-encoded lines, NEX is the number of excitations (also called Naq, for number of
acquisitions).
Each 180° pulse in a conventional SE sequence refocuses the MR signal at a time interval TE/2 after the center of
the pulse. At the center of the SE, the effects of static magnetic field inhomogeneities (like those occurring at
air-soft-tissue interfaces) are eliminated.1 Additional 180° pulses may be applied to generate additional
spin-echoes; each echo is used to create a single image.
2,3
Much faster imaging is possible with fast SE techniques. Fast SE (also called turbo SE) is an optimized version of
1 of 23 12-04-2010 00:11
4,5
rapid acquisition relaxation enhancement (RARE). It is a multi-echo SE sequence in which each SE signal is
separately phase encoded, rather than being used to produce a separate image. Like SE, fast SE has the
advantage over GRE that it uses 180° pulses to eliminate susceptibility artifacts. Multislice 2D fast SE acquisitions
can be completed within a period of a few tens of seconds to a few minutes. Thin-section T2-weighted 3D scans,
not practical with conventional SE, can be acquired using fast SE in 10 minutes or less.
Drawbacks of fast SE include increased RF deposition and magnetization transfer effects arising from the
application of multiple 180° pulses, increased signal intensity of fat, and blurring artifacts particularly at short TE.
Depending on the sequence implementation and choice of imaging parameters, some lesions may not be seen as
well with fast SE as with conventional SE techniques. In spite of these limitations, fast SE has largely replaced
conventional SE in most organ systems.
The scan time for a fast SE sequence is given by:
where ETL is echo train length (also called turbo factor).
The ETL (1 for standard SE, 3 to 256 for fast SE) is the number of echoes that are phase encoded after each 90°
RF excitation. For instance, if eight echoes are acquired (ETL = 8), the scan time is reduced by a factor of 8. A
large ETL provides the greatest reduction in scan time, but, beyond a factor of 16 or so, blurring artifacts may
become objectionable for some tissues. The blurring is attributable to T2-dependent signal loss that occurs over the
duration of the echo train.
page 188
page 189
Table 7-1. Examples of Fast Imaging Techniques

Typical scan
Technique times Contrast Uses/limitations
Pulse sequences
Fast spin-echo (e.g., Fast 1-10 min T1, Standard acquisition method for T2-, PD-, and,
SE, Turbo SE) PD,T2 to a lesser extent, T1-weighted images; high RF
power deposition, MTC effects
Fast-recovery fast 20 sec-10 PD,T2 Driven equilibrium fast spin-echo, improved T2
spin-echo min contrast for given TR; reduced breath-hold times
for abdominal MRI
Single shot fast spin-echo 200 msec-2.5 T2 Motion insensitive; useful for MRCP and MR
(e.g., SSFSE, HASTE) sec urography; suffers from image blurring at short
TE
Hyperechoes, TRAPS 1-10 min T2 Modification of fast SE; useful for 3T since less
RF power deposition; longer TE for similar fast
SE contrast; fewer MTC effects
GRASE 30 sec-10 PD,T2 Reduced RF power deposition; ghosting artifacts
min with poor fat suppression; high resolution for
brain scans
Incoherent gradient-echo 200 msec-30 PD,T1 Standard acquisition method for MRA,
(e.g., FLASH, SPGR) sec T1-weighted breath-hold abdominal MRI
Steady-state coherent 200 msec-30 T1/T2 Generally not used; sensitive to fluid motion
gradient-echo (e.g. sec
GRASS, FISP, FSE)
Cardiac cine, rapid scout imaging if single shot,
bright blood; velocity insensitive; RF intensive, B0
sensitive; shortest TR possible
Balanced gradient 200 msec-16
waveform (e.g., SSFP, sec
Fiesta, True FISP, Balance
FSE)
Magnetization-prepared 2D: 1-3 sec T1 Test bolus for contrast-enhanced MRA, motion-
gradient-echo(e.g. turbo 3D: 2-10 min T1,T2 insensitive T1-weighted breath-hold abdominal
FLASH, MP-RAGE) MRI; more blurring than standard GRE
3D scans with flexible contrast (T1 or T2)
2 of 23 12-04-2010 00:11
Single-shot EPI 50-200 msec T2*,T2 Perfusion/diffusion/BOLD MRI of brain

BURST <1 sec T2 Quiet scanning for fMRI; low image quality
k-space sampling methods to reduce scan time
Partial Fourier Reduced by - Shorten scan time for SE; fast SE; balanced
factor of SSFP; contrast-enhanced MRA
0.5-0.75
Keyhole - - Shorten scan time for contrast-enhanced MRI
Parallel imaging Reduced by - Shorter scan time requires appropriate phased
factorof 2-8 array coil; ghosting artifacts if inaccurate coil
map or inappropriate coil configuration
TRICKS Reduction 3-4 T1 Shorter scan time for CE-MRA
Special k-space sampling patterns
Spiral - - Effective coverage of k-space; good flow
properties; sensitive to B0 inhomogeneities and
gradientimperfections
Radial (projection- - - Can be used with variety of pulse sequences;
reconstructions) motion correction (PROPELLER); reduced
motion sensitivity; reduced scan time if
undersampling used
VIPR 5-30 sec T1 Undersampled 3D radial scan for multiphase
contrast-enhanced 3D MRA and phase-contrast
MRA; requires high vessel-to-background
contrast
page 189
page 190
BOLD, blood oxygen level dependent contrast; EPI, echo-planar imaging; FISP, fast imaging with steady-state
precession; FLASH, fast low-angle shot; GRASE, gradient and spin-echo; GRASS, gradient-recalled acquisition in
the steady state; HASTE, half-Fourier single-shot turbo spin-echo; MP-RAGE, magnetization prepared rapid
gradient-echo; MRA, magnetic resonance angiography; MRCP, magnetic resonance cholangiopancreaticography;
MTC, magnetization transfer contrast; PD, proton density weighted; PROPELLER, periodically rotated overlapping
parallel lines with enhanced reconstruction; RF, radiofrequency; SE, spin-echo; SPGR, spoiled gradient recalled;
SSFP, steady-state free precession; SSFSE, single-shot fast spin-echo; TE, echo time; TR, repetition time;
TRAPS, transition between pseudo steady states; TRICKS, time-resolved imaging of contrast kinetics; VIPR, vastly
undersampled isotropic projection reconstruction.
Add to lightbox
Figure 7-2 Edge enhancement on fast SE images. Left, Axial proton density-weighted fast spin-echo (SE) image
(4500/22) shows artifactual bright line (arrowhead) around the dark nerve roots. Right, Artifact is less obvious on
T2-weighted fast SE image (4500/85). The edge enhancement effect improves the conspicuity of the nerve roots.
3 of 23 12-04-2010 00:11
The echo train spacing (ETS) is also important. Typical ETS is on the order of 4 to 12 msec. With shorter spacing,
the duration of the echo train is reduced, thereby minimizing the amount of blurring. Sensitivity to magnetic
susceptibility artifact (e.g., from a hip prosthesis) is greatly reduced. The signal intensity from fat tends to increase
as the ETS is decreased.
It is possible to acquire as many as 80 or more echoes in less than 0.5 s. In conjunction with half-Fourier, one can
use this technique called HASTE (half-Fourier single-shot turbo spin-echo). HASTE images provide good depiction
of fluids, but other tissues with shorter T2 values tend to show some blurring artifacts. For making thick-slab
projection images of fluids such as bile (e.g. MR cholangiography, MR urography, MR myoelography), the
single-shot fast SE sequence is acquired with a TE in the order of 1 second and no half-Fourier. Typical ETL is then
in the order of 256.
Nominal Echo Time

As mentioned earlier, image contrast depends on the characteristics of the MR signals acquired during the central
phase-encoding lines. The central lines are selected simply by setting the amplitude of the phase-encoding gradient
to low values; this can be done for any of the echoes in the fast SE echo train, although the order must be
appropriate to avoid artifacts. The timing of the central phase-encoding lines determines the nominal TE. If the
central lines are timed to occur late, allowing T2 decay of the echo, the image appears T2 weighted; if it is timed to
occur early, the image appears proton density weighted (or T1 weighted if the TR is short).
One can split the echo train, for instance, to acquire proton density- and T2-weighted images simultaneously.
However, the reduction in scan time is less than it would be for a single echo acquisition. For faster scanning using
a dual-echo acquisition, an echo-sharing technique can be used. 6 Two images with different effective TE values are
acquired using the same high-frequency data. The central (low-frequency) lines are acquired twice at different TE
as part of the same echo train.
Phase-Encode Order
Because the MR signal strength decreases with each successive echo as a result of T2 decay, the image
appearance is dramatically altered with changes in the phase-encoding order. The T2-modulated time-dependent
decrease in signal intensity acts as a filter of high spatial frequencies, causing blurring. Phase-encoding order is a
critical determinant of image quality with fast SE sequences.7-9 The phase-encoding order must be such that abrupt
discontinuities in the signal intensity from line to line as well as gradient-induced eddy current effects are minimized.
Discontinuities can cause blurring and ghost artifacts.
The amount of blurring with fast SE imaging depends on the T2 of the tissue, because signal loss occurs over the
duration of the echo train as a result of T2 relaxation. For instance, T2 decay over the echo train is worse for white
matter than for cerebrospinal fluid (CSF). Thus, one could use a much longer echo train in conjunction with a
single-shot fast SE acquisition if a fluid is the tissue of interest than for imaging of white matter lesions.
Paradoxically, in addition to blurring, there can be an edge enhancement effect that in some circumstances
improves depiction of normal anatomy or lesion detection.10 For instance, blurring is minimal for CSF because of its
long T2. On the other hand, nerve roots are blurred because of their shorter T2. The net effect is that the interface
between nerve root and CSF is enhanced, thereby increasing the conspicuity of the root; however, there can be
artifactually low signal intensity within the root (Fig. 7-2).
page 190
page 191
For a given ETL, blurring is less for a short ETS than for a long one. 11 The drawbacks of short ETS are a reduced
SNR because the readout bandwidth must be increased (assuming that gradient ramp times are already minimized),
brighter fat signal, lower conspicuity of hemorrhagic lesions, and higher power deposition resulting from the use of a
shorter RF pulse.
Three-Dimensional Fast Spin-Echo

Although 3D imaging with conventional SE sequences is impractical because of the long scan times, 3D fast SE is
12
feasible with scan times less than 10 minutes. The sequence is similar to 2D fast SE but incorporates an extra
phase-encoding step in the slice-select direction. Multiple 3D volumes spanning the region of interest are acquired
within each TR interval. Compared with 2D fast SE imaging, thinner sections can be obtained and multiplanar
reformations can be done.
Fast Recovery Fast Spin-Echo

One way to reduce scan time with SE sequences is to decrease the excitation flip angle from 90°, a method known
13
as the fast low-angle multiecho (FLAME) technique. As with GRE sequences, the reduced flip angle decreases
4 of 23 12-04-2010 00:11
saturation effects with short TR. For this approach to work, an even number of 180° RF pulses (e.g., two) must be
used. The reduced flip angle excitation tilts the longitudinal magnetization partway into the transverse plane, so
there is a transverse component and a longitudinal component (along the +z direction). The first 180° pulse
refocuses the transverse component of the magnetization as an SE, which is read out. This 180° pulse also flips the
longitudinal component of the magnetization into the - z direction. The second 180° pulse flips the longitudinal
magnetization back along the +z direction. Using the FLAME technique, an SE sequence with a TR of 1500 msec
and a flip angle of 60° might generate T2 contrast similar to that obtained with a TR of 2500 msec and a 90° flip
angle. The advantage of the shorter TR is a reduced scan time; the drawback is a decreased multislice capability
because of the shorter TR.
Whereas the FLAME technique is outdated, a related but improved method called "fast recovery fast SE" is in
widespread use. Following the usual train of 180° refocusing RF pulses, the method applies an additional refocusing
180° RF pulse followed by a 90° RF pulse. The last RF pulse forces the longitudinal magnetization back to the +z
axis, thereby shortening the period needed for T1 recovery.14 The method can be used to shorten scan times in
general and to permit breath-hold fast SE imaging of the abdomen (Fig. 7-3).
Magnetization Transfer Effects

A slice-selective RF pulse excites protons over a range of hundreds to thousands of hertz. Although the frequency
of the pulse is selected to be on-resonance (at the Larmor frequency) for the particular slice being excited, it is
off-resonance for the other slices of a multislice acquisition. The result is that each slice in a multislice volume
experiences some degree of magnetization transfer effect from the RF excitations of each of the other slices,
particularly from the 180° pulses, which deposit more power than the 90° pulses. 15 Magnetization transfer effects,
which are relatively minor with conventional multislice SE sequences, become much more significant with fast SE
sequences because of the larger number of 180° pulses.16 The change in tissue contrast that results may or may
not be desirable. Magnetization transfer effects are more pronounced with long ETL and in tissues that contain both
mobile and restricted pools of protons. Thus, the signal intensities of brain, liver, or muscle are reduced by
magnetization transfer effects, but the signal intensities of tissues that contain only a small pool of restricted
protons, such as CSF, blood, or cavernous hemangiomas17 are affected less.
Stimulated Echoes
Any combination of three RF pulses has the potential to generate a stimulated echo. With all the RF pulses applied
in a fast SE sequence, stimulated echoes are unavoidable. In fact, the stimulated echoes are deliberately
18
overlapped with the SE signals to increase SNR, although some alteration in tissue contrast occurs as a result.
Hyperechoes
Fast SE scans are usually RF intensive. In order to reduce SAR, the refocusing angles of the fast SE train (180°)
are generally lowered (90° to 160°) to make it possible to scan an adequate number of slices when multislice
breath-hold measurements are performed. Although there may be a slight loss of T2 contrast, this can be
compensated by an increase in TE.
Lowering the refocusing angles will somewhat reduce the SNR. Furthermore, the risk of exceeding the SAR limits
increases at higher field strengths (e.g., SAR is 4 times higher at 3 T than at 1.5 T). Lower refocusing flip angles
and slightly longer RF pulses are used with 3 T fast SE implementations. However, lengthening the RF pulse to
reduce SAR constraints is done at the expense of an increased ETS (providing additional filtering from T2 decay)
and, at higher field strengths, additional signal loss in the presence of augmented magnetic field inhomogeneities
and chemical shift. Hennig and Scheffler19 have devised a new spin refocusing strategy that maintains the benefits
of fast SE by employing echoes of echoes (hence the name, hyperechoes), preserving SNR as in conventional fast
SE sequences but with dramatically reduced SAR values (reductions up to 70% to 90%; Fig. 7-4).
page 191
page 192
5 of 23 12-04-2010 00:11
Add to lightbox
Add to lightbox
Figure 7-3 Fast recovery fast spin-echo. A, Pulse sequence timing diagram illustrates the fast-recovery
modification. Once the final echo in the fast SE (FSE) echo train has been acquired, an additional 180° (180 deg)
pulse refocuses the residual magnetization in the transverse plane, and then a -90° (-90 deg) pulse flips it back to
the longitudinal axis instead of allowing it to recover with T1 processes. RF = radiofrequency pulses. B,
Comparison of fast SE techniques images in a patient with a small peripheral liver hemangioma (arrow). (Top)
respiratory-triggered fast SE (TR/TE = 3,600/90), (middle) breath-hold fast recovery fast SE (3,000/90), and
(bottom) half-Fourier single-shot fast SE (TE = 90). Depiction and sharpness of the lesion are best seen with the
fast recovery fast SE sequence. (Adapted from reference 14 with permission.)
The principle behind hyperechoes is fairly complex. In essence, a sequence of refocusing RF pulses, played with
arbitrary flip angles, phase and gradient pulses following an initial 90° RF excitation pulse, will refocus to a full echo
after the application of a 180° inversion pulse if the initial sequence of RF refocusing pulses is applied in reverse
order with the opposite amplitudes and phases and the same gradient pulsing scheme. The signal intensity of the
generated hyperecho can be thought of as the addition of all individual coherence pathways generated by each RF
pulse (mixing of transverse and longitudinal magnetization components and generation of stimulated echoes) at the
time of the hyperecho. In general, relaxation and diffusion effects accumulate between RF pulses, and these will
provide additional modulation of the amplitude of the hyperecho. To obtain similar T2 contrast as obtained with fast
SE, the effective TE is usually increased. Frank et al20 have taken advantage of the increased SNR with
hyperechoes to produce a sequence with greater diffusion contrast. In their experience, the incorporation of a 180°
RF refocusing pulse in a standard diffusion weighted stimulated echo sequence (resulting in the simplest hyperecho
sequence of 90° - [α1,ϕ1] - 180° - [-α1,-ϕ1] - hyperecho) retained the same diffusion weighting for tissues with small
T1/T2 ratios but with increased SNR compared with the 50% signal loss typically associated with the stimulated
echo approach.
Flow Effects with Fast Spin-Echo

Outflow of blood from the slice during the interval TE/2 between the 90° and 180° RF pulses (washout) tends to
make flowing blood appear dark in conventional SE images. Washout effects are prominent with fast SE
sequences, which use longer echo trains than conventional SE. Thus, fast SE imaging can be a useful way to create
6 of 23 12-04-2010 00:11
"dark blood" angiograms.

page 192
page 193
Add to lightbox
Add to lightbox
Figure 7-4 A, Diagram of the hyperecho pulse sequence showing RF pulse timings, phase and flip angles. B, High
resolution 3D turbo spin-echo imaging of the brain using hyperechoes at 3 T with spatial resolution of 0.5 × 0.5 ×
1.5 mm showing fine details of brain anatomy without limitation by SAR. (A, Reproduced with permission from
Hennig J, Scheffler K: Hyperechoes. Magn Reson Med 46:6-12, 2001; B, Courtesy of Dr. Juergen Hennig.)
Appearance of Hemorrhage with Fast Spin-Echo

Intracellular deoxyhemoglobin and methemoglobin are largely responsible for the signal loss in acute hemorrhage
21
seen on T2-weighted SE images. Iron storage products (ferritin and hemosiderin) cause the signal loss in chronic
22
bleeding. The paramagnetic iron in hemorrhage causes local magnetic field distortions (see Chapter 6,
Biochemical Basis of the MRI Appearance of Cerebral Hemorrhage). Although static field effects are canceled by
the refocusing pulse in SE sequences, diffusion of spins through these regions of inhomogeneous magnetic
susceptibility causes dephasing that is not completely refocused by the SE. This dephasing results in signal loss
within the hemorrhage.23 Fast SE is basically a Carr-Purcell-Meiboom-Gill (CPMG) sequence,24,25 so it also
minimizes signal loss resulting from diffusion. The greatest reduction in signal loss from diffusion occurs with fast SE
sequences that have a short ETS. The bottom line is that fast SE sequences are less sensitive to T2-dependent
signal loss in hemorrhage than are conventional T2-weighted SE sequences.
This insensitivity to susceptibility effects may also prove beneficial for other clinical applications, such as imaging in
the presence of metallic prostheses and imaging of the lung parenchyma.
Appearance of Fat on Fast Spin-Echo Images

One difference in the appearance of fast SE images compared with conventional SE images is the higher signal
intensity of fat. Several factors may contribute to this difference, the most important of which are diffusion effects
26,27
and J coupling. Other factors, such as stimulated echoes and steady-state effects, are considered to be less
important.
7 of 23 12-04-2010 00:11
page 193
page 194
Add to lightbox
Add to lightbox
Figure 7-5 Comparison of fast spin-echo (SE) with gradient and spin-echo (GRASE) in a case of a left temporal
cavernous hemangioma. A, SE: TR/TE = 3500/119, 23-cm FOV, 250 × 256 matrix, 3 minutes, 3 seconds
acquisition. B, GRASE: TR/TE = 3500/110, 24-cm FOV, 420 × 512 matrix, 2 minutes, 23 seconds acquisition. The
GRASE image appears slightly more sensitive to blood products. (Courtesy of NYU Faculty Practice.)
Diffusion of fat molecules normally causes some degree of signal loss in conventional SE images. Because the
sensitivity to diffusional effects is less with fast SE imaging, fat should appear brighter, although the scale of this
loss in vivo is uncertain. J coupling represents the intramolecular interactions of spins with different chemical shifts
through their electron bonds. It is responsible for splitting the spectrum of a molecule. Acquisition of a train of SEs
with a short inter-echo spacing minimizes J-coupling effects and thereby increases the signal intensity of fat. Again,
the relative importance of this effect in vivo is uncertain.
Fat suppression can be achieved in fast SE imaging by application of a chemical-shift-selective saturation pulse
before each repetition of the sequence. An alternative, though not widely applied, method to reduce the signal
intensity of fat is to vary the timing of odd and even echoes so that there is a 90° phase shift between fat and water
resonances. This gives a Carr-Purcell, rather than CPMG, condition for fat that increases dephasing.28
Gradient- and Spin-Echo Sequences

Fast SE offers a major improvement in acquisition speed over conventional SE, but still faster imaging is possible
29
using the gradient- and spin-echo (GRASE) sequence. With fast SE, the application of multiple 180° RF pulses is
time-consuming. Each 180° pulse has a finite duration, typically from 3 to 5 msec; moreover, these pulses are
SAR-intensive. If the ETS is shortened too much, the SAR limits may be exceeded, especially when the body coil is
used as the transmitter (power deposition is less for small transmit volume coils, such as the head coil). GRASE
solves this problem by acquiring several gradient-echoes (e.g., three) in place of the single SE of fast SE. Each of
these gradient-echoes is separately phase encoded. Because multiple phase-encoding steps are acquired after
each 180° pulse instead of just one, scan times are shorter than with fast SE. For instance, an entire brain study
can be completed in as little as 20 seconds, compared with 1 minute or longer with fast SE. There is a price to be
paid, however. Unlike SEs, GREs decay according to T2* and are more sensitive to variations in magnetic
susceptibility as well as chemical shift. In principle, GRASE sequences are more sensitive to the susceptibility
8 of 23 12-04-2010 00:11
effects of hemorrhage than fast SE, but the sensitivity depends on the particular sequence implementation (Fig.
7-5). Phase or amplitude corrections must be made to the data to eliminate abrupt jumps in phase or signal intensity
from one line of data to the next (these procedures are less crucial with fast SE). With GRASE, it is difficult to
achieve a smooth transition between the signal intensities of lines containing SE and GRE signals. These signal
discontinuities cause ringing artifacts that are more apparent than with fast SE. At present, we find GRASE of
limited value because of artifacts that occur if B0 homogeneity is suboptimal (e.g., for imaging near the dome of the
liver). Single-shot GRASE using a long ETL allows subsecond imaging, although with considerable blurring except
for tissues with long T2 values (i.e., fluids).
Gradient-Echo Pulse Sequences

A GRE sequence generates a gradient-echo (also called a gradient-recalled echo or field echo) simply by applying
a pair of balanced gradients of opposite sign along the frequency-encoding direction (see Chapter 5, Pulse
Sequence Design). The first gradient, called a dephasing or compensatory gradient, spreads out the transverse
magnetization over 180° (π) or more and thereby dephases the spins. The readout gradient exactly undoes the
dephasing at the TE so that there is no residual gradient dephasing effect. Unlike an SE sequence, which uses at
least two RF pulses to generate a spin-echo signal, a GRE sequence uses a single RF pulse to generate the
gradient-echo signal, usually with a flip angle less than 90°. A major advantage of GRE sequences is that good
image quality can be maintained even with quite short TR (e.g., 3 to 30 msec).
Several clinically useful variants of GRE sequences have been developed30-32 as will now be addressed.
Incoherent (Spoiled) Gradient-Echo

page 194
page 195
In incoherent GRE sequences such as FLASH or spoiled gradient recalled (SPGR), a spoiler gradient (i.e., a strong
gradient that is either constant or stepped through a different value after each excitation) is applied at the end of the
readout. The spoiler gradient disperses any residual transverse magnetization that remains after the signal has
been read and the analog-to-digital converter is turned off.33 If a spoiler is not applied, buildup of transverse
coherences over multiple RF excitations leads to horizontal striping in the image, particularly if short TR and large
34
flip angles are used. Spoiling is most effective if, in addition to the use of a spoiler gradient, the residual
magnetization is dispersed by randomizing the phases of the RF pulses, a method called RF spoiling.
In a spoiled GRE sequence, the transverse magnetization decays in much the same fashion as in an SE sequence.
The signal intensity in spoiled GRE images is given by:
where E1 = exp(-TR/T1), E* = exp(-TE/T2*), and α is the flip angle. S0 is the maximal obtainable signal, that is,
using a long TR, a short TE, and a 90° flip angle. The term E* contributes the T2* weighting. The T2* weighting is
somewhat different from the T2 weighting in an SE because, as denoted by the asterisk, magnetic field
inhomogeneity contributes to the loss of signal. The rate of signal loss is given by:
where ∆B0 is the magnetic field variation across a pixel and γ is the gyromagnetic ratio for hydrogen (approximately
42 MHz/T). Because artifacts resulting from magnetic field inhomogeneity worsen with long TE, strong T2*
weighting is problematic with spoiled GRE sequences. Susceptibility effects are minimized by the use of a short TE,
35
high readout bandwidth, and thin slice.
The amount of signal generated by an RF pulse is proportional to the longitudinal magnetization existing at the
moment before the pulse is applied. Now consider the effect of a 90° RF pulse in a SE or spoiled GRE sequence.
After applying this pulse, one must wait a considerable time, on the order of the tissue T1 relaxation time, before
repeating the excitation. Otherwise, as mentioned before, the tissue magnetization becomes saturated (i.e.,
approaches a value of zero) and insufficient signal is produced (the situation is completely different for steady-state
free precession sequences, below). If one uses a smaller flip angle, saturation effects are less and the TR can be
substantially shortened (Fig. 7-6). However, a small excitation flip angle will not work effectively with SE sequences
due to the confounding effects of the 180° refocusing pulse.
9 of 23 12-04-2010 00:11
Add to lightbox
Figure 7-6 Relationship of flip angle to signal for a spoiled gradient-echo (GRE) sequence. The time between α1
and α2 is TR. Top, Signal is directly proportional to spin density if α2 is much less than 2TR/T1. Only a small amount
of transverse magnetization (Mxy) and signal is generated from the first small flip angle RF pulse (α1), but the
protons are almost fully magnetized by the time of the second RF pulse (α2) and produce nearly the same amount of
signal as from the first excitation. Bottom, More transverse magnetization is initially generated with a large flip
angle excitation, but the protons have recovered only a portion of the magnetization by the time of the next RF pulse
and produce substantially less signal intensity from this excitation.
With a spoiled GRE sequence, the flip angle that produces the most signal for a given TR and T1 is known as the
Ernst angle (αE), as given by:
However, the Ernst angle is not necessarily the best choice for flip angle. More critical for lesion detection than the
SNR is the signal difference-to-noise or contrast-to-noise ratio (i.e., the difference in signal between two tissues,
such as a metastasis and the liver, divided by the standard deviation of the background noise). The flip angle is a
key determinant of GRE image contrast36,37 (Fig. 7-7). For T1-weighted spoiled GRE sequences, a flip angle larger
than the Ernst angle produces better T1 contrast; for proton density-weighted images, a smaller flip angle is
desirable. Note that the image contrast is very dependent on whether the GRE sequence is spoiled or not (Fig.
7-8).
Because short TR values are permissible with GRE sequences, 3D acquisitions can be completed in a reasonable
time period (e.g., 10 seconds to 10 minutes).38 With 3D, a thick volume of tissue rather than a thin 2D section is
excited by the RF pulse. Three-dimensional GRE imaging has the advantage of allowing quite thin (e.g., 1 mm or
less) contiguous sections with good SNR, because acquisition of multiple 3D partitions is comparable to using
multiple excitations in terms of boosting the SNR:
Coherent (Steady-State Free Precession) Gradient-Echo

page 195
page 196
Add to lightbox
10 of 23 12-04-2010 00:11
Add to lightbox
Add to lightbox
Add to lightbox
page 196
page 197
11 of 23 12-04-2010 00:11
Figure 7-7 Simulation of signal strengths from various tissues using different gradient-echo (GRE) methods. All
calculations obtained with TR/TE = 50/12. Vertical axis represents signal strength as a fraction of maximal signal
intensity (unweighted image). A, Spoiled GRE. T1 contrast is highest with large flip angles. However, signal is
maximal at smaller flip angles (Ernst angle), here shown to be approximately 15° to 30°. Note that signal curve in
region of Ernst angle is relatively flat, suggesting that there is considerable leeway in the choice of flip angles. B,
Steady-state coherent GRE obtained with phase alternation of RF pulses. Note marked enhancement of
cerebrospinal fluid because of its small T1/T2 ratio. C, Contrast-enhanced steady-state GRE sequence, first echo.
Contrast is similar to that in B. D, Contrast-enhanced steady-state GRE sequence (CE-FAST, PSIF). Note the
increased T2 contrast compared with A and B.
Add to lightbox
12 of 23 12-04-2010 00:11
Add to lightbox
page 197
page 198
Figure 7-8 Axial brain images showing comparisons of spoiled gradient-echo (GRE) and steady-state GRE
sequences. A, Spoiled GRE (FLASH). TR was 15 msec and the flip angle varied from 90° (upper left) to 10° (lower
right). B, Steady-state coherent GRE (FISP). TR was 15 msec and the flip angle varied from 90° (upper left) to 10°
(lower right). T1 contrast is inferior to that in A and the cerebrospinal fluid (CSF) signal intensity over the cerebral
convexities is high. Note that the lateral ventricles do not appear bright because CSF pulsation destroys the
steady-state signal.
Rather than discard the transverse magnetization that persists after readout, coherent (steady-state free
39,40
precession; SSFP) GRE sequences recycle this magnetization to increase the signal intensity of certain tissues.
41
One example of a coherent GRE sequence is fast imaging with steady-state precession (FISP), or gradient-
recalled acquisition in the steady state (GRASS). There is an extra phase-encoding gradient, of opposite sign to the
regular phase-encoding gradient, which has the opposite effect to a spoiler gradient. The purpose of this extra
gradient is to eliminate spoiling by the phase-encoding gradient of the transverse component of the steady-state
magnetization.
The difference in contrast behavior between spoiled and SSFP GRE is manifested only under the conditions of TR
[Lt ] T2 and large flip angle excitation (Fig. 7-9). When TR [Lt ] T2, the transverse magnetization persists long
enough after the readout that a portion of it can be tipped back onto the longitudinal axis by the next RF excitation; it
adds to the magnetization recovered by T1 relaxation over the TR interval. Subsequent RF pulses then tip this
enhanced magnetization back into the transverse plane, so that the signal intensity is greater than that produced by
a spoiled GRE sequence.
13 of 23 12-04-2010 00:11
Add to lightbox
Add to lightbox
Figure 7-9 Spoiled GRE (top) and steady-state coherent GRE (bottom) images show divergent contrast behavior
with short TR. Parasagittal images of a patient with a brain metastasis. TR/TE/FA = 30/14/90°. Note that the
spoiled GRE image appears T1 weighted, but with same scan parameters the steady-state coherent GRE image
shows marked enhancement of CSF (large arrow) and edema (small arrow) (small T1/T2 ratio).
Some tissues (particularly fluids) appear bright in SSFP GRE images, but the images are not T2 weighted in the
usual sense of T2-weighted SE images. The contrast depends on the ratio T1/T2; only tissues with a small T1/T2
ratio show high signal intensity. With longer TR (e.g., >250 msec), spoiled and SSFP GRE sequences produce
images that are more or less identical unless fluids are present (long T2). Moreover, with FISP and GRASS, motion
destroys the residual transverse magnetization so that moving fluids do not appear uniformly bright.42
Balanced Coherent Gradient-Echo

To compensate for signal loss due to motion, the balanced coherent or balanced SSFP sequence (e.g. true FISP or
FIESTA) was developed. This pulse sequence has a symmetrical gradient structure that eliminates phase shifts
caused by motion (see Chapter 5, Pulse Sequence Design). Fluids such as bile, blood, or CSF appear bright with
this sequence even when moving. Because it has a relatively small T1/T2 ratio, fat also appears bright. Because of
this property, the use of fat suppression can be helpful to differentiate flowing blood from fatty tissue.
With very short TR and TE (e.g. 3 msec and 1 msec), there are few if any artifacts; moreover, an image can be
acquired in as little as a few hundred milliseconds. These properties make balanced SSFP a motion-insensitive
technique that is ideal for scout imaging. Using cardiac gating, it is the method of choice for cine MRI43 (see
Chapter 32, Cardiac Imaging Techniques).
A drawback of this technique is the tendency to have dark stripes in the image, particularly in regions where the
magnetic field is inhomogeneous (see Chapter 22, Image Artifacts and Solutions). This "off-resonance effect" is
caused by shifts in the resonance frequency of the tissue protons in the presence of a nonuniform magnetic field.
The striping can be reduced by the use of a proper shim and frequency adjustment, along with a minimal TR and
TE. Alternatively, a maximal intensity projection (or the addition) of two acquisitions (one with alternating RF phase,
one with constant RF phase) can be done to eliminate the striping.44
14 of 23 12-04-2010 00:11
Power deposition is high with the large excitation flip angles and ultra-short TR required for this technique. The high
power deposition can become a limiting factor at 3 T, so that longer TR than the optimal may have to be used.
Under these circumstances, proper shimming is even more essential.
Contrast-Enhanced Steady-State Gradient-Echo

Another version of steady-state GRE is time-reversed FISP (PSIF) or contrast-enhanced Fourier-acquired
steady-state technique (CE-FAST). This pulse sequence has a structure that is reversed from a standard SSFP
GRE sequence.45,46 The result is that a heavily T2-weighted RF echo, rather than a GRE, is generated. The degree
of T2 weighting is exponentially dependent on twice the TR ≈ exp(-2TR/T2); thus, the effective TE is actually longer
than the TR. Contrast-enhanced steady-state GRE sequences render fluids and tumors bright but have the
drawback of being highly sensitive to motion (Fig. 7-10).
page 198
page 199
Add to lightbox
Add to lightbox
Figure 7-10 Patient with three cavernous hemangiomas (H) in the liver. Time-reversed fast imaging with
steady-state precession (PSIF) image (left) obtained with TR = 30 msec in less than 10 seconds provides similar
contrast to that obtained in 10 minutes using a T2-weighted spin-echo (SE) image (right, slightly different slice
position). However, sensitivity to motion (e.g., from pulsatile flow, cardiac pulsations transmitted to left lobe of liver)
is much worse with PSIF.
It is also possible to generate two echoes with different degrees of T2 weighting in the same acquisition. The first
echo has FISP-like contrast, the second is PSIF-like. These two echoes can be added together (double echo in the
steady state; DESS) to produce an image with a good SNR but the high signal intensity of fluid, which may be
helpful for the evaluation of hyaline cartilage in the knee when an effusion is present. However, interest in these
techniques has waned considerably with the advent of balanced SSFP and fast SE imaging, which respectively are
less motion sensitive and provide more robust T2 weighting than the contrast-enhanced GRE sequences.
Reaching the Steady State

Depending on the imaging parameters, the application of multiple RF pulses leads to the establishment of a
15 of 23 12-04-2010 00:11
steady-state or equilibrium condition. With both spoiled and SSFP GRE sequences, many (typically dozens of)
preparatory RF excitations must be applied before starting the acquisition of data for the spins to reach a
steady-state magnetization; otherwise, there are initial fluctuations in the tissue magnetization that produce
variations in the signal intensity and hence artifacts.47 With spoiled GRE, the approach to steady state is smooth for
any RF excitation flip angle and involves only longitudinal magnetization components.
However, for SSFP sequences where TR [Lt ] T2 and the flip angle is large, the approach to equilibrium is rather
complex and involves a mixture of transverse and longitudinal magnetization components that regain coherence at
every RF excitation, making the approach highly oscillatory with large flip angles (Fig. 7-11). The behavior of the
spins during the approach to steady state depends on the initial magnetization, the transient behavior between the
application of RF pulses (i.e., the T1 regrowth and T2 dephasing rate-which in turn depends on the field
inhomogeneity), the flip angle of the RF pulses, TE, and TR.
Initial attempts to drive spins into steady-state used an alpha/2 pulse played out at a time TR/2 before a train of
alpha pulses.48 In this case, the magnetization vector approximately follows an exponential decay toward the
49
desired steady-state conditions. However, this decay requires as many as 40 to 50 RF pulses to achieve steady-
state, which adds an undesirable amount of scan time to 2D exams. Nishimura has shown that the application of a
linearly increasing flip angle can achieve steady state magnetization in 10-15 pulses.50 Furthermore, the linear ramp
has shown to be less sensitive to the small off resonance effects that are responsible for artifact. The linear ramp
approach to steady-state has proven useful in cardiac MR imaging51 where prolonged magnetization
preparation-time could prohibit the use of SSFP acquisitions.
Three-Dimentional Gradient-Echo
Three-dimensional GRE acquisitions are routine. They are used for numerous clinical applications, including
contrast-enhanced MR angiography, brain imaging, and musculoskeletal imaging. The benefits include the ability to
acquire thin slices with adequate SNR. As addressed in Chapters 3, Practical Considerations and Image
Optimization and 23, Image Processing: Principles, Techniques, and Applications, the acquisition of thin, continuous
slices permits the application of various image processing techniques, including maximum intensity projection,
volume rendering, and multiplanar reformation.
page 199
page 200
Add to lightbox
16 of 23 12-04-2010 00:11
Add to lightbox
Figure 7-11 Approach to the steady state for spoiled and steady-state coherent gradient-echo (GRE) sequences.
Calculated signal intensity as a function of excitation number for TR/FA = 15/20°. A, Spoiled GRE sequence. Note
the smooth approach to the steady state. B, Steady-state coherent GRE sequence. Note the marked oscillation in
signal intensity as a function of excitation number, particularly for long-T2 tissues such as blood and cerebrospinal
fluid.
Slice Profile Effects in Fast 3D Gradient-Echo Imaging

With short TR values as used in most 3D GRE imaging, the profile of the 3D slab may be nonrectangular (i.e. the
flip angle is not uniform across the slab). The problem is compounded by the fact that RF pulses of short duration
(e.g. 1 to 2 msec) are often used for fast imaging. These short duration RF pulses have poor slab profiles
compared with lengthier pulses.
page 200
page 201
The degradation in the slice profile is most severe when the TR is short and the flip angle is large. Near the center
of the 3D slab, the flip angle is correct, but the edges of the slab experience smaller flip angles. Using a spoiled
GRE sequence, one finds that partitions (i.e. slices from a 3D acquisition) near the center of the 3D slab experience
the nominal flip angle and show the expected T1 contrast, whereas partitions near the edges of the slab experience
smaller flip angles and show less T1 contrast. (Additionally, a dynamic transformation of the slice profile occurs for
fast GRE during the approach to steady state.52 However, this additional problem is avoided by the use of
preparation scans.)
Turbo FLASH and MP-RAGE

With short TR values (e.g., <10 msec), GRE scan times less than 1 second are feasible. However, these images
show poor contrast and are best used as localizer scans. The addition of an inversion recovery (IR) magnetization
preparation alters tissue contrast so that the images can be clinically useful. This approach is called snapshot
FLASH,53 turbo FLASH, or, for 3D acquisitions, MP-RAGE.
In using turbo FLASH, one can make a particular tissue appear bright (e.g., liver parenchyma) and another appear
dark (e.g., metastasis) by choosing the proper inversion time (TI). There is an important difference from inversion
recovery-prepared fast SE. With turbo FLASH, where the data are obtained over a period of up to a second for 64
to 128 phase-encoding steps, it is obviously impossible to acquire the entire image at the zero crossing of the
longitudinal magnetization.
For instance, consider an IR turbo FLASH sequence with a TR of 7 msec and 128 phase-encoding steps. At 1.5 T,
the signal intensity from a metastasis might be minimized at a nominal TI of 700 msec. Using a standard phase-
encode order, the effective TI (i.e., the time from the 180° pulse to the central phase-encoding step) of this
sequence is equal to the nominal TI of 700 msec + (7 msec × 64 steps) = 1148 msec. On the other hand, it would
not be possible to null tissues having a short T1, such as fat or liver, with this sequence. Even with a nominal TI of 0
(effective TI of 448 msec), the longitudinal magnetization of these tissues would already have recovered well past
17 of 23 12-04-2010 00:11
zero by the time the central lines were acquired. This problem can be overcome by reordering the phase-encoding
54
steps to acquire the central lines first. For instance, the so-called centric approach acquires the steps as follows:
0, 1, -2, 2, -2, 3, -3,[mldr ].
Multishot or Segmented GRE

The data acquisition can be divided into groups or segments of phase-encoding steps. The advantage of
segmentation is that the data are acquired over a shorter time period. For cardiac imaging, the result is less blurring
from cardiac motion. Typically, 4 to 12 steps are acquired in each segment. With an inversion recovery preparation,
the method is ideal for studies of myocardial viability since it minimizes blurring from cardiac motion as well as
strong T1 contrast for detection of delayed enhancement (see Chapter 35, Myocardial Viability).
Miscellaneous Rapid Imaging Sequences

BURST
BURST is so named because it applies a burst of brief RF pulses. 55-57 Multiple echoes are generated, separated by
the interval that separates the RF pulses; each echo can be individually phase encoded. The data acquisition speed
is nearly as fast as for echo-planar imaging (EPI) but without the need for high-speed gradients. The whole brain
can be imaged using a 3D BURST sequence in just a few seconds. Moreover, the pulse sequence is very quiet
since rapid switching of the gradients is unnecessary.
URGE
Ultrarapid gradient-echo (URGE) uses a string of short, low-flip-angle nonselective RF pulses during the application
of a constant, negative gradient and generates a train of GREs that are equal in number to the RF pulses applied
using a constant, positive readout gradient, as in conventional GRE imaging. URGE departs from other
concepts-such as utilized in BURST, SUNBURST,58 optimized ultrafast imaging sequence (OUFIS),59 DANTE
60 61
ultrafast imaging sequence (DUFIS), or quick-echo split imaging technique (QUEST) -in which SEs rather than
GREs are acquired, so that one obtains maximal spin utilization and consequently better SNRs. Images can be
generated as quickly as in EPI but do not require the stringent gradient hardware necessary to achieve the
performance of the latter, because the data are acquired during a constant gradient readout. An interleaved
trajectory62 produces URGE images of good quality with small effects from field inhomogeneities, T2* decay, or
diffusion and flow. Only a 3D version of URGE has been implemented because of the nonselective nature of the RF
excitation process. Magnetization-prepared URGE has also been demonstrated with strong T1 weighting.
Fast STEAM
63,64 65
A stimulated-echo acquisition mode (STEAM) can be used for fast imaging. The magnetization is stored along
the - z axis using two 90° RF pulses and is then read out using a series of independently phase-encoded
low-flip-angle excitations. Subsecond acquisition times are possible with this approach. The images are T1
weighted according to the length of the interval between the second RF pulse and the central phase-encoding step.
Flowing blood is effectively dephased with STEAM imaging, so that it appears uniformly dark. Fast STEAM imaging
has been proposed for cardiac and perfusion imaging.66 However, image quality is less than can be obtained by
other fast imaging methods, so the method has not come into widespread use.
Echo-Planar Imaging
page 201
page 202
Add to lightbox
18 of 23 12-04-2010 00:11
Add to lightbox
Figure 7-12 The k-space trajectories for single-shot echo-planar imaging (EPI). Comparison of blipped (left) and
constant (right) phase-encoding.
Echo-planar imaging, proposed by Mansfield and Pykett in 1977,67 is a fast imaging technique that allows one to
collect all the data required to reconstruct an image in a brief interval, as short as the duration of a single readout
68-70
period (approximately 32 to 100 msec). With image acquisition times on the order of one tenth of a second, EPI
represents the fastest clinically useful imaging technique. A number of new clinical applications have become
feasible with the introduction of EPI, with the greatest use being for functional brain imaging (including imaging of
perfusion, diffusion, and cortical activation).
Single-Shot Echo-Planar Imaging

An EPI image can be acquired in a single measurement or shot, with all the data collected after one RF excitation,
or as multiple shots using repeated RF excitations, as discussed later. The single-shot approach differs from
standard SE and GRE sequences, with which only a portion of the data is collected after each RF excitation.
Echo-planar data are collected during the application of a rapidly oscillating gradient along the readout (frequency-
encoding) direction. With each reversal of the gradient polarity from positive to negative and vice versa, a GRE is
produced that is independently phase encoded, resulting in the generation of a train of echoes. One can use a
terminology for EPI similar to that used in fast SE imaging with respect to the echo train. The image acquisition time
is simply the duration of the echo train, typically from 30 to 150 msec. The acquisition is faster than for single shot
fast SE (e.g., HASTE, SSFSE).
Add to lightbox
Figure 7-13 Pulse sequence diagram for single-shot gradient-echo echo-planar imaging using a two-echo phase
correction to suppress n/2 artifacts. The correction data are acquired without phase-encoding immediately after the
90° RF excitation. A blipped phase-encoding approach is illustrated, but a constant phase-encoding gradient could
also be used. (Gp, phase-encoding gradient; Gr, readout gradient; Gs, slice-selection gradient; s(t), MR signal.)
Echo-planar imaging collects data in a markedly different way from standard pulse sequences, but differences exist
even among various EPI implementations (e.g., blipped versus constant phase-encoding; single shot versus
multishot). Figure 7-12 shows k-space trajectories for single-shot EPI. With single-shot EPI, the need for a TR
interval between data collections is eliminated. The TR is essentially infinite, eliminating any degree of T1 weighting.
T2 contrast is thereby improved, which is helpful for characterization of long-T1, long-T2 lesions such as cysts and
hemangiomas.
Radiofrequency Excitation for Echo-Planar Imaging
19 of 23 12-04-2010 00:11
Virtually any combination of RF pulses used in conventional pulse sequences can be used in EPI (Figs. 7-13 and
7-14). For instance, a 90° pulse and 180° RF pulse can be applied so that the EPI data are collected under the
T2-dependent decay envelope of an SE signal (SE EPI). Alternatively, a single RF pulse can be applied with a flip
angle of 90° or less so that the EPI data are collected under the T2*-dependent decay envelope of a GRE signal
(GRE EPI). Spin-echo EPI is routinely used for diffusion imaging.
Add to lightbox
page 202
page 203
Figure 7-14 Diffusion-weighted spin-echo echo-planar imaging sequence. A pair of diffusion gradients (GD) of
equal area and separated in time by ∆ are applied around the 180° RF pulse to dephase and cause signal loss
from diffusing protons but not from stationary spins.
Echo Planar-Capable Gradient Systems

To keep the duration of the echo train as short as possible (necessary because of signal loss resulting from T2*
decay), the ramp time (i.e., time from zero to maximum amplitude and vice versa) for the readout gradient must be
extremely short. For instance, if the gradient ramp time is 250 μs, one readout gradient pulse can be applied in as
little as 0.5 msec, and 128 echoes could be collected in 128 × 0.5 msec or 64 msec.
Add to lightbox
Add to lightbox
Add to lightbox
Figure 7-15 Uses of high slew rates. A, Standard gradient systems typically have gradient ramp times on the order
of 1 msec from zero to peak amplitude. B, Systems capable of echo-planar imaging have much higher slew rates,
20 of 23 12-04-2010 00:11
so that the duration of the gradient ramps can be shortened. This technique can be used to reduce TE without
increasing the signal bandwidth compared with that in A. C, Alternatively, the use of faster ramps allows the plateau
sampling period to be extended, thereby lowering the signal bandwidth and improving the signal-to-noise ratio
without increasing the total duration of the frequency-encoding gradient waveform compared with that in A.
Not only must the gradient rise times be several times shorter than for standard gradient systems, the gradients
must also be capable of reaching higher peak amplitudes. This is because the spatial resolution is determined by
the integral of the gradient waveform. Because the duration of each readout gradient pulse is so short, its amplitude
must be correspondingly high. The gradient performance may be characterized in terms of slew rate (Fig. 7-15),
which is the gradient magnetic field change per unit time, given in tesla per meter per second. For instance, for
trapezoidal gradients, a slew rate of 200 T/m/s could mean that the gradient rise time is 0.1 msec for a peak
gradient amplitude of 20 mT/m or 0.05 msec for a peak gradient amplitude of 10 mT/m. The definition of slew rate
is trickier for sinusoidal gradients, as the slew rate varies over the duration of the sinusoid. Potential advantages of
high slew rates include: 1. shorter TE for non-EPI applications; 2. reduction in the echo train duration, so that
susceptibility and T2* filtering effects are reduced; and 3. higher spatial resolution, because the gradient can be
ramped more rapidly. This allows more time at the plateau of peak gradient amplitude and a larger area under the
gradient. Drawbacks of high slew rates include greater risk of neuromuscular stimulation and more stringent
hardware requirements. With the use of high-slew-rate gradient pulses, good eddy current compensation is
essential to avoid image artifacts. Parallel imaging techniques (see Chapter 8, Parallel Imaging Methods) are
particularly beneficial for reducing artifact and improving spatial resolution for EPI.
Phase-Encoding in Echo-Planar Imaging

Because the data for EPI are collected as a series of very fast gradient-echoes, the techniques for phase-encoding
are different from those used for standard imaging. One of two methods for phase-encoding is generally used: 1. a
constant phase-encoding gradient is applied over the entire duration of the EPI readout; or 2. a phase-encoding
"blip" of short duration is applied at the end of each readout gradient pulse. The trajectory along the phase-encoding
direction is slightly different for each approach and is treated differently during the image reconstruction.46
Echo-planar images can also be acquired in a volumetric mode by applying a standard phase-encoding gradient
along the section-select direction, with much shorter image acquisition times (as little as 10 seconds for 128
partitions) than for standard 3D GRE sequences. However, the artifacts are then more severe than with 2D.
Artifacts in Echo-Planar Imaging

N/2 Ghosts
page 203
page 204
Ghost artifacts are typically caused by periodic fluctuations in image signal intensity. Because single-shot EPI data
are accumulated within an interval that is short compared with periods of physiologic motion, the images do not
suffer from ghost artifacts caused by breathing or cardiac pulsation. However, they can suffer from ghost artifacts
of a different kind. Odd and even EPI echoes are collected with readout gradients of alternating polarity.
Imperfections in the gradients, gradient-induced eddy currents, flow, or field inhomogeneities may cause slight
mismatches in the timings of the odd and even echoes, which translate into phase errors after the Fourier
transform. These phase errors are manifested in EPI as duplicate images of the object, called N/2 ghosts that
propagate along the phase-encoding direction. If N/2 ghosts are severe, they can degrade image quality. To
suppress N/2 ghosts, care must be taken to ensure gradient stability and minimize eddy currents. The N/2 ghosts
can also be suppressed by a phase correction determined from two echoes collected during a single bipolar
readout gradient (or three echoes).
Susceptibility and Chemical-Shift Artifacts

The amplitude of the phase-encoding gradient for EPI is much weaker than for standard pulse sequences. As a
result, EPI is sensitive to the effects of static magnetic field inhomogeneities (susceptibility and chemical shift
artifacts); however, the sensitivity is directed along the phase-encoding rather than the frequency-encoding
direction, as is typical of standard pulse sequences. Susceptibility artifacts may be severe at the skull base, near
the paranasal sinuses, at the anterior orbits, at the lung-heart interface, and near air-containing bowel loops.
Because chemical-shift artifacts with single-shot EPI are many times larger than with standard sequences, efficient
fat suppression is essential (Fig. 7-16).
Susceptibility artifacts are reduced by proper shimming. Susceptibility artifacts are further reduced by using thin
sections, a short TE (which can be achieved by reducing the ETL or by asymmetric k-space sampling along the
phase-encoding direction), or augmenting the amplitude of the phase-encoding gradient (e.g., using a rectangular
field-of-view; FOV). The TE also can be shortened by increasing the readout bandwidth (i.e., reducing the duration
of echo sampling), but this approach is ultimately limited by the gradient capabilities of the system.
21 of 23 12-04-2010 00:11
T2* Filtering
Add to lightbox
Figure 7-16 Abdominal echo-planar image acquired without fat suppression. Note the enormous chemical-shift
displacement of the subcutaneous fat (arrows).
Data collection with EPI is fast relative to conventional MRI techniques and to periods of physiologic motion; it is not
fast with respect to T2* decay. T2* decay during the echo train causes blurring (filtering of high spatial frequencies)
71
along the phase-encoding direction because the signal intensity changes substantially over this period. With
shorter T2* the blurring is more severe. This filtering effect can be minimized by shortening the echo train and TE.
In-Plane Spatial Resolution

In-plane spatial resolution along the x and y directions is determined, respectively, by the integrals of the readout
and phase-encoding gradient waveforms. Because the echo train must be kept short to minimize T2* decay and
resultant blurring, the FOV of single-shot EPI is ultimately limited by the peak gradient amplitude. As a result, the
minimal FOV for EPI using whole-body systems is on the order of 25 cm or greater for a 128 × 128 acquisition
matrix, giving a typical pixel size of 2 to 3 mm. Spatial resolution can be improved by use of small gradient insert
coils, parallel imaging, or multishot approaches.
Multishot Echo-Planar Imaging

One can substantially overcome the FOV limits and susceptibility artifacts that occur with single-shot EPI by
accumulating the data over multiple shots (each shot corresponds to one RF excitation). This procedure decreases
the ETL and duration of echo train in proportion to the number of shots. Multishot EPI greatly reduces the demands
on gradient performance and allows in-plane spatial resolution to be improved to a level comparable to that with
standard pulse sequences. There are several possibilities for multishot imaging. For instance, one could simply
acquire half of k-space in the first shot and concatenate the other half acquired in the second shot. When done
along the frequency-encoding direction, this is called mosaic imaging. 72 Half-Fourier imaging can also be used to
improve spatial resolution, at the expense of a loss in SNR and some increase in artifacts. Alternatively, a limited
portion or segment of data (e.g., 8 to 32 lines) is collected within each TR interval. To smooth out phase and
amplitude modulations in the raw data, which otherwise introduce artifacts into the image the technique of echo
shifting is used. With this technique, the timing of the echoes is varied slightly from shot to shot.73,74 Multishot EPI
75,76
images of diagnostic quality can be acquired in a single breath-hold period.
Signal-to-Noise Ratio in Echo-Planar Imaging

page 204
page 205
1/2
The SNR is proportional to 1/(readout bandwidth) . The duration of each GRE readout in the EPI echo train is
many times shorter than that of the readout of a conventional pulse sequence; therefore, the readout bandwidth is
correspondingly higher and the SNR lower. Despite this limitation, the SNR of single-shot EPI images can be
surprisingly good. One reason is the limited in-plane spatial resolution, as the SNR scales with the pixel area. For
instance, a 128 × 128 or 64 × 128 acquisition matrix is commonly used for EPI with pixel dimensions on the order of
2 to 4 mm, versus 1 to 2 mm with standard sequences. Furthermore, at least for single shot EPI, all echoes are
collected after a single RF excitation. Although the SNR of EPI images is lower than that of images acquired using
standard pulse sequences, the SNR per unit acquisition time is higher. An interesting feature of EPI is that image
quality may improve (up to a point), rather than worsen, as the section thickness is reduced. This is because T2*
22 of 23 12-04-2010 00:11
decay caused by intravoxel variations in magnetic susceptibility is reduced as the section thickness is decreased.
Safety of Echo-Planar Imaging

The RF power deposition with EPI is low because few RF pulses are applied. The gradients are another issue
entirely. A change in the magnetic field (B field) can create an electric field (E field) directly proportional to the rate
of change of the B field, thus giving rise to electric currents when the changes occur in a conductive surrounding.
The fast-changing magnetic field gradients needed for EPI can induce rapidly changing electric currents in the body
that are potentially hazardous. These currents can cause neuromuscular stimulation and have raised concerns about
the safety of EPI for human studies. Because of the different geometries of the three gradient coils, the stimulation
hazard is worst when the z gradient is used for readout (i.e., for sagittal or coronal EPI imaging). Swapping the
phase- and frequency-encoding gradients for an EPI axial acquisition increases the risk of neuromuscular stimulation
compared with an unswapped axial acquisition because of greater stimulation when the y gradient is used for
readout compared with the x gradient. Sinusoidal gradients may have a lower threshold for stimulation than
trapezoidal ones.77,78 Safety issues are reviewed in Chapter 24, Bioeffects, Safety, and Patient Management.
Studies have reported sensations ranging from mild twitching to pain, depending on the maximal change in B over
79,80
time when the magnetic field gradients oscillate during EPI data collection. The sensation is felt near the
physical extremes of the gradient coil, where the gradient field change (and therefore dB/dt) is largest. For
instance, the trunk muscles or buttocks are typically affected when the head is positioned at the center of the
magnet for brain imaging, because they lie near the edge of the magnet bore and thus the ends of the gradient
coils. Similarly, the bridge of the nose may be affected during abdominal imaging. Fortunately, the cardiac
stimulation threshold is at least 10-fold higher than for skeletal muscle.
The risk of stimulation can be reduced by shortening the gradient coil, which reduces the peak dB/dt at the ends of
the coil. Figure 7-17 shows field plots for long and short z gradient coils. A disadvantage of a short gradient coil is
that spatial distortions are seen for positions far off center. These distortions can be corrected, at least in part, if
the gradient field distribution is known.81
© 2010 Elsevier
23 of 23 12-04-2010 00:11
TIME-RESOLVED IMAGING
Add to lightbox
page 205
page 206
Figure 7-17 Field plot of a z-gradient coil. A long gradient coil has a higher dB/dt at its ends than does
a shorter gradient coil.
Add to lightbox
Figure 7-18 Illustration of raw data (k-space) showing the relationship of the central lines (weakest
phase-encoding gradient amplitudes and strongest signal) to image contrast and of the outer lines to
image detail. For a complete image, all (kx, ky) points must be covered by the acquisition.
It is obvious that an MR image must depict anatomy with high spatial resolution to be clinically useful.
However, it can also be the case that such static images do not by themselves provide all the
information that is needed and some level of temporal resolution may be required in addition.
Time-resolved imaging has proven to be an essential tool in several clinical applications, including
evaluation of bowel peristalsis, fetal motion, dynamic contrast-enhanced perfusion MRI in suspected
stroke, contrast agent kinetics in breast tumors, identifying the true and false lumens of an aortic
dissection, distinguishing arterial and venous pathology in the setting of rapid arteriovenous transit, and
measurement of transit times.82,83,84 In order to facilitate time-resolved imaging, new techniques have
been developed that either speed up the actual data acquisitions, such as echo planar and single shot
fast SE (reviewed earlier in this chapter), or reduce the amount of data that are acquired. The latter
approach is called "undersampling". In the following section, we will focus on undersampling techniques
using both Cartesian and non-Cartesian k-space trajectories and methods to minimize artifacts
1 of 18 12-04-2010 00:12
associated with them.85,86 Although parallel imaging techniques87,88 also use undersampling, they
involve a substantially different set of principles and are reviewed separately in Chapter 8.
k-Space and Undersampled Cartesian Imaging

A useful concept for understanding advanced imaging techniques is k-space.89 The raw data (i.e., the
data as collected before the 2D Fourier transform) can be depicted as a 2D matrix of spatial
frequencies (kx, ky) (Fig. 7-18). In generic form,
where γ is the gyromagnetic ratio and G(t) is the time-dependent gradient amplitude. Various parts of
k-space influence the appearance of the image in different ways. Because magnetic field gradients
cause dephasing, the MR signal is strongest when the weakest phase-encoding gradients are applied.
The center of k-space consists of data acquired during the weakest phase-encoding gradients.
Because the signals are so large in this part of k-space, these data control image contrast. The
weaker signals in the outer part of k-space (strongest phase-encoding gradients) contain the high
spatial frequencies that determine image detail (Fig. 7-19). The manner in which the various spatial
frequencies are collected determines the k-space trajectory. The MR data are collected to encode
both low and high spatial frequencies (i.e., the image should have good tissue contrast as well as high
spatial resolution). Independent of the acquisition matrix size selected, the relative speed at which
k-space is covered determines the time savings possible with a particular MRI technique.
Add to lightbox
Add to lightbox
2 of 18 12-04-2010 00:12
Add to lightbox
Figure 7-19 Raw data. A, The amount and manner in which data are acquired can dramatically alter
the appearance of the MR image. The raw data (left) are converted into an image (right) by the Fourier
transform. B, With removal of the outer lines of the raw data (high spatial frequencies), details are lost
but contrast is preserved. C, With removal of the central lines of the data (low spatial frequencies),
image contrast is lost, leaving a ghostly image showing tissue boundaries only.
The k-space trajectory for any pulse sequence must span all the (kx, ky) points to generate a complete
image. For instance, with standard GRE or SE sequences, one phase-encoding step is applied and
one line of data is collected (i.e., all kx points for one value of ky) after each RF excitation. The k-space
trajectory thus consists of a series of horizontal lines as shown in Figure 7-20. This is repeated as
many times as necessary to complete the full raw data. All frequency samples in k-space are acquired
in each readout period, corresponding to one pulse of the readout gradient. At the other extreme,
advanced sequences using the EPI encoding, for example, use a strong bipolar oscillating gradient
field to generate a train of gradient-echoes that are phase encoded independently after the application
of a single RF excitation, thus collecting the full k-space matrix in a single shot.
Partial Fourier
page 206
page 207
Add to lightbox
Figure 7-20 The k-space trajectory for standard gradient-echo (GRE) sequence. Data are acquired
one line at a time for each TR interval.
When the through-plane resolution of a 2D projection image exceeds a few millimeters, intravoxel
dephasing of signal reduces the image contrast. Therefore, in many regions of the anatomy, 2D
projection images are not an option and 3D images must be acquired. Since imaging 3D volumes
requires an order of magnitude more imaging time than 2D projection techniques, the need for
time-resolved MR acquisitions has prompted the use of k-space undersampling. A simple technique to
reduce the imaging time while minimizing the loss of resolution is to acquire only a fraction of the full
3 of 18 12-04-2010 00:12
set of k-space data. If half of the full set of MR image data is acquired, the image acquisition time is
cut in half.
Normally, the phase-encoding gradient is gradually increased from large negative amplitude, to zero, to
large positive amplitude (or vice versa) over the course of an acquisition. Using a partial-Fourier
acquisition, fewer data are collected and the remaining data are mathematically synthesized. In the
half-Fourier version, for instance, only the positive phase-encoding steps are acquired, along with the
86,90,91
zero step and a few negative ones. A mirror image copy of the positive data is created, and a
complete image is reconstructed from the acquired and synthetic data (Fig. 7-21). As only a little more
than half the data is acquired, scan time is reduced nearly two-fold. In practice, half-Fourier techniques
reduce scan time by slightly less than 50%, because the mirror image symmetry of the data is only
true for a hypothetic "ideal" acquisition. The actual MR data have phase offsets that distort the
symmetry of the matrix. Most half-Fourier techniques require the acquisition of some extra negative
phase-encoding steps to calculate phase corrections for the data (Fig. 7-22).
Partial-Fourier is a robust and routinely useful technique for time-resolved perfusion imaging and MRA.
Moreover, it is commonly used in conjuction with other fast imaging techniques such as HASTE and
balanced SSFP. For instance, with half-Fourier, the duration of the HASTE echo train and hence the
scan time is reduced by nearly half. Moreover, the nominal echo occurs near the beginning of the echo
train, so that relatively short TE can be acquired (i.e., 25 msec or less) even though the duration of the
echo train is a few hundred milliseconds.
There are several drawbacks to the half-Fourier method. The extra steps required for image
processing increase image reconstruction time. Other drawbacks include a reduction in SNR, which
may or may not be acceptable depending on the application, increased sensitivity to motion artifacts,
and image distortions caused by RF or magnetic field inhomogeneities.
Partial-Fourier methods work best when the echo is roughly symmetrical. Symmetrical echoes are
typical for SE and balanced SSFP gradient-echo, whereas the echo may be highly asymmetric for 3D
gradient-echo sequences used for MR angiography. Nonetheless, partial-Fourier methods will still be
acceptable so long as the TE is kept short (e.g., <2 msec). In some situations, the amount of
reconstruction artifact with half-Fourier is not clinically acceptable. The use of a ¾ (rather than ½)
Fourier acquisition will minimize the reconstruction artifact. In our experience, the amount of artifact is
often worse at higher field strengths (e.g., 3 T).
Add to lightbox
page 207
page 208
Figure 7-21 The k-space representation of conjugate synthesis in half-Fourier imaging. Compared
with a conventional image, only slightly more than half of the data are acquired. The rest are
synthesized from the acquired data, reducing imaging time almost by a factor of two.
4 of 18 12-04-2010 00:12
Add to lightbox
Add to lightbox
5 of 18 12-04-2010 00:12
Add to lightbox
page 208
page 209
Figure 7-22 Half-Fourier gradient-echo (GRE) images acquired with two extra negative projections (A)
and 16 extra projections (B). Note signal inhomogeneity in A, caused by spurious phase shifts. This
inhomogeneity is reduced by acquiring extra projections (B). Successful use of half-Fourier imaging
for GRE sequences requires a short TE to minimize sensitivity to susceptibility-induced phase shifts.
C, Comparison of standard SE images (left) and half-Fourier images (right) acquired in almost half
the time (eight extra projections). TR/TE = 3500/60. Image contrast and resolution are identical.
Despite the lower signal-to-noise ratio of half-Fourier images, images have comparable diagnostic
utility.
6 of 18 12-04-2010 00:12
Add to lightbox
Figure 7-23 The concept of partial-Fourier image acquisition has been extended to three dimensions.
A, Nominally, each point in a 3D k-space volume is acquired to form an MR image. These are shown
as filled points in the 3D Cartesian grid. B, Partial-Fourier acquisitions applied in 3D acquire the full
k-space data in octant only (filled points). The unacquired data (empty points) are filled with zeroes
prior to image reconstruction.
More robust partial-Fourier methods that are less sensitive to rapid local phase changes have been
developed, using an iterative approach to reconstruct the missing portion of k-space.92 Such methods
are more time-consuming for image reconstruction than partial-Fourier methods. Moreover, interest
has waned with the advent of alternative means to accelerate the data acquisition, such as parallel
imaging.
Partial-Fourier reconstructions can also be applied along the frequency-encoding direction in

conjunction with asymmetric sampling of the echo. Asymmetric sampling allows a reduced readout
period and thus a shorter TE, at the expense of increased blurring. By synthesizing the missing portion
of the data, half-Fourier reduces this blurring. Further reductions in imaging time may be possible using
constrained reconstruction techniques.93,94
When undersampling on all three axes (kx, ky, kz) is implemented, partial-Fourier acquisitions are
capable of exceedingly rapid image acquisitions. 95 When undersampling on three k-space axes, the
sampled data is fully sampled in one octant of k-space (Fig. 7-23). This has proven useful in evaluating
arteriovenous fistulae and other vascular abnormalities that cannot be dynamically imaged with other
3D MRA techniques. In one implementation, 3D images are acquired in less than 1 second per
image.96 Figure 7-24 shows a subsecond MRA exam acquired in the abdominal aorta. Undersampling
on all three axes allows the acquisition of 3D volumes in less than 1 second. In this case, the filling of a
false lumen is demonstrated.
Keyhole Imaging
7 of 18 12-04-2010 00:12
Add to lightbox
page 209
page 210
Figure 7-24 Partial-Fourier acquisition in 3D allows for imaging dynamic processes with subsecond
temporal resolution. A series of time frames, selected from a series of images acquired demonstrate
the filling of left anterior oblique (LAO) dissection using contrast-enhanced MRA. (Courtesy of Dr.
James Carr MD, Northwestern University.)
In many cases, the observation of contrast agent uptake in several organs is required after a rapid
bolus injection to observe deficits in tissue perfusion. Trading off coverage and resolution against time
resolution is sometimes not a desirable choice, depending on the time scale of the dynamic process
that is to be observed. Based on the fact that the center of k-space is responsible for the contrast in
the final image (the outer portion of k-space determines the finer detail), the keyhole technique
acquires only the low-spatial-frequency information during the dynamic process, thus reducing the
imaging time, and forms a composite data set that uses the high-spatial-frequency information
acquired from a base image before the dynamic acquisition to permit the reconstruction of a
97,98
high-resolution dynamic data set. The keyhole concept can be applied with any imaging sequence,
although it is commonly used with fast GRE sequences because the low-spatial-frequency information
can be obtained quickly.
A drawback of the keyhole technique is that the objects within the FOV of interest must remain
stationary during the experiment; otherwise, the high-spatial-frequency k-space data collected in the
high-resolution image do not match the corresponding low-spatial-frequency information and cause
blurring and spatial misregistration of the fine detail. The number of phase-encoding lines replaced
determines the size of the features that exhibit the changes. In general, a loss in contrast-to-noise ratio
is apparent in small enhanced structures and the quantification of the effects may be erroneous (small
structures have energy across the entire k-space data).99 Furthermore, dramatic changes in the signal
intensity between the full k-space data and the segment of data replaced can create significant signal
discontinuities that can lead to increased blurring and ghost artifacts along the phase-encoding
direction (as in cases in which a susceptibility-weighted sequence is used to obtain brain perfusion
maps).
Other methods have been explored that do not employ data-replacement strategies to maintain the
high spatial resolution with the reduced raw data collected dynamically. Instead, the higher resolution
image is utilized to impose constraints on the reconstruction of the high-spatial-frequency features of
the dynamic image. One of these methods, reduced encoding by generalized series reconstruction
(RIGR), uses a generalized series model to maintain data consistency when combining the dynamic
and the high-resolution reference data sets.100 This method has an advantage over keyhole imaging in
that the edge information and contrast in small features are retained at the resolution of the reference
image.
8 of 18 12-04-2010 00:12
TRICKS
Add to lightbox
Add to lightbox
Figure 7-25 Two images from a time-series of contrast-enhanced 3D time-resolved imaging of
contrast kinetics (TRICKS) images of the pulmonary arteries. The use of temporal undersampling
allows for improved spatial resolution without reducing the frame rate of the acquisition. A, Here the
subsegmental branched vessels are seen (arrows) in the arterial phase image. B, In a later time frame,
the pulmonary arteries are obscured by veins, but the aorta and renal arteries are well depicted.
The limitations of keyhole imaging have prompted the development of a pulse sequence that updates
both low- and high-spatial-frequency k-space views. This sequence, time-resolved imaging of contrast
kinetics (TRICKS), updates the low-spatial-frequency k-space lines at a higher frame rate than the high
spatial frequency lines.101 Since the high-spatial-frequency data are only acquired intermittently,
high-resolution images are formed through temporal interpolation of the peripheral k-space data.
TRICKS has been shown to be useful in several anatomic regions because it provides high-frame-rate
3D images. It has been particularly useful in anatomic regions such as the carotid arteries102 and
pulmonary vessels, where rapid artery to vein transit times exist. Figure 7-25 shows two phases of a
multiphase 3D MRA of the pulmonary vasculature. The use of TRICKS allows the imaging of distal
pulmonary branches without the need for dose-time scans. TRICKS has also been successfully applied
to imaging of the peripheral vasculature,103 where unpredictable contrast arrival time has prompted the
used of multiple dose-time exams.104 However, like other keyhole techniques the patient must remain
still during the entire duration of the scan.
The very nature of time-resolved acquisitions opens up a new dimension in the analysis of images, that
of the temporal domain. In addition to the spatial frequency or k-space dimension, time-resolved MRA
acquisitions possess temporal frequency information that may be exploited to improve image quality. In
one such method called UNFOLD, the wraparound artifact is given a well defined temporal frequency.
105
This results in the image artifact obtaining a high-frequency temporal "flicker". By filtering these
9 of 18 12-04-2010 00:12
images in the temporal-frequency domain, the artifact can be effectively removed from the image
without compromising spatial resolution.
A temporal filter has also been used in combination with parallel imaging reconstruction (see Chapter
8, Parallel Imaging Methods). One limitation of parallel imaging is that imperfections in sensitivity-
encoded reconstruction can result in residual wraparound artifact. Kellman et al106 derived an encoding
scheme called TSENSE that alternates the undersampled phase-encoding, which is central to the use
of parallel imaging. They intentionally imparted a high-frequency "flicker" to the residue reconstruction
artifact; then, by applying a bandpass temporal filter, the high-frequency artifact is removed. This
introduced a temporal blurring to the time series of images, but, since the physiologic changes in the
images are of lower frequency, the effect of the temporal filter is minimal.
page 210
page 211
Add to lightbox
Figure 7-26 A, Spin warp MR image acquisitions sample the k-space representation of an image on a
rectilinear grid. B, Projection-reconstruction acquisitions sample k-space on radial trajectories that
pass through the center of k-space and resemble the spokes of a bicycle wheel. These trajectories
sample both the low spatial frequencies (responsible for image contrast) and high spatial frequencies
(responsible for image detail) in every TR. Undersampling in the radial dimension does not cause
wrap artifact or reduced spatial resolution as in spin warp imaging.
Non-Cartesian Imaging
When designing an MRI pulse sequence, there is flexibility in the way in which the data are sampled
prior to creation of images. Most commercial MR scanners sample k-space directly on a 2D or 3D
Cartesian grid. Whether the data are acquired in a centric, sequential, or interleaved fashion, the data
points fall on a Cartesian grid. An advantage of the Cartesian grid is that image reconstruction can be
performed rapidly using the fast Fourier transform algorithm,107 making modern real-time imaging
possible.
The majority of the information required to form an MR image lies in the central region of k-space,
which contains the so-called low spatial frequency data. The remainder of k-space, the outer edges
containing high-spatial-frequency data, holds information which gives image detail. Concurrent
collection of low- and high-spatial-frequency data reduces the discontinuities in k-space and the
concomitant artifacts caused by, for example, patient motion or the passage of a bolus of contrast
agent.108 The need to reduce artifact in the presence of dynamic changes of the imaged object during
the scan has motivated the development of sampling schemes that eschew the conventional Cartesian
grid.
Projection Reconstruction
10 of 18 12-04-2010 00:12
One of the first k-space sampling patterns used in MRI was a 2D back projection technique109 similar
in principle to that used in CT scanning. Rather than sampling k-space on a rectilinear grid as is done
with conventional MR acquisitions, radial, or projection-reconstruction (PR), acquisitions sample
k-space on radial lines that pass through the center of k-space as can be seen in Figure 7-26. Radial
k-space trajectories sample both the center and periphery of k-space on each echo. This trajectory
has the benefit of minimizing the time between acquisition at the center of k-space and at high spatial
frequencies. The rapid update of the central region of k-space is particularly useful in time-resolved
applications, such as MRA, where image quality is strongly dependent on acquiring the central region
108,110
during peak contrast enhancement. The advantages of radial acquisitions in reducing motion
artifact are described in the section on motion correction.
The time to acquire PR images has limited their widespread use. For a given imaging matrix, the
acquisition of a PR image is π/2 (1.6) times longer than a comparable Cartesian sampled image,111,112
as a consequence of the sparse sampling density of a radial k-space trajectory. In radial sampling,
more "spokes" on the wheel are required to adequately sample the edges of k-space. Image
acquisition time can be reduced by acquiring fewer k-space samples and "violating" the Nyquist
sampling criterion for a given field of view. Undersampling in Cartesian acquisitions usually results in
the acquisition of low-spatial-resolution images, small FOVs, or highly anisotropic spatial resolution. By
way of contrast, undersampling in PR acquisitions, achieved by decreasing the number of angular
samples, does not decrease spatial resolution or FOV. Rather, decreasing the number of angular
111,113,114
samples results in a low-intensity streak artifact, similar to those seen in CT exams. The PR
undersampling artifact has proven to be unobtrusive in some types of MR imaging.
For example, in contrast-enhanced MRA, blood vessels are the dominant signal source, so the
undersampling artifact shows up as low-intensity streaks that lie away from the vessels. The
undersampling factor f is defined as:
where FOVfull is the prescribed field of view, FOVr is the artifact-free field of view, Np is the number of
sampled projection angles, and Nr is the number of readout points. In MRA exams of the lungs and
renal arteries,112,115 diagnostic quality images with angular undersampling factors as low as f = 0.39
have been acquired.
page 211
page 212
11 of 18 12-04-2010 00:12
Add to lightbox
Figure 7-27 An angularly undersampled projection reconstruction contrast-enhanced MR angiogram of
the renal arteries (TR/TE/Flip = 5.1 msec/1.3 msec/30°). In this example, 24, 3.5 mm slices and 64
projections were acquired with a 256 point readout. The angular undersampling allowed for a 60%
reduction in imaging time resulting in a frame rate of 2.56 seconds, with minimal artifact. (Courtesy of
Dr. Karl Vigen, PhD, University of Wisconsin-Madison.)
Figure 7-27 shows a contrast-enhanced 3D MR angiogram acquired using a hybrid undersampled PR

image where the undersampled PR trajectory was used to sample in-plane, and conventional partition
encoding was employed; 256 radial samples (Nr) but only 64 projections (Np) were acquired. The set
of sampled angles was alternated for each partition; this has the effect of reducing the coherence and
intensity of the streak artifact.
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
page 212
page 213
Figure 7-28 In phase contrast MR angiograms acquired with Cartesian phase-encoding, pulsatility in
blood vessels causes severe ghosting artifact. This is demonstrated in A, a pulsatile flow phantom and
B, a coronal projection phase-contrast MRA in the circle of Willis. In contrast, using projection-
12 of 18 12-04-2010 00:12
reconstruction acquisitions, pulsatility artifact appears as low-intensity streaks emanating from the
object. This is shown C, in a phantom and D, in the circle of Willis. Note the improved depiction of the
vessels and absence of ghosting in D. (Courtesy of Dr. Andrew Barger MD, PhD, University of
Wisconsin-Madison.)
Add to lightbox
Add to lightbox
Figure 7-29 Undersampled projection-reconstruction imaging produces streak artifact that emanates
from all objects in the image. A simple simulation of the artifact (A) correlates well with the observed
artifact seen in a short-axis view of the heart (B). (Courtesy of Dr. Andrew Larson, PhD, Northwestern
University.)
Three-dimensional phase contrast images can provide clinically relevant information for flow
quantification, but require prohibitively long scan times. Barger et al116 implemented an undersampled
PR version of a 3D phase contrast image acquisition. They demonstrated a 3-4 fold improvement in
scan time while maintaining the same spatial resolution acquired of a Cartesian acquisition. Figure 7-28
demonstrates how pulsatility artifact manifests in both Cartesian and undersampled PR acquisitions.
The ghosting observed in the Cartesian acquisition (Fig. 7-28A) is a result of time-varying signal
intensity caused by pulsatile flow. The PR version exhibits no ghosting but rather a radial streak artifact
(Fig. 7-28B). The differences can also be seen in the phase-contrast angiograms (Fig. 7-28C and D).
13 of 18 12-04-2010 00:12
Note that with phase-contrast MRA undersampling artifact from static tissue is suppressed through the
bipolar phase cancellation inherent in the phase-contrast image acquisition.
A disadvantage of projection reconstruction imaging is streak artifact. The acquisition of an insufficient

number of projections leads to a star-like artifact.113,116 These streaks are demonstrated in Figure
7-29 where simulated streaks are shown in an undersampled PR cardiac image. The intensity and
spatial distribution of the streaks is a function of the degree of undersampling on the periphery of
k-space. In many cases, the intensity of these streaks can be reduced through the application of
117
apodizing filters to the raw MR data prior to reconstruction.
3D Radial Projection Imaging

The examples above demonstrate radial sampling in which partitions are acquired using normal
Cartesian sampling. Radial sampling has also been applied in 3D. In this approach, the k-space
sampling line runs outward from a point in 3D at equally spaced angles. The trajectories have the
shape of a spherical pin-cushion, or toy koosh ball. The koosh ball trajectory has the advantage of
acquiring truly isotropic spatial resolution in 3D, at the expense of increased scan time. However, by
undersampling the koosh ball trajectory, the benefits of isotropic k-space coverage have been
achieved without prolonging the scan time. This sequence is called vastly undersampled isotropic
projection reconstruction (VIPR).118 Figure 7-30 shows the VIPR kooshball trajectory as well as
contrast-enhanced coronal and sagittal views of the pulmonary arteries.
Spiral Scanning
Spiral scan techniques were developed as a means to rapidly acquire MR images similar to
echo-planar imaging.119,120 Spiral imaging samples k-space on spiral trajectory (see Fig. 7-31). In
Figure 7-31A, a single spiral k-space trajectory is shown. The spiral trajectory begins at the center of
k-space and samples the lowest spatial frequencies immediately after the excitation pulses. This has
the advantage that much of the low-spatial-frequency k-space data is acquired before T2* decay or
flow phase can accrue. For this reason spiral trajectories are less susceptible to flow artifacts than
other rapid imaging techniques.119 Spiral scanning has been employed in several applications, including
fMRI of the brain, dynamic contrast-enhanced MRI of the breast, and coronary MR angiography (see
Fig. 7-31D).
Reconstruction of both projection reconstruction and spiral acquisitions is more complex than with
Cartesian sampling. Several steps are required prior to Fourier transformation.121 A complete
description of the reconstruction for projection reconstruction and spiral images is beyond the scope of
this manuscript, and has been provided elsewhere.122 Two key concepts that are central to the
reconstruction of non-Cartesian acquisitions are density compensation and regridding. Density
compensation accounts for nonuniform sampling density throughout k-space. Usually, more data points
are acquired near the center of k-space. If these images were reconstructed without density
compensation, the dominance of low-spatial-frequency data would result in images that have low
spatial resolution (i.e., blurry images). Density compensation in spiral images can be accomplished, in
part, through a judicious choice of the spiral trajectory. By slowing down the k-space velocity at large
radii, some measure of density compensation is ensured.
page 213
page 214
14 of 18 12-04-2010 00:12
Add to lightbox
Add to lightbox
Add to lightbox
Figure 7-30 Radial projection acquisitions have been extended to 3D for the acquisition of isotropic
spatial resolution. A, Each projection terminates on the surface of a sphere in 3D. B, A contrast-
enhanced MR angiogram of the pulmonary arteries. C, Isotropic spatial resolution allows for sagittal
display of the image volume to clearly depict a coarctation of the aorta. This demonstrates the high
frame rates that allow capturing of an arterial phase, and the coverage with isotropic spatial resolution.
(Courtesy of Dr. Walter F. Block PhD, University of Wisconsin-Madison.)
15 of 18 12-04-2010 00:12
Spiral scanning requires very long readouts compared with Cartesian acquisitions. The prolonged spiral
readouts become comparable to the T2*, so that the signal readout suffers from modulation. Radial
symmetric blurring of the image results-the degree depends on the tissue T2*. Furthermore, offsets in
resonant frequency also cause image artifact. The offsets can be caused by chemical shift or by slight
inhomogeneity in B0.123 Spectrally selective preparation pulses are useful in reducing the contribution of
off-resonant spins.124
As with radial projections, spiral k-space trajectories sample the central region of k-space immediately
after excitation. This minimize the effect of both T2* decay and flow effects. However traversing large
regions of k-space within the T2*-decay time can be very demanding on a scanner's gradient system.
For this reason, segmented acquisitions are usually employed. In segmented spiral acquisitions,
multiple interleaved spirals are acquired. Each spiral arm is rotated with respect to the others to avoid
redundancy in the sampling. Figure 7-31B shows two such interleaved spiral arms.
Wavelet-Encoded Imaging
Wavelet-encoded MRI is a new encoding modality that departs radically from the traditional
Fourier-based method.125,126 The Fourier transform is a global transform and the entire k-space must
be acquired to resolve images with a particular pixel resolution across the entire FOV. By using
wavelets, the resolution in a particular portion of the FOV can be selectively enhanced.127 This is
possible because wavelets are spatially localized and the spatial information is contained within each
wavelet coefficient, which encodes a specific portion of the FOV. A wavelet-based encoding technique
has been implemented in the form of a line-scanning sequence using an SE technique in which the
frequency encoding is acquired as in conventional imaging; however, the second dimension is wavelet
encoded, using a wavelet-shaped RF profile that changes from one excitation to the next based on the
Battle-Lemarie wavelet basis.128 Although the NEX required to reconstruct the same resolution across
the entire FOV is the same as in the Fourier transform case, wavelet encoding has a worse SNR.
Nonetheless, the use of wavelet encoding can be advantageous in some cases, as it has better
immunity to motion degradation and, because of the spatial selectivity within the FOV, the NEX
necessary to resolve a particular region of the FOV decreases. This feature allows one to update only
information that significantly changes over time within the FOV. Wavelet encoding can be used to
speed up the acquisition process for dynamically changing features without a loss in resolution, or to
track the motion of edge information without having to search over the entire image.129
page 214
page 215
16 of 18 12-04-2010 00:12
Add to lightbox
Add to lightbox
Add to lightbox
17 of 18 12-04-2010 00:12
Add to lightbox
Figure 7-31 Spiral k-space trajectories efficiently acquire a large fraction of k-space in a single
acquisition. Single-arm spiral (A) and interleaved spiral (B) acquisitions have been implemented.
Multiple interleaved spirals have shown to reduce T2 decay that occurs during prolonged readouts. C,
Pulse sequence diagram for spiral imaging, showing simultaneous oscillation of two gradients as well
as use of a spectral-spatial water excitation. D, Interleaved spiral coronary MR angiogram shows
uniform signal intensity of the cardiac blood pool-a consequence of the motion insensitivity of the spiral
technique. (Courtesy of National Naval Medical Center/USUHS.)
© 2010 Elsevier
18 of 18 12-04-2010 00:12
METHODS FOR MOTION CORRECTION

Patient motion leads to artifacts such as ghosting and blurring that can limit the resolution and accuracy
of an exam.130,131 Motion correction techniques aim to prevent or diminish such artifacts. The efficacy
of any technique depends, in large part, on the type of motion. Motion associated with blood flow,
circulation, and diffusion is not considered here, and the techniques used for cardiac imaging are
discussed elsewhere (see Chapter 32, Cardiac Imaging Techniques). The focus here is on techniques
designed to minimize effects due to respiration and due to gross patient motion.
Respiratory Triggering
page 215
page 216
Breath-holding and respiratory triggering are simple methods for minimizing motion artifacts in
abdominal scans, but they have limitations. Not all patients are willing or able to perform a breath-hold.
Even with a capable and cooperative patient, the scan must be performed rapidly, thus limiting the type
of scan that can be used. Slow involuntary relaxation of the diaphragm over the course of the
132
breath-hold can corrupt images. Respiratory gating is effective but can extend the duration of the
scan considerably. Breathing can also be irregular, leading to inconsistent repetition times and, hence,
unpredictable T1 weighting.
Phase-Encode Reordering
Respiratory motion leads to blurring and ghosting. Blurring is produced as tissue moves during the
scan. Ghosting arises because the periodic motion associated with breathing modulates the signal.
Although blurring affects only the moving tissue, ghost artifacts lie at different distances from the
moving tissue along the phase-encoding direction. The ghost artifact can thus degrade signal in
stationary tissue. One way to minimize the ghost artifact is to choose the acquisition order of phase-
encode values based on the point in the respiratory cycle at which they are acquired. If the phase-
encoding acquisition is properly ordered, the periodic modulation of raw image data can be reduced.
This is effective in reducing ghosting artifacts. Acquisitions are taken throughout the respiratory cycle,
making the exam faster than one based on respiratory triggering. The resulting image spares the
stationary tissue from ghost artifacts.
In respiratory ordered phase-encoding (ROPE),133 the most negative phase-encode value is taken at
full expiration while the most positive phase-encode value is taken at full inspiration. Intermediate
phase-encode values are taken so that the degree of extension with respiration increases
monotonically with phase-encode value. This approach reduces the periodic modulation of signal
among phase-encode values, thus minimizing ghosting. Related methods (centrically ordered phase-
encoding, COPE; and hybrid ordered phase-encoding, HOPE) associate full expiration with the central
region of k-space because more time is usually spent at full expiration and because most of the image
intensity lies near the center of k-space.134,135 The ordering can also be chosen to maximize the
modulation frequency, which puts artifacts out to the edge of the FOV, away from the region of
interest.136
Non-Rectilinear k-Space Trajectories

Scans using certain trajectories in k-space are inherently less susceptible to motion artifact than the
standard two-dimensional Fourier transform (2DFT, or spin-warp) acquisition. Projection-reconstruction
and spiral scans oversample the center of k-space so that effects of motion are averaged and thus
reduced.113,137 Furthermore, motion artifacts in a PR scan can be less obtrusive than those associated
with 2DFT. Motion during a PR scan leads to streaks that are arranged perpendicular to the motion,
and which have low intensity near the moving structures. The bulk of the artifact intensity may, in some
cases, lie entirely outside the region of interest. By way of contrast, the ghosting associated with 2DFT
imaging has a high intensity near the moving structure and spreads throughout the phase-encode
direction.
1 of 6 12-04-2010 00:13
Rigid-Body Motion
Motion of the abdomen is complicated because the object stretches, rather than moving "en masse";
different parts of the image move by different amounts during a scan. Movement of the head or of the
limbs can often be considered as rigid-body motion; the entire object of interest moves as a whole
without any distortion in shape. In the case of rigid-body motion, artifacts can be characterized, and
hence corrected, in a systematic fashion. Even though the head and limbs can be immobilized to a
much greater extent than the abdomen, lack of patient cooperation can make immobilization difficult.
Submillimeter movements can also corrupt images despite effective immobilization. A number of
techniques have been devised to correct for rigid-body motion. To a certain extent, they are also able
improve image quality in the presence of stretching.
View-to-View and Intraview Motion

Two time scales are important when considering the effects of motion on the image in a standard
spin-warp image. Motion can occur during the acquisition of a single line in k-space, or phase-encode
value (intraview motion), or between acquisitions of different phase-encode values (view-to-view
motion).138 The time of acquisition of a single line is typically under 10 milliseconds while the time
between acquisitions of different lines (TR) can exceed 1 second. As a result, view-to-view motion
typically causes more severe degradation of the image because larger displacements can occur over
the longer time period.
The effect of intraview motion is an accumulation of phase in the MR k-space data above and beyond
that which would be acquired in the absence of motion. This phenomenon is used in phase-contrast
MRA in order to measure the velocity of flow. Gradient-moment nulling139 (Fig. 7-32) adds additional
rephasing gradients to a standard gradient-echo in order to force the additional phase of moving spins
to zero at the echo, regardless of the velocity of the spins. Navigator echoes, described below, can
also be used to correct for intraview motion.
Effects of Translation and Rotation in k-Space

page 216
page 217
Add to lightbox
Figure 7-32 Spin-echo sequence with standard readout gradients and moment-nulled readout
gradients. The moment-nulled gradients have a negative-going component that rephases spins that
are stationary and that have constant velocity.
View-to-view translations and rotations of a rigid body have distinct effects on the image that are best
understood in k-space. By the Fourier transform shift theorem, an in-plane translation in image space
leads to an overall phase shift in k-space.140 Therefore, if the object moves along a vector between
views, the complex signal after the translation is related to its value had the object not moved by:
where γ is the gyromagnetic ratio and G is the gradient. Effects of translations can, therefore, be
removed by determining and applying a phase correction to the complex signal.
2 of 6 12-04-2010 00:13
The effect of in-plane translation is a phase shift at each point in k-space. There is a one-to-one
correspondence between each data point in k-space before and after an in-plane translation. Because
rotations in image space cause a rotation in k-space, the one-to-one correspondence no longer
applies, and the intensity as well as phase of each point in k-space is affected. Figure 7-33 shows how
141
a rotation leads to misregistration between data points in k-space taken before and after rotation.
Because of discrete sampling, rotation shifts intensity values in k-space to positions off the initial grid
of sampled data. To correct for a rotation, interpolation is required to reconstruct data on the grid in
k-space.
Navigator Gating
Add to lightbox
Figure 7-33 Illustration of effects of rotation of image (top) on acquisition in k-space (bottom). Dark
spots represent signal intensity in the frame of the object before (left) and after (right) rotation. Circles
represent where acquired data is placed in k-space. The misregistration between the two sets of data
makes interpolation necessary for corrections of rotational motion.
Motion can be measured by adding an extra echo to each phase-encoding line. Each of these extra
echoes is acquired along the same trajectory in k-space. In the absence of motion, each navigator
echo142 is identical. In-plane translations in the phase-encode direction produce phase shifts between
navigator echoes. By measuring the phase shift and then applying a corresponding phase correction to
140
each view, the effect of translation can be removed. Navigator echoes can be acquired in an
interleaved fashion to track translations in the horizontal and vertical directions of the image plane.143
144
Orbital navigator echoes, which traverse a circle in k-space, correct for translations in both
horizontal and vertical directions in the image plane, while lending themselves to straightforward
analysis for correction of rotations. Helical144 and spherical145 navigator echoes can correct for motion
in all directions in a three dimensional acquisition. Navigator echoes have also been introduced in spiral
trajectories.146-148
Corrections based on navigator echoes can be applied retrospectively, in post-processing, or

prospectively, during the scan. The diminishing variance algorithm (DVA),149 which has been applied to
ungated, free-breathing cardiac and abdominal scans, continuously acquires views with navigator
echoes. Because the Fourier transform of the navigator echo is the projection of the image, the center
of the projection can be used to track overall motion. For example, in an axial view of the abdomen,
respiratory motion is largely anterior-posterior. If the frequency encode direction of the navigator echo
is aligned in the left-right direction, the Fourier transform of the navigator has a center of mass that
moves, tracking the respiratory motion. The degree of extension of the body cavity with respiration can
thus be tracked in real time. In the DVA approach, phase encode values are revisited throughout the
scan until the degree of extension as determined from each navigator echo is nearly the same.
3 of 6 12-04-2010 00:13
Self-Calibration and Self-Consistency

page 217
page 218
Add to lightbox
Add to lightbox
Figure 7-34 A, k-space trajectory for "periodically rotated overlapping parallel lines with enhanced
reconstruction" (PROPELLER) scan for a single strip (left) and for entire scan (right). The center of
k-space is heavily oversampled, and the central overlapping region of each strip is compared with the
average among all strips to correct for motion artifacts. B, Raw data (left) and resultant brain image
(right) for two differently oriented PROPELLER strips (top, middle) and for combination of all strips
(bottom).
Navigator echoes require special pulse sequences that are not universally available. The addition of
extra echoes may lead to undesirable additional scan time or signal saturation. Properties of the image
itself may be used to detect and correct for motion in post-processing without the addition of echoes.
For example, a large change in intensity along the phase-encode direction as detected by a Student's
4 of 6 12-04-2010 00:13
t-test150 can indicate motion. In a multislice acquisition,151 the Fourier transform of lines acquired near
the center of k-space for each slice gives the projection of the slice. Motion can be measured by
determining the relative shift between the center of mass of each slice, assuming a relatively
symmetric object.
145
In the orthogonal k-space phase difference (ORKPHAD) technique, two images with orthogonal
phase-encode directions are taken. The difference in phase between the two images is measured at
each point in k-space. Each phase difference has a different functional dependence on the set of
view-to-view translations, leading to a highly over determined set of linear equations relating phases
and translations. The equations can be solved to determine the set of translations and undo motion
artifact. Scan time can, in principle, be the same as that for a single image by taking partial Fourier
acquisitions.
Periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) acquires
152
rectangular strips of k-space that are rotated with respect to one another. As shown in Figure 7-34
this trajectory oversamples the center of k-space, leading to the signal averaging that helps reduce
motion artifact in PR and spiral acquisition images. In the central region of k-space, where the strips
overlap, the image intensity and phase of each strip is compared with the average among all the strips.
Corrections corresponding to translations and rotations are then applied to each strip. Furthermore,
the overall weight of the corrected strip is inversely proportional to the degree to which its central
region matches the average. This weighting strategy reduces contributions from strips that may have
been corrupted by out-of-plane motion.
121,153,154
Other constraints on the image have been pointed out in the context of PR imaging. The
moments of each projection have characteristic symmetry properties for a motionless object. For
example, the zero-order moment of each projection corresponds to the entire image intensity and
should be the same for each projection. The first-order moment of each projection corresponds to the
center of mass of the object and will be the same for each projection if the center of mass is in the
center of the FOV, but will vary sinusoidally with direction otherwise. Higher order moments have
similar constraints. Motion can be measured from deviations of the projections from the constraints
and, hence, corrected by applying appropriate phase shifts and rotations to the data in k-space.
Image-Metric-Based Methods
Artifacts can be removed by minimizing a quantitative measure of image quality, and image metric. In a
diffusion-weighted imaging study, the intensity of the ghosting outside the region of interest was
measured.155 Different sets of translations and rotations were applied to each line in k-space until the
ghosting intensity was minimized. Other methods use an entropy-like image metric that is large when
there is a high degree of blur or ghosting but minimized when the image is sharp.156,157 Again, different
translations and rotations are applied to each line in k-space until the image metric is minimized. Such
an approach can be applied entirely in post-processing but can be costly in computational time.
page 218
page 219
Marker-Based Methods
Markers can be applied to the object of interest, and the signal from those markers can be used to
track the motion of the object. Corrections are relatively simple because each marker is
well-approximated by a point. Different strategies have been used to distinguish the marker signal from
that of the phantom or patient. Marker shapes can be chosen so that the marker signal lies primarily at
141
the edges of k-space, where signal from typical phantoms or the patient is weak. The marker may
158,159
be placed in a tuned circuit that amplifies the strength of B1 and the signal of the marker, thus
easing differentiation between the marker and the phantom or patient. Each marker may also have its
160
own detection coil, which is plugged into the scanner directly. Most of the methods to date correct
primarily for in-plane motions using the same types of analysis described above. However, a plane-
tracking method160 updates the orientation of the image plane based on signal acquired from the
5 of 6 12-04-2010 00:13
markers, allowing for correction of large changes in position.
Summary
Despite improvements in MRI hardware, patient motion still presents problems in the acquisition of
high-quality images. A number of prospective and retrospective techniques have been developed to
reduce motion-related artifacts. Corrections of in-plane rigid-body translation has been explored to a
great degree, but more general motion including rotation, out-of-plane motion, and dilation is less easily
accounted for. The methods require special equipment (marker-based methods) or pulse sequences
(navigator echoes) or they may be computationally intensive (image-metric-based methods). The
approach must, therefore, be chosen depending on the parameters of the imaging situation at hand.
© 2010 Elsevier
6 of 6 12-04-2010 00:13
METHODS FOR ALTERING TISSUE CONTRAST

Fat Suppression
The vast majority of clinical MRI exams aim to image water-containing tissue. Adipose tissue,
containing primarily fat, is of less interest, but the signal from fat can overwhelm an image.
Fat-suppression techniques are commonly used to eliminate the signal from fat. Without suppression,
chemical-shift artifacts from the fat can reduce image quality in rapid imaging by echo-planar
acquisitions. T1-weighted sequences that measure contrast uptake benefit from the suppression of the
nonspecific signal from fat. Fat can be distinguished from water-containing tissue through the
difference in longitudinal relaxation rate (1/T1) and resonance frequency.
Add to lightbox
Figure 7-35 Simulated longitudinal magnetization recovery curves for tissue with T1 = 300 msec (solid
curve) and T1 = 800 msec (dashed curve) after the application of a 180° inversion pulse. The signal
from lipids, which have a T1 of approximately 300 msec, will be nulled if imaged at the time when the
magnetization passes through zero, the so-called "null point."
Inversion Recovery
The longitudinal relaxation rate of fat is much faster (i.e. has a shorter T1) than that of water in tissue.
An inversion pulse can, therefore, be used to selectively null the signal from fat. Figure 7-35 shows
typical signal recovery curves for a variety of T1 values. These curves represent the longitudinal
component of magnetization after the application of a 180° RF inversion pulse. The magnetization of
the short T1 species crosses through the null point before that of tissues with longer T1. If an
excitation pulse is applied at the point in time when the fat is passing through zero, the fat signal is
nulled. This approach is known as short tau inversion recovery (STIR). Tissues with longer or shorter
T1 will have a non-zero longitudinal magnetization at the time of excitation and so their signals will not
be nulled.
The signal recovers after the application of an inversion pulse as:
where Mz(TI) is the longitudinal magnetization at inversion time TI, M0 is the equilibrium value of Mz
and T1 is the longitudinal relaxation time constant. From this expression for signal recovery, we can
determine that the zero crossing, or "null point" of a tissue is given by T1/ln(2), which is approximately
0.69 ×T1. For adipose tissue with a T1 of 300 msec the null point is at 207 msec.
page 219
page 220
1 of 21 12-04-2010 00:14
Add to lightbox
Figure 7-36 The resonance frequency of fat is shifted from that of water by approximately 220 Hz at
1.5 T, allowing frequency selective RF pulses to selectively saturate the signal from fat.
The use of inversion recovery is effective in suppressing (or "nulling") fat signal. However, the inversion
pulse affects water as well as fat, resulting in signal loss from the tissue compared with sequences
without the inversion preparation. Tissues which coincidentally have a T1 value comparable to that of
fat will suffer considerable signal loss. Furthermore, because image brightness typically depends on
the signal magnitude, not sign, contrast between tissue with T1 shorter than fat (inverted signal) and
tissue with T1 longer than fat (uninverted signal) will be diminished by STIR.
Chemical-Shift-Selective Saturation
A second property that is useful in the suppression of the signal of adipose tissue is its distinct
resonance frequency. Adipose tissue is primarily made up of triglycerides, in which the resonance
frequency of protons has a chemical shift of 3.4 parts per million (ppm; 220 Hz at 1.5 T and 440 Hz at
3.0T) with respect to water (Fig. 7-36). This allows for selective saturation of the fat signal by
adjusting the transmitter frequency and the bandwidth of the RF excitation pulse.161 Note that the
frequency shift grows with field; the consequently larger frequency separation between fat and water
signal at higher field strength should allow for more effective saturation of the fat peak.
Add to lightbox
2 of 21 12-04-2010 00:14
Add to lightbox
Figure 7-37 A, Off-resonant fat signal appears shifted in a single-shot spin-echo echo-planar image
(TR/TE = 2090/88 msec; matrix 128 × 128; FOV 300 mm × 300 mm) acquired without fat saturation.
The expected displacement (220 Hz × 0.088 seconds = 19.3 pixels) agrees with the observed 20 pixel
shift (47 mm in this image). B, Chemically selective fat-saturation pulses eliminate the signal from fat
and the resulting chemical-shift artifact.
The additional RF pulses used for chemical-shift-selective saturation increase the TR of the acquisition,
thus limiting their use in rapid imaging techniques. Furthermore, the size of the shift and the width of the
fat resonance depend strongly on the homogeneity of the main magnetic field B0. Therefore, the
efficiency of the saturation pulses will depend on the homogeneity of B0. Incomplete saturation can be
observed in regions which are imaged far from the isocenter, where the strength and homogeneity of
the B0 is diminished. In addition, local field inhomogeneity will limit the effectiveness of the pulse.
The difference in resonance frequency between water and fat can cause misregistration of the position
of fat and water within the image. However, this effect is exacerbated in echo-planar imaging
techniques where off-resonance effects accumulate in the phase-encoding direction throughout the
readout. The location of the fat image can be misregistered by tens of pixels. The size of the shift
depends on the echo time and the resonant offset of the fat. At 1.5 T the frequency offset is ∆ν = 220
Hz. For a TE of 100 msec, the fat will be misregistered by ∆ν × TE = 220 Hz × 0.100 seconds = 22
pixels. Figure 7-37 shows two spin-echo EPI images acquired in the head of a volunteer. These
images show a significant fat artifact, which is eliminated by application of a chemical-shift-selective
saturation pulse.
Dixon Methods
page 220
page 221
Add to lightbox
Figure 7-38 The difference in resonance frequency between fat at water protons results in a difference
in precession frequency that induces a time-dependent phase shift between the two. Immediately after
application of an RF excitation pulse, fat and water are in-phase and the total magnetization in the
3 of 21 12-04-2010 00:14
voxel containing fat and water is the sum of their individual magnetization vectors. At longer times after
excitation, the accrued phases cause destructive interference between fat and water, reducing the
signal intensity.
Another method of nulling fat signal developed by Dixon162 also exploits the difference in resonance
frequency between water and fat.163,164 The resonance frequency of fat is slightly higher than that of
water protons. As a result, the protons in fat precess faster than water protons. Immediately after a
nonselective excitation pulse, the transverse component of the water and fat magnetization will be in
phase (Fig. 7-38). At short times after the excitation (t [Lt ] T2) T2 decay can be ignored, and we can
consider the net magnetization of the voxel as the vector sum of the water and fat spins. The fat and
water spins will accrue a phase difference, which increases over time. Eventually, this phase will be
equal to 180°, at which point the signal will be small due to destructive interference of the
magnetization vectors. At a later time, the magnetization vectors come back into phase, and
constructive interference between the water and fat signal leads to a large overall signal in the voxel. In
the Dixon method, two acquisitions with appropriately chosen echo times, one in which the fat and
water signal are in phase and one in which they are 180° out of phase, are taken:
Addition and subtraction of the resulting images leads to separate water and fat images.
Figure 7-39 shows a coronal examination of the abdomen of a patient acquired with two echo times
(TE1 = 2.38 msec, TE2 = 4.76 msec). Water (i.e., fat-suppressed) and fat (i.e., water-suppressed)
images were calculated by adding and subtracting the images acquired at different echo times.
At 1.5 T the frequency shift between water and fat is 220 Hz. The time for the phase difference to go
through one 360° rotation is 4.6 msec. Therefore, fat and water magnetization will be in-phase for
echo times of 4.6 msec, 9.2 msec, 13.2 msec, etc. Similarly, fat and water will be out of phase by
180° at echo times of 2.3 msec, 6.9 msec, 11.9 msec, etc.
Once again, since these values depend on magnetic field strength, for a 3.0 T scanner, the frequency
shift is twice that of the 1.5 T shift, halving the above times. The shorter time scale is an advantage
because T2 weighting of the signals is minimized.
Add to lightbox
Add to lightbox
4 of 21 12-04-2010 00:14
Add to lightbox
Add to lightbox
Figure 7-39 By acquiring images with different echo times, the effect of the phase difference between
fat and water can be optimized. Coronal abdominal images with A, TE = 4.76 msec (fat and water
in-phase) and B, TE = 2.38 msec (fat and water out-of-phase) show a marked difference in contrast.
By combining these images as per Equations 7-12 and 7-13, water-only fat-suppressed (C) and
fat-only (D) images can be formed. (Courtesy of Dr. Frank Miller MD, Northwestern University.)
Add to lightbox
Add to lightbox
Add to lightbox
5 of 21 12-04-2010 00:14
Add to lightbox
Figure 7-40 The ability to create lipid images allows for tissue characterization using in-phase (A) and
out-of-phase (B) MR images. An example of unusual deposition of focal fat in the liver can be seen as
signal drop off on the out-of-phase images relative to in-phase, proving it is fat. This is seen more
clearly in the water image (C) where the suspected mass is hypointense and the fat image (D) where
the bright signal confirms the mass as fat. (Courtesy of Dr. Frank Miller MD, Northwestern
University.)
An additional benefit of Dixon methods over inversion recovery or chemical-shift-selective saturation is

the ability of in-phase/out-of-phase images to characterize tissue. Figure 7-40 shows an example of
unusual deposition of focal fat in the liver, which mimics a mass on CT. Signal drop-off on the
out-of-phase image relative to in-phase indicates the mass is rich in lipid. Water and fat images were
synthesized from the in-phase and out-of-phase images, improving the contrast of the focal fatty
region.
page 221
page 222
Add to lightbox
Add to lightbox
Figure 7-41 Alternating the echo time to induce 180° phase shifts in a multi-echo radial k-space
acquisition produces inherent lipid suppression. A two-point Dixon acquisition that acquired 128 views
in separate images (A) compared with an alternating TE image in which the views were acquired in an
interleaved fashion (B). Note that the images were acquired in the same amount of time. (Courtesy of
Dr. Chris Flask Ph.D., Case Western Reserve University.)
6 of 21 12-04-2010 00:14
The two-point Dixon method is robust when perfect B0 field homogeneity leads to a well-defined
frequency shift between fat and water. However, B0 inhomogeneity and shim errors cause variations in
the frequency shift. Furthermore, when phase information is not included in the two-point Dixon
165
technique, voxels that contain more lipid than water appear bright in water images. Field
inhomogeneities can be sizable at air-tissue and bone-tissue interfaces, thus leading to fat
suppression. By extending the Dixon two-point method to include an additional image, it has been
shown that the effects of field inhomogeneity can be reduced.165-167 Three-point Dixon techniques
acquire two images in which lipid is out of phase, where their echo times are slightly shifted relative to
the two-point, out-of-phase echo time. The two out-of-phase images can be combined to remove the
effect of phase errors and local field inhomogeneity.
Although Dixon methods have proven useful in the suppression of lipid, they are inefficient due to the
requirement that multiple images are acquired. The power deposition of the magnetization preparation
pulses in chemical-shift-selective saturation or inversion recovery leads to unacceptably high SAR at
high field (>3.0 T) in fast imaging applications. Methods of modifying Dixon methods with either
keyhole168 or radial k-space acquisitions have been developed to address these problems. Flask et
al169 have shown that alternating the echo time to induce 180° phase shifts in a multiecho, interleaved
radial k-space acquisition produces inherent lipid suppression without increasing image acquisition
time. Figure 7-41 shows a two-point Dixon acquisition that acquired 128 views in separate images (A)
compared with alternating TE images in which the views were acquired in an interleaved fashion (B).
Note that the images were acquired in the same amount of time, but, because the alternating TE
images acquired more views per image, the undersampling artifact is reduced in Figure 7-41B.
Magnetization Transfer
Add to lightbox
Figure 7-42 The frequency spectra of bound and free water are shown. The bound-water spectrum,
shown in blue, is much broader than the free-water spectrum, concomitant with the shorter T2 of the
bound water. Magnetization-transfer saturation pulses are applied away from the water peak to
saturate the pool of bound water molecules.
Beyond the proton density, T1, and T2 weighting that are commonly used in MRI, magnetization
transfer offers a distinct type of contrast.170 Magnetization transfer takes advantage of the differences
between the resonances of freely moving water and water undergoing restricted motion. Because
water bound to macromolecules experiences restricted motion, the methods described here typically
improve contrast between tissues with different concentrations of macromolecules. Applications include
7 of 21 12-04-2010 00:14
improved resolution of small vessels in time-of-flight (TOF) angiography,171,172 coronary artery

173 174
imaging, and cartilage imaging.
The resonances of the "free" and "bound" pools of water differ in two distinct ways. The T2 of the free
pool is longer than that of the bound pool. Concomitantly, the width in frequency of the resonance of
the free pool is smaller than that of the bound pool (Fig. 7-42). The differences arise because
molecules in the free pool tumble rapidly, leading to a time-average of magnetic fields experienced on
the microscopic length scale. The averaging results in a smaller net range of resonant frequencies;
hence, there is a narrower line shape for molecules in the free pool. The magnetic fluctuations that
cause transverse relaxation are also averaged, leading to a longer T2 for the free pool.
page 222
page 223
Add to lightbox
Figure 7-43 Frequency spectrum of signal from free and bound pools. A fraction of the bound pool
exchanges with the free pool, contributing to the total signal. Because of its short T2, the contribution of
the bound pool signal to the detected signal is typically only seen through the exchanging protons.
The breadth of the resonance and the short T2 of the bound pool make it difficult to directly detect
signal from this group of molecules. However, exchange occurs between the free and bound pools.
Exchange can be direct, in which case the water in the bound pool breaks free from a restrictive
neighborhood in a macromolecule. The exchange can also occur indirectly, through dipole-dipole
interactions. In the indirect case, the net magnetic polarization among the bound molecules equilibrates
with the net magnetic polarization of nearby free molecules. Overall, the effect of exchange is that the
narrow resonance of the free pool is actually made up of two components. One component is made up
of water molecules that lie near macromolecules and so can experience exchange during the MR
measurement. The other component is made up of molecules that remain in the free pool for the
duration of the experiment (Fig. 7-43).
The differences between the resonances of the free and bound pools can be exploited to improve
contrast. For example, RF pulses can be tailored to selectively saturate the resonance of the bound
pool, effectively leaving the bound pool with no net magnetization. This can be achieved by setting the
center frequency and bandwidth of the RF pulses so that only the bound pool is irradiated by the
pulses. Through exchange, the resonances of free-pool molecules that lie near bound-pool molecules
are indirectly saturated. In other words, the magnetization transfer occurs between the free pool and
the bound pool.
It is, in general, difficult to efficiently selectively saturate a broad resonance while leaving an
overlapping narrow resonance unperturbed. An alternative strategy takes advantage of the short T2 of
8 of 21 12-04-2010 00:14
bound-pool resonances. In this approach, a series of so-called on-resonance RF pulses of large

bandwidth are applied to the entire spectrum.172,175 The net rotation induced in the spin magnetization
by these pulses is zero degrees, but there is a short delay between each pulse. During these short
delays, a short T2 leads to the rapid decay, and hence saturation, of the bound pool, while a long T2
leaves the free pool unperturbed. A typical on-resonance magnetization transfer pulse train consists of
a series of pulses of opposite phase (90°x - delay - 90°-x - delay - 90°-x - delay - 90°x; where delay =
0.2 to 0.3 msec). The signal of the bound pool decays during the delays while the signal of the free
pool remains nearly unperturbed along the z axis, ready for readout by any given MR imaging
sequence.
Typical values of signal attenuation by magnetization transfer are as high as 80% in skeletal muscle,
40% to 70% in white and gray matter, and less than 5% in blood.176 Improving contrast for
visualization of small blood vessels is, therefore, a common application of magnetization transfer. In
TOF angiography, image contrast depends on the difference in magnetization between inflowing blood
and background signal. Multiple short TR pulses saturate the signal of tissue within the imaged slices
but do not saturate the signal of blood in vessels that flow in from outside the imaged region. Large
vessels are well depicted, but incomplete saturation of stationary tissue often leaves the smaller
vessels indistinct. One of the first clinical applications of magnetization transfer was to improve the
depiction of small vessels in TOF angiography.171,172 Because gray and white matter exhibit strong
attenuation by magnetization while blood is attenuated only slightly, intracranial MRA is an ideal
environment for use of the technique.
173
Li et al have shown that magnetization-transfer pulses can effectively saturate signal from the
myocardium, thus improving contrast in bright-blood coronary-artery imaging. Magnetization-transfer
pulses are added to a cardiac-gated 3D gradient-echo acquisition. A trigger delay is required between
the detection of the R wave and the acquisition of MR image data so that image data is acquired at
mid-diastole. This approach minimizes the effect of cardiac motion while maximizing the inflow of
unsaturated blood. Magnetization-transfer and fat-saturation pulses are applied during the delay to
suppress background signal. The delay between the detection of the R wave and mid-diastole is large
enough to allow for multiple preparation pulses. In this study, 25 magnetization-transfer and
fat-saturation pulses were applied before image acquisition, resulting in a 10-fold improvement of
contrast-to-noise ratio between blood and myocardium. Contrast for left ventricular volume acquisitions
was also improved.
Spin-Lock Imaging
High-field magnets are desirable in MRI because SNR is proportional to magnetic polarization, which in
turn is larger at high fields. Contrast based on differences in relaxation rates, however, can suffer at
high fields. For example, water molecules in homogeneous protein solutions have spin-lattice relaxation
177
rates (1/T1) that are 50% higher than that of free water at 0.5 T but are five-fold higher at 200 μT.
Substantial gains in contrast can be achieved by using lower fields, but at a cost of vanishingly small
SNR. Spin-lock imaging provides the contrast of low fields while maintaining the large magnetic
polarization and SNR of high fields. The technique has been found to improve contrast in situations in
which the presence of macromolecular proteins is important, such as imaging of cartilage, 178,179
180 181,182 183
cerebral ischemia, tumors, and cardiac function.
page 223
page 224
Before considering the implementation of spin-lock imaging, it is important to consider why spin-lattice
relaxation rates vary with field. Spin-lattice relaxation is driven by magnetic fluctuations on the
microscopic length scale. The rate is maximized, and T1 minimized, when the inverse of the correlation
time (1/Tc) of the fluctuations is comparable to the resonance frequency184:
At 1.5 T, the resonance frequency of protons is 64 MHz, corresponding to a correlation time of 16

nanoseconds. The correlation times of processes that dominate spin-lattice relaxation are considerably
longer, given the dramatically higher relaxation rates at lower field and resonance frequency.
9 of 21 12-04-2010 00:14
A typical spin-lock magnetization preparation scheme is shown in Figure 7-44. A 90°x pulse is applied,
placing the spin magnetization along the y axis in the rotating frame (A). The spins diphase slightly (B)
before a RF magnetic field of magnitude B1L is applied along the y axis in the rotating frame (C). A
subsequent 90°-x places the weighted magnetization along the z axis for readout by the imaging
sequence of choice. The spin magnetization precesses around B1L at a rate νL = γB1L. As the
magnetization remains on a cone aligned along B1L instead of dephasing, the B1L is known as the
"spin-lock" field. Spin-lattice relaxation occurs in the rotating frame so that the spin magnetization
decays as:
where M is the spin magnetization, TL is the duration of the spin-lock field, and T1ρ is the spin-lattice
relaxation rate in the rotating frame.
Add to lightbox
Figure 7-44 A spin-lock magnetization preparation RF pulse of duration T1L and strength B1L is
applied between the 90° excitation and 180° refocusing pulses. In the rotating frame, the 90° pulse of
magnitude B1 oriented along the x axis rotates the magnetization M along the y axis. Spin
magnetization dephases. Spin lock pulse of magnitude B1L along the y axis is applied, and spin
magnetization precesses around B1L (not shown). A 90° pulse oriented along the negative x axis
places the prepared magnetization along the positive z axis for readout.
Because B1L is typically 10 μT, the precession frequency around B1L is correspondingly smaller than
that around B0. For example, the precession frequency is 430 Hz at B1L = 10 μT as opposed to 64
MHz at B0 = 1.5 T. The magnetic field fluctuations that determine T1ρ have correlation times that are
correspondingly longer, typical of those that dominate relaxation of resonances of water near
macromolecules. Magnetization preparation with a spin lock field therefore enhances contrast between
freely moving water and water that is in contact or in proximity with macromolecules. Furthermore,
note that the magnitude of B1L can be changed far more easily than B0, allowing one to observe the
effects of physical processes with different correlation times.
An important consideration in the addition of spin-lock pulses to any pulse sequence is SAR. For a
spin-lock pulse sequence with a locking pulse of strength B1L, main field of B0, locking time of TL and
repetition time of TR the power deposited is proportional to both B1L and TL:
10 of 21 12-04-2010 00:14
However, significant improvement in image quality has been demonstrated with low locking field
strengths in the range of 1 to 9μT178,181
page 224
page 225
The sensitivity to the molecular environment has proven useful in the imaging of articular cartilage.
178,179
Duvvuri at al have compared T1ρ imaging of cartilage to standard clinical T2-weighted images.
This study demonstrates that T1ρ contrast is achievable using relatively low B1L. They implemented
spin-locking pulses of 1.4, 3.0, 5.8 and 8.8 μT to measure T1ρ dispersion (the degree to which T1ρ
varies with B1L) in fast spin-echo pulse images of the human knee at 1.5 T. They found that T1ρ
dispersion exists in both muscle and cartilage, but is significantly higher in cartilage. The improved
contrast in the case of degenerative cartilage was particularly noticeable in T1ρ-weighted images using
an 8.8 μT locking pulse of 38 msec duration.
T1ρ imaging has shown correlations between changes in T1ρ and irreversible ischemic damage
caused by acute stroke. The stratification of patients for thrombolytic therapy is driven by the time
since the onset of symptoms and by the existence of an ischemic penumbra. Irreversibly infarcted
182
tissue is normally defined as being hyperintense on diffusion-weighted MRI images. Grohn et al
have measured changes in T1ρ in response to an induced stroke and compared these changes with
diffusion MRI images. They find that T1ρ increases at the values of cerebral blood flow that result in
hyperintensity on diffusion-weighted images. However, upon reperfusion of irreversibly infarcted tissue,
T1ρ dispersion remains high, whereas the diffusion abnormalities diminish. T1ρ may therefore prove to
be a more specific indicator of irreversible damage in cases of early reperfusion.
Because of the inherent sensitivity to changes in the macromolecular environment T1ρ imaging has
shown potential for identifying therapeutic response to chemotherapy in tumors. Duvvuri et al183 have
documented changes in T1ρ in an animal model that indicate that these may more accurately reflect
response to treatment than changes in T2. They measured T1ρ values using a fast spin-echo readout
and B1L = 11.7 μT. They found that T1ρ values were significantly longer in cyclophosphamide treated
RIF-1 tumors, whereas T2 changes were not significant in the same tumors.
T1ρ contrast has been successfully implemented in cardiac MRI to suppress signal from the
myocardium in bright-blood images. In bright-blood cardiac MRI, Dixon et al185 have shown that
spin-locking pulses can be used to improve contrast between myocardium and blood even in patients
with compromised cardiac function. These improvements were seen in these patients with reduced
ventricular function, cases in which the flow effects typically responsible for blood-myocardium contrast
are reduced.
Zero-Quantum Imaging
The picture of nuclear spins precessing in a magnetic field and flipping in response to RF excitation is a
classical representation of underlying quantum mechanical processes. The motion and relaxation of
nuclear spins is more accurately described by transitions between quantum-mechanical states. In
standard MRI measurements, the transitions are between states that differ by a single spin-flip, known
as single-quantum transitions.
In general, a nuclear spin will exert a small magnetic field on a neighboring nucleus, leading to an
internuclear coupling. The coupling is weak but can be used to induce what are called "zero-quantum
transitions." An example of a zero-quantum transition in a system of two coupled spins, A and B,
involves a transition from a state in which A is up and B is down (|AB〉 = |↑↓〉) to a state in which A is
down and B is up (|AB〉 = |↑↓〉). The net change in spin is zero and cannot be observed directly. The
so-called zero-quantum coherence does, however, modulate the detected MRI signal and can,
therefore, be detected indirectly. The approach is used for structure determination of molecules in
multi-dimensional MR spectroscopy.
185
Warren et al have shown that zero-quantum transitions can be measured in liquids even when the
11 of 21 12-04-2010 00:14
coupling between nuclei in the same or nearby molecules is zero. They find measurable effects due to
coupling between distant nuclei separated by as much as a millimeter. Although the coupling between
individual nuclei that are a millimeter apart is negligible, there is a tremendous number of spins lying at
that distance, leading to detectable zero-quantum coherence. Furthermore, the length scale to which
the zero-quantum coherence is sensitive can be tuned by the user from 10 μm to 1 mm.
Acquisition of zero-quantum coherence MRI requires only slight modification to an MR pulse sequence
and no special hardware (Fig. 7-45). The pulse sequence starts with a spin-magnetization preparation
component followed by standard readout. The preparation component consists of two pulses, α and β,
separated by a "mixing time," τZQ. During the mixing time, "correlation gradients" of strength G and
duration T are applied.
The characteristic length scale (d) detected by the zero-quantum coherence is approximately d =
π/γGT. The signal detected by the standard readout primarily comes from spins separated by the
characteristic length scale.
In Vivo Application
Zero-quantum coherence can be used as a contrast mechanism in vivo in MRI.186 The contrast arises
from field inhomogeneities of a length scale ranging from 10 μm to 1 mm, with the length scale being a
tunable parameter. The inhomogeneity arises from differing oxygen concentrations. The contrast in
187
images taken in humans differs from contrast in other imaging modalities. Contrast enhancement
has been seen in tumors in rat brain.186 Although the detected signal is small187 compared with
188,189
standard imaging, higher-order coherences may exhibit similar, tunable contrast with higher SNR.
Electric Current Imaging

page 225
page 226
Add to lightbox
12 of 21 12-04-2010 00:14
Add to lightbox
Figure 7-45 A, Magnetization preparation for zero-quantum imaging. Readout is a standard spin-echo.
The degree of weighting by zero-quantum processes is determined by the flip angles, α and β, and by
the mixing time, τZQ. B, Conventional spin-echo EPI image (top left) is compared to four different
iZQC images using spin-echo EPI (right). The iZQC intensity and contrast are different from any
conventional image, particularly in regions with large susceptibility variations. For short values of τzq,
the iZQC has a magnetization-squared weighting. The figure at the bottom left shows a "map" of iZQC
relaxation times by pixel (white > 60 ms, black <10 ms). Other imaging parameters include FOV = 24
× 24 cm, matrix size 128 × 64, slice thickness of 10 mm, TE = 60 ms, NEX = 16, and TR = 4 s. (From
Rizi RR, Ahn S, Alsop DC, et al. Intermolecular zero-quantum coherence imaging of the human
brain. Magnetic Resonance in Medicine 43(5):627-632, 2000. Reproduced with permission of
Wiley-Liss, Inc., a subsidiary of John Wiley & Sons, Inc.)
Work on electric current imaging has shown, in phantoms and in vivo, that the magnetic field
associated with small applied currents can be measured by MRI. A basic implementation190 inserts
two current pulses of opposite polarity before and after the 180° RF pulse in a spin-echo acquisition
(Fig. 7-46). The current pulses create fields in addition to B0 that induce phase shifts in the acquired
signal. The phase shifts are in turn used to create a field map. Using Ampere's law, the current
distribution is derived.
Current density imaging has been performed in a number of systems of biological interest. Current
profiles in bone191 may aid in diagnosis of osteoporosis. Assessment of spinal cord function and
192 193
viability and of epilepsy may also be possible. Tumors in mice have been detected with currents
typical of those used in electrochemotherapy.191 Phantom studies194 suggest that intrinsic currents
associated with event-related potentials of large clusters of neurons can also be measured.
The sensitivity of the technique depends on the size of the current, which is limited by stimulation
thresholds. Enhanced sensitivity has been found by using RF currents195,196 and multiple pulse
197
sequences. The use of RF currents also avoids stimulation effects while eliminating the need for
multiple orientations of the subject.195 Improved post-processing algorithms198-200 also eliminate the
need for multiple orientations without the use of RF currents.
page 226
page 227
13 of 21 12-04-2010 00:14
Add to lightbox
Add to lightbox
Figure 7-46 A, In electrical current imaging, electrical current pulses of duration Tc/2 and opposite
polarity separated in time by a delay TD, are applied to the subject. B, The experiment was performed
with a 3.0 tesla MRI system on a phantom with an inhomogeneous conductivity distribution and
injected electrical current. Left: Magnitude image of the phantom obtained with an injected electrical
current of 24 mA. Right: Corresponding Bz (z-component of the internal magnetic field density) image
demonstrating spatial variations in conductivity not apparent in the magnitude image. (From Oh SH,
Lee BI, Park TS, et al. Magnetic resonance electrical impedance tomography at 3 tesla field
strength. Magnetic Resonance in Medicine 51(6):1292-1296, 2004. Reproduced with permission of
REFERENCES
2. Melki PS, Jolesz FA, Mulkern RV: Partial RF echo-planar imaging with the FAISE method. II. Contrast equivalence with
spin-echo sequences. Magn Reson Med 26:342-354, 1992. Medline Similar articles
3. Melki PS, Mulkern RV, Panych LP, Jolesz FA: Comparing the FAISE method with conventional dual-echo sequences. J
Magn Reson Imaging 1:319-36, 1991. Medline Similar articles
5. Hennig J, Friedburg H: Clinical applications and methodological developments of the RARE technique. Magn Reson
14 of 21 12-04-2010 00:14
6. Johnson BA, Fram EK, Drayer BP, et al: Evaluation of shared-view acquisition using repeated echoes (SHARE): a
dual-echo fast spin-echo MR technique. Am J Neuroradiol 15:667-673, 1994. Medline Similar articles
7. Mulkern RV, Wong ST, Winalski C, Jolesz FA: Contrast manipulation and artifact assessment of 2D and 3D RARE
sequences. Magn Reson Imaging 8:557-566, 1990.
8. Melki PS, Jolesz FA, Mulkern RV: Partial RF echo planar imaging with the FAISE method. I. Experimental and theoretical
assessment of artifact. Magn Reson Med 26:328-341, 1992. Medline Similar articles
9. Mulkern RV, Melki PS, Jakab P, et al: Phase-encode order and its effect on contrast and artifact in single-shot RARE
sequences. Med Phys 18:1032-1037, 1991. Medline Similar articles
10. Listerud J, Einstein S, Outwater E, Kressel HY: First principles of fast spin echo. Magn Reson Q 8:199-244, 1992.
11. Vinitski S, Mitchell DG, Einstein SG, et al: Conventional and fast spin-echo MR imaging: minimizing echo time. J Magn
12. Oshio K, Jolesz FA, Melki PS, Mulkern RV: T2-weighted thin-section imaging with the multislab three-dimensional RARE
technique. J Magn Reson Imaging 1:695-700, 1991. Medline Similar articles
13. Hackney DB, Lenkinski RE, Grossman RI, et al: Initial experience with fast low-angle multiecho (FLAME) imaging of the
central nervous system. J Comput Assist Tomogr 12:171-174, 1988. Medline Similar articles
14. Augui J, Vignaux O, Argaud C, et al: Liver: T2-weighted MR imaging with breath-hold fast-recovery optimized fast
spin-echo compared with breath-hold half-Fourier and non-breath-hold respiratory-triggered fast spin-echo pulse sequences.
15. Dixon WT, Engels H, Castillo M, Sardashti M: Incidental magnetization transfer contrast in standard multislice imaging.
16. Melki PS Mulkern RV: Magnetization transfer effects in multislice RARE sequences. Magn Reson Med 24:189-195, 1992.
17. Outwater E, Schnall MD, Braitman LE, et al: Magnetization transfer of hepatic lesions: evaluation of a novel contrast
technique in the abdomen. Radiology 182:535-540, 1992. Medline Similar articles
18. Constable RT, Smith RC, Gore JC: Signal-to-noise and contrast in fast spin echo (FSE) and inversion recovery FSE
imaging. J Comput Assist Tomogr 16:41-47, 1992. Medline Similar articles
19. Hennig J, Scheffler K: Hyperechoes. Magn Reson Med 46:6-12, 2001. Medline Similar articles
20. Frank LR, Wong EC, Liu TT, Buxton RB: Increased diffusion sensitivity with hyperechoes. Magn Reson Med 49:1096-1105,
2003.
21. Gomori JM, Grossman RI, Goldberg HI, et al: Intracranial hematomas: imaging by high-field MR. Radiology 157:87-93,
22. Thulborn KR, Waterton JC, Matthews PM, Radda GK: Oxygenation dependence of the transverse relaxation time of water
page 227
page 228
23. Gomori JM, Grossman RI, Yu-Ip C, Asakura T: NMR relaxation times of blood: dependence on field strength, oxidation
state, and cell integrity. J Comput Assist Tomogr 11:684-690, 1987. Medline Similar articles
24. Carr HY, Purcell EM: Effects of diffusion on free precession in nuclear magnetic resonance. Phys Rev 94:630-638, 1954.
25. Meiboom S, Gill D: Modified spin-echo method for measuring nuclear relaxation times. Rev Sci Instr 29:688-691, 1958.
26. Henkelman RM, Hardy PA, Bishop JE, et al: Why fat is bright in RARE and fast spin-echo imaging. J Magn Reson Imaging
27. Constable RT, Anderson AW, Zhong J, Gore JC: Factors influencing contrast in fast spin-echo MR imaging. Magn Reson
28. Higuchi N, Hiramatsu K, Mulkern RV: A novel method for fat suppression in RARE sequences. Magn Reson Med
29. Oshio K, Feinberg DA: GRASE (Gradient- and spin-echo) imaging: a novel fast MRI technique. Magn Reson Med
30. Chien D, Edelman RR: Ultrafast imaging using gradient echoes. Magn Reson Q 7:31-56, 1991. Medline
Similar articles
31. Haacke EM, Wieloposki PA, Tkach AJ: A comprehensive technical review of short TR, fast, magnetic resonance imaging.
Rev Magn Reson Med 3:53-170, 1991.
32. Haacke EM, Tkach JA: A review of fast imaging techniques and applications. Am J Roentgenol 155:951-964, 1990.
33. Crawley AP, Wood ML, Henkelman RM: Elimination of transverse coherences in FLASH MRI. Magn Reson Med
34. Frahm J, Merboldt KD, Hanicke W: Transverse coherence in rapid FLASH NMR imaging. J Magn Reson 27:307-314,
1987.
35. Haacke EM, Tkach JA, Parrish TB: Reduction of T2* dephasing in gradient field-echo imaging. Radiology 170:457-462,
15 of 21 12-04-2010 00:14

36. Buxton RB, Fisel CR, Chien D, Brandy TJ: Signal intensity in fast imaging. J Magn Reson 83:576-585, 1989.
37. van der Meulen P, Groen JP, Tinus AM, Bruntink G: Fast Field Echo imaging: an overview and contrast calculations. Magn
38. Frahm J, Haase A, Matthaei D: Rapid three-dimensional MR imaging using the FLASH technique. J Comput Assist
39. Parz S: Steady-state free precession: an overview of basic concepts and applications. Adv Magn Reson Imaging
1:73-102, 1989.
40. Gyngell ML: The steady-state signals in short repetition-time sequences. J Magn Reson 81:474-483, 1988.
41. Oppelt A, Graumann R, Barfuss H, et al: FISP. A new fast MRI sequence. Electromedica 3:15-318, 1986.
42. Hawkes RC, Patz S: Some factors that influence the steady state in steady-state free precession. Magn Reson Imaging
5:117-127, 1989.
43. Carr JC, Simonetti O, Bundy J, et al: Cine MR angiography of the heart with segmented true fast imaging with steady-state
precession. Radiology 291:828-823, 2001.
44. Haacke EM, Wielopolski PA, Tkach JA, Modic MT: Steady-state free precession imaging in the presence of motion:
application for improved visualization of the cerebrospinal fluid. Radiology 175:545-552, 1990. Medline Similar articles
45. Gyngell ML: The application of steady-state free precession in rapid 2DFT NMR imaging: FAST and CE-FAST sequences.
46. Bruder H, Fischer H, Reinfelder HE, Schmitt F: Image reconstruction for echo planar imaging with nonequidistant k-space
sampling. Magn Reson Med 23:311-323, 1992. Medline Similar articles
47. Hanicke W, Merboldt KD, Chien D, et al: Signal-to-noise in subsecond FLASH NMR imaging. Considerations of the
dynamic approach to the steady state by successive saturation. Med Phys 17:1004-1010, 1991.
48. Deimling M, Heid O: Magnetization prepared true FISP imaging. Proc Intl Soc Magn Reson Med 2:495, 1994.
49. Scheffler K: On the transient phase of balanced SSFP sequences. Magn Reson Med 49:781-783, 2003. Medline
Similar articles
50. Nishimura DG, Vasanawala S: Analysis and reduction of the transient response in SSFP imaging. Proc Intl Soc Magn
Reson Med 8:301, 2000.
51. Deshpande V, Li D: Contrast-enhanced coronary artery imaging using 3D trueFISP. Magn Reson Med 50:570-577, 2003.
52. Hanicke W, Merboldt KD, Chien D, et al: Signal strength in subsecond FLASH magnetic resonance imaging: the dynamic
approach to steady state. Med Phys 17:1004-1010, 1990. Medline Similar articles
53. Haase A: Snapshot FLASH MRI. Applications to T1, T2, and chemical-shift imaging. Magn Reson Med 13:77-89, 1990.
54. Chien D, Atkinson D, Edelman RR: Strategies to improve contrast in k-space coverage: reordered phase encoding and
k-space segmentation. J Magn Reson Imaging 1:63-70, 1991. Medline Similar articles
55. Hennig J: Burst Imaging on a clinical whole body system. Proc Intl Soc Magn Reson Med 11:101, 1992.
56. Hennig J, Hoddap M: BURST Imaging. MAGMA 1:39-48, 1993.
57. Duyn JH, van Gelderen P, Liu G, Moonen CT: Fast volume scanning with frequency-shifted BURST MRI. Magn Reson Med
58. Wong EC: Magnetic resonance functional neuroimaging using interleaved SUNBURST. Proc Intl Soc Magn Reson Med
2:25, 1994.
59. Zha L, Lowe IJ: Optimized ultra-fast imaging sequence (OUFIS). Magn Reson Med 33:377-395, 1995. Medline
Similar articles
60. Lowe IJ, Wysong RE: DANTA ultrafast imaging sequence (DUFIS). J Magn Reson B101:106-109, 1993.
61. Heid O, Deimling M: Quest a quick echo split imaging technique. Proceedings of the 11th Annual Meeting of the Society of
Magnetic Resonance in Medicine. 11:433, 1992.
62. Heid O, Deimling M, Huk WJ: Ultra-rapid gradient echo imaging. Magn Reson Med 33:143-149, 1995. Medline
Similar articles
63. Sattin W, Mareci TH, Scott KN: Exploiting the stimulated echo in nuclear magnetic-resonance imaging. 1. Method. J Magn
Res 64:177-182, 1985.
64. Sattin W, Mareci TH, Scott KN: Exploiting the stimulated echo in nuclear magnetic-resonance imaging. 2. Applications. J
Magn Res 65:298-307, 1985.
65. Frahm J, Haase A, Matthaei D, et al: Rapid NMR imaging using stimulated echoes. J Magn Res 65:130-135, 1985.
66. Warach S, Chien D, Li W, et al: Fast magnetic resonance diffusion-weighted imaging of acute human stroke. Neurology
67. Mansfield P, Pykett IL: Biological and medical imaging by NMR. J Magn Res 29:355-373, 1978.
68. Stehling MK, Turner R, Mansfield P: Echo-planar imaging-magnetic-resonance-imaging in a fraction of a second. Science
16 of 21 12-04-2010 00:14

69. Edelman RR, Wielopolski P, Schmitt F: Echo-planar MR imaging. Radiology 192:600-612, 1994. Medline
Similar articles
70. DeLaPaz RL: Echo-planar imaging. Radiographics 14:1045-1058, 1994. Medline Similar articles
71. Farzaneh F, Riederer SJ, Pelc NJ: Analysis of T2 limitations and off-resonance effects on spatial resolution and artifacts in
echo-planar imaging. Magn Reson Med 14:123-139, 1990. Medline Similar articles
72. Cohen MS, Weisskoff RM: Ultra-fast imaging. Magn Reson Imaging 9:1-37, 1991. Medline Similar articles
73. Farzaneh F, Reiderer SJ: Hybrid imaging with gradient recalled sliding echoes. Proc Intl Soc Magn Reson Med 7:7, 1989.
74. Feinberg D, Oshio K: Gradient-echo shifting in fast MRI techniques (GRASE imaging) for correction of field inhomogeneity
and chemical shift. J Magn Reson 97:177-183, 1992.
75. Butts K, Riederer SJ, Ehman RL, et al: Interleaved echo planar imaging on a standard MRI system. Magn Reson Med
76. McKinnon GC: Ultrafast interleaved gradient-echo-planar imaging on a standard scanner. Magn Reson Med 30:609-616,
77. Mansfield P, Harvey PR: Limits to neural stimulation in echo-planar imaging. Magn Reson Med 29:746-758, 1993.
78. Schmitt F, Wielopolski P, Fischer H, Edelman R: Peripheral stimulations and their relation to gradient pulse shapes. Proc
Intl Soc Magn Reson Med 2:102, 1994.
79. Cohen MS, Weisskoff RM, Rzedzian RR, Kantor HL: Sensory stimulation by time-varying magnetic fields. Magn Reson
80. Budinger TF, Fischer H, Hentschel D, et al: Physiological effects of fast oscillating magnetic field gradients. J Comput
81. Schmitt F: Correction of geometric distortions in MR-images. Proc Intl Sym Comp Assist Radiol 2:14-23, 1985
82. Francois CJ, Shors SM, Bonow RO, Finn JP: Analysis of cardiopulmonary transit times at contrast material-enhanced MR
imaging in patients with heart disease. Radiology 227:447-452, 2003. Medline Similar articles
83. Van Hoe L, De Jaegere T, Bosmans H, et al: Breath-hold contrast-enhanced three-dimensional MR angiography of the
abdomen: Time-resolved imaging versus single-phase imaging. Radiology 214:149-156, 2000. Medline
Similar articles
84. Herborn CU, Goyen M, Lauenstein TC, et al: Comprehensive time-resolved MRI of peripheral vascular malformations. Am J
Roentgenol 181:729-735, 2003.
85. Feinberg DA, Hale JD, Watts JC, et al: Halving MR imaging time by conjugation: demonstration at 3.5 kG. Radiology
161:527-531, 1986.
86. Amartur S, Haacke EM: Modified iterative model based on data extrapolation method to reduce Gibbs ringing. J Magn
87. Sodickson DK, Manning WJ: Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency
coil arrays. Magn Reson Med 38:591-603, 1997. Medline Similar articles
88. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P: SENSE: sensitivity encoding for fast MRI. Magn Reson Med
89. Twieg DB: The k trajectory formulation of the NMR imaging process with applications in analysis and synthesis of imaging
90. Margosian PM: Faster MR imaging: imaging with half the data. Health Care Instrum 161:527-531, 1986.
91. Haacke EM, Mitchell J, Lee D: Improved contrast using half-Fourier imaging: application to spin-echo and angiographic
imaging. Magn Reson Imaging 8:79-90, 1990. Medline Similar articles
92. Haacke EM, Lindskog ED, Lin W: A Fast, iterative, partial-fourier technique capable of local phase recovery. J Magn Res
92:126-145, 1991.
93. Haacke EM, Liang ZP, Boada F: Image-reconstruction using projection onto convex-sets, model constraints, and linear
prediction-theory for the removal of phase, motion, and Gibbs artifacts in magnetic-resonance and ultrasound imaging. Optical
Eng 29:555-566, 1990.
94. Liang ZP, Haacke EM, Thomas CW: High resolution inversion of finite Fourier transform data through localized polynomial
approximation. Inverse Problems 5:831-838, 1989.
95. Goldfarb JW, Prasad PV, Griswold MA, Edelman RR: Dynamic three-dimensional magnetic resonance abdominal
angiography and perfusion: implementation and preliminary experience. J Magn Reson Imaging 11:201-207, 2000.
page 228
page 229
96. Finn JP, Baskaran V, Carr JC, et al: Thorax: low-dose contrast-enhanced three-dimensional MR angiography with
subsecond temporal resolution-initial results. Radiology 224:896-904, 2002. Medline Similar articles
97. van Vaals JJ, Tuithof HH, Dixon WT: Increased time resolution in dynamic imaging. Proc Soc Magn Reson Image 10:1108,
17 of 21 12-04-2010 00:14
1992.
98. Chenevert TL, Pipe JG: Dynamic 3D imaging at high temporal resolution via reduced k-space sampling. Proc Soc Magn
Reson Image 12:1262, 1993.
99. Spraggins TA: Simulation of spatial and contrast distortions in keyhole imaging. Magn Reson Med 31:320-322, 1994.
100. Liang ZP, Hanson JM, Potter CS, Lauterbur PC: Efficient high-resolution dynamic imaging with explicit boundary
constraints. Proc Soc Magn Reson 2:53, 1994.
101. Korosec FR, Frayne R, Grist TM, Mistretta CA: Time-resolved contrast-enhanced 3D MR angiography. Magn Reson Med
36:345-351, 1996.
102. Carroll TJ, Korosec FR, Petermann GM, et al: Carotid bifurcation: evaluation of time-resolved three-dimensional contrast-
enhanced MR angiography. Radiology 220:525-532, 2001. Medline Similar articles
103. Swan JS, Carroll TJ, Kennell TW, et al: Time-resolved three-dimensional contrast-enhanced MR angiography of the
peripheral vessels. Radiology 225:43-52, 2002. Medline Similar articles
104. Morasch MD, Collins J, Pereles FS, et al: Lower extremity stepping-table magnetic resonance angiography with multilevel
contrast timing and segmented contrast infusion. J Vasc Surg 37:62-71, 2003. Medline Similar articles
105. Madore B, Glover GH, Pelc NJ: Unaliasing by fourier-encoding the overlaps using the temporal dimension (UNFOLD),
applied to cardiac imaging and fMRI. Magn Reson Med 42:813-28, 1999. Medline Similar articles
106. Kellman P, Epstein FH, McVeigh ER: Adaptive sensitivity encoding incorporating temporal filtering (TSENSE). Magn
107. Cooley JW, Tukey JW: An algorithm for the machine computation of complex Fourier series. Math Comput 19:297-301,
1965.
108. Maki JH, Chenevert TL, Prince MR: Three-dimensional contrast-enhanced MR angiography. Top Magn Reson Imaging
109. Lauterbur PC: Progress in N.M.R. zeugmatography imaging. Philos Trans R Soc Lond B Biol Sci 289:483-487, 1980.
110. Wilman AH, Riederer SJ: Performance of an elliptical centric view order for signal enhancement and motion artifact
suppression in breath-hold three-dimensional gradient echo imaging. Magn Reson Med 38:793-802, 1997. Medline
Similar articles
111. Scheffler K, Hennig J: Reduced circular field-of-view imaging. Magn Reson Med 40:474-480, 1998. Medline
Similar articles
112. Peters DC, Korosec FR, Grist TM, et al: Undersampled projection reconstruction applied to MR angiography. Magn
113. Glover GH, Pauly JM: Projection reconstruction techniques for reduction of motion effects in MRI. Magn Reson Med
114. Lauzon ML, Rutt BK: Polar sampling in k-space: reconstruction effects. Magn Reson Med 40:769-782, 1998. Medline
Similar articles
115. Vigen KK, Peters DC, Grist TM, et al: Undersampled projection-reconstruction imaging for time-resolved contrast-
enhanced imaging. Magn Reson Med 43:170-176, 2000. Medline Similar articles
116. Barger AV et al: Phase-contrast with interleaved undersampled projections. Magn Reson Med 43:503-509, 2000.
117. Pipe JG: Reconstructing MR images from undersampled data: data-weighting considerations. Magn Reson Med
118. Barger AV, Block WF, Toropov Y, et al: Time-resolved contrast-enhanced imaging with isotropic resolution and broad
coverage using an undersampled 3D projection trajectory. Magn Reson Med 48:297-305, 2002.
119. Nishimura DG, Irarrazabal P, Meyer CH: A velocity k-space analysis of flow effects in echo-planar and spiral imaging.
120. Meyer CH, Hu BS, Nishimura DG, Macovski A: Fast spiral coronary artery imaging. Magn Reson Med 28:202-213, 1992.
121. Glover GH, Noll DC: Consistent projection reconstruction (CPR) techniques for MRI. Magn Reson Med 29:345-351, 1993.
122. Jackson JI, Meyer CH, Nishimura DG: Selection of a convolution function for Fourier inversion using gridding. IEEE Trans
Med Imaging 10:473-478, 1991. Medline Similar articles
123. Noll DC, Pauly JM, Meyer CH, et al: Deblurring for non-2D Fourier transform magnetic resonance imaging. Magn Reson
124. Meyer CH, Pauly JM, Macovski A, Nishimura DG: Simultaneous spatial and spectral selective excitation. Magn Reson
125. Healy DM, Weaver JB: Two applications of wavelet transforms in magnetic resonance imaging. IEEE Trans Inform Theory
38:840-862, 1992.
18 of 21 12-04-2010 00:14
126. Weaver JB, Xu Y, Healy DM, Driscoll JR: Wavelet-encoded MR imaging. Magn Reson Med 24:275-287, 1992.
127. Panych LP, Jakab PD, Jolesz FA: Implementation of wavelet-encoded MR-imaging. J Magn Res Imaging 3:649-655,
1993.
128. Battle G: Cardinal spline interpolation and the block spin construction of wavelets. In Chui CK (ed): Wavelets: a tutorial in
theory and applications, San Diego: Academic Press, 1992, pp73-90.
129. Panych LP, Jolesz FA: A dynamically adaptive imaging algorithm for wavelet-encoded MRI. Magn Res Med 32:738-748,
1994.
130. Wood ML, Henkelman RM: MR image artifacts from periodic motion. Med Phys 12:143-151, 1985. Medline
Similar articles
131. Kumar AW, Welrti D, Ernst RR: NMR Fourier zeugmatography. J Magn Reson 18:69-83, 1975.
132. Runge VM, Clanton JA, Partain CL, James AE, Jr: Respiratory gating in magnetic resonance imaging at 0.5 tesla.
Radiology 151:521-523, 1984.
133. Bailes DR, Gilderdale DJ, Bydder GM, et al: Respiratory ordered phase encoding (ROPE): a method for reducing
respiratory motion artefacts in MR imaging. J Comput Assist Tomogr 9:835-838, 1985. Medline Similar articles
134. Haacke EM, Patrick JL: Reducing motion artifacts in two-dimensional Fourier transform imaging. Magn Reson Imaging
135. Jhooti P, Wiesmann F, Taylor AM, et al: Hybrid ordered phase encoding (HOPE): an improved approach for respiratory
artifact reduction. J Magn Reson Imaging 8:968-980, 1998. Medline Similar articles
136. Korin HW, Riederer SJ, Bampton AE, Ehman RL: Altered phase-encoding order for reduced sensitivity to motion in three-
dimensional MR imaging. J Magn Reson Imaging 2:687-693, 1992. Medline Similar articles
137. Glover GH, Lee AT: Motion artifacts in fMRI: comparison of 2DFT with PR and spiral scan methods. Magn Reson Med
33:624-635, 1995.
Roentgenol 148:1251-1258, 1987.
140. Korin HW, Farzaneh F, Wright RC, Riederer SJ: Compensation for effects of linear motion in MR imaging. Magn Reson
141. Korin HW, Felmlee JP, Riederer SJ, Ehman RL: Spatial-frequency-tuned markers and adaptive correction for rotational
motion. Magn Reson Med 33:663-669, 1995. Medline Similar articles
142. Ehman RL, Felmlee JP: Adaptive technique for high-definition MR imaging of moving structures. Radiology 173:255-263,
143. Song HK, Wehrli FW: In vivo micro-imaging using alternating navigator echoes with applications to cancellous bone
structural analysis. Magn Reson Med 41:947-953, 1999. Medline Similar articles
144. Fu ZW, Wang Y, Grimm RC, et al: Orbital navigator echoes for motion measurements in magnetic resonance imaging.
145. Welch EB, Felmlee JP, Ehman RL, Manduca A: Motion correction using the k-space phase difference of orthogonal
acquisitions. Magn Reson Med 48:147-156, 2002. Medline Similar articles
146. Sachs TS, Meyer CH, Hu BS, et al: Real-time motion detection in spiral MRI using navigators. Magn Reson Med
147. Moriguchi H, Lewin JS, Duerk JL: Novel interleaved spiral imaging motion correction technique using orbital navigators.
148. Heberlein KS, Hu XS: Motion correction in segmented spiral imaging using dual-density sampling. Proc Int Soc Magn
Reson Med 10:375, 2002.
149. Sachs TS, Meyer CH, Irarrazabal P, et al: The diminishing variance algorithm for real-time reduction of motion artifacts in
MRI. Magn Reson Med 34:412-422, 1995. Medline Similar articles
150. Wood ML, Shivji MJ, Stanchev PL: Planar-motion correction with use of k-space data acquired in Fourier MR imaging. J
151. Zang LH, Fielden J, Wilbrink J, et al: Correction of translational motion artifacts in multi-slice spin-echo imaging using
self-calibration. Magn Reson Med 29:327-334, 1993. Medline Similar articles
152. Pipe JG: Motion correction with PROPELLER MRI: application to head motion and free-breathing cardiac imaging. Magn
153. Shankaranarayanan A, Wendt M, Lewin JS, Duerk JL: Two-step navigatorless correction algorithm for radial k-space MRI
acquisitions. Magn Reson Med 45:277-288, 2001. Medline Similar articles
154. Gai N, Axel L: Correction of motion artifacts in linogram and projection reconstruction MRI using geometry and
consistency constraints. Med Phys 23:251-262, 1996. Medline Similar articles
155. Robson MD, Anderson AW, Gore JC: Diffusion-weighted multiple shot echo planar imaging of humans without navigation.
19 of 21 12-04-2010 00:14

156. Manduca A, McGee KP, Welch EB, et al: Autocorrection in MR imaging: adaptive motion correction without navigator
echoes. Radiology 215:904-909, 2000. Medline Similar articles
157. Atkinson D, Hill DL, Stoyle PN, et al: Automatic compensation of motion artifacts in MRI. Magn Reson Med 41:163-170,
158. Flask C, Elgort D, Wong E, et al: A method for fast 3D tracking using tuned fiducial markers and a limited projection
reconstruction FISP (LPR-FISP) sequence. J Magn Reson Imaging 14:617-627, 2001.
159. Bernstein MA, Shu Y, Elliott AM: RINGLET motion correction for 3D MRI acquired with the elliptical centric view order.
Magn Reson Med 50:802-812, 2003.
160. Derbyshire JA, Wright GA, Henkelman RM, Hinks RS: Dynamic scan-plane tracking using MR position monitoring. J
161. Keller PJ, Hunter WW Jr, Schmalbrock P: Multisection fat-water imaging with chemical shift selective presaturation.
163. Bydder GM, Young IR: Clinical use of the partial saturation and saturation recovery sequences in MR imaging. J Comput
164. Smith RC, Constable RT, Reinhold C, et al: Fast spin echo STIR imaging. J Comput Assist Tomogr 18:209-213, 1994.
165. Lodes CC, Felmlee JP, Ehman RL, et al: Proton MR chemical shift imaging using double and triple phase contrast
acquisition methods. J Comput Assist Tomogr 13:855-861, 1989. Medline Similar articles
page 229
page 230
166. Glover GH: Multipoint Dixon technique for water and fat proton and susceptibility imaging. J Magn Reson Imaging
167. Glover GH, Schneider E: Three-point Dixon technique for true water/fat decomposition with B0 inhomogeneity correction.
168. Flask CA, Salem KA, Moriguchi H, et al: Keyhole Dixon method for faster, perceptually equivalent fat suppression. J Magn
169. Flask CA, Dale B, Lewin JS, Duerk JL: Radial alternating TE sequence for faster fat suppression. Magn Reson Med
170. Wolff SD, Balaban RS: Magnetization transfer contrast (MTC) and tissue water proton relaxation in vivo. Magn Reson
172. Pike GB, Hu BS, Glover GH, Enzmann DR: Magnetization transfer time-of-flight magnetic resonance angiography. Magn
173. Li D, Paschal CB, Haacke EM, Adler LP: Coronary arteries: three-dimensional MR imaging with fat saturation and
magnetization transfer contrast. Radiology 187:401-406, 1993. Medline Similar articles
174. Wolff SD, Chesnick S, Frank JA, et al: Magnetization transfer contrast: MR imaging of the knee. Radiology 179:623-628,
175. Hua J, Hurst GC: Analysis of on- and off-resonance magnetization transfer techniques. J Magn Reson Imaging 5:113-120,
176. Vlaardingerbroek M, den Boer J: Magnetic Resonance Imaging. Berlin: Springer Verlag. 1999, p 481.
177. Koenig SH, Brown RD, 3rd, Adams D, et al: Magnetic field dependence of 1/T1 of protons in tissue. Invest Radiol
19:76-81, 1984.
178. Duvvuri U, Charagundla SR, Kudchodkar SB, et al: Human knee: in vivo T1(rho)-weighted MR imaging at 1.5
T-preliminary experience. Radiology 220:822-826, 2001.
179. Duvvuri U, Kudchodkar S, Reddy R, Leigh JS: T(1rho) relaxation can assess longitudinal proteoglycan loss from articular
cartilage in vitro. Osteoarthritis Cartilage 10:838-844, 2002. Medline Similar articles
180. Grohn OHJ, Kettunen MI, Makela HI, et al: Early detection of irreversible cerebral ischemia in the rat using dispersion of
the magnetic resonance imaging relaxation time, T1rho. J Cereb Blood Flow Metab 20:1457-1466, 2000. Medline
Similar articles
181. Duvvuri U, Poptani H, Feldman M, et al: Quantitative T1rho magnetic resonance imaging of RIF-1 tumors in vivo: detection
of early response to cyclophosphamide therapy. Cancer Res 61:7747-7753, 2001. Medline Similar articles
182. Charagundla S, Duvvuri U, Rizi R, et al: Dynamic 17O imaging with fast T1, rho dispersion MRI. Proc ISMRM 3:2106,
1999.
183. Dixon WT, Oshinski JN, Trudeau JD, et al: Myocardial suppression in vivo by spin locking with composite pulses. Magn
20 of 21 12-04-2010 00:14
184. Slichter CP: Principles of Magnetic Resonance. 3rd ed. Springer Series in Solid-State Sciences. Vol. 1. Berlin: Springer-
Verlag, 1990.
185. Warren WS, Richter W, Andreotti AH, Farmer BT, 2nd: Generation of impossible cross-peaks between bulk water and
biomolecules in solution NMR. Science 262:2005-2009, 1993.
186. Warren WS, Ahn S, Mescher M, et al: MR imaging contrast enhancement based on intermolecular zero quantum
coherences. Science 281:247-251, 1998. Medline Similar articles
187. Rizi RR, Ahn S, Alsop DC, et al: Intermolecular zero-quantum coherence imaging of the human brain. Magn Reson Med
188. Zhong J, Chen Z, Kwok E: New image contrast mechanisms in intermolecular double-quantum coherence human MR
imaging. J Magn Reson Imaging 12:311-320, 2000. Medline Similar articles
189. Zhong J, Chen Z, Kwok E: In vivo intermolecular double-quantum imaging on a clinical 1.5 T MR scanner. Magn Reson
Med 43:335-341, 2000.
190. Joy M, Scott G, Henkelman M: In vivo detection of applied electric currents by magnetic resonance imaging. Magn Reson
191. Beravs K, White D, Sersa I, Demsar F: Electric current density imaging of bone by MRI. Magn Reson Imaging
192. Bossart EL, Wirth ED, Silver XS, et al: Current density magnetic resonance imaging of excised normal and injured rat
spinal cord. Proc Int Soc Magn Reson Med 6:197, 1998.
193. Beravs K, Frangez R, Gerkis AN, Demsar F: Radiofrequency current density imaging of kainite-evoked depolarization.
194. Bodurka J, Bandettini PA: Toward direct mapping of neuronal activity: MRI detection of ultraweak, transient magnetic field
changes. Magn Reson Med 47:1052-1058, 2002. Medline Similar articles
195. Scott GC, Joy ML, Armstrong RL, Henkelman RM: RF current density imaging in homogeneous media. Magn Reson Med
196. Scott GC, Joy ML, Armstrong RL, Henkelman RM: Rotating frame RF current density imaging. Magn Reson Med
197. Mikac U, Demsar F, Beravs K, Sersa I: Magnetic resonance imaging of alternating electric currents. Magn Reson Imaging
198. Birgul O, Eyuboglu BM, Ider YZ: Experimental results for 2D magnetic resonance electrical impedance tomography
(MR-EIT) using magnetic flux density in one direction. Phys Med Biol 48:3485-504, 2003.
199. Lee BI, Oh SH, Woo EJ, et al: Static resistivity image of a cubic saline phantom in magnetic resonance electrical
impedance tomography (MREIT). Physiol Meas 24:579-589, 2003. Medline Similar articles
200. Khang HS, Lee BI, Oh SH, et al: J-substitution algorithm in magnetic resonance electrical impedance tomography
(MREIT): phantom experiments for static resistivity images. IEEE Trans Med Imaging 21:695-702, 2002. Medline
Similar articles
page 230
page 231
*Portions of this chapter appeared in abbreviated form as an article in the January 2004 issue of Diagnostic Imaging.
© 2010 Elsevier
21 of 21 12-04-2010 00:14
ARALLEL MAGING ETHODS

Daniel K. Sodickson
Imaging speed is a crucial consideration in clinical and research applications of magnetic resonance
imaging. High frame rates are required for accurate tracking of dynamic physiologic processes such as
cardiac contraction and tissue perfusion. The comparatively short time intervals associated with
passage of vascular contrast agents also render rapid imaging essential, as does the need to minimize
effects of respiratory or peristaltic motion in clinical imaging examinations.
Unfortunately, imaging speed is fundamentally limited in conventional MRI studies by the underlying
mechanism of spatial encoding. Radiofrequency (RF) pulses excite magnetization in the imaged region,
after which signal data are acquired in sequential bursts or "readouts" as field gradients of varying
amplitude, direction, and/or duration are applied. However, the maximum rate of gradient switching is
limited by the inductance of gradient coils and, in the case of in vivo imaging, by the need to avoid
neuromuscular stimulation from currents induced by the rapidly changing fields. Safety limits on tissue
heating also limit the rate of application of RF pulses. These physical and physiologic constraints on
gradient switching rate and RF power deposition limit the rate at which imaging sequences may be
played out and, consequently, the rate at which new image data may be acquired. Thus, in some ways
the most advanced MR scanner in its traditional mode of operation resembles a more quotidian
scanning device: the fax machine. As is painfully clear to anyone who has waited impatiently for either
a multi-page fax or a lengthy MR scan to conclude, each line of information can be recorded only after
scanning of the previous line is complete.
Since the late 1990s, a new mechanism of spatial encoding has been available to practitioners of MRI.
Parallel MRI uses arrays of radiofrequency detector coils to acquire multiple data points
simultaneously rather than one after the other and thereby circumvents the fundamental speed limits
associated with traditional sequentially acquired studies. The resulting improvements in imaging speed
can be used to extend the capabilities of MRI across a broad range of imaging applications.
Numerous examples of parallel imaging devices exist in nature and in the technology of daily life. The
retina of the eye is one of the most impressive biological examples, capable of capturing images of an
extensive field of view nearly instantaneously in a large array of photodetector cells. The camera,
whether analogue or digital, is a technological example. Figure 8-1 illustrates schematically the
distinction between sequential scanning and parallel imaging, using the familiar optical examples of the
fax machine and the video camera. Despite the existence of such venerable models of parallel imaging,
the medical imaging modalities of CT and MRI are relative newcomers to parallelism, with the required
conceptual and technological advances having appeared only in the past decade. The advent of
multidetector CT has revolutionized imaging speed for existing applications of X-ray computed
tomography and has enabled new applications previously precluded by long image acquisition times.
MRI is currently experiencing the beginnings of a similar revolution.
© 2010 Elsevier
1 of 1 12-04-2010 00:16
HISTORY OF PARALLEL MAGNETIC RESONANCE IMAGING

page 231
page 232
Add to lightbox
1 of 2 12-04-2010 00:16
Add to lightbox
Figure 8-1 Optical examples of sequential scanning and parallel imaging. In a sequential scan (left), a
single detector is moved across the field of the image in order to capture multiple views of the imaged
object (or else the object is moved with respect to the detector). This form of imaging is exemplified by
a fax machine, in which a page to be scanned advances past a line of photodetectors and an image of
the entire page is accumulated line by line. By contrast, a parallel imaging device such as a video
camera (right) acquires all data simultaneously in multiple distributed detectors, allowing rapid image
formation. (Strictly speaking, the fax machine may be viewed as a hybrid device, which operates in
parallel along the line of photodetectors and in a sequential mode along the direction of paper motion.)
The first proposals for parallel acquisition in MRI actually predate the development of multidetector CT
by a number of years. In the late 1980s, use of multiple RF coils was proposed either for simultaneous
imaging of multiple separated regions1 or for imaging of a shared region with increased SNR.2,3 As
early as 1987, even before the use of coil arrays had become commonplace, suggestions for
4
reconstruction algorithms using multiple coils for faster imaging of shared regions had appeared. It
was also recognized early on that signal detection with large arrays of small coils could in principle
replace gradient-phase-encoding,5-7 although both practical and theoretical obstacles to this ambitious
goal precluded extensive development. By the early 1990s, predecessors of the image-domain
Cartesian SENSE technique had been proposed, in which coil arrays were used to remove aliasing
8,9
from undersampled data sets. A k-space based expansion technique using nested volume coils to
approximate multiple lines in k-space simultaneously was also described.10 For these latter
proposals,8-10 accelerated imaging was demonstrated in phantoms. However, whether as a result of
the limited availability of multiple-receiver systems at the time or as a result of limitations in the early
algorithms and results, these proposals did not catch the general attention.
Parallel MRI saw in vivo implementation with the SMASH technique in 1996-199711 and a phase of
rapid development followed. The first self-calibrating parallel imaging technique, AUTO-SMASH, 12
appeared the following year. The popular SENSE technique was introduced in 1998-199913 and
various alternative or hybrid approaches have been described since then.14-32 A selection of these
techniques will be touched on in the sections to follow.
Today, commercial implementations are available on most major MR manufacturers' platforms and
parallel MRI is used routinely at numerous centers. At the time of writing, parallel MRI has begun to
move from the realm of research and development into the mainstream clinical arena. It is the aim of
this chapter to introduce some of the basic principles of parallel MRI, to illustrate its broad utility as a
practical tool, and to suggest certain changes in imaging paradigms that may be enabled by future
developments, particularly at high magnetic field strengths and at high levels of acceleration.
© 2010 Elsevier
2 of 2 12-04-2010 00:16
BASIC PRINCIPLES OF PARALLEL MAGNETIC RESONANCE IMAGING

Spatial Encoding and Decoding
page 232
page 233
Any parallel imaging device relies for its operation upon a distributed array of detectors (such as the rods and
cones in the retina or the photosensitive grains in photographic film). In MRI, RF coils constitute the elementary
detector units that receive the signal from precessing nuclear spins. One might imagine, then, that a parallel MRI
system would involve a set of RF coils arrayed around a patient, with each coil providing information about a
distinct portion of the imaged field of view.
Extraction of the appropriate spatial information from such an array is the job of parallel image reconstruction
algorithms. Two principal challenges exist for image reconstruction in parallel MRI as opposed to the optical
examples of parallel imaging previously cited. The first is the comparatively long wavelengths of RF radiation at the
Larmor frequencies associated with typical magnetic field strengths (e.g., 1.5 Tesla) used in clinical MRI. Since
these wavelengths are of the same order as the dimensions of the human body as a whole, basic principles of
optics dictate that electromagnetic radiation from the spins themselves cannot be focused well enough to achieve
anything resembling the millimeter or submillimeter resolution typically associated with MRI. This is in fact why the
advance recognized by the awarding of the 2003 Nobel Prize in Medicine or Physiology-the concept of "induced
local interactions" with magnetic field gradients33-was required to initiate the modern era of MRI.
Since the MR signal itself is not well focused, parallel MRI must rely upon the properties of the detectors
themselves and in particular upon their spatial distribution of sensitivity to MR signal. Figure 8-2A shows a
photograph of a sample RF coil array with four adjacent loop elements. Simulated spatial sensitivity patterns for
each of the elements of the array are shown beneath a schematic depiction of the array in Figure 8-2B. The
individual sensitivity patterns are peaked beneath each array element and fall off smoothly with distance from each
coil. The fall-off is so smooth, in fact, that the sensitivities are broad and overlapping at any appreciable distance
from the array and this leads to the second challenge for parallel MR image reconstruction: adjacent RF coils do
not generally contain highly distinct spatial information over small spatial scales.
Add to lightbox
Add to lightbox
Figure 8-2 A, Photograph of a four-element array with the underlying coil geometry exposed to view (courtesy of
Randy Giaquinto, GE Global Research Center, Niskayuna, NY, USA). B, Simulated coil sensitivity patterns in a
plane parallel to the array (color map reflects vertical scale).
1 of 6 12-04-2010 00:17
The solution to this problem lies in combining coil-based spatial encoding with the existing mechanism of spatial
encoding with gradients. Figure 8-3 illustrates the approach of the SMASH (SiMultaneous Acquisition of Spatial
Harmonics) technique.11 The function of encoding gradients is to produce sinusoidal modulations of magnetization
components across the field of view: each distinct gradient step corresponds to a modulation with a different
spatial frequency and the signal acquired in the presence of each modulation corresponds to a particular Fourier
component of the image content. Figure 8-3A displays the spatial modulations associated with various phase-
encoding gradient strengths in a traditional sequential scan using a single RF coil. (By comparison with the optical
scanning example of Figure 8-1, in MRI it is of course the gradient strength, rather than the detector position,
which is incremented to yield the scanned signal data set.) As is demonstrated in Figure 8-3B, appropriate linear
combinations of the sensitivity patterns of individual coils in an array may be formed to mimic such sinusoidal
modulations. The result is that a reduced number of time-consuming gradient steps may be applied (schematized
by the thick data line in Fig. 8-3B) and the "missing" data (thin gray lines in Fig. 8-3B) may be filled in by
appropriate linear combinations of signal from the array elements. The acquisition time is therefore reduced by the
ratio of the number of acquired data points to the number of points which would have been required for a fully
sampled data set.
page 233
page 234
Add to lightbox
Add to lightbox
Figure 8-3 Conventional spatial encoding with gradients and parallel spatial encoding with the SMASH technique.
A, In a conventional sequential MR imaging study, evolution of nuclear spins in phase-encoding gradients of
varying amplitude creates linear phase modulations, or sinusoidal modulations of the transverse components of
magnetization, across the object. The signal detected in any given RF coil represents the underlying magnetization
density projected against these sinusoidal modulations, yielding various spatial Fourier components of the object
content. Each line in the matrix of acquired data contains all measured coefficients with the same Fourier index, or
2 of 6 12-04-2010 00:17
k-space index, along a given direction in the image plane. An image is created by Fourier transformation of the
signal data. B, In the SMASH parallel image reconstruction procedure, individual coil sensitivities (shown beneath
an eight-element array) are weighted and combined to approximate sinusoidal modulations across the field of
view. These modulations serve to emulate the effects of phase-encoding gradients. The result is that a reduced set
of phase-encoding steps (thick lines) may be applied in a reduced time and the remaining data (thin lines) may
be synthesized after the fact by combinations of component coil data that emulate the missing gradient
modulations.
An alternative picture of parallel image reconstruction may be formed by appealing to the concept of Nyquist
aliasing. When an undersampled data set is acquired, the corresponding images after Fourier transformation are
"folded" onto themselves (see Fig. 8-4). Basic principles of digital signal processing prevent unfolding of the
aliased positions in any individual coil's image. However, each of the coils in an array provides a distinct "view" of
the aliasing, with the combination of aliased points weighted by that particular coil's sensitivity pattern. Therefore,
using knowledge of the coil sensitivity patterns, appropriate linear combinations of aliased component images may
be formed to produce the single unaliased image which would have resulted from a fully sampled data set. This is
the concept behind the original Cartesian SENSE (SENSitivity Encoding) reconstruction.13
page 234
page 235
Add to lightbox
page 235
3 of 6 12-04-2010 00:17
page 236
Figure 8-4 Schematic illustration of the Cartesian SENSE parallel image reconstruction procedure. When a
reduced set of phase-encoding gradient steps is used, individual coil images suffer from Nyquist aliasing. For
example, intensities from the red and blue pixel positions are superimposed at the aliased yellow pixel position.
Differential shading of the images by each coil's sensitivity pattern changes the relative intensities of the
superimposed positions. In SENSE, knowledge of the coil sensitivities is then used to produce an "unfolded"
image in which the aliased positions have been separated.
Add to lightbox
4 of 6 12-04-2010 00:17
Add to lightbox
Figure 8-5 Generalized projections in X-ray CT (A) and in parallel MRI (B). Though the apparatus (top) and the
particular projection functions (bottom) differ between the modalities, the process of image reconstruction from
projections is analogous.
Clearly, the common theme of SMASH and SENSE approaches is the weighted combination of individual coil data
based on knowledge of the component coil sensitivities. Various other parallel image reconstruction techniques that
have been proposed may generally be classified as SMASH-like, SENSE-like or hybrid approaches based on their
particular algorithms for deriving and applying the necessary weightings of component coil data. Meanwhile, a
generalized picture of the encoding process in parallel imaging has emerged, which suggests intriguing analogies
with X-ray CT. This picture is illustrated in Figure 8-5. In a CT scan (Fig. 8-5A), various projections of the imaged
region are recorded during rotation of an X-ray gantry and the internal image content is reconstructed through
algebraic manipulation of the set of projections. In an MR scan using an array of RF coils (Fig. 8-5B), the signal
detected in each coil at each setting of an encoding gradient actually represents a generalized projection of the
object content against complex modulations containing both sinusoidal gradient modulations and coil sensitivity
modulations (see Fig. 8-5B, bottom right). Though the nature of these generalized projections differs from the CT
case, nevertheless they may be manipulated in a similar fashion to form images. This is the approach of the
generalized algorithms used in non-Cartesian SENSE, 13,23 SPACE-RIP,16 Generalized Encoding Matrix (GEM),19
and Generalized SMASH25 reconstructions. Since data are acquired simultaneously in all array elements, multiple
projections are available in parallel and the number of time-consuming gradient steps can be reduced (by a factor
less than or equal to the number of independent coils) while still preserving the full image information.
The Mathematics of Parallel Magnetic Resonance Imaging

The fast Fourier transform (FFT) algorithm is the mainstay of MR image reconstruction for standard gradient-
encoded imaging (supplemented as needed by gridding algorithms34 to bring irregularly spaced data into a form
compatible with the FFT). For parallel image reconstruction, this algorithmic repertoire must be expanded to
include weighted combinations of k-space lines, pixels or encoding functions, along with procedures to determine
the correct weightings. Such operations are most conveniently expressed in the language of linear algebra.
5 of 6 12-04-2010 00:17
For SMASH-like approaches, the FFT is preceded by a linear combination of acquired k-space lines from various
coil array elements. The weights for the combination are determined in advance by fitting known sensitivity
functions to the target harmonic modulations-a procedure which may be performed compactly using a matrix
15 12,24,29
pseudoinverse algorithm. For certain self-calibrating approaches, the fitting procedure uses reference
signal data in place of coil sensitivity data but the basic mathematics are similar. In image-domain SENSE,13 the
FFT of individual component coil signals to yield component images is followed by linear combinations of pixel
values, with weights determined beforehand by inversion of a matrix containing known coil sensitivity values at the
various aliased positions to be separated. In generalized approaches,13,16,19,23,25 an encoding matrix is formed by
combining measured coil sensitivities with spatial modulations from known gradient-encoding steps and arranging
these generalized encoding functions into a suitably discretized set. The encoding matrix is then inverted and
multiplied by suitably rearranged signal data to yield the reconstructed image.
page 236
page 237
Additional details of the various reconstruction algorithms, beyond these basic building blocks, may be found in the
parallel imaging literature. The original SENSE article,13 for example, contains a rigorous exposition of the
mathematical principles underlying reconstruction of images from coil- and gradient-encoded data. Detailed
connections between SMASH-like and SENSE-like reconstructions are explored in subsequent articles. 19,35,36
Generally speaking, the use of algebraic techniques in parallel image reconstructions entails a loss of certain
convenient properties of the FFT, including efficient computation (though efficient parallel image reconstruction
algorithms do exist) and unitarity (a property of the simple sinusoidal functions produced by gradient encoding).
This departure has certain practical implications, particularly for signal-to-noise ratio (SNR).
© 2010 Elsevier
6 of 6 12-04-2010 00:17
PRACTICAL CONSIDERATIONS
Signal-to-Noise Ratio
The increased speed associated with parallel imaging does not come without a price. Most notably, as a
result of both the data acquisition strategies and the image reconstruction algorithms used in parallel MRI,
the SNR of a parallel imaging study is always reduced as compared with an unaccelerated study obtained
using the same coil array.13,37 The scaling of SNR may be expressed as follows:
Here, R represents the acceleration factor and the square-root dependence in Equation 8-1 reflects the
reduced noise averaging resulting from a reduced number of acquired data points in an accelerated scan.
The so-called "geometry factor", g, on the other hand, is an additional SNR reduction factor which results
from the fact that the linear combinations used in a parallel image reconstruction may amplify noise out of
proportion to signal. This noise amplification depends sensitively upon the choice of image reconstruction
algorithm and upon coil array design, though it is also subject to certain fundamental electrodynamic
limits.38,39 The value of g varies spatially across the image plane, meaning that the noise background is
generally non-uniform in a parallel imaging study.
Add to lightbox
1 of 6 12-04-2010 00:17
Figure 8-6 Noise amplification with increasing acceleration. A four-element array is shown at the top with
its component coil sensitivities. Surface plots of g factor for acceleration factors R = 1 to 4 using this array
are shown to the left of simulated images. The noise patterns in the images match the g-factor patterns.
Figure 8-6 illustrates the amplification of noise for increasing acceleration factors along the principal axis
of a four-element array (shown at the top with its component coil sensitivities). In the left-hand column,
surface plots of the g-factor map across the image plane are shown, with corresponding "phantom"
images shown to the right. As one can appreciate from the figure, both the overall magnitude and the
inhomogeneity of the g-factor increase with increasing acceleration. Peaks in the g-factor map correspond
to areas of increased noise, and hence decreased SNR, in the images. This structured noise background
may be an unfamiliar feature for those unaccustomed to parallel imaging and it should be taken into
account in image interpretation. The particular location and shape of noise peaks and valleys are
influenced by the choice of image plane and the acceleration factor. In Figure 8-6, for example,
particularly sharp features are seen in the g-factor map for an acceleration factor of four-the maximum
possible acceleration for a four-element array. Generally speaking, g-factor maps display a repeating
structure at spatial intervals equal to the field of view divided by the acceleration factor. Noise
amplification may be controlled to some extent by numerical conditioning in reconstruction algorithms-an
area which has been the subject of active research in the field of parallel image reconstruction. SNR
behavior is also influenced powerfully by coil array design.
page 237
page 238
Coil Array Design

3
The traditional "MR phased array" design of partially overlapped loops is generally well suited for parallel
acquisition and most commercially available coil arrays are compatible with some form of parallel MRI.
Designs that yield optimal SNR for traditional gradient-encoded imaging, however, are not necessarily
optimal for parallel imaging. Various array designs tailored for parallel imaging have been reported in the
literature40-45 and a rich body of ongoing work is represented in yearly conference proceedings and
commercial offerings.
As was outlined in the previous section, optimization of net SNR with parallel imaging in mind requires
attention not only to the intrinsic SNR of the array but also to the noise amplification for target image
40-42,44
planes and accelerations. A variety of loop-element designs have been proposed with these
42
features in mind. In one study, a small number of parameters, including individual element size and
spacing, were varied to maximize total SNR and to minimize noise hot-spots in areas of interest within the
prospective field of view. The result of this empirical optimization procedure was a design with gapped,
rather than overlapped, adjacent elements. Alternatively, one may seek in a design to yield flexible
performance for a range of imaging situations. In this case, it is important to provide elements with distinct
sensitivity variations along multiple directions.
In an effort to produce distinct encoding functions, designs using alternative element geometries departing
from the traditional loop structure have also been explored, including arrays made up of isolated
strips,43,45-47 arrays with overlying crossed elements,48-50 and volume arrays with tailored interior field
shapes.51-53 Another promising approach is the use of arrays with large numbers of elements. 54,55
Although this latter approach presents a number of practical and conceptual challenges, it may be shown
that, at least in principle, such highly subdivided arrays can approach the optimum achievable SNR while
increasing the overall flexibility of parallel imaging performance. We will return to the question of multiple-
element arrays and ultimate SNR limits at the end of this chapter.
Though a great deal more can be said about coil array design for parallel imaging, the bottom line for
prospective users is this: arguably even more than is the case for traditional sequential imaging, there is
no single optimal coil design for all parallel imaging applications. Though many modern coil arrays may be
advertised as "Optimized for parallel MRI", it is important to assess the performance of any candidate
array in image planes and imaging situations that match your needs.
Coil Sensitivity Calibration
2 of 6 12-04-2010 00:17
One practical consequence of the use of RF coil arrays for spatial encoding is that calibration of the
component coil sensitivity profiles is required. This is analogous to the calibration of field gradient profiles
that is required for optimal performance of commercial MR scanners. For parallel MRI, both the gradient
profiles and the coil sensitivities must be known accurately in order to reconstruct accurate images.
However, since coil arrays may be flexible and may conform differently to different patients, and since
some of their electrical characteristics can change from study to study, coil sensitivity calibration is
typically performed separately for each patient. Two principal calibration strategies have been described
and are schematized in Figure 8-7: external calibration and self-calibration.
For externally calibrated studies (Fig. 8-7A), a short low-resolution scan is obtained, generally of a large
volume containing all image planes of potential interest, prior to the execution of accelerated scans. Coil
sensitivities in particular target image planes are then derived as needed by post-processing of the
acquired calibration volumes. Various approaches have been proposed, both in the primary literature and
in conference proceedings, for the acquisition and postprocessing of sensitivity calibrations. Underlying
magnetization density variations may be removed from in vivo sensitivity estimates through division by a
body coil or composite reference image with subsequent filtering and extrapolation. 13 Alternatively, direct
filtering of the in vivo calibration images may be used56 or else the target image intensity may be tailored
to render the reconstruction insensitive to underlying magnetization density variations. 15,19
One potential difficulty associated with external calibrations is that motion of the patient and/or the coil
array between the time of the calibration and the accelerated scan can result in calibration errors and
12,24,26,29
corresponding image artifacts. In order to address this difficulty, self-calibrating approaches use
a small set of calibration lines placed within the accelerated scan itself (see Fig. 8-7B). Typically, a small
central region of k-space is sampled fully while the remaining data set is undersampled and the fully
12
sampled portion is processed to yield a sensitivity reference. In the case of the AUTO-SMASH,
24
Variable-Density AUTO-SMASH, and GRAPPA (GeneRalized Autocalibrating Partially Parallel
29
Acquisitions) techniques, calibration information is obtained by fitting acquired data lines to one another
within the fully sampled region. Alternatively, the central fully sampled region may be Fourier-transformed
directly to yield low-resolution sensitivity reference images which may then be used with a parallel image
reconstruction technique of choice.26 The self-calibrating approach eliminates the need for a separate
calibration scan and improves the registration of the calibration data with the remaining accelerated data.
The trade-off is that full sampling of the central portion of k-space may require a sacrifice either in total
acceleration or in SNR as compared with an externally calibrated study, except in the case of certain
non-Cartesian k-space trajectories with inherently dense central sampling.57
Imaging Sequences and k-Space Trajectories

page 238
page 239
3 of 6 12-04-2010 00:17
Add to lightbox
page 239
page 240
Figure 8-7 Coil sensitivity calibration strategies for parallel imaging. A, External calibration. A
low-resolution calibration data volume is acquired separately from the accelerated scan and is processed
4 of 6 12-04-2010 00:17
to yield sensitivity estimates in the target image plane. These sensitivities are then used for parallel image
reconstruction. B, Self-calibration. Additional calibration lines (shown in gray) are acquired along with the
undersampled accelerated data sets. These calibration lines are generally placed so as to produce a fully
sampled region in the center of k-space. The fully sampled region for each component coil's data set may
then be Fourier transformed or otherwise processed to yield the sensitivity information necessary for
parallel image reconstruction.
As an acquisition strategy rather than an imaging sequence per se, parallel imaging may be combined
productively with most existing pulse sequences. The full range of MR contrast options are therefore
available for accelerated scans. The choice of sequence may affect both the details of image
reconstruction and the behavior of image contrast, however, and certain sequences have particular
advantages when combined with parallel imaging.
The straightforward aliasing relationships illustrated in Figure 8-4 only apply when the undersampled signal
data are regularly spaced on a Cartesian grid, as is the case for the bulk of Fourier imaging applications.
When irregular data spacings are used, for example with non-Cartesian k-space trajectories such as
spiral or radial trajectories, each voxel position in an undersampled data set may be connected with all
other voxel positions in the image, rather than with a discrete set of well-defined positions as in Figure
8-4. Parallel image reconstruction may still be performed but generalized techniques must be used, with a
potentially dramatic increase in computational burden. Fortunately, efficient iterative techniques have been
described23 which render non-Cartesian parallel image reconstruction practical. Hybrid techniques have
also been proposed, which divide the reconstruction into multiple smaller reconstruction problems for
58,59
distinct regions of k-space.
For 3D imaging sequences or other sequences with multiple phase-encoded directions, acceleration may
be applied along multiple directions simultaneously, in which case the net acceleration factor is the product
of individual acceleration factors along each dimension. When a coil array suitable for multidimensional
encoding is available, such a multidimensional approach has been shown to have SNR advantages as
compared with one-dimensional accelerations using the same net acceleration factor.28 This is a result of
the spreading of the spatial encoding burden along more than one dimension; it is easier, for example, to
separate four corners of a square (2 × 2 acceleration) than to distinguish four points along one side of the
same square (4 × 1 acceleration).
For those who aim at ultrashort imaging times, it is a fortunate turn of events that some of the fastest
existing pulse sequences-namely single-shot sequences-tend to benefit most dramatically from parallel
imaging accelerations. Simply by virtue of allowing faster acquisitions, parallel imaging limits relaxation-
related signal attenuations that can degrade image quality for prolonged echo trains. Accelerated fast
spin-echo sequences, for example, may show improved spatial resolution as compared with their
unaccelerated counterparts.60 Echo-planar imaging (EPI) sequences also behave synergistically with
acceleration, since time-dependent phase accumulations responsible for susceptibility artifacts and other
61,62
imperfections in these sequences are also limited when echo train length is reduced.
Even for multishot sequences, careful sequence design can result in potentially unexpected benefits. For
example, appropriate modifications of bandwidth, TR, and flip angle in accelerated gradient-echo
sequences can more than make up for the SNR losses associated with parallel imaging and can actually
yield improved SNR for a fixed acquisition time.63
Scan Planning, Image Artifacts, and Image Interpretation

As is the case for traditional sequential imaging, coil arrays in a parallel imaging study should be placed so
as to bring the target anatomy within "reach" of the array's sensitivity pattern. One additional constraint
particular to parallel imaging is that the phase-encoding direction/s targeted for acceleration should be
aligned as far as possible with axes along which the array elements have substantially distinct sensitivity
variations. By way of example, for a one-dimensional array with adjacent elements disposed from left to
right, selection of a left-right phase-encode direction will yield superior accelerated image quality as
compared with an anterior-posterior choice.
One additional constraint on scan prescription that is particular to parallel imaging is the requirement that
5 of 6 12-04-2010 00:17
the target "full" field of view (FOV) after reconstruction be free of aliasing along any direction to be
accelerated. This requirement can, in some cases, limit the effective acceleration of a parallel imaging
study if the unaccelerated comparison technique uses a significantly aliased rectangular FOV. At least for
image-domain SENSE reconstructions, overlap of structures in the target field of view leads to ambiguities
in the partitioning of intensities among aliased positions, resulting in image artifacts. 64 (The effect of "full
FOV" aliasing in other reconstructions such as GRAPPA is not as clear and some relaxation of the FOV
constraint may prove to be possible for these reconstructions.)
Artifacts may also result from errors in sensitivity calibration. At least for regularly undersampled data
sets, these artifacts tend to take the form of residual aliasing, with spurious intensity appearing at image
positions which were formerly folded onto one another in the undersampled source data sets. Figure 8-8
compares an "artifact-free" fourfold accelerated abdominal image (Fig. 8-8A) with an image for which the
coil sensitivity references were deliberately shifted by several pixels (Fig. 8-8B). The locations of residual
aliasing artifacts (indicated by arrows in Fig. 8-8B) are predictable given the acceleration factor. If
anomalous intensities appear in these locations, calibration error or some other reconstruction error
should be considered as a possible cause.
© 2010 Elsevier
6 of 6 12-04-2010 00:17
MANUFACTURER IMPLEMENTATIONS
page 240
page 241
Add to lightbox
Add to lightbox
Figure 8-8 Artifacts that may be encountered in parallel imaging studies. A, Fourfold-accelerated axial
image of the abdomen of a healthy adult subject, obtained with an encircling eight-element array. Note
the regions of reduced SNR due to selective noise amplification towards the center of the image. B,
Image reconstructed from the same data set but with coil sensitivity references deliberately shifted by
several pixels in the anterior-posterior phase-encoding direction. Residual aliasing artifacts from the
back and the chest wall are indicated with arrows. Any mismatches between sensitivity references and
accelerated data may result in artifacts of this sort.
1 of 2 12-04-2010 00:18
In practice, the following steps are required for a parallel imaging study:
calibration scan (if external calibration is used)

image plane and parameter selection for the accelerated scan
computation of necessary weight factors for image reconstruction (generally using either a
direct or an iterative procedure for matrix inversion based on derived coil sensitivity information)
execution of the accelerated scan
weighted combination of acquired data, and
display of the reconstructed images.
Most of these steps can be made transparent to the user. In existing commercial implementations of
parallel imaging, a comparatively small number of parameters is required to plan and execute
accelerated scans.
Figure 8-9 shows sample graphical display screens from the consoles of three major manufacturers'
systems in the year 2003. Philips Medical Systems (Fig. 8-9A) has maintained the SENSE acronym
and a SENSE algorithm is used for image reconstruction. A volumetric external sensitivity reference is
used, which may be selected from a set of standard protocols. The approach implemented by General
Electric Medical Systems (Fig. 8-9B) is the Array Spatial Sensitivity Encoding Technique (ASSET),
which also uses a SENSE-like algorithm with a volumetric external sensitivity reference. A scan may be
identified as either a sensitivity reference or an accelerated ASSET scan in the list of scan options.
Siemens Medical Solutions' integrated Parallel Acquisition Technology (iPAT) gives the user a choice
between two image reconstruction strategies: the SMASH-like hybrid GRAPPA technique29 or a
modified SENSE (mSENSE) technique (Fig. 8-9C). Both GRAPPA and mSENSE employ
self-calibration so no separate sensitivity calibration scan is required.
Parallel imaging products are also available from other manufacturers, such as Toshiba (SPEEDER)
and Hitachi (RAPID). Most MR system and RF coil manufacturers also offer coil arrays designed with
parallel imaging applications in mind. The capabilities of commercial products are evolving continually
and this chapter is not intended as a detailed or complete review of available platforms-the interested
reader is encouraged to ask his or her manufacturer about current offerings.
© 2010 Elsevier
2 of 2 12-04-2010 00:18
CLINICAL APPLICATIONS AND IMPACT

Within the constraints of available SNR, the increased speed and efficiency associated with parallel MRI may be
invested in various ways, including:
shortening long exams

improving spatial resolution
improving temporal resolution
improving image quality.
Figure 8-10 shows images from a sampling of application areas, illustrating some of these strategies. Figures
8-10A and B demonstrate the benefits of reduced breath-hold duration for abdominal imaging. Twofold acceleration
of a volumetric liver examination using the hybrid Parallel Imaging with Augmented Radius in k-Space (PARS)
technique58 resulted in reduced motion artifact for a patient incapable of long breath-holds. Figures 8-10C and D
demonstrate the use of ASSET for increased spatial resolution at fixed imaging time in contrast-enhanced breast
studies. Figures 8-10E-H are time-resolved frames from an accelerated realtime cardiac imaging study obtained
using a self-calibrated GEM technique. Figures 8-10I-J illustrate the beneficial effects of SENSE accelerations for
EPI imaging of the brain at 3 Tesla field strength. The susceptibility artifacts which can plague single-shot EPI
studies, particularly at high field strength, are markedly reduced by shortening the image acquisition time.
page 241
page 242
Add to lightbox
Add to lightbox
1 of 5 12-04-2010 00:19
Add to lightbox
page 242
page 243
Figure 8-9 Screen shots of graphical user interfaces demonstrating vendor implementations of parallel imaging by
(A) Philips (SENSE), (B) GE (ASSET), and (C) Siemens (iPAT). (Images courtesy of Dr Johan van den Brink,
Philips Medical Systems, Best, The Netherlands (A); Dr Kevin King, GE Medical Systems, Waukesha, WI, USA
(B); and Dr Berthold Kiefer, Siemens Medical Solutions, Erlangen, Germany (C).)
Add to lightbox
Add to lightbox
2 of 5 12-04-2010 00:19
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
3 of 5 12-04-2010 00:19
Add to lightbox
Add to lightbox
Add to lightbox
page 243
page 244
Figure 8-10 Selected applications of parallel imaging. A,B, Volumetric Interpolated Breath-hold Examination
(VIBE) images of the abdomen at 1.5 T, from 3D data sets obtained without (A) and with (B) parallel imaging. In A,
the patient was unable to tolerate the breath-hold duration of 26 seconds, resulting in significant respiratory motion
artifact, whereas in B, twofold acceleration using a hybrid reconstruction technique resulted in a successful
13-second breath-hold. C,D, Contrast-enhanced breast images from 3D data sets, obtained without (C) and with
(D) parallel imaging. In D, an ASSET acceleration factor of two was used to increase in-plane matrix size from 256
× 256 to 320 × 320 in the same total imaging time of 1 minute per volume. Note the improved distinctness of
vessels in the accelerated scan. E-H, Successive time-resolved frames from a realtime cardiac imaging study,
accelerated by a factor of 1.5 using a self-calibrating GEM technique. Motion of the heart, chest wall, and
diaphragm is captured effectively with high temporal resolution. I-J, diffusion-weighted echo-planar images
obtained at 3 Tesla, showing artifactual image stretching from global field distortion, along with hyperintense foci
related to more localized susceptibility effects. These artifacts, which are pronounced in the unaccelerated study (I),
decrease markedly with a SENSE factor of three (J), due to the decreased echo train lengths associated with a
reduced number of phase-encode steps. (Images in A-D courtesy of Dr Charles A McKenzie and Dr Neil M
Rofsky, Beth Israel Deaconess Medical Center, Boston, MA, USA. Images in I-J courtesy of Dr Johan van den
Brink, Philips Medical Systems, Best, The Netherlands.)
65-68 15,20,32,69-73 74,75

In addition to the body, cardiac, and brain imaging applications surveyed in Figure 8-10,
particularly promising areas of application for parallel imaging include contrast-enhanced MR angiography, 76-82
functional MRI,83-87 and diffusion-tensor imaging.88-91 Other areas of application explored in the literature include
pulmonary imaging,82,92,93 spectroscopic imaging,94,95 musculoskeletal imaging,96-98 pediatric imaging,99 and
dynamic/interactive MRI.20,32,72,73
A number of uses of parallel imaging for more than mere acquisition speed have also been described. As
60
mentioned in the preceding section, spatial resolution can be enhanced and susceptibility artifacts can be
61,62
reduced by using parallel imaging to decrease the duration of long echo trains. Parallel imaging has also been
used to reduce gradient acoustic noise.100 Use of redundant information in fully sampled data sets has been shown
to allow detection and correction of motion during the acquisition.101,102 Application of parallel imaging principles to
transmission with multi-element arrays has been used to enable reductions in transmit pulse durations.31,103
In summary, the currency of imaging speed associated with parallel MRI may be spent wherever speed is required
or else the extra information inherent in the use of RF coil arrays may be invested in other aspects of the imaging
protocol. The interested reader may find additional information on parallel imaging methods and applications in
104,105
review articles covering both the early development period and the more recent era of clinical
4 of 5 12-04-2010 00:19
106-108
implementation.
© 2010 Elsevier
5 of 5 12-04-2010 00:19
LIMITS OF PARALLEL MAGNETIC RESONANCE IMAGING AND FUTURE

DIRECTIONS
Some of the earliest proposals for parallel MRI envisioned massively parallel implementations in which
spatial encoding using multiple-element coil arrays might entirely replace gradient-phase-encoding. 5,6
Until recently, however, relatively modest accelerations, typically by factors of 2-4 or, in special cases,
6-8, have been achieved with parallel imaging approaches. The large gap between these
implementations and the vision of massively parallel imaging results in part from practical
considerations (e.g., the complexity and expense of multiple-element arrays and multiple-receiver
systems) and in part from theoretical SNR limitations. Based on empirical observations of the scaling
of SNR with increasing acceleration factor, it has been suggested over the years that practical in vivo
accelerations would never greatly exceed current moderate levels, at least for typical body imaging
applications. Several recent developments in the field of parallel imaging have either challenged or
clarified such a contention. In light of these developments, it seems appropriate, then, to conclude this
chapter with an oft-asked question: "How fast can we (really) go?".
On the one hand, the elimination of phase-encoding has recently been demonstrated with the SEA
technique,109 which used a 64-element array55 to acquire entire images in a single echo. This work
hearkens back to the early models of massively parallel imaging and throws down the gauntlet to an
MR engineering community who might begin to contemplate large-scale designs for in vivo massively
parallel imaging. On the other hand, numerical computations have established the existence of
fundamental electrodynamic limits on parallel imaging performance, indicating that SNR will always
suffer dramatic degradation for high accelerations at any appreciable depth within the imaged
body.38,39,110-112 Indeed, the high accelerations obtained with SEA were possible only at extremely
shallow depths beneath the finely segmented linear coil array. What freedom exists within the
theoretical limits of parallel imaging performance to achieve routinely high accelerations for practical in
vivo applications? How can we optimize our parallel imaging apparatus to approach these limits of
performance?
Figure 8-11 shows computed values of ultimate intrinsic SNR at the center of an elliptical volume
(modeled after the human torso) as a function of acceleration factor for various acceleration strategies
and magnetic field strengths. The curves in the figure were obtained by using a large basis set of
possible solutions to Maxwell's equations in order to calculate the maximum achievable SNR and g
factor independent of any particular coil array design.38 The most prominent feature of all the curves in
Figure 8-11 is a rapid decrease in achievable SNR beyond a comparatively low threshold level of
acceleration.
Some leeway clearly remains, however. For example, as shown in Figure 8-11A, the use of
multidimensional accelerations can dramatically mitigate ultimate SNR losses-a result which is
consistent with experimental experience to date. 28 Large multidimensional accelerations call for
multiple-element arrays with multidimensional encoding capabilities. One practical concern for the
construction of such arrays involves the effect of inductive couplings between elements, which might be
expected to spoil the distinctness of individual coil sensitivities. Inductive coupling, however, has been
113,114
shown to have a smaller effect on parallel imaging performance than was originally imagined.
Another common concern is that depth penetration will decrease as the number of elements covering a
given total surface area increases and the size of each individual element decreases correspondingly.
However, it has been demonstrated that the use of large arrays of small elements does not in itself
constitute a limit on depth penetration, contrary to rules of thumb developed for single coil elements.
High accelerations still hurt the SNR bottom line but superposition of signals from many elements can
actually improve depth penetration, as least so long as the noise from individual coil circuits can be
controlled. This principle was originally demonstrated for standard sequential imaging115-117 and has
118
been verified in the context of parallel imaging. Indeed, the principle that an infinite number of
(idealized) infinitesimal array elements will provide the best available SNR is the very principle upon
which ultimate intrinsic SNR calculations are based.
page 244
1 of 7 12-04-2010 00:19
page 245
Add to lightbox
Add to lightbox
Add to lightbox
Figure 8-11 Scaled ultimate intrinsic SNR as a function of acceleration factor for various acceleration
strategies and magnetic field strengths (adapted with permission from reference 38). SNR was
calculated at the center of an elliptical volume simulating the human torso. In all cases, the square-root
dependence on acceleration factor has been removed from the curves, so as to emphasize the
constraints particular to parallel imaging. A, Relative SNR (normalized to unity at an acceleration factor
R of 1) for 1D versus 2D acceleration. The 2D accelerations (solid line), which represent the product
of equal acceleration factors along each of two phase-encoding directions, yield significantly higher
intrinsic SNR than their 1D counterparts (dashed line) at the same net acceleration factor. B, SNR at
various field strengths. The increase in baseline SNR with increasing field strength may be
2 of 7 12-04-2010 00:19
appreciated from these plots, which are scaled relative to the 1.5 T SNR which was normalized to unity
at R = 1. Line types are given by the legend in C. C, The same data as in B, but with the curves for all
field strengths normalized to unity at R = 1. The critical turning point in ultimate intrinsic SNR shifts to
higher acceleration as field strength increases.
At the time of writing, acceleration factors as high as 24 have been demonstrated in vivo using a
32-element array designed for multidimensional spatial encoding on an experimental whole-body 1.5 T
MR system with 32 channels54; 12- to 16-fold accelerated acquisitions were demonstrated repeatably
for 3D contrast-enhanced MR angiography studies over large and clinically relevant imaging
54,119
volumes. Figure 8-12A-K shows several examples of large-volume highly accelerated MR
angiograms using the 32-element array pictured in Figure 8-12L-M. In the absence of parallel imaging,
the volumetric acquisitions would have required 54 seconds (A-E), 58 seconds (F-I) or 4 minutes 24
seconds (J-K) respectively, intervals clearly too prolonged for breath-holding by the vast majority of
patients. With an acceleration factor of 12 to 16, multiple dynamic acquisitions with full volumetric
coverage were possible at 3.5- to 4.5-second intervals (A-I) or else high spatial resolution could be
obtained over the full volume in a single 22-second breath-hold (J-K).
Volumetric imaging is a particularly appealing candidate for highly parallel imaging, not only because of
the availability of multiple directions suitable for acceleration but also because SNR in 3D sequences
increases with the quantity of acquired data. Highly parallel MRI enables large volumetric acquisitions
with otherwise prohibitive imaging times and the resulting gains in baseline SNR serve to offset, at
least in part, the SNR losses associated with parallel imaging. At the same time, large volumetric
acquisitions allow a simplification of scan prescription, enabling, for example, visualization of full
vascular trees at the press of a button, as an alternative to the careful and patient-specific planning of
limited vascular studies. As many-receiver imaging systems become more widely available, one might
envision a shift in the paradigm of routine MR scanning, moving away from targeted imaging of limited
regions of interest and towards rapid volumetric imaging in the style of multidetector CT. Recent
moves by several MR manufacturers towards commercial systems with 32 or more receiver channels
will clearly pave the way for such possibilities. The practical design and operational requirements of
multiple-element arrays promise to motivate advances in conductor arrangement, cabling methods,
substrate materials, and image reconstruction hardware.
In the meantime, two areas of active research also promise to shift the SNR balance for highly parallel
imaging and perhaps to pave the way for more massively parallel approaches. First, the move to
progressively higher magnetic field strength represents a significant opportunity for parallel imaging
applications. It is well known that parallel MRI can improve high-field imaging, reducing susceptibility
artifacts (see Fig. 8-10I-J) and also addressing specific absorption ratio (SAR) constraints. As may be
appreciated from Figure 8-11B-C, high-field strength also has a beneficial effect on parallel imaging
performance. To begin with, the increased baseline SNR at high-field (Fig. 8-11B) enables higher
accelerations. In addition, the RF wavelength decreases as field strength increases and the resulting
improvement in RF focusing capability can in fact reduce the noise amplification effects associated with
parallel image reconstruction, thereby increasing achievable accelerations still further. 38,39 This effect is
evident from the normalized SNR curves in Figure 8-11C, which show the critical SNR turning point
moving out to higher accelerations as field strength increases. In short, there is a powerful synergy
between parallel MRI and high-field strength that remains to be exploited for clinical imaging.
page 245
page 246
3 of 7 12-04-2010 00:19
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
4 of 7 12-04-2010 00:19
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
5 of 7 12-04-2010 00:19
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
6 of 7 12-04-2010 00:19
Add to lightbox
Figure 8-12 Contrast-enhanced MR angiograms acquired at an acceleration factor of 12 to 16 in three
healthy adult subjects.54,119 A-D, Coronal maximum intensity projections (MIPs) of multiple dynamic
4.5-second volumetric acquisitions, reduced in duration by a factor of 12 from 54-second acquisitions,
are shown as successive time-resolved frames, identified by time delay following contrast bolus
injection. (Contrast in the renal collecting system in A is residual from a previous injection.) Given the
high temporal resolution of this accelerated study, no timing bolus was required. Yellow dashed lines in
E show typical restricted slab dimensions for a conventional unaccelerated renal MR angiography
study, as compared to the full volumetric coverage (blue outline) of the accelerated study. F-I, Various
rotations of volume-rendered data from the pulmonary phase of another time-resolved angiographic
study, accelerated by a factor of 16 (58 s to 3.6 s). J,K, Volume-rendered data with high spatial
resolution shown in various rotated frames. Full volumetric coverage of the imaged structures was
obtained in a 22-second breath-hold, reduced by a factor of 12 from an impractical 4 minute 24
seconds. The image in J was obtained in the first postcontrast phase, timed for optimal arterial
enhancement. The image in K (showing a zoomed segment of the full acquired volume) was
generated from a second postcontrast acquisition obtained immediately following the first acquisition.
L,M, Photographs of the prototype 32-element multidimensional array used for these studies.
120,121
A second area of opportunity lies in the development of new hyperpolarized contrast agents. The
dramatic improvements in baseline SNR associated with these contrast agents may also be synergistic
with parallel imaging approaches, allowing highly accelerated acquisition of large data volumes during
the comparatively short duration of enhanced polarization.
© 2010 Elsevier
7 of 7 12-04-2010 00:19
CONCLUSION
To return, in closing, to optical analogies, MRI has begun to depart from its roots as a sequential
scanning modality and to take on some of the characteristics of a modern camera. While low-level
accelerations enabled by parallel acquisition techniques offer a wide range of benefits for routine
clinical use, order-of-magnitude accelerations have the potential to change the paradigm of clinical MR
imaging, allowing rapid volumetric imaging in the style of multidetector CT while preserving the wide
range of contrast options and biochemical selectivity associated with MRI.
REFERENCES
1. Hyde JS, Jesmanowicz A, Froncisz W, et al: Parallel image acquisition from noninteracting local coils. J Magn Reson
70(3):512-517, 1986.
2. Hyde JS, Froncisz W, Jesmanowicz A, et al: Simultaneous image acquisition from the head (or body) coil and a surface coil.
Magn Reson Med 6(2):235-239, 1988.
page 246
page 247
3. Roemer PB, Edelstein WA, Hayes CE, et al: The NMR phased array. Magn Reson Med 16(2):192-225, 1990.
4. Carlson JW: An algorithm for NMR imaging reconstruction based on multiple RF receiver coils. J Magn Reson 74:376-380,
1987.
5. Hutchinson M, Raff U: Fast MRI data acquisition using multiple detectors. Magn Reson Med 6(1):87-91, 1988.
6. Kwiat D, Einav S, Navon G: A decoupled coil detector array for fast image acquisition in magnetic resonance imaging. Med
Phys 18(2):251-265, 1991.
7. Kwiat D, Einav S: Preliminary experimental evaluation of an inverse source imaging procedure using a decoupled coil
detector array in magnetic resonance imaging. Med Eng Phys 17(4):257-263, 1995.
8. Kelton JR, Magin RL, Wright SM: An algorithm for rapid image acquisition using multiple receiver coils. Eighth Annual
Meeting of the Society for Magnetic Resonance in Medicine. Amsterdam, Netherlands:1172, 1989.
9. Ra JB, Rim CY: Fast imaging using subencoding data sets from multiple detectors. Magn Reson Med 30(1):142-145, 1993.
10. Carlson JW, Minemura T: Imaging time reduction through multiple receiver coil data acquisition and image reconstruction.
Magn Reson Med 29(5):681-687, 1993.
coil arrays. Magn Reson Med 38(4):591-603, 1997.
12. Jakob PM, Griswold MA, Edelman RR, et al: AUTO-SMASH: a self-calibrating technique for SMASH imaging. MAGMA
13. Pruessmann KP, Weiger M, Scheidegger MB, et al: SENSE: sensitivity encoding for fast MRI. Magn Reson Med
42(5):952-962, 1999.
14. Lee RF, Westgate CR, Weiss RG, et al: An analytical SMASH procedure (ASP) for sensitivity-encoded MRI. Magn Reson
Med 43(5):716-725, 2000.
15. Sodickson DK: Tailored SMASH image reconstructions for robust in vivo parallel MR imaging. Magn Reson Med
16. Kyriakos WE, Panych LP, Kacher DF, et al: Sensitivity profiles from an array of coils for encoding and reconstruction in
parallel (SPACE RIP). Magn Reson Med 44(2):301-308, 2000.
17. Griswold MA, Jakob PM, Nittka M, et al: Partially parallel imaging with localized sensitivities (PILS). Magn Reson Med
44(4):602-609, 2000.
18. Larkman DJ, Hajnal JV, Herlihy AH, et al: Use of multicoil arrays for separation of signal from multiple slices simultaneously
excited. J Magn Reson Imag 13(2):313-317, 2001.
19. Sodickson DK, McKenzie CA: A generalized approach to parallel magnetic resonance imaging. Med Phys
28(8):1629-1643, 2001.
20. Kellman P, Epstein FH, McVeigh ER: Adaptive sensitivity encoding incorporating temporal filtering (TSENSE). Magn
Reson Med 45(5):846-852, 2001.
21. McKenzie CA, Ohliger MA, Yeh EN, et al: Coil-by-coil image reconstruction with SMASH. Magn Reson Med
46(3):619-623, 2001.
22. McKenzie CA, Yeh EN, Sodickson DK: Improved spatial harmonic selection for SMASH image reconstructions. Magn
Reson Med 46(4):831-836, 2001.
23. Pruessmann KP, Weiger M, Bornert P, et al: Advances in sensitivity encoding with arbitrary k-space trajectories. Magn
Reson Med 46(4):638-651, 2001.
24. Heidemann RM, Griswold MA, Haase A, et al: VD-AUTO-SMASH imaging. Magn Reson Med 45(6):1066-1074, 2001.
25. Bydder M, Larkman DJ, Hajnal JV: Generalized SMASH imaging. Magn Reson Med 47(1):160-170, 2002.
1 of 5 12-04-2010 00:20
26. McKenzie CA, Yeh EN, Ohliger MA, et al: Self-calibrating parallel imaging with automatic coil sensitivity extraction. Magn
Reson Med 47(3):529-538, 2002.
27. Sodickson DK, McKenzie CA, Ohliger MA, et al: Recent advances in image reconstruction, coil sensitivity calibration, and
coil array design for SMASH and generalized parallel MRI. MAGMA 13(3):158-163, 2002.
28. Weiger M, Pruessmann KP, Boesiger P: 2D SENSE for faster 3D MRI. MAGMA 14(1):10-19, 2002.
29. Griswold MA, Jakob PM, Heidemann RM, et al: Generalized autocalibrating partially parallel acquisitions (GRAPPA).
Magn Reson Med 47(6):1202-1210, 2002.
30. Azhari H, Sodickson DK, Edelman RR: Rapid MR imaging by sensitivity profile indexing and deconvolution reconstruction
(SPID). Magn Reson Imag 21(6):575-584, 2003.
31. Katscher U, Bornert P, Leussler C, et al: Transmit SENSE. Magn Reson Med 49(1):144-150, 2003.
32. Tsao J, Boesiger P, Pruessmann KP: k-t BLAST and k-t SENSE: dynamic MRI with high frame rate exploiting
spatiotemporal correlations. Magn Reson Med 50(5):1031-1042, 2003.
33. Lauterbur P: Image formation by induced local interactions: examples employing nuclear magnetic resonance. Nature
242:190-191, 1973.
34. Jackson JI, Meyer CH, Nishimura DG, et al: Selection of a convolution function for Fourier inversion using gridding. IEEE
Trans Med Imag 10(3):473-478, 1991.
35. Wang Y: Description of parallel imaging in MRI using multiple coils. Magn Reson Med 44(3):495-499, 2000.
36. Madore B, Pelc NJ: SMASH and SENSE: experimental and numerical comparisons. Magn Reson Med 45(6):1103-1111,
2001.
37. Sodickson DK, Griswold MA, Jakob PM, et al: Signal-to-noise ratio and signal-to-noise efficiency in SMASH imaging.
38. Ohliger MA, Grant AK, Sodickson DK: Ultimate intrinsic signal-to-noise ratio for parallel MRI: electromagnetic field
considerations. Magn Reson Med 50(5):1018-1030, 2003.
39. Wiesinger F, Boesiger P, Pruessmann KP: Electrodynamics and ultimate SNR in parallel MR imaging. Magn Reson Med
52(2):376-390, 2004.
40. Griswold MA, Jakob PM, Edelman RR, et al: A multicoil array designed for cardiac SMASH imaging. MAGMA 10:105-113,
41. Bankson JA, Griswold MA, Wright SM, et al: SMASH imaging with an eight element multiplexed RF coil array. MAGMA
42. Weiger M, Pruessmann KP, Leussler C, et al: Specific coil design for SENSE: a six-element cardiac array. Magn Reson
Med 45(3):495-504, 2001.
43. Lee RF, Westgate CR, Weiss RG, et al: Planar strip array (PSA) for MRI. Magn Reson Med 45(4):673-683, 2001.
44. de Zwart JA, Ledden PJ, Kellman P, et al: Design of a SENSE-optimized high-sensitivity MRI receive coil for brain
imaging. Magn Reson Med 47(6):1218-1227, 2002.
45. Lee RF, Hardy CJ, Sodickson DK, et al: Lumped-element planar strip array (LPSA) for parallel MRI. Magn Reson Med
51(1):172-183, 2004.
46. Lee RF, Boskamp EB, Giaquinto RO, et al: A 16-channel transmit/receive volume lattice array (VLA) for high acceleration
in parallel imaging. Eleventh Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Toronto:467,
2003.
47. Adriany G, van de Moortele P-F, Wiesinger F, et al: Transceive stripline arrays for ultra high field parallel imaging
applications. Eleventh Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Toronto: 474, 2003.
48. Hajnal JV, Larkman DJ, Herlihy DJ: An array that exploits phase for SENSE imaging. Eighth Scientific Meeting of the
International Society for Magnetic Resonance in Medicine. Denver, Colorado, USA:1719, 2000.
49. Ohliger MA, Greenman R, McKenzie CA, et al: Concentric coil arrays for spatial encoding in parallel MRI. Ninth Scientific
Meeting of the International Society for Magnetic Resonance in Medicine. Glasgow, Scotland:21, 2001.
50. Fujita H, Spence DK: A novel 8-channel "saddle-train" array coil for abdominal SENSE imaging at 1. 5T. Tenth Scientific
Meeting of the International Society for Magnetic Resonance in Medicine. Honolulu, HI, USA:326, 2002.
51. Willig J, Brown R, Eagan T, et al: Perfectly sinusoidal SMASH field shapes from birdcage sectors. Ninth Scientific Meeting
of the International Society for Magnetic Resonance in Medicine. Glasgow, Scotland:696, 2001.
52. Duensing GR, Gotshal U, King S, et al: N-dimensional orthogonality of volume coil arrays. Tenth Scientific Meeting of the
International Society for Magnetic Resonance in Medicine. Honolulu, HI, USA:771, 2002.
53. Seeber D, Pikelja V, Jevtic J: New RF coil topology for high performance SENSE in 3D. Eleventh Scientific Meeting of the
International Society for Magnetic Resonance in Medicine. Toronto:465, 2003.
54. Zhu Y, Hardy CJ, Sodickson DK, et al: Highly parallel volumetric imaging with a 32-element RF coil array. Magn Reson Med
52(4):869-877, 2004.
55. McDougall MP, Wright SM, Brown DG: A 64 channel planar RF coil array for parallel imaging at 4. 7 Tesla. Eleventh
Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Toronto:472, 2003.
56. Lin FH, Chen YJ, Belliveau JW, et al: A wavelet-based approximation of surface coil sensitivity profiles for correction of
2 of 5 12-04-2010 00:20
image intensity inhomogeneity and parallel imaging reconstruction. Hum Brain Mapp 19(2):96-111, 2003.
57. Yeh E, Stuber M, McKenzie C, et al: Self-calibrated spiral parallel imaging. Tenth Scientific Meeting of the International
Society for Magnetic Resonance in Medicine. Honolulu, HI, USA:2390, 2002.
58. Yeh E, McKenzie C, Lim D, et al: Parallel imaging with Augmented Radius in k-Space (PARS). Tenth Scientific Meeting of
the International Society for Magnetic Resonance in Medicine. Honolulu, HI, USA: 2399, 2002.
59. Griswold MA, Heidemann R, Jakob PM: Direct parallel imaging reconstruction of radially sampled data using GRAPPA
with relative shifts. Eleventh Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Toronto: 2349,
2003.
60. Griswold MA, Jakob PM, Chen Q, et al: Resolution enhancement in single-shot imaging using simultaneous acquisition of
spatial harmonics (SMASH). Magn Reson Med 41(6):1236-1245, 1999.
61. Griswold MA, Sodickson DK, Jakob PM, et al: Partially parallel acquisition strategies to increase image quality in single
shot imaging. Fast MRI Workshop of the International Society for Magnetic Resonance in Medicine. Asilomar, CA, USA:186,
1997.
62. Bammer R, Stollberger R, Hartung HP, et al: Parallel imaging strategies for reduction of resonance offset induced artifacts
and k-space filtering effects. Eighth Scientific Meeting of the International Society for Magnetic Resonance in Medicine.
Denver, Colorado, USA:1503, 2000.
63. Weiger M, Pruessmann KP, Hilfiker PR, et al: Increasing the SNR in steady-state imaging by sensitivity encoding. Ninth
Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Glasgow, Scotland:1798, 2001.
64. Goldfarb JW, Shinnar M: Field-of-view restrictions for artifact-free SENSE imaging. Tenth Scientific Meeting of the
International Society for Magnetic Resonance in Medicine. Honolulu, HI, USA:2412, 2002.
65. Yoshioka H, Takahashi N, Yamaguchi M, et al: Double arterial phase dynamic MRI with sensitivity encoding (SENSE) for
hypervascular hepatocellular carcinomas, J Magn Reson Imag 16(3):259-266, 2002.
66. Dobritz M, Radkow T, Nittka M, et al: [VIBE with parallel acquisition technique-a novel approach to dynamic contrast-
enhanced MR imaging of the liver.] Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 174(6):738-741, 2002.
page 247
page 248
67. Hori M, Ichikawa T, Sou H, et al: [Improving diffusion-weighted imaging of liver with SENSE technique: a preliminary study.]
Nippon Igaku Hoshasen Gakkai Zasshi 63(4):177-179, 2003.
68. McKenzie CA, Lim D, Ransil BJ, et al: Shortening MR image acquisition time for volumetric interpolated breath-hold
examination with a recently developed parallel imaging reconstruction technique: clinical feasibility. Radiology 230(2):589-594,
2004.
69. Jakob PM, Griswold MA, Edelman RR, et al: Accelerated cardiac imaging using the SMASH technique. J Cardiovasc
Magn Reson 1(2):153-157, 1999.
70. Pruessmann KP, Weiger M, Boesiger P: Sensitivity encoded cardiac MRI. J Cardiovasc Magn Reson 3(1):1-9, 2001.
71. Weiger M, Pruessmann KP, Boesiger P: Cardiac real-time imaging using SENSE. SENSitivity Encoding scheme. Magn
Reson Med 43(2):177-184, 2000.
72. Madore B: Using UNFOLD to remove artifacts in parallel imaging and in partial-Fourier imaging. Magn Reson Med
48(3):493-501, 2002.
73. Guttman MA, Kellman P, Dick AJ, et al: Real-time accelerated interactive MRI with adaptive TSENSE and UNFOLD. Magn
Reson Med 50(2):315-321, 2003.
74. Bammer R, Keeling SL, Augustin M, et al: Improved diffusion-weighted single-shot echo-planar imaging (EPI) in stroke
using sensitivity encoding (SENSE). Magn Reson Med 46(3):548-554, 2001.
75. Willinek WA, Gieseke J, von Falkenhausen M, et al: Sensitivity encoding for fast MR imaging of the brain in patients with
stroke. Radiology 228(3):669-675, 2003.
76. Sodickson DK, McKenzie CA, Li W, et al: Contrast-enhanced 3D MR angiography with simultaneous acquisition of spatial
harmonics: a pilot study. Radiology 217(1):284-289, 2000.
77. Weiger M, Pruessmann KP, Kassner A, et al: Contrast-enhanced 3D MRA using SENSE. J Magn Reson Imag
12:671-677, 2000.
78. Larkman DJ, deSouza NM, Bydder M, et al: An investigation into the use of sensitivity-encoded techniques to increase
temporal resolution in dynamic contrast-enhanced breast imaging. J Magn Reson Imag 14(3):329-335, 2001.
79. Golay X, Brown SJ, Itoh R, et al: Time-resolved contrast-enhanced carotid MR angiography using sensitivity encoding
(SENSE). Am J Neuroradiol 22(8):1615-1619, 2001.
80. Maki JH, Wilson GJ, Eubank WB, et al: Utilizing SENSE to achieve lower station sub-millimeter isotropic resolution and
minimal venous enhancement in peripheral MR angiography. J Magn Reson Imag 15(4):484-491, 2002.
81. Walter C, Philippi G, Westerhausen R, et al: [High resolution contrast-enhanced 3D MR-angiography of renal arteries using
parallel imaging (SENSE).] Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 175(9):1244-1250, 2003.
82. Ohno Y, Kawamitsu H, Higashino T, et al: Time-resolved contrast-enhanced pulmonary MR angiography using sensitivity
encoding (SENSE). J Magn Reson Imag 17(3):330-336, 2003.
83. Golay X, Pruessmann KP, Weiger M, et al: PRESTO-SENSE: an ultrafast whole-brain fMRI technique. Magn Reson Med
43(6):779-786, 2000.
3 of 5 12-04-2010 00:20
84. Weiger M, Pruessmann KP, Osterbauer R, et al: Sensitivity-encoded single-shot spiral imaging for reduced susceptibility
artifacts in BOLD fMRI. Magn Reson Med 48(5):860-866, 2002.
85. De Zwart JA, van Gelderen P, Kellman P, et al: Application of sensitivity-encoded echo-planar imaging for blood oxygen
level-dependent functional brain imaging dagger. Magn Reson Med 48(6):1011-1020, 2002.
86. Klarhofer M, Dilharreguy B, van Gelderen P, et al: A PRESTO-SENSE sequence with alternating partial-Fourier encoding
for rapid susceptibility-weighted 3D MRI time series. Magn Reson Med 50(4):830-838, 2003.
87. Preibisch C, Pilatus U, Bunke J, et al: Functional MRI using sensitivity-encoded echo planar imaging (SENSE-EPI).
Neuroimage 19(2 Pt 1):412-421, 2003.
88. Bammer R, Auer M, Keeling SL, et al: Diffusion tensor imaging using single-shot SENSE-EPI. Magn Reson Med
48(1):128-136, 2002.
89. Yamada K, Kizu O, Mori S, et al: Brain fiber tracking with clinically feasible diffusion-tensor MR imaging: initial experience.
Radiology 227(1):295-301, 2003.
90. Stahl R, Dietrich O, Teipel S, et al: [Assessment of axonal degeneration on Alzheimer's disease with diffusion tensor MRI.]
Radiologe 43(7):566-575, 2003.
91. Cercignani M, Horsfield MA, Agosta F, et al: Sensitivity-encoded diffusion tensor MR imaging of the cervical cord. Am J
Neuroradiol 24(6):1254-1256, 2003.
92. Heidemann RM, Griswold MA, Kiefer B, et al: Resolution enhancement in lung 1H imaging using parallel imaging methods.
Magn Reson Med 49(2):391-394, 2003.
93. Fink C, Bock M, Puderbach M, et al: Partially parallel three-dimensional magnetic resonance imaging for the assessment
of lung perfusion-initial results. Invest Radiol 38(8):482-488, 2003.
94. Dydak U, Weiger M, Pruessmann KP, et al: Sensitivity-encoded spectroscopic imaging. Magn Reson Med 46(4):713-722,
2001.
95. Dydak U, Pruessmann KP, Weiger M, et al: Parallel spectroscopic imaging with spin-echo trains. Magn Reson Med
50(1):196-200, 2003.
96. Romaneehsen B, Oberholzer K, Muller LP, et al: [Rapid musculoskeletal magnetic resonance imaging using integrated
parallel acquisition techniques (IPAT)-initial experiences.] Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr
175(9):1193-1197, 2003.
97. Magee T, Shapiro M, Williams D, et al: Usefulness of the simultaneous acquisition of spatial harmonics technique during
MRI of the shoulder. Am J Roentgenol 181(4):961-964, 2003.
98. Niitsu M, Ikeda K: Routine MR examination of the knee using parallel imaging. Clin Radiol 58(10):801-807, 2003.
99. Beerbaum P, Korperich H, Gieseke J, et al: Rapid left-to-right shunt quantification in children by phase-contrast magnetic
resonance imaging combined with sensitivity encoding (SENSE). Circulation 108(11):1355-1361, 2003.
100. de Zwart JA, van Gelderen P, Kellman P, et al: Reduction of gradient acoustic noise in MRI using SENSE-EPI.
Neuroimage 16(4):1151-1155, 2002.
101. Bydder M, Larkman DJ, Hajnal JV: Detection and elimination of motion artifacts by regeneration of k-space. Magn Reson
Med 47(4):677-686, 2002.
102. Bydder M, Atkinson D, Larkman DJ, et al: SMASH navigators. Magn Reson Med 49(3):493-500, 2003.
103. Zhu Y: Parallel excitation with an array of transmit coils. Magn Reson Med 51(4):775-784, 2004.
104. Sodickson DK, Griswold MA, Jakob PM: SMASH imaging. Magn Reson Imag Clin N Am 7(2):1-18, 1999.
105. Sodickson DK: Spatial encoding using multiple RF coils: SMASH imaging and parallel MRI. In Young IR (ed): Methods in
biomedical magnetic resonance imaging and spectroscopy. Wiley, Chichester, 2000.
106. van den Brink JS, Watanabe Y, Kuhl CK, et al: Implications of SENSE MR in routine clinical practice. Eur J Radiol
46(1):3-27, 2003.
107. Heidemann RM, Ozsarlak O, Parizel PM, et al: A brief review of parallel magnetic resonance imaging. Eur Radiol
13(10):2323-2337, 2003.
108. Sodickson A: Breaking the magnetic resonance imaging acquisition speed barrier: clinical implications of parallel
imaging. Appl Radiol 33(1) (suppl):6-17, 2004.
109. Wright SM, McDougall MP, Brown DG: Single Echo Acquisition (SEA) MR imaging. Eleventh Scientific Meeting of the
International Society for Magnetic Resonance in Medicine. Toronto:23, 2003.
110. Ohliger M, Yeh E, McKenzie C, et al: Fundamental physical constraints on the performance of parallel magnetic
resonance imaging. Tenth Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Honolulu, HI,
USA:2387, 2002.
111. Reykowski A, Schnell W, Wang J: Simulation of SNR limit for SENSE related reconstruction techniques. Tenth Scientific
Meeting of the International Society for Magnetic Resonance in Medicine. Honolulu, HI, USA:2385, 2002.
112. Griswold M, Lanz T, Haase A, et al: A brute-force estimation of the SNR limits of sensitivity encoding in a cylindrical
geometry. Tenth Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Honolulu, HI, USA:2386,
2002.
113. Ohliger MA, Ledden P, McKenzie CA, Sodickson DK: Effects of inductive coupling on parallel MR image reconstructions.
Magn Reson Med 52(3):628-639, 2004.
4 of 5 12-04-2010 00:20
114. Pruessmann KP, Weiger M, Wiesinger F, et al: An investigation into the role of coil coupling in parallel imaging. Tenth
Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Honolulu, HI, USA:196, 2002.
115. Hayes CE, Hattes N, Roemer PB: Volume imaging with MR phased arrays. Magn Reson Med 18(2):309-319, 1991.
116. Wang J, Reykowski A, Dickas J: Calculation of the signal-to-noise ratio for simple surface coils and arrays of coils. IEEE
Trans Biomed Eng 42(9):908-917, 1995.
117. Porter JR, Wright SM, Reykowski A: A 16-element phased-array head coil. Magn Reson Med 40(2):272-279, 1998.
118. Sodickson DK, Lee RF, Giaquinto RO, et al: Depth penetration of RF coil arrays for sequential and parallel MR imaging.
Eleventh Scientific Meeting of the International Society for Magnetic Resonance in Medicine. Toronto:469, 2003.
119. Sodickson DK, Hardy C, Zhu Y, et al: Highly accelerated parallel MRI on an MR scanner with 32 receiver channels.
Eighty-Ninth Scientific Assembly and Annual Meeting of the Radiological Society of North America. Chicago:172, 2003.
120. Golman K, Axelsson O, Johannesson H, et al: Parahydrogen-induced polarization in imaging: subsecond (13)C
angiography. Magn Reson Med 46(1):1-5, 2001.
121. Svensson J, Mansson S, Johansson E, et al: Hyperpolarized 13C MR angiography using trueFISP. Magn Reson Med
50(2):256-262, 2003.
© 2010 Elsevier
5 of 5 12-04-2010 00:20
ASIC RINCIPLES OF UNCTIONAL

Kâmil Uluda[gbreve]*
David J. Dubowitz*
Richard B. Buxton*
*The authors are supported by NIH grants NS-36722 and NS-042069.
page 249
page 250
INTRODUCTION TO FUNCTIONAL MAGNETIC RESONANCE IMAGING (fMRI)

The MRI Signal is Sensitive to Changes in Blood Oxygenation
One of the remarkable developments in recent work on MRI is the recognition that changes in the
metabolic state of the brain affect the MR signal in a detectable fashion and therefore provide an
intrinsic mechanism of contrast for brain activation studies. The origin of this effect is that the magnetic
state of hemoglobin (Hb) depends upon its oxygenation, so that changes in oxygen saturation of the
hemoglobin produce a small change in the local MR signal, the blood oxygenation level-dependent
(BOLD) effect. Specifically, deoxygenated hemoglobin is paramagnetic and tends to reduce the local
MR signal by creating microscopic field gradients within and around the blood vessels. If the local
oxygen extraction fraction (E) always remained constant, the local oxygenation of the blood would not
change, and the BOLD effect would simply be an interesting, but not particularly useful, biophysical
effect. However, when combined with an unexpected physiologic phenomenon, this becomes a
powerful tool for mapping brain activation. Following increased neural activity in the brain, the local
cerebral blood flow (CBF) increases much more than the cerebral metabolic rate of oxygen (CMRO 2),
and as a result E decreases with activation. Because the local blood is more oxygenated, there is less
deoxyhemoglobin present and the local MR signal increases slightly.
Brain activation studies based upon BOLD contrast typically employ an experimental paradigm in which
a subject alternates between periods of stimulation and rest while a rapid series of MR images is
collected. The time series for each image voxel is then analyzed to determine if the signal shows a
significant correlation with the stimulus, i.e., increasing when the stimulus was applied and decreasing
when the stimulus was removed. Those pixels that do show a correlation are displayed in color on a
regular anatomical MR image as the areas activated by the stimulus.
The Origins of fMRI

The fact that the magnetic state of hemoglobin changes with its state of oxygenation was discovered in
1936 by Pauling and Coryell, before the discovery of nuclear magnetic resonance (NMR) itself.1 In
1982 Thulborn and colleagues demonstrated relaxation rate (T2) changes in blood samples due to the
magnetic susceptibility changes caused by the presence of paramagnetic deoxyhemoglobin.2 However,
it was not until the 1990s that the potential significance of this effect for functional neuroimaging was
realized.3-7
The first demonstration that changes in blood oxygenation had a measurable effect on the MR signal in
vivo was not an activation study but rather a physiologic manipulation in which the inspired oxygen was
varied. Ogawa et al imaged the brains of mice at high magnetic fields (7 and 8.4 T) with gradient-echo
4
imaging. They found that the veins became noticeably darker in the MR image when the oxygen in the
inspired air was reduced. The reduction of the blood signal was consistent with the earlier in vitro NMR
2
studies that had demonstrated the effect of oxygenation on T2. But, in addition, Ogawa and
colleagues made the key observation that the signal from the tissue surrounding the veins also was
reduced, and they proposed that the cause of this effect was a change in the magnetic susceptibility of
the blood. Furthermore, the effect was greatly reduced in spin-echo images. Both these observations
were consistent with the source of the effect being related to magnetic susceptibility changes (to which
gradient-echo images are highly sensitive) brought on by the presence of the deoxygenated
hemoglobin.
1 of 3 12-04-2010 00:20
Ogawa and colleagues suggested that this phenomenon could form the basis for monitoring regional
oxygen use in the brain, and speculated that during activation more oxygen would be removed from the
blood and the deoxyhemoglobin concentration would increase. The reality of the situation turned out to
be the opposite of this-deoxyhemoglobin concentration decreases with activation because of the large
CBF change-but the insight that this NMR effect could be used to measure brain function was the
critical beginning of fMRI. Subsequently, Turner and colleagues imaged cat brains under the controlled
conditions of anoxia and apnea with an echo-planar imaging (EPI) pulse sequence on a 2 T system.8
They also found MR signal changes that were dependent on the oxygenation of the blood.
The detectability of blood oxygenation effects in these well-controlled animal experiments at least
suggested the possibility that such effects might be seen in humans performing tasks that alter the
oxygen utilization in the brain. Kwong and colleagues acquired images of a normal human subject
during visual stimulation with a gradient-echo EPI sequence using a long echo time (40 ms) in order to
enhance the susceptibility effects.3 Temporally resolved images acquired during and in the absence of
stimulation showed clear differences in signal intensity, with the signal increasing during stimulation,
suggesting that the deoxyhemoglobin concentration decreased with activation. In short order several
other studies confirmed this finding.5-7 Other fMRI studies in the visual cortex with a high field (4 T)
scanner and the motor cortex with an echo-planar system with specially designed gradient coils also
9
demonstrated signal increases with activation. Similar results were soon obtained on a conventional
10
clinical scanner as well.
fMRI Has Become an Important Tool in Neuroscience Research

page 250
page 251
From its origins in basic MRI research described above, fMRI has grown explosively to become a
standard and indispensable tool in neuroscience research. Previously, positron emission tomography
(PET) methods of measuring CBF change were the standard for mapping functional activity in the
human brain. While PET studies are still done for a number of applications, most human brain mapping
studies are now done with fMRI. Over the last decade, the sophistication of the techniques has
improved enormously. For example, techniques for retinotopic mapping have become standard in
studies of the visual system. The visual image produced on the retina is mapped in a spatially coherent
way onto the visual cortex, and this coherent retinotopic map is repeated in many sub-regions of the
visual cortex. By mapping progressive waves of activation as a subject views expanding rings or
rotating wedges, the boundaries of these different functional regions of the visual cortex can be
11,12
mapped. Because the functional organization does not always match up in the same way with the
anatomical organization, fMRI studies provide an enormous advance in the ability to characterize the
working human brain by identifying these functional sub-divisions of the visual cortex in addition to
anatomical sub-divisions.
Overview of the Chapter

The remainder of the chapter introduces the basic principles, mechanisms, and techniques that underlie
fMRI. In "The Physiologic Basis of fMRI," the physiologic mechanisms linking blood flow, oxygen
metabolism, and neural activity are described. In fact, in recent years fMRI techniques have become
useful tools for exploring these links. In "The BOLD Effect," the biophysics underlying the BOLD effect
is described, including mathematical models for how the BOLD signal depends on the local change in
the O2 extraction fraction E and the venous blood volume V. In addition, because fMRI techniques
measure dynamic changes, and the dynamics of E and V may have different time constants, there is
the possibility for a range of transient effects in the measured BOLD response. Because the BOLD
signal changes are small-typically only a few percent-the design and analysis of fMRI experiments to
detect these subtle effects is a critical component of fMRI; this is introduced in "Design and Analysis of
BOLD-fMRI Experiments." In addition, because the BOLD signal changes are small, artifacts that
would have little impact on the diagnostic utility of clinical MR images nevertheless can severely
degrade fMRI data, and some of these effects and possible remedies are discussed in "Artifacts and
Noise." One of the powerful features of MRI is its flexibility, and other MR-based techniques have been
2 of 3 12-04-2010 00:20
developed to measure different physiologic aspects of brain activation that complement and enhance
the standard measurements of the BOLD effect. These other methods, and how they can be combined
with BOLD-fMRI, are introduced in "Measuring Cerebral Blood Flow, Cerebral Metabolic Rate of
Oxygen, and Cerebral Blood Volume." Finally, in "Exploring the Hemodynamic Response to Brain
Activation with MRI," we describe the ways in which fMRI is being used in current research on the
physiology of brain activation, and some of the notable open questions.
© 2010 Elsevier
3 of 3 12-04-2010 00:20
THE PHYSIOLOGIC BASIS OF fMRI

In a typical fMRI experiment the goal is to map patterns of neuronal activation in the subject's brain
while he or she performs specific tasks. However, fMRI does not measure the neuronal activity itself.
Instead, the BOLD effect in response to activation is sensitive to the concentration change of
deoxygenated hemoglobin, which in turn is dependent on cerebral blood flow (CBF), cerebral blood
volume (CBV), and cerebral metabolic rate of oxygen (CMRO 2), illustrated in Figure 9-1.
A critical goal for interpreting fMRI data is to understand the underlying link between neuronal activity
and the hemodynamic response. This is still an area of active research, and in this section we outline
the current thinking.
Neuronal Signaling and Energy Metabolism
Add to lightbox
page 251
page 252
Figure 9-1 The path of changes linking an applied stimulus to the measured local BOLD signal change
in a subject's brain during an fMRI experiment. In this illustration, a flickering checkerboard stimulus
triggers increased neuronal activity in the visual cortex. This is accompanied by increased blood flow,
blood volume, and oxygen metabolism, and these physiologic changes combine to alter the local
deoxyhemoglobin, which in turn alters the local MR signal.
In the brain, neurons are maintained in a thermodynamic state far from equilibrium. The sodium ion
(Na+) concentration outside the neuron is much higher than that inside the cell. Given the more negative
+
potential inside the cell, it is a strongly downhill reaction for Na to move into the cell. On the
presynaptic side, calcium ions (Ca2+) are also concentrated outside the cell, and neurotransmitters are
highly concentrated in small vesicles within the presynaptic terminal waiting for release. The arrival of
2+
an action potential triggers a cascade that includes Ca influx, neurotransmitter release into the
synaptic cleft, binding of neurotransmitter on the post-synaptic side, and opening of ion channels for
Na+ and potassium (K+) currents. This signaling process is all thermodynamically downhill, so no
energy is required. The energy costs of neural activity come mainly in the recovery from this signaling:
+ + 2+
Na, K , and Ca must be pumped against their gradients to restore the original ion distributions, and
neurotransmitter must be cleared from the synaptic cleft and re-packaged in vesicles in preparation for
the arrival of the next action potential. The source of thermodynamic free energy to power these uphill
processes is the pool of adenosine triphosphate (ATP) and adenosine diphosphate (ADP). The
ATP/ADP system is far from equilibrium, with approximately ten times more ATP than ADP.13 For this
reason, the conversion of ATP to ADP carries a large negative free energy (∆G) that can drive other
reactions uphill. In addition to direct use of ATP, some uphill processes, such as the clearance of the
+
neurotransmitter glutamate from the synaptic cleft, are driven by co-transport of Na down its gradient
+
from the extracellular to intracellular space. The degraded Na gradient, in turn, is restored by the
+ + + +
Na /K pump, which pumps both Na and K against their gradients at the expense of ATP. It has
14
been estimated that at least half of the energy consumed in the brain is due to the action of the
+ +
Na /K pump.
1 of 9 12-04-2010 00:22
In short, one can think of the brain as containing two stores of free energy-two batteries-that can be
used to drive all of the energy-consuming reactions in the cell: the ATP/ADP system, and the Na+
+ +
gradient across the cell membrane. These two systems are in close communication through the Na /K
pump. From a biochemical perspective, the available free energy is defined by a ratio of
concentrations: [ATP]/[ADP] in one case, and extracellular/intracellular Na+ in the other. That is, it is
not the ATP itself that carries the energy; it is the high ratio of [ATP]/[ADP] that is far from equilibrium
that endows the conversion of ATP to ADP with a large negative free energy.
Recent studies have reported estimates of the energy budget for brain processing by tallying up the
number of ATP/ADP conversions needed to fuel the different processes involved.15 Neurons and glia
require energy to maintain their resting membrane potential and carry out other non-signaling functions
within the cell. One can break down the signaling costs by considering a single action potential, which
starts in one cell and travels to thousands of synapses on other cells, where presynaptic transmitter
release triggers post-synaptic membrane potential changes. A key question is: how does the energy
cost of generating and propagating an action potential compare with the energy costs of synaptic
activity?
Add to lightbox
page 252
page 253
Figure 9-2 Relative metabolic costs of each process per neural signaling event (first number: rodents,
16
second number: primates). Post-synaptic activity is estimated to consume most of the energy in
humans. Action potentials only require 10% of the total energy. In contrast, in rodents the action
potentials account for almost half of the total energy. Maintaining resting potentials, presynaptic
processes, and the glia utilize only small amounts of energy. (Adapted, with permission, from Attwell
D, Iadecola C: The neurol basis of functional brain imaging signals. Trends Neurosci 25:621-625,
2002. © 2002, with permission from Elsevier.)
In Figure 9-2 the energy estimates of each process per signaling event are shown (the first number
refers to rodents, the second number to primates). Post-synaptic activity, such as uptake of
neurotransmitters by the neurons and astrocytes and restoration of the ionic gradients, is estimated to
consume most of the energy in humans.16 The spiking itself (action potentials) only requires 10% of the
2 of 9 12-04-2010 00:22
total energy. In contrast, in rodents the action potentials account for almost half of the total energy.
Maintaining resting potentials, presynaptic processes, and the glia utilize only small amounts of energy.
In "Exploring the Hemodynamic Response to Brain Activation with MRI," we discuss what implications
this estimation has for the interpretation of the measured fMRI signals.
The Brain is Fueled by the Oxidative Metabolism of Glucose

The ATP/ADP system that fuels the recovery from neural activity must be restored by coupling the
uphill conversion of ADP to ATP to an even more downhill reaction: the oxidative metabolism of glucose
and oxygen to carbon dioxide and water. The complete conversion of one molecule of glucose and
six molecules of O2 generates 38 ATP molecules from ADP. The full metabolism is illustrated in Figure
9-3.
This fundamental energy metabolism happens in two stages. In the cytoplasm, glycolysis converts the
glucose molecule to two molecules of pyruvate, stores some of the energy in the conversion of two
nicotinamide adenine dinucleotide ions (NAD+) to NADH,17 and generates two ATP molecules from
ADP, all without using O2. Although the ATP yield of glycolysis is low, it is very fast. For this reason,
exercising muscle relies on glycolysis to generate the ATP needed for short bursts of intense activity
(e.g., sprinting), and it has been suggested that speed of production of ATP may also be important in
the brain.18
Add to lightbox
Figure 9-3 Non-oxidative and oxidative metabolism of glucose . The non-oxidative metabolism of
glucose (glycolysis) generates two ATP molecules from ADP by converting the glucose molecule
to two molecules of pyruvate. The TCA cycle and the electron transfer chain in the mitochondria
metabolize the two pyruvate molecules and 6 O2 molecules to 6 CO2 and 6 H2O molecules, producing
36 ATP from ADP. In the cytosol, pyruvate and lactate are in near equilibrium, and if the glucose
metabolic rate exceeds the oxygen metabolic rate, the lactate concentration will rise.
Much more ATP is generated in the second stage of energy metabolism when pyruvate and O2 diffuse
into the mitochondria and enter the tricarboxylic acid (TCA) cycle. The end products of mitochondrial
energy metabolism are six molecules each of H2O and CO2, and the conversion of 36 ADP molecules
to ATP. As part of this net metabolism, the NADH produced by glycolysis also is shuttled into the
mitochondria in exchange for NAD+, restoring the cytosolic balance.
It is important to recognize that, because energy metabolism occurs in two stages with glycolysis
feeding the TCA cycle, it is possible for the rate of glycolysis to exceed the rate of pyruvate
metabolism in the mitochondria. In this case the cerebral metabolic rate of glucose (CMRGlc) and the
cerebral metabolic rate of oxygen (CMRO 2) are not matched. Fully matched oxidative metabolism
3 of 9 12-04-2010 00:22
requires that six O2 molecules are consumed for each glucose molecule metabolized, often described
as an oxygen-glucose index (OGI) of 6.0. A typical experimental value is OGI = 5.5 at rest, but it is
interesting to note that this quantity appears to decrease with activation: the CMRGlc change with
18
activation exceeds the CMRO2 change (for further discussion see "Exploring the Hemodynamic
Response to Brain Activation with MRI").
If the glucose and oxygen metabolic rates are not matched, pyruvate and NADH would accumulate in
the cell and ultimately disrupt glycolysis. However, an important enzyme called lactate dehydrogenase
catalyzes the conversion of pyruvate and NADH to lactate and NAD+.17 This restores the NADH/NAD+
balance but leads to the accumulation of lactate, which ultimately diffuses out of the cell and is carried
away in the blood. The accumulation of lactate in the tissue, which can be measured with MR
spectroscopy techniques (see "Measuring Cerebral Blood Flow, Cerebral Metabolic Rate of Oxygen,
and Cerebral Blood Volume" and "Exploring the Hemodynamic Response to Brain Activation with
MRI"), is thus a sign of a mismatch of CMRGlc and CMRO2.
Cerebral Blood Flow Delivers O2 and Glucose and Clears CO2

For the brain to continue functioning, glucose and oxygen must be supplied and CO2 cleared from
each tissue element, and this is accomplished by blood flow. In order to understand hemodynamics, it
is important to keep in mind that CBF and CBV are two distinct physiologic quantities. The CBV
describes the total volume of the vasculature: the sum of the volumes of the arteries, arterioles,
capillaries, venules, and veins in a volume of tissue (Fig. 9-4). The CBV is usually expressed as a
dimensionless quantity, the fraction of the tissue volume occupied by blood. A typical value of CBV in
the brain is 4%.
page 253
page 254
Add to lightbox
Figure 9-4 Vascular structure of a chinchilla measured with corrosion casts. The feeding arteriole (red)
delivers oxygen-saturated blood to the capillaries (orange and green). The capillaries end in draining
venules (blue) having only approximately 60% oxygen saturation, that is, brain tissue extracts 40% of
the oxygen from the capillaries. All these vascular compartments contribute to the cerebral blood
volume (CBV). Cerebral blood flow (CBF) is the rate of blood delivery of blood to the capillary bed.
(Adapted from Harrison RV, et al: Blood capillary distribution correlates with hemodynamic-based
functional imaging in cerebral cortex. Cereb Cortex 12:225-233, 2002, by permission of Oxford
University Press.)
In contrast, CBF is the volume of arterial blood delivered to an element of tissue in a specified time,
i.e., only blood flowing through arteries, arterioles, and capillaries is counted for CBF. Thus, blood
flowing through a volume of tissue but destined for another is not counted for CBF of this volume. The
usual units are mL/100 g/min, and a typical value in the human brain is 60 mL/100 g/min. If we refer the
4 of 9 12-04-2010 00:22
CBF to 1 mL of tissue (with a density of about 1 g/mL), the units become mL/mL/min, or simply inverse
time. Expressed this way, we could write a standard CBF of 60 mL/100 g/min as 0.01 s-1. This
formulation of the units emphasizes that CBF often acts like a rate constant. In particular, the
metabolic rate of oxygen metabolism can always be written as:
where Ca is the arterial concentration of O2, and E is the net oxygen extraction fraction. This is simply
the total rate at which O2 is being delivered to tissue (Ca CBF) times the fraction of the delivered
oxygen that is extracted and metabolized.
In general, it is helpful to think of CBF and CBV as independent quantities: in plumbing terms, the
volume of the pipes and the flow being delivered to the pipes. At some level, however, there is a
connection. The CBF increase associated with neural activity is triggered by relaxation of the smooth
muscle in the wall of the arterioles. The arterioles provide most of the resistance in the vascular tree
and provide a way to quickly decrease the vascular resistance (by expanding). In this sense, an
increased blood volume is part of the mechanism of increasing CBF (see Fig. 9-4).
However, if the arteriolar volume fraction is small, this may be only a small change in total CBV. As the
resistance of the arterioles decreases, the pressure drop across these vessels also decreases, raising
the pressure in the capillaries and veins. These vessels may also expand due to the increased
19
pressure, further increasing the CBV. Experimental studies have indicated that the steady-state
relationship between CBF and CBV can be described with a power law:
where the exponent is approximately α= 0.38, i.e., a CBF increase of 50% corresponds to a CBV
increase of approximately 18%. This empirical relationship applies to the entire cerebral blood volume,
and only after a steady state has been reached.
The temporal dynamics of the total CBV will be a weighted composite of the changes in each of its
compartments. For example, during functional activation the initial change will be dominated by the
arterioles, but it has been postulated that the venous vessels may be slower to expand and slower to
contract back to baseline after CBF has returned to normal.20,21 In general, the CBV and CBF ratio
during the transients is as yet not well described, but measuring this ratio dynamically promises to
provide some insights into these basic physiologic variables and into neurovascular coupling that may
be altered in disease.
Neuronal Activation is Followed by a Hemodynamic Response

A large number of positron emission tomography (PET) studies, and more recently fMRI studies, have
measured the changes in CBF, CBV, CMRGlc, and CMRO2 accompanying neural activity, see for
example references 19 and 22 to 25. There is of course a good deal of experimental variation, but in
rough numbers the basic pattern for a strong stimulus is that CBF increases dramatically (40%),
CMRGlc also increases by about the same amount, CMRO2 increases much less (<20%), and CBV
increases by a modest amount (15%). It is useful to summarize this complex of physiologic changes in
terms of two key dimensionless numbers defined above, the oxygen extraction fraction (E) and the
oxygen-glucose index (OGI). The key results are that with activation E and OGI both decrease. Trying
to understand this unexpected pattern is the focus of much current research (see "Exploring the
Hemodynamic Response to Brain Activation with MRI").
The decrease of E with activation is the primary cause of the BOLD effect. Oxygenated blood is
diamagnetic and deoxygenated blood is paramagnetic, so the increase in blood oxygenation during
activation changes the magnetic properties of the blood and the tissue and causes the MR signal to
increase. Because the temporal resolution of fMRI is much better than with PET techniques, the BOLD
signal provides a window on the temporal dynamics of the hemodynamic response. Typical
experimental responses to a very short stimulation (approximately 1 second in duration), called the
impulse response, and to a long stimulation (20 seconds in duration) measured with fMRI are shown in
Figure 9-5.
page 254
page 255
5 of 9 12-04-2010 00:22
Add to lightbox
Figure 9-5 Typical experimental hemodynamic responses to a very short stimulation (approximately 1
second duration, blue), called the impulse response, and to a long stimulation (20 seconds duration,
red) measured with fMRI.
The BOLD response is typically delayed by 1 to 2 seconds and reaches its maximum after
approximately 8 s (typically between 5 and 10 s). After the end of the stimulus a post-stimulus
undershoot often can be seen which lasts typically 20 s or more. More rarely, and not highly
reproducibly, an initial undershoot of the BOLD signal is observed (for an overview see reference 26),
called the "initial dip." For further discussion and implications of the transients see "Exploring the
Hemodynamic Response to Brain Activation with MRI".
Multiple Agents Mediate Neurovascular Coupling

The translation of increased neuronal activity to increased CBF-neurovascular coupling-can be
considered from two different viewpoints: 1. what is the function served by neurovascular coupling?;
and 2. what is the mechanism that accomplishes this function? Somewhat surprisingly, the mechanism
is better understood than the function. A number of vasoactive agents are produced in association with
neural activity, and it appears that the particular mechanisms employed to raise CBF vary in different
27 +
parts of the brain. Some of the key vasoactive agents are nitric oxide (NO), potassium ions (K ),
+
hydrogen ions (H ), adenosine , and carbon dioxide (CO2). This is only a partial list, but it already
includes a highly diffusible gas linked to G-protein activation (NO), a key player in ion homeostasis
(K+), the local pH, a neurotransmitter (adenosine ), a key element of the energetic stores of the tissue
(adenosine again, as the final form of degradation of ATP), and a key product of energy metabolism
(CO2). This suggests that multiple mechanisms exist to increase blood flow in association with
increased neural activity. Indeed, the experimental data indicate that there is no single mechanism
controlling CBF in the brain, and the mechanism of CBF control may be quite adaptive during
development.
At this point, there is no shortage of possible mechanisms although little is known about how the full
integrated system for neurovascular coupling works.27 Two interrelated themes have received growing
attention in recent years, one focused on the central role of glutamate and the other focused on the
role of the non-neuronal (glial) cells in neuronal signaling and neurovascular coupling.28-31 Glutamate is
the primary excitatory neurotransmitter, and the majority of synapses are glutamatergic. When
glutamate binds to a post-synaptic receptor it opens a Na+ channel that allows Na+ to diffuse down its
gradient, and eventually it must be pumped back by the Na+/K+ pump. For this reason, glutamate itself
would serve as a good signal for the energy-consuming processes associated with neural activity, and
there is some evidence that glutamate has a direct vasodilatory effect. 32 But the more significant
6 of 9 12-04-2010 00:22
connection is likely to be that glutamate is cleared from the synaptic cleft by astrocytes.29
The current view of the role of the glia has expanded substantially in the last few years.31,33 It has
been known for some time that astrocytes are both positioned near synapses for recycling of
glutamate and also have numerous projections to blood vessels by their end-feet, suggesting at least
the anatomical connections for linking synaptic activity to blood flow. Recent work has indicated that
astrocytes have very well-developed and essentially non-overlapping territories, further supporting a
key role for the astrocytes in assessing the level of activity of the neurons and in some way
communicating this to the blood vessels. In addition, astrocytes play a role in modulating neuronal
activity.31 It has been hypothesized that glia provide the energy substrate for the neurons by shuttling
lactate to the neurons,29 although this is controversial.34
The Function of Neurovascular Coupling and the Oxygen Limitation

Model
The mechanisms of neurovascular coupling do not necessarily clarify the function served; for example,
even if NO is the mechanism that triggers increased blood flow, why is the resulting CBF increase so
large? For the description of the function of neurovascular coupling, the details and the pathways of
the CBF regulation are less important.
Because CBF increases much more than CMRO2, so that E decreases with activation, this physiologic
35
phenomenon was originally called an "uncoupling" of blood flow and oxygen metabolism. However, it
is possible that the large change in CBF is an integral part of regulating oxygen delivery, so that the
decrease of E is necessary for increasing CMRO2, an idea we referred to as the oxygen limitation
26,36-40
model.
page 255
page 256
Add to lightbox
Figure 9-6 Illustration of the oxygen limitation model. The oxygen concentration, measured as a partial
pressure PO2, varies from approximately 100 torr in the arteries to approximately 30 torr in the veins,
with a mean capillary value of approximately 45 torr. The PO2 in the mitochondria is thought to be low
(approximately 5 torr). Oxygen diffuses from the high concentration in the capillary to a low
concentration in the mitochondria, and to increase the O2 flux the gradient must be increased. If
mitochondrial PO2 is already low, and there is no capillary recruitment to bring the blood closer to the
mitochondria, then mean capillary PO2 must be increased. This requires that the oxygen extraction
fraction E must be reduced, so CBF must increase more than CMRO2.
In the context of this model the large increase in CBF is required to support the smaller increase of
oxygen metabolism (Fig. 9-6). Although this idea has by no means been proven, it is the only
quantitative explanation that has been proposed for the function served by a large CBF increase. The
model is based on experiments showing two effects: 1. at rest a large fraction of the delivered oxygen
never leaves the capillary, but nearly all of the oxygen that does enter the extravascular tissue space is
7 of 9 12-04-2010 00:22
metabolized; and 2. CBF increases by increasing capillary velocity rather than by opening new
capillaries (i.e., there is no capillary recruitment).
The implications of these results can be understood from the viewpoint of diffusion down a
concentration gradient, with oxygen diffusing from a high concentration in the capillary to a low
concentration in the mitochondria. If the net extraction is nearly equal to the unidirectional extraction,
then backflux of O2 from tissue to capillary must be small, implying that the tissue PO2 is near zero.
With no capillary recruitment the diffusion distance from capillary to mitochondria is fixed, so to
increase the O2 flux either the mean capillary PO2 must be increased or the mitochondrial PO2 must be
decreased. If the mitochondrial PO2 is already near zero, then the only option is to raise capillary PO2
by raising the venous O2 content, and this means that E must decrease. At steady state, the product
of E and CBF must match this increased flux from capillary to mitochondria (Eq. 9-1), so CBF must
increase by a larger amount to overcome the decrease of E.
The original model for the effects of limited oxygen delivery assumed the extreme form of equal
unidirectional and net extraction fractions, and the prediction of that model was that the fractional
change in CBF would need to be approximately 5 times larger than the fractional change in CMRO2.
Measurements in the awake human brain with PET and calibrated fMRI have found this ratio to lie in
the range n = 2 to 6,35,41-46 so the predicted ratio lies near the high end of the experimental results.
However, this form of the model is undoubtedly too simple. Other factors can influence the diffusibility
of O2, such as capillary dilation and shifts of the binding curve of O2 and hemoglobin, and these effects
should be included in the modeling. 37 An increase of diffusivity through these mechanisms would soften
the requirement for the CBF increase, but detailed modeling of these effects in this context has not
been reported. Furthermore, in the original simple form of the oxygen limitation model the assumption
of equal unidirectional and net extraction fractions is equivalent to assuming no backflux of O 2 and a
tissue PO2 of zero. This is certainly an oversimplification, and is incompatible with studies of the
oxidation state of cytochrome oxidase, which indicate that at rest oxygen concentration is not a limiting
47
factor in determining the rate of oxidative metabolism in the mitochondria.
These observations can be reconciled with the oxygen limitation model if the mitochondrial PO2 is
greater than zero but still significantly less than the average capillary PO2, and the blood flow increase
serves to maintain a constant mitochondrial oxygen tension (see Fig. 9-6). Calculations indicate that a
CBF/CMRO2 ratio of n = 4 is required to maintain mitochondrial PO2 at a constant value while
increasing the oxygen flux.37 Other factors such as the Bohr effect and perhaps capillary dilation could
reduce the value of n required. In this case the mitochondrial PO2 would be held at a high enough level
that O2 availability would not become limiting in the mitochondria.
In short, the tissue PO2 would constitute a buffer that is normally not used with physiologic activation,
but which could come into play under some conditions at the beginning of stimulation or in hypoxia.
However, oxygen delivery is still fundamentally limited, in the sense that maintenance of this
mitochondrial PO2 requires a steady flux of O2 from the capillaries. Variability of this oxygen buffer
(e.g., through alterations of inspired oxygen content or resting CBF) is a possible source of the
variability between laboratories of detection of the initial deoxygenation26 (see "Exploring the
Hemodynamic Response to Brain Activation with MRI").
page 256
page 257
Is Neurovascular Coupling a Feed-Forward Mechanism?

By the model described above, the function served by a large CBF increase following neural activity is
to maintain mitochondrial PO2 at a constant value so that oxidative metabolism can proceed without
being limited by oxygen availability. Two possibilities arise for how CBF is regulated: 1. a feedback
model in which the amount of oxygen already available determines the increase in CBF following
functional activity; 2. a feed-forward model in which the change in neuronal activity determines the
change in CBF independent of oxygen availability and hence from the baseline CBF.
8 of 9 12-04-2010 00:22
A feedback system that responded to a drop in cytosolic PO2 would not work because the goal is to
increase cytosolic PO2 above baseline in an activated state and would only produce transient changes
in CBF. However, a feed-forward system is consistent with these theoretical ideas and with current
experiments using vasoactive agents. That is, the neural activity starts a cascade of biochemical
processes that leads to increased CBF, with no direct feedback about whether such a CBF change is
needed to support the increase in CMRO2. This idea is prompted by several recent studies that
measured the CBF or deoxygenation (i.e., BOLD) response to activation under conditions where the
48-50
baseline CBF was altered. Baseline CBF can be increased by breath-holding, inhalation of CO 2, or
acetazolamide and decreased by ingestion of caffeine. The remarkable finding from most of these
studies is that the increment of CBF change due to the activation is unaffected. The idea that the
neurons simply call out for more oxygen without sensing whether they have enough to begin with
seems inefficient. However, given the critical importance of oxygen for continued energy metabolism,
the links between hypoxia and the initiation of apoptosis, and the theoretical reasons above for why
oxygen itself is a poor error signal for a control system, a feed-forward system begins to seem more
plausible.
© 2010 Elsevier
9 of 9 12-04-2010 00:22
THE BOLD EFFECT

The primary MRI technique used for fMRI exploits the blood oxygenation level-dependent (BOLD)
effect. The physiologic basis of the BOLD effect is that CBF increases much more than CMRO 2 during
increased neural activity, and as a result the oxygen extraction fraction E is reduced and the venous
blood is more oxygenated. The reason this physiologic change is detectable with MRI is that the MR
signal is sensitive to microscopic magnetic field gradients, and deoxyhemoglobin and oxyhemoglobin
have different magnetic properties. In this section we consider this biophysical basis of the BOLD
effect in some detail, with the goal of deriving basic mathematical models that relate the underlying
physiologic changes to the measured signal. This is not a simple transformation, and indeed one of the
complexities involved in interpreting the BOLD signal is that it depends on several physiologic variables.
Specifically, the oxygen extraction fraction E governs the oxygen saturation of blood leaving the
capillary, but the blood volume (CBV) also affects how much deoxyhemoglobin is present in an image
voxel. In the following sections we consider how local changes in deoxyhemoglobin alter both the
intravascular and extravascular MR signals for gradient-recalled echo (GRE) and spin-echo (SE)
acquisitions, and then use this as the basis for modeling the BOLD signal in terms of E and CBV.
Magnetic Susceptibility Variations Distort the Local Magnetic Field and

Often Create Image Artifacts
The central physical idea at the heart of the BOLD effect is that deoxyhemoglobin alters the magnetic
susceptibility of blood and creates magnetic field gradients around the vessels. In clinical MRI
magnetic susceptibility effects are usually familiar as a source of artifacts. In BOLD imaging we exploit
these effects, and indeed choose pulse sequences that are especially sensitive to magnetic
susceptibility, in order to allow us to detect the BOLD signal changes. To understand this phenomenon,
it is helpful to review the basic physics of magnetic susceptibility.
When any material is placed in a uniform magnetic field B0, the intrinsic magnetic moments within the
material partially align with the field, so that the material becomes slightly magnetized. The magnitude
of this induced magnetization is described by the magnetic susceptibility χ. The magnetization M = χ·
B0 is the equivalent dipole density of the material, and so depends on both the intrinsic dipole density
and the degree of alignment of the dipoles with B0. Because the material is now magnetized, it creates
an additional magnetic field that adds to B0. It is important to note that a uniformly magnetized object
generally produces a distinctly nonuniform magnetic field that depends strongly on the geometric shape
of the object. For example, a uniformly magnetized cylinder produces a uniform field inside, but a
dipole-like field outside (Fig. 9-7). This idealized geometry can serve as a model for the field
distortions around a blood vessel containing deoxyhemoglobin.
page 257
page 258
Add to lightbox
Figure 9-7 Magnetic field distortions around a cylinder due to a magnetic susceptibility difference
1 of 11 12-04-2010 00:23
between the inside of the cylinder and the surrounding medium. The cylinder is oriented perpendicular
to the main magnetic field B0, creating a dipole pattern of the Bz component in the space around the
cylinder.
To understand the effects of microscopic field distortions on the MR signal, it is helpful to think of the
field distortion (illustrated in Fig. 9-7) on two distinct spatial scales. On the macroscopic scale the full
frame of the figure corresponds to a normal image, and the field distortions are due to large structures
such as sinus cavities or the petrous bones. The field gradients due to these susceptibility differences
between tissues add to the imaging gradients and can create image distortions. On the microscopic
scale, Figure 9-7 corresponds to a single image voxel, with microscopic field distortions around venules
and small veins containing deoxyhemoglobin. In a gradient-recalled echo (GRE) MR experiment a 90°
radiofrequency (RF) pulse tips the longitudinal magnetization into the transverse plane where it begins
to precess with a frequency proportional to the field at the location of that spin. At time TE the signal is
measured, and the image intensity can then be viewed as a snapshot of the net signal from the voxel
at TE. If the signals from each sub-region of the voxel are out of phase due to local field offsets, the
net signal will be reduced due to the local phase dispersion. As TE increases so does the degree of
phase dispersion, and so the signal is more attenuated. Microscopic field distortions around blood
vessels thus attenuate the local signal.
On the other hand, if there was no microscopic variation in magnetic susceptibility, so that within a
voxel all spins experienced the same field offset, then the magnitude of the local MR signal would not
be reduced but the phase of that signal would reflect the field offset of the voxel. In other words, the
phase map of a GRE image can be taken as a map of large-scale magnetic field distortions within the
imaged object. The phase image in Figure 9-8 was collected in this way, and the light to dark
transitions, as the phase cycles from 359° to 0°, can be thought of as contour lines of the magnetic
field.
Add to lightbox
2 of 11 12-04-2010 00:23
Add to lightbox
Figure 9-8 Gradient-recalled echo (GRE) phase maps showing magnetic field distortions, with
magnitude images displayed on the top and phase images on the bottom: left, in a coronal section
through a human head, and right, around a cylinder (as in Fig. 9-7). The local phase is proportional to
the local field offset, and the transitions from black to white (from 359° to 0°) can be viewed as contour
lines of the magnetic field. Note the field distortion in the brain due to the sinus cavity.
Thus the local GRE signal is affected in two ways: macroscopic field variations across the brain affect
the phase of the signal, while microscopic field offsets within a voxel reduce the magnitude of the
signal. As noted above, magnetic susceptibility variations also can interfere with mapping the local
signal to its proper location in space. MRI relies on using magnetic field gradients to encode the spatial
origin of the MR signal within the signal itself. Specifically this means that the basic assumption of MRI
is that the magnetic field is perfectly uniform until the gradients are applied, so that any phase offsets
in the signal are due just to those applied gradients. Magnetic susceptibility differences between
tissues (e.g., brain, bone, and air) create broad static field gradients that add to the applied pulsed
field gradients used in imaging. The result is that the MR signals are mismapped, creating distortions in
the image. In Figure 9-8 the large-scale field distortions due to the sinus cavities are readily apparent.
In general, these large-scale effects are a nuisance, distorting the image and sometimes creating
signal dropouts. In addition, signals from spins lying in different fields are mapped into the same voxel.
When this happens the individual signals add incoherently and so partially (or completely) cancel out. In
functional brain imaging, this is most prominent in the prefrontal cortex and midbrain.
Magnetic Susceptibility Changes due to Blood Oxygenation Create the

BOLD Effect
page 258
page 259
The field distortions around a magnetized blood vessel due to deoxyhemoglobin (see Fig. 9-7) occur on
a much finer spatial scale than the broad field distortions that produce image distortions. For modeling
the effect of these microscopic field distortions, our primary goal is to understand how they produce
phase dispersion within the voxel and a reduction of the net signal. Initially we will consider the GRE
pulse sequence, the technique most sensitive to field offsets. Empirically, the signal decay with
increasing TE in a GRE image is described in terms of an apparent transverse relaxation time T2*. In
contrast, with a spin-echo (SE) pulse sequence, a 180° RF refocusing pulse is applied at time TE/2.
By reversing the phase of each local signal halfway through, the phase accumulation due to a constant
field offset is perfectly refocused for each local spin group. As a result, the effects of local field offsets
3 of 11 12-04-2010 00:23
are compensated, and the signal decay with increasing TE is described by the true intrinsic T2, with T2
being greater than T2*. This simple picture must be modified to consider how diffusion makes the
refocusing incomplete, so that the SE signal does have some sensitivity to the BOLD effect. This is
discussed below, after first describing the more pronounced BOLD effect seen with GRE imaging.
The basic effect of increased deoxyhemoglobin creating field offsets around the vessels and reducing
the MR signal can be described as a reduction of T2*. With activation there is less deoxyhemoglobin
present, and so the BOLD effect is really a partial lifting of this signal reduction. That is, the signal
increase associated with the BOLD effect can be described as an increase in the local T2*. In the
literature, the decay of the GRE signal is often written as e-TE·R2*, where R2*, the apparent transverse
rate constant, is simply the inverse of T2*. This formulation is simpler to work with in the modeling
(considered in more detail below). At 1.5 T, a typical value for T2* in the brain is 50 ms (R2* = 20 s-1).
With activation, the increase of T2* is described as a decrease of R2*: a negative ∆R2*. A strong
activation measured at 1.5 T will create a change in R2* due to the BOLD effect that is on the order of
∆R2* = -1.0 s-1. With TE = 40 ms, the fractional change of the MR signal is then about 4%.
From the basic arguments above, the field distortions around vessels due to deoxyhemoglobin should
affect the GRE signal, but not the SE signal. However, the physical picture is not yet complete; we
must still consider the effects of diffusion and the changes of the intravascular signal, and these effects
make the resulting GRE and SE signals more complex than the basic picture above.
Diffusion of Water Molecules Moderates the GRE-BOLD Effect and

Creates the SE-BOLD Effect
Water molecules continually tumble and move about due to random thermal motions, and molecules
starting at the same location will spread out over time like a drop of ink in water. Specifically, for pure
diffusion a group of spins starting at x = 0 at t = 0 will spread into a gaussian distribution centered on x
= 0 (there is no shift of the mean position) with a variance that grows with time. This variance is σ2=
2Dt, where D is the diffusion coefficient (approximately 10-5 cm2/s, or 1 μm2/ms in brain). Note that this
means that the characteristic width of the distribution, σ, grows only with the square root of the time.
For example, in brain the characteristic distance moved due to diffusion in 10 ms is approximately 4.5
μm, and in 100 ms it is approximately 14 μm.
Add to lightbox
Figure 9-9 Diffusion through magnetic field gradients affects the GRE signal. Water molecules
undergo a random walk, shown as a jagged path, through the magnetic field gradients around a
vessel. The spatial scale of the field gradient is proportional to the vessel radius, shown as a contour
map on the left. The effects of diffusion are important only if the random walk carries the spin through a
range of field offsets during the duration of the experiment TE. On the right is shown a random walk
corresponding to a 40 ms duration compared with the spatial scale of capillaries, venules, and small
veins. Motion due to diffusion will tend to sample the full range of field offsets for a capillary but will
have little effect for a small vein.
This diffusion of water molecules is slow, but it does have a measurable effect on the MR signal. To
4 of 11 12-04-2010 00:23
understand the effect, we start by considering a simple GRE pulse sequence used to image a voxel
containing magnetized vessels. An excitation pulse is applied at t = 0, and the image data are acquired
at t = TE. To keep the argument simple, imagine that the data acquisition time is very short, so that we
are essentially taking a snapshot of the net signal at the time TE. The signal attenuation is directly
related to the dispersion of the spin phases at time TE. If there is no diffusion each spin stays at the
same location, and precesses in the same field offset, for the entire duration TE. In this case the
distribution of phases is directly proportional to the distribution of field offsets. If we now introduce
diffusion, each spin undergoes a random walk, moving around and sampling different locations (Fig.
9-9). For any one spin the final phase will correspond to the average field offset experienced by the
spin in its random walk-diffusion effectively smooths the distribution of field offsets. In the limiting case
of very rapid diffusion, each spin wanders all around the vessel, sampling the full range of field offsets.
This reduces the phase dispersion because all the spins have similar histories, regardless of where
they started out at t = 0. In this limit there is little phase dispersion and so little signal attenuation. This
is called motional narrowing, meaning that the random motions have narrowed the distribution of
phases at the time of data acquisition.
page 259
page 260
Add to lightbox
Figure 9-10 Change in the transverse relaxation rates ∆R2 (for an SE experiment with TE of 100 ms)
and ∆R2* (for a GRE experiment with TE of 40 ms) of the extravascular signal as a function of the
radius of the vessel creating the magnetic field distortion. For large vessels (the static broadening
regime) the gradient-echo (GRE) effect is large, but the SE effect is small because the 180° pulse is
effective in refocusing the signal. For very small vessels (the motional narrowing regime) diffusion
ensures that each spin samples the full range of field offsets around the vessel, and there is little phase
dispersion for either GRE or SE signals. The intermediate regime occurs when the typical distance
moved due to diffusion is comparable to the size of the vessel, and only in this regime is there an
appreciable SE effect. This regime happens to correspond to the capillary and small venule size
scale. (Curves adapted from Weisskoff.56 Weisskoff RM: Basic theoretical models of BOLD signal
change. In Bandettini PA, Moonen CTW (eds): Functional MRI. Berlin: Springer, 1999.)
In short, diffusion moderates the effects of the field offsets, so the GRE signal is most attenuated
when diffusion effects are negligible, and much less attenuated when diffusion effects are prominent
(Fig. 9-10). The key physical parameter that defines whether diffusion is important is τ, the typical time
2
required for a spin to diffuse a distance equal to the radius of the magnetized vessel (τ = r / 2D,
where r is the radius of the vessel). For example, for a capillary with radius 3 mm, τ is approximately
4.5 ms in the brain, for a venule with radius 15 mm τ is approximately 112 ms, and for a small vein with
a radius of 100 mm τ is approximately 5 s (Fig. 9-9 illustrates the relative scale of a random walk path
for TE = 40 ms compared with different vessels). If τ is much less than TE, so that each spin is able to
sample all of the field offsets around a magnetized vessel, then diffusion effects are important and
there is little signal loss due to the field offsets. On the other hand, if τ is much greater than TE, then
each spin essentially samples only the relatively uniform field near its starting location, and the larger
phase dispersion between spins attenuates the signal. From these arguments we expect that for a
5 of 11 12-04-2010 00:23
typical TE = 40 ms, diffusion reduces the GRE-BOLD effect around capillaries but has little effect on
venules and small veins.
For SE measurements, the same ideas apply, but the physics is a bit more complicated because of
the 180° refocusing pulse. One can think of the action of a 180° pulse as flipping the sign of the
precessional phase of a spin. If a spin has acquired a phase ϕ1 during the interval TE/2 due to field
offsets, then after the 180° pulse the phase will be -ϕ1. During the next interval TE/2, the spin will pick
up an additional phase ϕ2, so the net phase angle at the time of data acquisition (TE) is ϕ2 - ϕ1. If
these accumulated phases are due just to precession with a constant field offset, then -ϕ1 = ϕ2 and the
net phase is zero-all spins will be coherent at the time of data collection. When diffusion is added to
this simple picture, the two phases ϕ1 and ϕ2 will not be the same, because each is the result of an
independent random walk. In short, if field distortions are on a small enough scale that diffusion
becomes important, the spin-echo process does not fully refocus the signal, and so there is attenuation
of the SE signal.
In general, the resulting attenuation of the SE signal is hard to model because it depends on the
geometry of the field offsets and the interaction of this geometry with the diffusion process. 51-56
However, two limiting cases bracket the effect, and define the character of the signal attenuation (see
Figs 9-9 and 9-10). At one extreme is the case TE is much less than τ, in which the available diffusion
time is much less than the time required for a diffusing spin to sample all of the field offsets. In this
"static broadening" regime the effects of diffusion are small, the 180° pulse works well to refocus
phase offsets, and there is little signal attenuation. At the other extreme, TE is much greater than τ,
each spin samples all of the field offsets and acquires approximately the same phase in the two
periods before and after the 180° pulse. In this "motional narrowing" case there is again little
attenuation, although it is not because of the 180° pulse; as described above, in this regime there is
little attenuation of the GRE signal either. The result is that it is only in the "intermediate regime," when
TE is approximately equal to τ, that the SE signal is attenuated. Interestingly, diffusion of water around
brain capillaries appears to be in this intermediate regime, and for this reason the SE signal has often
been viewed as being particularly sensitive to susceptibility changes in the capillary bed rather than
larger veins. A particular concern is that a vein draining an activated region may be displaced from the
actual site of activity, producing errors in localization, and that for this reason SE-BOLD activations,
although weaker, may provide better localization. However, there is one more complicating factor that
must be considered: the intrinsic signal change of the blood itself.
Intravascular BOLD Effects Make a Strong Contribution to the Net

Signal Change
page 260
page 261
The discussion above has dealt primarily with the extravascular signal change. The blood has been
treated as a uniform medium whose susceptibility depends on hemoglobin saturation. This is a
reasonable model for the field offsets produced outside the vessel, as in Figure 9-7. But the reality, of
course, is that the blood is a much more complex medium. The deoxyhemoglobin is sequestered in the
erythrocytes, disks a few μm across, and the water molecules of blood diffuse around and through the
erythrocytes. Because the intravascular water is much closer to the source of the paramagnetic
inhomogeneity-the deoxyhemoglobin-the phase changes produced would be expected to be much
larger than for the extravascular water. This effect is partly tempered by diffusion effects for GRE
signals, but the net signal from blood is much more strongly dependent on the oxygenation of the
hemoglobin. While the net signal change from extravascular spins may change by only approximately
1% with activation, the intravascular signal may change by 40%. If the venous blood volume fraction is
approximately 3%, a 40% change of the intrinsic intravascular signal would contribute a change of
1.2% to the net signal change. For this reason, at least half of the net BOLD signal change at 1.5 T
57
has been estimated to be due to the intrinsic change of the intravascular component.
Recently, modeling of the SE signal of blood has suggested that most of the signal change seen with a
SE experiment at 1.5 T is due to the change of the blood signal.58,59 This complicates the argument
6 of 11 12-04-2010 00:23
that the SE signal is more sensitive to changes at the capillary level. The largest signal change in blood
should be in the veins, where the greatest change in oxygenation occurs. Because the extravascular
signal change with a SE experiment is small, this suggests that the SE signal may be even more
dominated by draining veins than the GRE signal. Note, however, that the contribution of the
intravascular signal change becomes smaller at higher magnetic fields. As the magnetic field increases,
the T2* of venous blood becomes very small, so that even with activation the intravascular signal
becomes unimportant. In this high field limit, the SE signal should only depend on extravascular signal
changes, and these in turn should depend primarily on changes in oxygenation of the capillary blood.
Thus, at high magnetic fields the SE signal, although intrinsically less sensitive to the BOLD effect than
the GRE signal, may nevertheless be more specific to the capillary bed and less sensitive to large
draining veins.
Modeling the BOLD Effect

Davis and colleagues introduced a model for the BOLD signal change based on reasonable
approximations and the results of numerical simulations.43 Because of its simplicity, the model has
proven to be a useful tool for understanding the BOLD effect in a quantitative way. The model starts
from the simple picture of how the BOLD effect arises, and relates the signal change to the underlying
physiologic variables and a few parameters that describe the local tissue.
The MR signal is modeled in the usual way as a simple exponential dependence on the echo time TE,
and can be written as:
where S0 is the effective spin density (the signal that would be measured if TE could be reduced to
zero). The transverse relaxation rate constant R2* is written as a sum of two terms: R0 is the value of
R2* if no deoxyhemoglobin is present, and R describes the additional relaxation produced by
deoxyhemoglobin. Note that typically R0 is much larger than R, i.e., the local T2* that describes the
decay of the signal is largely determined by the intrinsic T2 and large-scale field gradients through the
voxel, and the additional effect of deoxyhemoglobin is minor.
We now assume that with activation R is the only parameter that changes. Using the subscript "rest" to
denote the resting value, and "act" to denote the activated value, the BOLD signal change with
activation, ∆S = Sact - Srest, is then:
The key question is: how does ∆R depend on blood oxygenation and volume? The magnitude of the
field offset ∆B near a magnetized vessel is proportional to ∆χ?B0, the magnetic susceptibility
difference between the blood and the surrounding extravascular space multiplied with the main
magnetic field. Experiments indicate that ∆χ, in turn, can be accurately modeled as having a linear
dependence on the local deoxyhemoglobin concentration in blood, and this quantity in turn can be
expressed in terms of the change in the oxygen extraction fraction E. Specifically, the concentration of
deoxyhemoglobin is proportional to E. Then the local field offsets produced by deoxyhemoglobin should
vary as:
However, the scaling of the field offsets does not necessarily define the scaling of the signal
attenuation. As spins evolve in an inhomogeneous field a distribution of phase angles develops, and it is
this phase dispersion at the time of data collection that determines the signal change and thus ∆R2*. In
particular, as noted above, diffusion effectively smooths the field distribution to create a narrower
spread of phases in a GRE experiment. To model this in an approximate way we follow Davis et al and
assume a power law relationship between R and ∆B, the magnitude of the field distortions:
7 of 11 12-04-2010 00:23
Numerical simulations54,60 and theoretical analyses55 suggest that when diffusion is not important, β
approximates to 1. Numerical simulations with diffusion suggest that β approximates to 2 gives a better
description around the smallest vessels when diffusion effects are significant. In reality we are dealing
with both intravascular and extravascular effects. Note that with GRE methods, the signal change is
likely to be largest in the vicinity of small veins, where the change in deoxyhemoglobin is largest. For
the extravascular signal we would expect that diffusion is not important, and so β ≈ 1 would be
appropriate. However, within the vessel diffusion is very important, and so β ≈ 2 would be more
appropriate for the intravascular signal change. Numerical simulations suggest that β ≈ 1.5 is a good
approximation for the net change in R at 1.5 T. At higher magnetic fields, the intrinsic signal of the
intravascular compartment is reduced, and the BOLD signal is dominated by the extravascular effect,
and β should approach one.
page 261
page 262
In addition to the change in E with activation, a change in blood volume also affects R. For example,
even if the oxygenation of the blood did not change but the venous blood volume increased, the total
deoxyhemoglobin would be increased, and we would expect this to increase R and decrease the net
MR signal. Numerical simulations suggest that a reasonable approximation is to assume that R is
proportional to V, the venous blood volume.
Combining these arguments, we can model the BOLD signal as43:
where k is the net proportionality constant from all of the proportionalities above, V is blood volume,
and A is a local scaling factor given by:
This final equation for the BOLD signal change is quite simple, depending on two physiologic changes
(the change in blood volume V and oxygen extraction fraction E) and two additional parameters β and
A. The form of the signal equation directly describes the ceiling effect on the BOLD signal. In simple
terms, A is the maximum BOLD signal change that could occur, corresponding to complete removal of
deoxyhemoglobin from the voxel. The parameter β should be primarily field dependent, and we can
assume that it is not a function of brain region. The parameter A, however, is a local parameter and so
may vary across different voxels in the brain. Note that this parameter is proportional to the value of R
at rest, the relaxation rate produced by deoxyhemoglobin in the baseline state. This means that the
more deoxyhemoglobin is present at rest, the larger the BOLD signal change will be for the same
fractional change in V and E with activation. We will come back to this later in the chapter when we
consider the effect of the baseline condition on the magnitude of the BOLD effect.
At 1.5 T, measured values of A range from 0.08 to 0.22.43,50 In practice, the observed BOLD signal
change, ∆S, in brain tissues is somewhat smaller. Signal changes of 1.9 ± 0.7% are seen in brain
parenchyma of visual cortex at 1.5 T. This increases superlinearly with increasing B0 field strength to
3.3 ± 0.2% in visual cortex at 4 T.61 In addition to measuring the BOLD signal, the underlying
hemodynamic changes in CBF and CBV can also be measured independently. This is important for
calculating the local CMRO2 (see "Measuring Cerebral Blood Flow, Cerebral Metabolic Rate of
Oxygen, and Cerebral Blood Volume"). Additionally, these individual components of the BOLD
response can have a superior spatial correspondence with the underlying neural activity than the
measured BOLD signal itself (see "Exploring the Hemodynamic Response to Brain Activation with
MRI").
Although this is a very useful model for the BOLD signal, the reader should bear in mind that it does
not describe all of the effects that may contribute to the measured signal change in an activation
experiment. Specifically, small direct effects of CBF and CBV changes on the MR signal that are
8 of 11 12-04-2010 00:23
independent of the BOLD effect are likely present in real data. For example, if the repetition time TR is
shorter than the T1 of blood, and the flip angle is large (e.g., 90°), the increased delivery of fresh
unsaturated blood due to increased CBF could increase the net signal slightly. In addition, the intrinsic
signal from arterial blood is larger than the intrinsic signal of the extravascular space, so increasing the
arterial blood volume fraction of the voxel also could produce a slight signal increase. Note that both of
these effects are due to arterial blood changes, where deoxyhemoglobin is negligible, so these are
effects in addition to the BOLD effect. In most applications these effects are thought to be small
compared to the BOLD effect, especially at higher magnetic fields, but they may not be negligible.
Dynamics of the BOLD Signal

Equation 9-7 relates the BOLD signal change to the change in oxygen extraction fraction E and blood
volume V. We can go one step further and try to relate these two physiologic quantities to the change
in CBF. The CBF increase associated with neural activity is triggered by a relaxation of the smooth
muscle in the wall of the arterioles. The arterioles provide most of the resistance in the vascular tree
and provide a way to quickly decrease the vascular resistance by dilating. As the resistance of the
arterioles decreases, the pressure drop across these vessels also decreases, raising the pressure in
the capillaries and veins. These vessels may also expand due to the increased pressure, further
increasing the CBV. As noted earlier, experimental studies have suggested that the relationship
between CBF and CBV can be modeled as a power law (Eq. 9-2) with an exponent that is
approximately α = 0.38.19
The oxygen extraction fraction E depends on the change in CBF relative to the change in CMRO2.
From Equation 9-1 at steady state, this relationship is:
We can adopt a more compact notation by introducing dimensionless variables in which each
physiologic variable is normalized to its value at rest: f = CBFact/CBFrest, v = Vact/Vrest, and m =
CMRO2act/CMRO2rest. Then for the resting state f = v = m = 1, and with activation all three variables
α
increase to different degrees, and Equation 9-2 becomes v = f . The BOLD signal can then be
expressed as:
page 262
page 263
Equation 9-10 can be used to illustrate the possible dynamics of the BOLD effect. Each of the
physiological parameters v, m, and f is a function of time. When a stimulus is applied the state of the
brain changes to a new steady state, characterized by particular relationships between these
physiologic variables. Empirically, these relationships are:

where the first equation is simply Equation 9-2 in our notation with normalized physiologic variables.
The second equation represents the empirical finding that CBF and CMRO2 increase together, but the
fractional change in CBF is larger than the fractional change in CMRO2 by a factor of n. Most
43,44
experimental studies have found n = 2 to 3, although larger values also have been reported.
Equations 9-11 can be taken as steady-state relationships between the physiologic variables.
However, during the transitions between one steady state and another the variables may change at
different rates, even though they eventually settle into the relationships defined by Equations 9-11. This
possibility for different dynamics of the underlying physiologic variables can introduce transient features
into the BOLD signal response (Fig. 9-11).
9 of 11 12-04-2010 00:23
Add to lightbox
Add to lightbox
Figure 9-11 Dynamic curves for CBF, CMRO2, CBV, and the resulting BOLD signal, showing left, a
simple response in which the physiologic variables co-vary with similar time courses, and right,
transients due to a CBV change that is slow to return to baseline, creating a post-stimulus undershoot.
In addition the CBF response lags behind the CMRO2 by 1 s, producing an initial dip.
For example, a common observation is a dip of the BOLD signal below baseline after the end of the
stimulus, referred to as a post-stimulus undershoot, that can take a long time to resolve (more than 30
s). In block design experiments in which the rest period is not sufficiently long to allow the undershoot
to resolve, the effect appears as an apparent lowering of the baseline after the first stimulus block.
When the CBF response itself is measured with arterial spin labeling techniques, the post-stimulus
undershoot is smaller (or not evident) and resolves more quickly. For this reason, the post-stimulus
undershoot of the BOLD signal cannot be explained in terms of an undershoot of CBF. Two similar
models, the balloon model20 and the windkessel model,21 have been proposed to explain this
phenomenon as a slow return of CBV to baseline after the stimulus. These models were motivated by
the observation of such a slow return of CBV in a rat model, 62 corresponding to the duration of the
BOLD post-stimulus undershoot, and occurring despite a more rapid return of CBF to baseline. The
basic idea is that if CBF and CMRO2 return to baseline values, the oxygenation of venous blood also
returns to normal. However, if the total venous blood volume remains elevated, there is more
deoxyhemoglobin present, so the signal dips below baseline. Figure 9-11 illustrates this phenomenon
with balloon model calculations.
A less common finding is a brief initial dip of the BOLD signal prior to the larger signal increase.63-68
26
The effect is small and not always present, but it has stirred interest because it may reflect a rapid
increase of CMRO2 prior to the CBF increase,69 and this phenomenon may be better localized to the
area of increased metabolism (that is, the CBF increase may cover a wider area). Optical studies have
found a brief initial increase of deoxyhemoglobin, consistent with the initial dip of the BOLD signal.69-72
10 of 11 12-04-2010 00:23
This was interpreted as an increase of CMRO2 that precedes the increase in CBF, creating a short
increase in the oxygen extraction fraction E. In principle, however, an initial increase in
deoxyhemoglobin also could occur due to a rapid rise in blood volume V. Early experimental studies
usually showed an increase in deoxyhemoglobin, but no corresponding decrease of oxyhemoglobin,
suggesting a combined change of E and V. A more recent study, however, found a corresponding
initial decrease in oxyhemoglobin in conjunction with an increase in deoxyhemoglobin, more clearly
73
suggesting a change in E. Figure 9-11 shows how a delay of 1 s between the CMRO2 response and
the CBF response can produce an initial dip in the BOLD signal.
© 2010 Elsevier
11 of 11 12-04-2010 00:23
DESIGN AND ANALYSIS OF BOLD-fMRI EXPERIMENTS

Statistical Analysis is Required to Detect Small Signal Changes
The basic data analysis done in fMRI is to examine the time course from each voxel of the image and
identify those voxels in which the signal intensity increases when the stimulus is on. The intrinsic
difficulty with this analysis is that the effect is subtle, with a CBF change of 30% causing a BOLD
signal change of only approximately 1% at 1.5 T.
page 263
page 264
Even without any stimulation the MR signal varies, due to a number of physical and physiologic
perturbations. Because the BOLD signal change is small, it is necessary to take these perturbations
into account when performing a statistical analysis. The key challenge is to distinguish the signal
changes due to neural activity in response to the stimulus from all the other random and systematic
sources of variation of the MR signal. Hence, the general idea of the statistical analysis is to estimate
whether a response to the stimulus pattern accounts for a sufficiently large fraction of the measured
signal variance to reach statistical significance. Because the statistical analysis is an integral part of an
fMRI experiment, in this section a short introduction to the general principles is presented.
The MR signal always contains random thermal noise, and this adds random fluctuations to the signal
that are also on the order of 1%. A primary goal of MR technology development is to improve the
intrinsic signal-to-noise ratio (SNR), either by moving to higher magnetic fields or by improvements in
RF coil design, such as multichannel coils. However, random thermal noise is not the only source of
signal variability, and indeed often is not the primary problem. A number of systematic effects also
create signal fluctuations, including instrumental effects, such as coil heating, but also physiologic
fluctuations. The most important physiologic fluctuations are due to heartbeat (approximately 1 Hz),
respiration (approximately 0.3 Hz), and the so-called "slow oscillations" (approximately 0.1 Hz, and
discussed further in "Exploring the Hemodynamic Response to Brain Activation with MRI"). In addition,
the appearance of these fluctuations in the MR signal is complicated because the data sampling rate
(1/TR) is typically not fast enough to resolve the cardiac frequency, and so these fluctuations are
aliased to lower frequencies in the data. For these reasons, the observed fluctuations in the MR signal
are a composite of physical and physiologic perturbations.
The required statistical analysis is essentially analysis of variance. The signal exhibits fluctuations
around a mean value, and the variance of these fluctuations has several sources: 1. the BOLD
fluctuations driven by the stimulus, the effect of interest; 2. systematic physiologic and instrumental
fluctuations, described as confounding effects; and 3. random errors that are taken to be gaussian
noise. The aim of the statistical analysis is to determine if there is sufficient evidence to say with
confidence that the contribution of the effect of interest is different from zero. This is done by
calculating a statistic, such as a correlation coefficient, a t-statistic, or a z-score, that essentially
measures the ratio of the signal variance that can be attributed to the effect of interest (stimulus-driven
BOLD signal changes) to the variance due to noise, after taking into account the variance due to the
confounding effects.
Note that in this formulation we have specifically separated systematic effects from gaussian noise.
This is essentially a physics approach, aimed at identifying and measuring all systematic fluctuations in
order to remove them from the data. For example, key physiologic variables such as the cardiac and
respiratory cycles can be measured during data acquisition, and their contributions to the total signal
variance can be included in the analysis (see "Artifacts and Noise"). A more general statistical
approach is to consider all of the error signals together as one source of error, but with a non-gaussian
distribution. Then the data can be analyzed to empirically derive the statistics of the noise, or
smoothed (temporally and/or spatially) to force the distribution closer to gaussian. The combination of
these two approaches, the physical and the statistical, provides the best promise for achieving the
maximum sensitivity for detecting subtle activations, by carefully removing as many physiologic sources
of variance as possible, and then analyzing variance that remains to determine the distribution. For the
rest of this section, which is meant to be simply an introduction, we will assume that the total error
1 of 9 12-04-2010 00:23
consists of random gaussian noise and low-frequency systematic variations of the MR signal.
The General Linear Model Provides a Statistical Framework for

Incorporating All the Components of the Signal
The most used statistical method in analyzing fMRI data is the general linear model (GLM).74
Essentially, the GLM regards the measured signal as the weighted sum of different model functions,
with unknown weights, and random gaussian noise. The analysis then amounts to fitting the collection
of models to the data to estimate the weighting factors, or more generally, the unknown parameters
that define the model functions. An illustration of the GLM is shown in Figure 9-12, with model functions
shown for the stimulus response, a systematic cardiac signal, and a linear drift.
The stimulus response function is usually obtained by convolving the stimulus waveform with an
assumed ideal impulse hemodynamic response, which can be represented by one or more of the
so-called "basis functions". As an example, in Figure 9-13 a gamma-variate hemodynamic response
function it with a full width at half maximum (FWHM) of about 4 s and its derivative are shown.
The mathematical expression of the gamma-variate h(t) is:
Typical values used are k = 3, and τh = 1.2. In addition, to cover the delay of the hemodynamic
response, a time lag of 1 s is used. Ideally, this delay is simply another parameter of the model
function, and should be estimated along with the amplitude. However, the signal does not depend on
this parameter in a linear way, so it cannot fit into the GLM analysis. A shift of the hemodynamic
response function can be approximated in a linear way by including the derivative of it as a second
model function. Then to a first approximation, a non-zero weight for this additional model function shifts
it in time. That is, the weighted sum of h(t) and its derivative is similar to a shifted hemodynamic
response function (HRF), so by using both functions hemodynamic responses with different delays can
be modeled. This is very useful because it is known that the HRFs in different brain areas vary in their
delays.75
In mathematical terms, the total MR signal time course from a voxel can be modeled as:
page 264
page 265
2 of 9 12-04-2010 00:23
Add to lightbox
Figure 9-12 Illustration of the general linear model (GLM). Model functions such as an ideal stimulus
response, physiologic noise, and scanner drift are fitted to measured data assuming that the residual
is purely random noise. For each voxel the GLM analysis provides estimates of the weights for each
model function that give the best fit to the data. The GLM analysis tests whether the estimated
amplitude of the stimulus-driven model function is significantly different from zero for the estimated
level of noise.
The term Y denotes BOLD signal, the terms Xi the model functions, β their weightings, and e the
residual noise assumed to be gaussian. In a more compact way this equation can be written in matrix
terms:
The matrix X is called the design matrix, the vector Y the observed variable, the term β is the weighting
vector, and e the noise vector. For each voxel the GLM analysis provides estimates of the weights for
each model function that give the best fit to the data. The best fit is defined in a least-squares sense,
such that the sum of the squared residuals, obtained after the model fit is subtracted from the data, is
minimized. Provided that the noise has a gaussian distribution, the least-squares solution to Equation
9-13 is:
The superscript T denotes the transpose of the matrix.
Identifying Activated Voxels
3 of 9 12-04-2010 00:23
Add to lightbox
page 265
page 266
Figure 9-13 A gamma-variate hemodynamic response function (HRF) and its derivative. The ideal
stimulus response function used in the GLM formalism is usually obtained by convolving the stimulus
waveform with an assumed ideal impulse hemodynamic response. A shift of the HRF can be
approximated in a linear way by including the derivative of HRF as a second model function.
The statistical goal is now to determine if the voxel time course shows a stimulus response, by testing
whether the amplitude of the stimulus-driven model function estimated from Equation 9-14 is
significantly different from zero. That is, even if there was no stimulus-driven activity whatsoever in the
voxel, the GLM analysis could produce a non-zero amplitude for the stimulus model function just
because of the randomness of the noise. The key question is whether the estimated amplitude is
sufficiently larger than the range of amplitudes we would expect due to just noise alone. For this
purpose the signal contribution due to the stimulus response has to be compared with the noise level.
The higher the signal-to-noise ratio, the more likely it is that the voxel is truly activated. This can be
quantified in terms of a t-statistic, whose distribution is known, so that the probability that the fitted
amplitude could result from chance-the p-value-can be estimated. For example, a p-value of 0.05
means that by chance there is a 5% probability that this t-value or larger is obtained. The higher the
t-value, the higher the significance and the less likely that this is a chance event, and thus the lower the
p-value.
An alternative way of describing the signal-to-noise value is the correlation between the model stimulus
response and the measured response.76 The correlation coefficient r varies between -1 and 1, with the
value 1 being a total correspondence between the reference function and the measured response and
the value -1 corresponding to a perfect inverse correlation between the responses. The value r = 0
means that there is no correspondence between the model response and the data. Analyzing the data
in terms of either a t-value or a correlation coefficient is equivalent, because r can be transformed to a
corresponding t-value using the formula:
The variable ν represents the "degrees of freedom" (dof), which is the number of time points reduced
by the number of model functions. The source of this term is that we begin with N independent
measurements in the time series, and from these data we estimate k model parameters. This leaves ν
= N-k measurements to estimate the variance of the random gaussian noise (i.e., the variance of what
remains after subtracting out the best-fit model functions). For this reason the significance of a given
correlation coefficient depends on the degrees of freedom; the higher the dof, the higher is the
statistical significance.
If the stimulus is presented in a periodic fashion, another approach for determining if there is a
statistically significant activation is to work in the frequency domain and compare the amplitude at the
stimulus frequency with the amplitudes at other frequencies. A useful feature of this approach is that
the Fourier transform also provides a phase for each frequency component, and this phase translates
4 of 9 12-04-2010 00:23
directly into the time delay between the frequency components. For a periodic design the main
frequency representing the stimulation is the inverse of the time of the stimulation cycle. Hence, the
magnitude of this component can be statistically compared with the magnitude of the other frequency
components using the t-statistic as described above in order to determine the contribution of the
stimulus to the total variance. An additional feature of this frequency analysis is that it can give
information about other neuronal populations that have an atypical response to the stimulus pattern.
For example, neurons that respond while the stimulus is on will show a strong response at the
fundamental frequency of the stimulus, while neurons that respond to a change of stimulus, firing at the
beginning and end of the stimulus, will show a strong response at twice the fundamental frequency.
While the implementation of the Fourier approach seems different from the GLM approach, it is really
just a GLM analysis with sine waves as the model functions. The particular advantage of sines and
cosines is that they are derivatives of each other, and a shifted sine wave, representing a delayed
response, is precisely modeled as a linear combination of a sine and cosine wave with the same
frequency. For the more general shape of response described above, the inclusion of the derivative
function as another linear model function is only an approximate way to include a shift of the response.
For sine waves it is exact.
Statistical Parametrical Maps are Used to Display Activated Voxels

After the time course of each voxel has been analyzed to determine a t-value (or correlation
coefficient), the data are displayed as a statistical parametric map. This is typically done by choosing a
threshold for the statistic and overlaying the map of voxels passing the threshold on a high-resolution,
black and white anatomic image. The colored image is then presented as a map of the brain activation
pattern associated with the stimulus. However, it is important to note that this approach to deciding
whether a voxel is activated or not is based on a signal-to-noise evaluation, not a pure signal
evaluation. For example, it is conceivable that two voxels in different parts of the brain would exhibit the
same neural activation and BOLD signal change (say, 1%), but the other sources of signal fluctuations
are much higher in one voxel than the other. Because the statistical test for whether a voxel is
activated measures the signal-to-noise ratio, one voxel could pass the threshold and be identified as
"activated" while the other does not, despite identical levels of BOLD signal change. To put this more
precisely, this statistical parametric mapping approach identifies voxels in which the signal time course
shows a fluctuation due to the stimulus that is sufficiently large that it is unlikely to happen randomly
due to the noise fluctuations of the voxel. Thus this is a test that is designed to protect against false
positives, but not against false negatives. For this reason, the correct interpretation of a statistical
parametric map is that we can say with confidence that the colored voxels were activated, but we
cannot make that statement for the other voxels. In other words, we cannot conclude from this analysis
alone that the other areas were not activated unless we know that the noise level is similar across the
brain.
Limitations of the General Linear Model

page 266
page 267
Potentially, there are several problems with the GLM approach deriving from the basic assumptions,
which are not necessarily valid for fMRI data obtained on the living human brain. The assumptions and
prerequisites of the GLM are:
1. Model functions have to be specified. This requires prior knowledge about which functions-
hemodynamic response and confounds-should be included in the design matrix. The
hemodynamic response is still incompletely understood (see "Exploring the Hemodynamic
Response to Brain Activation with MRI") and may vary across the brain and with age and
75
disease. In order to be flexible with the assumed response, two or more hemodynamic model
functions (see above) can be used enabling modeling of different hemodynamic responses.74
The other model functions used in the analysis are usually linear and quadratic drifts due to
scanner instabilities. The physiologic confounds are more difficult to treat, and it is still an area
of active investigation to determine how they translate into fluctuations in MR signal time series
5 of 9 12-04-2010 00:23
(see "Artifacts and Noise"). Several correction schemes have been proposed for removing
cardiac and respiratory fluctuations (for an overview, see reference 77).
2. The residual noise is assumed to be gaussian, although real fMRI data show a non-gaussian
distribution. One approach to deal with this problem is to "whiten" the noise into a gaussian form
by filtering.78 Another approach is to use the data themselves to estimate the distribution.
3. The physical and physiologic noise levels are non-uniformly distributed over the brain areas.
Because the statistical significance is essentially determined by the signal-to-noise ratio, the
GLM favors areas in which the noise level is low and the signal sensitivity is high. Smoothing the
data spatially reduces the non-uniform distribution of the noise. However, this leads to an
effective lowering of the spatial resolution. As a recommendation, if a hypothesis about the
activity of one area is specified or discussed, a close look at its time series and averaging data
over time and space should be done; this is also the case when the statistical analysis fails to
result in significant voxels in this area. This is especially important when different groups are
compared, because their spatial noise distribution and noise level might differ.
4. The statistical analysis is performed voxel-wise. Because of the huge number of voxels (64 × 64
× 24 approximates to 100,000 voxels in a typical experiment) some voxels can be active by
chance. For example, if p = 0.01 was chosen as the threshold, approximately 1000 voxels
would be classified as activated by chance, referred to as the "multiple comparison" problem.
The most conservative approach is to take the multiple comparisons into account with the
Bonferroni correction, choosing a much more conservative p-value to determine the threshold.
By dividing the desired p-value (e.g., p = 0.01) by the number of voxels, the probability of
detecting a false activation anywhere in the brain is reduced to the desired p-value. Because the
resulting threshold is usually too strict, several other correction schemes have been proposed.
However, the multiple comparison problem cannot be regarded as solved. One way of reducing
the problem is to look at clusters of activated voxels, because it is less probable that several
voxels are activated by chance than that only one voxel is activated by chance. The bigger the
cluster size, the lower is the probability of recording a false activation. A drawback of this
approach is that it biases the analysis to regions where the expected activity is more spatially
distributed, such as sensory areas, than to regions of more focal activation, such as small
nuclei.
Block Designs versus Event-Related Designs
6 of 9 12-04-2010 00:23
Add to lightbox
page 267
page 268
Figure 9-14 Block designs and event-related designs of fMRI experiments. Top, Two possible
hemodynamic response functions (HRFs), constructed to have the same area. The lower figures show
the resulting time courses when the HRFs are convolved with a block design (middle) and a
randomized event-related design (bottom). The block design is more efficient for detecting a response
from the variance of the signal, while the event-related design is more efficient for estimating the
shape of the HRF.
In nearly all fMRI experiments the hemodynamic response is treated as a linear smoothing filter. The
simplest experimental design is to block the stimuli in 20 to 60 s blocks alternated with equal blocks of
the rest condition, and repeat this pattern for several cycles. As described above, the typical data
analysis strategy is to convolve the stimulus pattern with an assumed impulse response function, and
then correlate this estimated response with the time course of the signal from each voxel.76 The BOLD
impulse response function is often taken as a gamma-variate function with a width of 4 to 6 s.
However, for a block design the assumption of the shape of the impulse response is not critical: if the
block length is significantly longer than the width of the impulse response, then the amplitude of the
BOLD response on the plateau of a block will depend just on the area under the impulse response, and
not on details of the shape (Fig. 9-14). Most of the fMRI studies done to date have used block
designs, and so the measured quantity that is used for mapping activation has been primarily the area
of the impulse response. For most of these studies the goal was to detect whether a local piece of
tissue was activated, not to characterize the shape of the BOLD response.
For many applications a block design is not feasible, or at the very least introduces an artificial quality
into the stimulus that makes the results difficult to interpret. A trial-based (or event-related) design
significantly broadens the types of neural processes that can be investigated. 79-88 A block design by
definition presents similar stimuli together, which makes it difficult to study processes where
predictability of the stimulus is an important consideration. For example, studies of recognition using
familiar stimuli and novel stimuli are hampered if all of the familiar stimuli are presented together. A
trial-based design allows randomization of different stimuli and a more sophisticated experimental
7 of 9 12-04-2010 00:23
design.
To analyze event-related data one could use the same correlation approach used for block designs, by
convolving the stimulus pattern with an assumed hemodynamic response and correlating the resulting
model curve with each voxel time course. However, for an event-related design the resulting model
curve is much more dependent on the exact shape of the assumed BOLD response, as illustrated in
Figure 9-14. Two impulse response functions were constructed to have the same area but different
shapes, and then convolved with a block of stimuli and a random presentation of stimuli. For the block
design the two predicted responses are similar, but for the event-related design there is a greater
difference. The reason for this is that the expected BOLD response depends strongly on the shape of
the impulse response only during transitions (e.g., the beginning and end of a block), while the steady
state (e.g., the plateau of a block) depends only on the area of the impulse response. With long blocks
there are long steady-state periods but only a few transition periods, but with an event-related design
there are many transition periods and typically no steady-state periods.
The introduction of event-related designs into fMRI experiments has significantly expanded the field,
but it has also made our lack of understanding of the hemodynamic response a more acute and
significant problem. A fruitful approach for dealing with the uncertainty about the exact shape of the
82
hemodynamic response is to estimate the response from the data. Specifically, instead of assuming
a given shape for the impulse response, with only a single unknown amplitude, the impulse response is
treated as set of unknown amplitudes at different time lags after the initiation of a stimulus. That is, the
amplitudes of the response 1 s after the stimulus, 2 s after the stimulus, etc., are treated as separate
model functions within the GLM. For example, if the impulse response is assumed to last 15 s and it
will be estimated at 1 s intervals, this is equivalent to including 15 model functions corresponding to
each of the time lags. A hemodynamic response is then estimated for each voxel, and the full response
(or a specified range) is then examined to determine if it differs significantly from zero.
Detection versus Estimation

An interesting feature that emerges from consideration of block and event-related designs is that there
are actually two distinct criteria that can be used to evaluate different experimental designs: the
89
detection sensitivity and the estimation sensitivity. If we are only interested in detecting whether a
response is present, and do not care what the shape of the response is, then the primary measure of
efficiency is the signal variance produced by the stimulus. For example, Figure 9-14 illustrates two
distinct patterns for presenting the same 48 stimuli. For each one we can convolve the pattern with a
simple assumed response and then calculate the variance of the resulting model curve. The statistical
analysis will then compare this variance to the variance produced by other noise sources to determine
if the voxel is activated, so the larger this intrinsic variance the greater the probability of detecting the
response. As illustrated in Figure 9-14, blocking the stimuli creates a variance approximately four times
larger than the variance for the randomized pattern, and so creates a more detectable response.
However, if the goal is to estimate the shape of the response itself, a different criterion is needed to
evaluate different stimulus patterns. Dale and colleagues introduced the figure of merit for this
purpose; it is calculated just from the pattern of stimuli, and is independent of the shape of the
hemodynamic response.82 In the example in Figure 9-14, this figure of merit is approximately four
times larger for the random design than for the block design, so the random design is much more
efficient for estimating the impulse response.
The differences in detection efficiency and estimation efficiency can be traced to the width of the
impulse response, which creates overlap of responses from different events. Such overlap can be
good or bad, depending on the type of sensitivity of interest. The variability of the estimation efficiency
comes primarily from differences in the degree of systematic overlap of responses from different
stimuli, such as the poor efficiency of a periodic closely spaced single trial design with perfectly regular
overlap of responses. The variability of the detection power can be traced to the constructive effects
of overlapping responses to different events, as in a block design, where the addition of responses to
closely spaced events serves to increase the final variance of the signal. Both of these sources of
8 of 9 12-04-2010 00:23
variability in the sensitivity of stimulus designs would be removed if we restricted the minimum time
between events to be greater than the width of the impulse response. In that case there would be no
overlap of responses from different events, and all designs with the same number of events would
have the same detection efficiency and estimation efficiency. Because this is not the case, the design
of fMRI experiments is a richer and more complex field.
page 268
page 269
In general, event-related designs are a powerful and robust approach for fMRI experimental design.
Many experiments do not fit into a block design, and the variability of the hemodynamic response
across brain regions, subjects, and experimental conditions is an important factor to take into account.
Event-related designs allow one to do this in a systematic way, and the efficiency of potential designs
can be evaluated numerically to optimize the experiment in advance. 81,82,90-92 Finally, block and random
event-related designs represent two ends of a spectrum, with one of the key differences being
predictability of the stimulus. An intermediate design, with some pseudo-random portions and some
pseudo-block portions, can make the stimulus less predictable while retaining some of the detection
86,89
efficiency of a block design.
© 2010 Elsevier
9 of 9 12-04-2010 00:23
ARTIFACTS AND NOISE

fMRI is More Sensitive to Imaging Artifacts than Clinical MRI
Functional MRI of brain activation is one of the more demanding applications for an MRI scanner. It
relies on detecting MR signal changes of only a few percent (or less) over a relatively prolonged
measurement period, typically 20 to 40 minutes. There are several sources of noise in fMRI; the
primary sources are noise introduced by the MRI scanner or associated hardware, and physiologic
noise introduced by the subject (e.g., changes in the MR signal due to thermodynamic or electrical
noise, or changes in MR signal due to cardiac or respiratory motion). In addition to noise, distortions in
the image (which degrade image quality) as well as susceptibility dropouts will also reduce the
available SNR. To detect the small perturbations in the MR signal which are associated with fMRI it is
critically important that the noise and scan-to-scan variability are well below the signal changes being
investigated. This section addresses some of these sources of artifacts and noise and how their
influence can be minimized.
Minimizing and Correcting Image Distortions
Add to lightbox
Figure 9-15 Mean registered B0 maps calculated following synthetic linear shimming from five
subjects with head position: chin down (upper panel), head neutral (middle panel), and chin up (lower
panel). The improvement in homogeneity as the head tilts up is most evident in the inferior regions of
the frontal and temporal lobes, as well as in occipital cortex. (Modified from Tyszka JM, Mamelak AN:
Quantification of B0 homogeneity variation with head pitch by registered three-dimensional field
mapping. J Magn Reson 159:213-218, 2002.)
MRI relies on a homogeneous magnetic field to correctly encode the gradient-induced distribution in
precessional frequencies of nuclear spins into an image. Inhomogeneities in the static magnetic field
will result in distortions and signal loss in the final MR image. fMRI utilizes rapid gradient-echo pulse
sequences that are particularly sensitive to the microscopic susceptibility perturbations induced by
deoxyhemoglobin. This is particularly evident in single-shot EPI or spiral gradient-echo sequences,
where the read-out window is of the same order as the echo time; these sequences will also be
sensitive to bulk susceptibility changes. There is a greater need in fMRI than in most conventional
clinical MRI applications to minimize these macroscopic inhomogeneities in the magnetic field.
Functional MRI studies frequently rely on the use of additional materials for monitoring subjects (eye
tracking, EEG electrodes, headphones) or to stabilize the head (bite bar, plastic pads). Even though
these materials are non-metallic they may have susceptibility different from the adjacent tissues and
cause local field distortions.93 Materials used inside the MR scanner during fMRI studies must thus be
chosen carefully.
1 of 9 12-04-2010 00:24
Even without extraneous materials in the scanner, the patient or subject alone will cause
inhomogeneities in the magnetic field. The close juxtaposition of air, fat, bone, and tissue in the human
head is not conducive to maintaining a homogeneous magnetic field. The susceptibility distortions
around the sinuses and petrous parts of the temporal bones are often the most evident and can be
difficult to correct, resulting in signal loss in the prefrontal cortex and temporal lobes. The use of
high-order shims to minimize bulk susceptibility effects is particularly important, particularly with the use
of high-field MRI (3 T and above). In addition to shimming, the position of the head within the static
magnetic field is also important. Tyszka and Mamelak demonstrated that the normal anatomic position
used for radiographic positioning for almost a century might actually worsen the inhomogeneities
around the prefrontal and temporal cortices. By tipping the head upwards so that the position of the
sinus is moved more cranially, the orientation of the induced gradients relative to the B0 field lines
94
tends to reduce the effects on local field inhomogeneity. Figure 9-15 shows the dependence of B0
94
homogeneity on head position.
page 269
page 270
Add to lightbox
Add to lightbox
Figure 9-16 Improved B0 homogeneity (upper panel) and image quality on GRE echo-planar images
(lower panel) from strategic placement of intraoral local shims (made from continuously nucleated
pyrolytic graphite). The B0 range is -0.8 ppm (white) to +0.8 ppm (black). Note the improved B0
homogeneity in the inferior frontal cortex, and corresponding reduced distortion and improved gray
matter/white matter interface on the EPI images. (Modified from Wilson JL, Jenkinson M, Jezzard P:
Protocol to determine the optimal intraoral passive shim for minimisation of susceptibility artifact in
human inferior frontal cortex. Neuroimage 19:1802-1811, 2003. © 2003, with permission from
Elsevier.)
In addition to externally applied shims, other approaches have been suggested to minimize the local
95,96
magnetic field inhomogeneity around the paranasal sinuses. Wilson et al proposed the use of local
shims in the mouth to reduce the local field inhomogeneities in inferior frontal lobes from the paranasal
sinuses and skull base. Pyrolytic graphite is strongly diamagnetic and can be used to change the local
magnetic field to minimize local susceptibility-induced gradients. Figure 9-16 shows the improved B0
2 of 9 12-04-2010 00:24
homogeneity and image quality from strategic placement of local shims against the hard palate.
In addition to signal loss, susceptibility differences between tissues also cause distortion in the image
as the spatial mapping of the MR signal assumes the magnetic field is homogeneous. The
mislocalization of a particular pixel is predictable if the local magnetic environment is known. An
additional measurement made of the actual local field (the "field map") acquired at the same time the
EPI or spiral acquisition is made can then be used to reposition pixels and reduce distortions. 97,98
fMRI is Sensitive to Subject Motion

Functional MRI studies are extremely sensitive to subject motion. This is most evident at tissue or
contrast boundaries where even sub-pixel movement can generate large perturbations in the MR
signal, frequently several times greater than the underlying BOLD signal itself. The best approach to
removing motion artifacts in the MR data is to try to prevent them at the outset. The comfort of the
subject in the MR scanner is of paramount importance. Subjects should be dissuaded from drinking
excessively prior to the examination, and allowed to empty their bladder immediately before the
examination. Although minor discomfort (insufficient padding around the head, knees not well
supported, hard examination table, neck over or under extended) may seem trivial at the beginning of
the examination, after 30 minutes it may no longer be tolerable, resulting in voluntary and involuntary
movement. Many groups also employ a bite-bar or a vacuum pad to further stabilize the head.
Although very effective, this requires careful set-up and subject cooperation or it too can become
uncomfortable part way through the examination.
Subject motion results in displacement or rotation of the image in successive acquisitions. Small
degrees of motion can be corrected with image registration. The most common of these are based on
99,100
the automated image registration (AIR) algorithm. Slow patient drift within the magnet may
appear as a linear drift in the MR time series and can also be compensated for in the general linear
model during post-processing. Subject motion that is most difficult to remove is motion that is
temporally correlated with the stimulus. This most often occurs with poorly trained subjects or with
strong stimuli that produce a startle response coincident with the onset of stimulation (e.g., loud
noises, bright lights, or painful somatosensory stimuli). A careful examination of the raw data is needed
to ensure that apparent BOLD activation is not biased by patient motion. Excessively large signal
changes out of proportion to the expected fMRI response, and apparent activation at contrast and
tissue boundaries or in areas not expected to be associated with fMRI activation (e.g., the scalp) are
signs of patient motion rather than true activation. Changes on the statistical map that are out of phase
on either side of a tissue boundary (i.e., positively correlated signal change on one side of the head,
and negatively correlated signal change on the other side of the head) are also signs of gross subject
movement.
The fMRI Signal Includes Contribution from Physiologic Fluctuations

page 270
page 271
A significant source of artifact in fMRI results from the periodic modulation of the MRI signal by
physiologic processes, particularly from respiration and cardiac pulsations.101,102 These periodic
103
modulations increase noise and reduce the statistical significance of the activation signals, and are
collectively referred to as physiologic noise. Artifacts due to physiologic effects are well recognized in
clinical MRI and a number of strategies have been suggested to try to minimize them, in particular
breath-holding, novel k-space trajectories,104 reordering k-space to time lock it with respiration or
cardiac motion,105 navigator echoes,106 cardiac and respiratory gating, and flow compensation.
However, not all the techniques used in clinical imaging are applicable for functional MRI. Use of
navigator acquisitions in fMRI requires some additional modification since there is only a short time
available between excitation and data acquisition.107 Additionally, gated acquisitions are difficult to
implement in fMRI as this results in a variable TR. This introduces additional signal fluctuations unless
the TR is sufficiently long.108 Also signal variations related to motion, such as respiration, are
incompletely addressed with flow compensation.107 The problem of physiologic noise requires some
3 of 9 12-04-2010 00:24
additional techniques more specifically suited to fMRI. We present here an overview of the effects of
this physiologic noise on the MR signal and T2*-weighted images, and consequently on the fMRI
statistical power. Strategies to reduce its impact are also addressed.
Physiologic motion that occurs during the mapping of MR signals into k-space will result in phase
offsets and ghosts in the T2*-weighted images. This is most evident for multi-shot techniques (fast
low-angle shot [FLASH] imaging, multi-shot EPI). The ghosts can be identified during a careful
inspection of the raw images prior to making statistical maps. Physiologic variation that occurs
between successive repetitions of snapshot EPI or spiral acquisitions will tend to lead to periodic
variations in signal intensity synchronous with the cardiac or respiratory cycle. The most effective way
to observe these effects is by analysis of the frequency distribution by Fourier transform of the fMRI
time-series data (the power spectrum). Figure 9-17 shows a power spectrum acquired during a visual
task with a stimulus repetition rate of 0.016 Hz. The fundamental frequency of the stimulus is clearly
seen in the spectrum. Additionally, periodic variations in MRI signal are also seen at 0.85 to 0.95 Hz
and at 0.15 to 0.25 Hz, corresponding to cardiac and respiratory components.
A number of processes lead to this periodic variation in the MR signal. For respiratory pulsations, the
dominant effect is from susceptibility changes originating in the chest due to the cyclical change in the
proportion of water and air. This leads to a magnetic gradient in the z-direction, and results in small
amounts of intra-slice dephasing and minor misregistration of slice position (for axial images) or
distortions (for sagittal and coronal images).77 In addition to the periodic modulation of the field
homogeneity, there is also evidence that the brain physically moves between 110 and 950 μm during
the respiratory and cardiac cycle.109 Changes in intrathoracic pressure are transmitted to the head as
changes in intravascular and CSF pressure, with secondary cyclical effects on cerebral blood flow.
Since the volume of the calvaria is fixed, this results in displacement of brain parenchyma. In addition,
depending on the positioning of the subject, movement in the anterior chest wall may be directly
transmitted as a pitching head motion.
Add to lightbox
3
Figure 9-17 Frequency distribution of the MR signal from a 4 mm region in primary visual cortex
during a visual functional MRI experiment (alternating 20 s flickering checkerboard, 40 s darkness).
Note spectral power in the 0.15 to 0.25 Hz band corresponding to respiration, in the 0.85 to 0.95 Hz
band corresponding to cardiac pulsations, and at 0.016 Hz corresponding to the fundamental
frequency of the stimulus. Slow oscillations are seen around 0.1 Hz. (The relative power at the different
frequency bands varies with choice of region of interest and subject.)
Cardiac-induced susceptibility changes are significantly smaller, and the dominant transmission of
cardiac artifacts to the brain is a direct result of fluctuation in cerebral blood volume and oxygenation
(and consequently brain and CSF volume) within the calvaria. An additional source of physiologic noise
4 of 9 12-04-2010 00:24
that has received increasing attention in recent years is slow oscillations, synonymously referred to as
vasomotion, V-signal, spontaneous oscillation, and low-frequency waves (see reference 110 for
overview). These oscillations are centered on 0.1 Hz, and their origin is still under debate (metabolic,
vascular, or neuronal). Even lower frequency drifts have been described, and these may also
represent physiologic noise or patient drift. However as these effects are also seen in cadaveric
studies,111 they can be very difficult to distinguish from signal drifts originating within the MRI
hardware.
Correcting for Physiologic Noise

page 271
page 272
Strategies to reduce the effects of physiologic noise can be broadly divided into methods to remove
the effects beforehand, and methods to remove the effects using post-processing. Training subjects to
maintain steady breathing and to avoid excessive, large excursions in chest motion will help reduce
some of the larger signal changes from respiratory motion. Careful positioning and head restraint
(ideally using a bite-bar) will also help to minimize pitching motions of the head during thoracic wall
motion. Since the predominant effect of physiologic noise is to cause a phase shift in successive image
107
acquisitions, Hu and Kim proposed applying navigator echoes to track these induced phase
variations. Acquiring an additional line at the center of k-space (for gradient-echo [FLASH] acquisitions)
or an additional data point at the center of k-space (for EPI acquisitions) before the phase-encode
gradient allows the phase offset due to motion or B0 fluctuation to be quantified and thus accounted
for. This can reduce the spectral power due to physiologic noise by 30% to 50% (depending on exact
location). Imaging sequences that collect additional measurements of the center of k-space can also
be used to remove the phase offsets from periodic noise. For example, a rosette trajectory that
over-samples the center of k-space allows a 25% to 62% reduction in the standard deviation of the
noise.104 Removing respiratory and cardiac motion with gated acquisitions is not widely used in fMRI
as this induces a variation in TR, introducing signal modulation due to variable longitudinal relaxation
and thus increasing the variance in the fMRI signal. For imaging the brainstem, where there is
significant brain motion with the cardiorespiratory cycle, Guimaraes et al proposed a combination of
prospective cardiac gating and retrospective correction for the variable longitudinal relaxation based on
recordings of the actual imaging TR for each acquisition.112
Several strategies have also been proposed to remove physiologic noise retrospectively. An
examination of the power spectrum (see Fig. 9-17) shows that the physiologic noise often falls into
well-defined frequency bands. It would appear that a simple approach to remove physiologic noise is
selective digital filtering of a specific frequency band. The sampling rate for the physiologic fluctuations
is determined by the imaging TR. For a typical MR experiment this is typically 1 to 2 s, corresponding
to a temporal resolution of between 0.5 and 0.25 Hz. As the cardiac rate is around 1 to 2 Hz, this will
be under-sampled at this TR and the signal will be aliased.113 The aliased signal will no longer appear
in the same well-defined frequency band and may also be superimposed at the fundamental frequency
of the stimulus/response, making it more difficult to selectively filter. Several approaches have been
proposed to address these aliased noise signals. By repeating the task at a different rate, the effective
sampling rate is changed and the degree of aliasing may be reduced or the position in the frequency
spectrum where physiologic noise signals become superimposed can be shifted. Frank et al proposed
a technique that utilizes the fact that physiologic noise is spatially correlated across large areas of the
brain and treats adjacent slices as phase-shifted additional samples, increasing the effective sampling
rate of physiologic noise and allowing previously aliased signals to be retrieved and removed.114 Biswal
and colleagues measured the cardiac and respiratory contributions separately by resampling
contemporaneous pulse oximeter waveforms recordings at the same sampling rate as the imaging TR,
time-locked to the fMRI signal acquisition.113 The frequency location of aliased physiologic signals
could then be determined, allowing them to be selectively filtered.
The relative under-sampling of noise data is a significant limitation to identifying and characterizing
physiologic noise. An alternative approach is to record physiologic and cardiac signals as a separate
5 of 9 12-04-2010 00:24
time series using a photoplethysmograph and pneumatic belt to record cardiac pulsations and
respiratory excursions respectively. The cardiac and respiratory changes can then be sampled at a
considerably higher rate, and their contribution to the MRI signal estimated and removed. Mitra and
Pesaran propose that since cardiac and respiratory effects are approximately periodic in nature, they
can be modeled in terms of slow amplitude- and frequency-modulated sinusoids. 115 The contributions
to the MR signal can be removed from the data in the frequency domain. Other techniques take a
similar approach and characterize the cardiorespiratory contribution from a separate time series, but
remove the noise using image-based techniques and are thus less reliant on the periodicity of the
noise. A second-order Fourier series is expanded in terms of cardiac and respiratory phases
(calculated from the separate cardiac and respiratory activity measured during the fMRI study). The
weights of the Fourier terms can be estimated with a general linear model and are then applied to the
116 108
image or the k-space data. Another approach involves decomposition of the imaging data into its
115,117,118
principal or independent components This differs from the image and frequency domain
techniques described so far in being data driven (rather than being constrained to a particular model,
waveform, or time course). However, removing selected components with such a decomposition of the
data must be done cautiously as the noise component is not necessarily uniquely contained in any
single component.115
No single technique will remove all the physiologic noise from every data acquisition. Thus, whichever
approach is employed, it remains imperative to carefully examine the raw data visually (looking at the
raw images, or a movie of the images and the power spectra of the data series) both as a crude
check of the quality of data collection and to direct further analysis.
Scanner Stability and Thermal Noise

The fidelity and stability demanded for good fMRI are not typically included in the quality assurance
(QA) specifications of a routine clinical MR scanner. However, with the more widespread use of fMRI
in clinical and neuroscience applications, it is now more important than ever to address the
performance capabilities of the MR instrument and hardware chain. Ensuring homogeneous scanner
performance also provides quality assurance for longitudinal investigations, studies across multiple
subjects, or comparison of measurements from different institutions. A number of techniques have
appeared in the literature in recent years to address MR scanner stability and baseline performance,
but their implementation has traditionally been guided by the needs of individual scanner operators. A
more widespread initiative is required to standardize these measures and facilitate comparison of
scanner performance from different manufacturers, from different sites, or obtained at different time
points.
page 272
page 273
The Biomedical Informatics Research Network (BIRN) ( http://www.nbirn.org) is an initiative sponsored

by the National Institutes of Health (NIH) and National Center for Research Resources (NCRR) that
facilitates large-scale biomedical science collaborations. Standardization of MR scanner performance
has been one of the foci of attention of BIRN. As well as addressing stability of a particular scanner,
BIRN also proposes to standardize levels of performance for fMRI across centers, allowing studies
from different sites to be compared. To address the noise and variation introduced into an fMRI study
by the MR instrument itself, four basic measures have been identified by BIRN to measure scanner
performance and ongoing stability: 1. the root mean square (RMS) signal stability over a region of
interest (typically 20 × 20 pixels) expressed as percent fluctuation; 2. the signal-to-fluctuation-noise
ratio (SFNR) calculated for each pixel and averaged over the same region of interest; 3. the correlation
diameter (rdc); and 4. the drift in the signal over time, calculated per pixel and expressed as a percent
change. The basis of these tests, and a suitable protocol for making the measurements, are presented
here.
Quality assurance measurements are done on a 17 cm spherical phantom filled with agar gel using a
standard head coil. It is recommended that the tests be performed at least weekly (ideally, daily) to
quickly identify adverse trends that may affect scanner performance. The imaging parameters for the
6 of 9 12-04-2010 00:24
test are similar to those used in a typical fMRI study. Thirty-five contiguous 4 mm slices are acquired
axially through the phantom using a single-shot EPI or spiral gradient-echo imaging sequence. The
acquisition is repeated for 200 time frames with field of view (FOV) of 22 cm, matrix of 64 × 64, TR of
3000 ms, TE of 30 ms (at 3 T) or 40 ms (at 1.5 T), and flip angle of 90°. The bandwidth (BW) is 100
kHz or higher. Acquisition time is typically 10 minutes. Through-time measurements of MR signal
variation in the central slice are used to calculate RMS stability, drift, mean signal, and SNR. The first
two time points are ignored to allow longitudinal magnetization to reach steady state.
The RMS stability is measured as the fluctuation over time of the average signal from a 20 × 20 pixel
region of interest (ROI). Trends are removed by fitting a second-order polynomial. The residual
fluctuation is expressed as a percentage. Typically, this should be less than 0.15%. This provides a
measure of the fluctuation over time of the average signal across 400 pixels. This is useful to detect
global fluctuations in the MR signal, for example, due to instabilities in the RF amplifiers, receiver gain
instabilities, shim instabilities, or timing errors in the effective echo time.
The SFNR is calculated from maps of signal mean and signal standard deviation across the 198 time
points. For each pixel within a 20 × 20 ROI the ratio of mean to standard deviation is calculated and
the average value of this ratio is reported as SFNR. The SFNR should be 200 or greater. The SFNR is
an important metric for fMRI. It describes the variance in a single pixel independent of any functional
activation and thus directly influences the statistical power of any activation-induced changes. The
RMS and SFNR measures are both used to examine spatially varying effects induced by the MR
scanner hardware. SFNR is different from the RMS measurement in that this is the average of the
fluctuations in 400 pixels (rather than the fluctuation of the average). Effects that act globally such as
hardware instabilities will tend to influence both measures. Effects that act more locally such as
random noise in the image will be more evident in the SFNR than the RMS measurement.
The rdc (correlation diameter) is a measure of the spatial auto-correlation of the noise in the MR signal
over the 20-pixel diameter of the sampled ROI. If the variation in the MRI signal was purely
thermodynamic noise (i.e., uncorrelated noise), then this could be reduced by simple averaging (the
basis for improving imaging SNR in clinical scanning by increasing the NEX value). In this case,
averaging adjacent pixels together would decrease the relative noise standard deviation in proportion
to the square root of the number of pixels averaged (i.e., a four-fold increase in pixels would decrease
the relative noise standard deviation two-fold). Other sources of correlated noise, such as fluctuations
in amplifier performance or spikes from the gradients, will tend to affect large numbers of pixels
119
similarly. Thus averaging across several pixels will not reduce this spatially correlated noise. For an
ideal scanner, a plot of noise standard deviation against number of pixels would follow an inverse
square root relationship. Any deviation from this relationship indicates the presence of correlated noise.
The point at which the plot of noise versus ROI width deviates from this ideal negative square root
relationship is a quantitative indication of the degree of correlated noise present in the MR system.
This is expressed as a pixel diameter-the correlation diameter or rdc. For an ideal scanner, this value
will be 20 (or higher if larger ROI measurements were taken) indicating that even up to a diameter of
20 (or more) pixels, there is still an SNR benefit from pixel averaging (i.e., presence of uncorrelated
noise). A poorly performing scanner will have a low value of 1 or 2 pixels, indicating that the noise is
strongly correlated. A well-performing scanner should have a rdc value of 5 or higher; a value lower
than this is an indication of larger correlated noise contributions from the system hardware.
Deterioration in the rdc value indicates potential instabilities in subsystem hardware components such
as gradients, RF amplifiers, or resistive shims.
Drifts in the MR signal over time are also important. These may be caused by problems in the
amplifiers or shims, or may indicate drifts of the B0 field itself. Problems in the gradients, which cause
excessive heating, can also be seen as drifts in the MR signal and the heating may also change the
temperature of the resistive shims, exacerbating any drifts in signal. Drift is calculated on a pixel-
by-pixel basis and expressed as a percent change over a defined time interval.
Figure 9-18 shows plots of signal fluctuation and frequency spectra for a clinical 3 T scanner while
operating at a baseline level for that scanner, and subsequently when image quality was degraded by
7 of 9 12-04-2010 00:24
large noise contributions.
EPI acquisitions are prone to ghosting artifacts, an apparent displacement of some of the image
intensity by half of the field of view (FOV). Ghosting in EPI images is another source of noise in fMRI,
and needs to be tightly controlled (typically less than 2% to 3% signal power in the ghosts). An
analysis of the frequency spectrum of any scanner noise or ghosting is valuable in determining its origin
and significance. Thermodynamic noise is uncorrelated and has equal amplitude at all frequencies.
Hardware noise often occurs at a particular frequency and will thus appear as individual spikes in the
frequency spectrum (see Fig. 9-18).
page 273
page 274
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 9-18 Two examples of measurements of scanner stability taken on the same 3 T MR scanner
on different days (day 1: A and B; day 2: C and D). On day 1, signal fluctuation (A) is less than 0.15%
(0.08%) and there is minor drift in the baseline signal over the 200 acquisitions (600 s). SFNR is
greater than 200, and the spectrum (B) shows the noise power is low and white. The RDC measure
was 2.2 pixels (data not shown). These numbers can then be used as the baseline indicators for
scanner performance on subsequent days. On day 2, signal drift (C) over 600 seconds has now
8 of 9 12-04-2010 00:24
increased (-6.58), and the noise fluctuation is now greater than 0.15% (0.25%). The reason for this is
apparent on the noise power spectrum (D), which is no longer white and shows large spectral power at
0.02 Hz (excessive low-frequency oscillatory noise). The RDC value was lower (1.2 pixels), indicating
the noise is more spatially correlated across pixels than previously.
Additional measurements that should be recorded in the regular quality assurance are the transmitter
gain and receiver gain used for the EPI measurements on the standard phantom and the center
frequency of the scanner. Minor statistical fluctuations in all of these parameters are expected on a
day-to-day basis, however the ability to monitor trends can identify emerging problems with scanner
fidelity even before image quality is adversely affected.
© 2010 Elsevier
9 of 9 12-04-2010 00:24
MEASURING CEREBRAL BLOOD FLOW, CEREBRAL METABOLIC RATE OF

OXYGEN, AND CEREBRAL BLOOD VOLUME
Cerebral Blood Flow
Cerebral blood flow (CBF) studies using 15O-labeled water as the contrast agent have been used
120
extensively for functional mapping with positron emission tomography (PET). The magnetic
resonance counterpart to this PET technique is to magnetically label ("spin label") arterial water with an
applied RF pulse.121 In a typical implementation, the magnetization of blood is inverted on the proximal
side of the slice, and after a delay of 1 to 1.5 s to allow labeled spin to flow into the slice, the image is
acquired. The experiment is then repeated without the inversion, and a subtraction of these tagged and
non-tagged images then yields a net signal proportional to the arterial spin delivered to the voxel. For
this reason the difference image directly reflects CBF.
page 274
page 275
Arterial spin labeling (ASL) can be used to measure the baseline CBF, as well as activation-dependent
changes in CBF. Because most of the water molecules delivered to a capillary bed in the brain are
extracted, and because the inflow time in these experiments is very short, few if any of the tagged
spins reach the venous side of the vasculature. For this reason, the ASL signal reflects delivery rather
than clearance, so the ASL signal is localized in parenchyma rather than draining veins. Images of CBF
using the ASL technique result in improved spatial correlation with brain parenchyma compared with
122 123
conventional BOLD fMRI in both human and animal studies. ASL also provides a quantitative
means to study the mechanisms underlying the BOLD technique itself. 50,124 A number of ASL
techniques have been developed; they can be broadly divided into continuous and pulsed tagging
techniques. In pulsed techniques the inversion is done with a slice selective 180° pulse, and with
continuous labeling the inversion is done adiabatically as the blood moves through a gradient field
during a continuous RF pulse. Both classes of ASL have been used for measuring CBF changes in
20
humans following neuronal activation. The durations of CBF and BOLD signals are similar (see Fig.
9-11) and the same stimulus design can be adapted for both BOLD and flow activation
measurements.125
Calibrating the BOLD Signal to Measure CMRO2 Changes

As noted above, ASL techniques have been useful for unraveling the possible sources of transients in
the BOLD experiment. But if we now return to the consideration of steady-state changes, Equation
9-10 can be used as the basis for calibrating the BOLD effect and estimating the CMRO2 change with
activation. Consider a block design experiment in which the response reaches a plateau steady-state
value and both the BOLD signal change ∆S/S and the CBF change (f) are measured for each voxel. If
we assume that v is related to f by the empirical relationship of Equation 9-11 there are then only two
unknown quantities in Equation 9-10: A, the local scaling factor, and m, the normalized change in
43
CMRO2. The innovation of Davis et al was to show that the local scaling factor A can be measured
with a separate experiment, and the combination of measured values of A, ∆S/S, and f then provides
an estimate of the change in CMRO2 (m) through Equation 9-10. The trick for measuring A is to make
use of the fact that inhalation of CO2 causes a pronounced increase in CBF, but apparently no change
in CMRO2.126 The experiment is then to use the same ASL/BOLD imaging with CO 2 as a challenge.
Equation 9-10 is then applied with the assumption that m = 1 (no CMRO2 change), so the measured
BOLD signal and CBF changes with CO2 allow calculation of the local value of A. This calibrated BOLD
approach has been used very effectively to show that the CBF increase with activation is 2 to 3 times
larger than the CMRO2 increase.43,44
The limitation of this technique is that, currently, only a limited number of slices can be imaged at a
time, and the slices chosen need to be as orthogonal as possible to the feeding artery for optimal
122
spin-tagging. Although the magnitude of the physiologic change is higher, the MR signal, and hence
1 of 5 12-04-2010 00:25
the available SNR, is lower for quantitative CBF measures.122 The CBF and BOLD signals from this
combined technique are not without some "crosstalk." The T1 of blood is known to be longer than that
of tissue, so there is some residual negative flow weighting in the average due to exchange of blood
water with tissue water. Typically, the flow signal in the average of tag and control images is
approximately 18% of the flow signal in the difference signal between tag and control states. This flow
contribution has the opposite sign to that of the BOLD signal, and thus causes a small underestimation
of the BOLD signal. Conversely, the BOLD effects can also contaminate perfusion measurements. This
effect is smaller than the effect of flow on the BOLD signal. To a first approximation, the BOLD effect
causes a simple scaling of the MR signal on the order of 2% to 4% during activation, which is just at
122
the limit of detectability.
Assessment of Cerebral Blood Volume Using an Exogenous Contrast

Agent
The earliest implementations of functional MRI used contrast based on cerebral blood volume (CBV)
rather than blood oxygenation. These utilized the intensity of signal change during the first-pass
passage of an intravenous bolus of paramagnetic contrast agent through the brain and tracer kinetics
analysis to estimate changes in CBV.127 The passage takes approximately a minute to complete, and
additional time is required for clearance of the contrast agent before the measurement can be
repeated. Temporal resolution is thus limited, and this approach provides only a "snapshot" measure of
cerebral blood volume. These first-pass measurements of dynamic susceptibility contrast using
gadolinium can be useful for assessing resting blood volume or abnormal CBV/CBF associated with
128
tumor neovascularization or cerebral ischemia, but they lack the temporal resolution necessary to
examine CBV changes during neuronal activation. Greater temporal resolution can be achieved with
intravascular contrast agents that maintain a prolonged steady-state concentration. Depending on the
type of contrast agent used, the sensitivity to CBV can alter the T1 or T2 relaxation time; however, the
use of T2 agents appears to give better sensitivity to CBV changes than does the use of T1 agents. 62
Dextran-coated monocrystalline iron oxide nanoparticles (MION) affect T2 relaxation and have a
129,130
prolonged blood half-life (ranging from 3 to 15 hours in monkeys ), making them very suitable as
susceptibility agents for measuring CBV. At a sufficiently large dose the contribution to blood
magnetization from the contrast agent far exceeds that from deoxyhemoglobin, so T2*-weighted MR
sequences are rendered insensitive to changes in deoxyhemoglobin concentration. Flow changes also
have no effect (the concentration of paramagnetic agent is essentially the same in the arterial and
venous circulation, so increased flow does not affect susceptibility). The imaging contrast is thus only
sensitive to the concentration of paramagnetic agent in the imaging voxel. As local CBV increases, the
vascular proportion of the voxel (and hence the observed amount of contrast agent) increases. Since
T2 agents shorten T2 and T2*, the effect of increased blood volume (and increased iron concentration)
is to reduce the signal in the voxel, thus neuronal activation is associated with a decrease in MR signal
for these CBV-weighted images.129
page 275
page 276
The signal advantage of the CBV measurement over BOLD for fMRI is most evident at low B0 field
strengths. Several factors contribute to this. First, the effect of the BOLD signal (which increases with
functional activation) is opposite to the CBV signal (which reduces with activation). A BOLD signal is
still present, but the larger negative signal change from CBV measurement dominates any positive
signal change due to BOLD. BOLD contamination of the CBV signal is more pronounced at higher B0
62
field strength for a given dose of contrast agent. Second, as the effect of the contrast agent is to
decrease the MR signal, the injected dose of contrast agent needs to be reduced at large B0 fields, or
the MR signal becomes too small to measure above the noise. Finally, the maximum CBV signal
change is constrained by the actual physiologic CBV change (which, unlike BOLD, is field
independent). The signal boosting effect of contrast agent will be less when the BOLD effect is highest
62
(at high B0 field), and the maximum improvement in SNR is thus seen at lower field strengths.
129,130
Threefold increases in signal at 1.5 T and 3 T have been reported compared to BOLD studies.
Since the BOLD signal and contrast-enhanced CBV signal have opposite sign, enough contrast agent
2 of 5 12-04-2010 00:25
must be used to overcome the BOLD signal. Using high-dose contrast and short echo times at 1.5 T,
fivefold increases in SNR have been observed.131 This increased SNR allows higher resolution imaging,
131
and voxel dimensions of 2 mm × 2 mm × 2 mm are achievable at 1.5 T with this method. Signals
from large veins relax very quickly and do not contribute significantly to the image, so the CBV signals
are thus better spatially correlated with brain parenchyma than BOLD fMRI. 62
In addition to providing improved SNR for functional MRI studies, this method also allows the actual
CBV change to be calculated, and like CBF measures of functional activity, the CBV change provides
quantitative information. By measuring the iron concentration in venous blood it is possible to calibrate
the R2* relaxation rate, and thus get an absolute measure of CBV change with neuronal activation (the
CBV can be calculated from the slope of R2* change with [Fe]blood (for derivation see references 62
and 129).
R2* is the relaxation rate at rest (r), during activation (a), and without MION prior to the start of the
experiment (0).
For photic stimulation from a 6-second duration 8 Hz flickering checkerboard, the CBV change in
macaque primary visual cortex using the iron oxide contrast agent method is 32%. This value
127
corresponds closely to the CBV changes measure in humans using the first-pass gadolinium
contrast agent method.
Iron oxide contrast media enter the normal iron stores in the body and are excreted by the liver. The
large dose needed to overcome the signal from BOLD (approximately 5 to 7 mg Fe/kg at 1.5 T) would
rapidly lead to overload of the normal iron stores. This has limited these iron oxide-based methods to
studies in animal models.
Assessment of Cerebral Blood Volume Using an Endogenous Contrast

Agent
In recent years, noninvasive techniques have been proposed using the blood itself as the contrast
agent to determine the CBV. These are based on nulling the blood pool signal using flow-weighting
gradients132,133 or inversion recovery.134
Determining the activation-induced change in CBV from the nulled blood pool signal is not trivial. The
flow-weighting method is based on the observation that the application of flow-weighting gradients
leads to a greater attenuation of blood signal than of brain parenchyma because of the intravoxel
dephasing of randomly oriented vessels in the microvasculature. Thus, the difference of images
acquired with and without flow-weighting gradients yields an image that is proportional to CBV.
Confounding factors include slight diffusion weighting of brain parenchyma, artifacts due to the
pulsatility of CSF, and activation-related changes in the T2 of venous blood. To address these
confounds, Liu et al132 used a T2-FLAIR preparation sequence135 followed by a spin-echo sequence
with a spiral readout to null out the CSF signal and compensate for blood T2 changes. The diffusion
weighting of parenchyma was minimal and could be compensated for by a multiplicative term based on
the average diffusion coefficient for brain tissue. A remaining confounding factor is the increased
attenuation of blood signal due to the activation-related increase of microvascular blood velocities,
which can result in an overestimate of the CBV change. Calculations based on current knowledge of
the microvasculature architecture indicate that this factor is small; however, a comparison of the
flow-weighting method with CBV measurements from the MION measurement described above would
provide an independent validation of the accuracy of the method.
The vascular space occupancy (VASO) method proposed by Lu et al134 uses a nonselective inversion
pulse to null out the blood signal. Because the T1 of blood has been shown to be independent of
3 of 5 12-04-2010 00:25
oxygenation, this pulse nulls out blood in all compartments of the microvasculature. The remaining
VASO signal is composed of brain parenchyma and CSF. As CBV increases, the VASO signal should
decrease as parenchyma is displaced by blood. In experiments with healthy human volunteers, Lu et al
found that the VASO signal decreases with hypercapnia, increases with hypocapnia, and decreases
with visual stimulation. The temporal dynamics of the VASO signal with visual activation were
consistent with a rapid increase in arteriolar volume upon onset of activation and a slow decrease in
venous volume at the end of activation. Although the VASO signal provides an indication of the
dynamics of CBV, it does not currently provide a quantitative measurement of either the percent
change in CBV or the absolute values of CBV in the rest and active conditions. A comparison of the
VASO signal to CBV measurements obtained with MION has the potential to provide a calibration
method for the VASO method. At present these techniques for fMRI based on CBV change are less
widely used for human subjects than BOLD and CBF studies.
page 276
page 277
Magnetic Resonance Spectroscopy

MR methods other than fMRI can be utilized to measure functional activity and also provide measures
of additional aspects of cerebral physiology. A number of studies have demonstrated stimulus-related
changes in cerebral metabolites, in particular reduced glucose and elevated lactate levels, following
visual,136,137 somatosensory,138-140 motor,141 auditory,142 and cognitive stimulation.143,144 Recent
145,146
functional MR spectroscopy studies have improved the temporal resolution using time-resolved
142,147
or EPI-based techniques. These show good spatial concordance of spectroscopic changes with
147
BOLD fMRI. The SNR for spectroscopic techniques is lower than with BOLD, and thus there is
reduced spatial and/or temporal resolution. However, they allow investigation of additional cerebral
parameters associated with neuronal activation, and thus provide further insight into control of cerebral
metabolism.
Diffusion
Diffusion tensor MRI (DTI) provides quantitative maps of natural microscopic displacements of water
molecules that occur in tissues as part of the physical diffusion process. During typical diffusion times
148
of 50 to 100 ms, unrestricted water molecules move approximately 15 μm. Measurement of any
restriction in water diffusion reveals microscopic architectural details about tissue structure. Although
more widely used to diagnose cerebral ischemia149-151 or to map white matter tract architecture,152
153
functional changes in diffusion anisotropy may also be used to image cerebral activation. Direct
optical measurements in animals demonstrate an early change in photon scattering following neuronal
activation. Changes in neuronal volume, particularly at the axon hillock, have been proposed as a
possible mechanism for this observation.154,155 Small changes (less than 1%) seen in the apparent
diffusion coefficient (ADC) following visual stimulation may reflect this change in neuronal and glial cell
156
shape. This MRI technique thus provides another way of interrogating changes in cellular physiology
associated with neuronal activation.
Manganese Tract Tracing
4 of 5 12-04-2010 00:25
Add to lightbox
Figure 9-19 T1-weighted axial-oblique image of the brain of a mouse at 7 tesla following injection of 1
μL of manganese chloride into the anterior chamber of the right eye. Manganese is a T1 contrast
agent that is transported both antegradely and retrogradely along activated neurons. The globe, retina,
and optic tracts are enhanced by the contrast agent. (Image courtesy of Miriam Scadeng and James
Lindsey.)
As described above, measures of diffusion anisotropy can provide architectural detail of white matter
tracts. The fiber maps generated using this DTI technique impart structural connectivity information
based on diffusion of water, but do not necessarily detail functional connectivity. Manganese tracking
is an emerging MRI technique that maps functional white matter connectivity by tracking the transport
of manganese chloride contrast agent.157-160 Manganese is a T1 contrast agent which acts as a
calcium analog. It is transported both anterogradely and retrogradely along neuronal tracts and
161
crosses synaptic junctions. Synaptic transport of manganese parallels neuronal activity, thus
providing images based on actual functional connectivity. Figure 9-19 shows retinal and optic tract
enhancement in a mouse after injection of manganese chloride into the anterior chamber of the eye.
At present the technique is limited to animal studies due to the need to deliver the manganese directly
into the tract under investigation and also toxicity associated with manganese . With further
improvements in mechanisms for delivery and safety, this technique holds promise for studying
functional connectivity in normal human subjects and patients.
© 2010 Elsevier
5 of 5 12-04-2010 00:25
EXPLORING THE HEMODYNAMIC RESPONSE TO BRAIN ACTIVATION WITH

MRI
In the previous sections we introduced and reviewed the current thinking about fMRI signals, their
physiologic basis, and the analytic methods typically employed. In this section we discuss concepts
and issues that are currently being debated and researched. By its nature, the selection of the topics
discussed is somewhat arbitrary.
page 277
page 278
Are Oxygen and Glucose Metabolism Linked during Increased Neural

Activity?
As mentioned earlier (see "The Physiologic Basis of fMRI"), recent evidence suggests that CBF and
CMRO2 change in a graded fashion with different neural activity states, the CBF change being
approximately twice as large as the CMRO2 change. These data provide some support for the general
idea that neural activity leads to increased energy demand and that CBF increases to provide the
oxygen and glucose for energy metabolism, and suggests that neuronal activity changes are
appropriately reflected by CBF changes. However, the link between oxygen metabolism, glucose
metabolism, and neuronal activity is yet not well established.
The primary expenditure of energy is required to restore the ion gradients degraded during neural
activation. The intracellular/extracellular Na+ gradient is far from equilibrium, so pumping Na+ against
this gradient is a strongly uphill reaction in a thermodynamic sense. For this reason, the most costly
aspect of neural activity is likely to be excitatory synaptic activity in which glutamate opens sodium
+ +
channels. Indeed, the action of the Na /K pump is thought to consume a large fraction of the ATP
energy budget in the brain.14 In a recent animal experiment, blocking voltage-dependent sodium
162
channels substantially reduced the CBF response, supporting the idea that the dominant energy-
consuming process in the brain is recovery from excitatory activity.
Glucose metabolism has attracted less attention than oxygen metabolism from the MRI community,
13
although C spectroscopy techniques show considerable promise for addressing critical
questions.163,164 The delivery of glucose is not as limited as that of oxygen (see "The Function of
Neurovascular Coupling and the Oxygen Limitation Model"). In fact, almost half of the glucose
extracted from the blood vessels is not metabolized and diffuses back into the venous blood. This
means that, at least in principle, an increase in glucose metabolism can be accomplished without a
concomitant increase in CBF. In animal experiments in which the CBF response to activation was
blocked, the glucose metabolic rate nevertheless increased by the same amount as it did when the
165
CBF change was not blocked. Thus, within certain physiologic limits, the glucose metabolism
change might be functionally independent of the CBF.
Despite this independence during activation, at baseline CBF and CMRGlc appear to be well matched.
The glucose transporters Glut1 and Glut3 are mainly responsible for the transport process in the
brain by facilitated diffusion.165 As reported by Duelli and Kuschinsky, the density of these transporters
matches the local glucose utilization measured with the 2-[14C]deoxyglucose method. In addition, it
was found that the transporter density was highly correlated with the capillary density. The transporter
density can be regulated upward or downward by a chronic (approximately a few days) increase or
decrease in glucose metabolism. This finding, together with the hypothesis of Harder et al166 that
astrocytes sense neuronal activity and lead to angiogenesis as observed in cell cultures, leads to an
intriguing speculation about the function of hyperemia following stimulation: that the CBF increase is
needed rather for the formation of new capillaries than for nutritional support. While angiogenesis
following stimulation is most likely true for chronic hyperemia, it is, however, unlikely that this could
serve as a general explanation for the large transient CBF changes associated with the short
stimulation periods (less than a few minutes) used in fMRI studies.
1 of 18 12-04-2010 00:26
The close correspondence in the baseline state between the enzymes associated with glucose
metabolism and CBF is similar to the situation for oxygen metabolism. Stains for cytochrome oxidase,
a key enzyme in the mitochondrial respiratory chain, correlate closely with capillary density.167 In
167a
addition, Raichle and colleagues reported that from many PET studies the oxygen extraction
fraction E is nearly uniform across the brain with a value of approximately 40%. This suggests a close
coupling of CBF and CMRO2 in the baseline state. Also, the oxygen glucose index (OGI), the ratio of
CMRO2 to CMRGlc, is typically approximately 5.5 at rest, close to the theoretical ratio of 6 for full
oxidative metabolism of glucose .18
In short, a number of measures support a close association of CBF, CMRO2, and CMRGlc at rest,
reflected by a uniform value of E and an OGI near 6. What is then surprising is that this pattern is not
followed with activation. As we have described in detail, E decreases with activation, and produces the
BOLD effect. The OGI also decreases with activation, meaning that CMRGlc increases more than
CMRO2.
Why Does Glucose Metabolism Increase more than Oxygen

Metabolism with Brain Activation?
page 278
page 279
As noted above, the OGI is approximately 5.5 during rest and deviates from the expected value of full
18
oxygen metabolism of 6 (see "The Physiologic Basis of fMRI"). In addition, the OGI even decreases
168
to a lower value during activation. It is currently not known why this OGI pattern during rest and
activation occurs. Some possible explanations are:
1. Glucose is metabolized non-oxidatively to pyruvate and then lactate, which is partly cleared by
venous flow and is lost for full oxidative metabolism. MR spectroscopy can be utilized to
measure the increase in lactate concentration in the brain137,169,170 and thus allows the study of
the fate of the glucose used.
2. Glucose is used to produce other chemical products (e.g., glutamate, glutamine) and hence
171
serves other functions than metabolism alone. The products are then metabolized after
stimulation ends, replacing glucose and reducing the glucose extraction fraction until the
resting concentrations are reached. An MR method to measure these metabolites is [13C]-MR
28
spectroscopy.
3. Another hypothesis for the extensive use of glucose is that the non-oxidative metabolism is
needed for fast processes. Glycolysis generates ATP much faster than full oxidative
metabolism.170 Shulman and colleagues proposed a "glycogen shunt" model in which glucose
is metabolized to glycogen during rest to fill up the glycogen pool in astrocytes, which in turn
during stimulation is metabolized quickly within milliseconds in order to deliver the energy
needed to recycle the neurotransmitters released during neuronal activity. However, this model
would only explain transient changes in OGI.
4. In some structures in the brain oxidative metabolism may not be possible, e.g., due to missing
172
mitochondria in some dendritic spines. However, energy metabolism is still required. No
estimate has been presented of how much of the glucose metabolism needs to be
non-oxidative and if this mechanism can explain the experimental findings during rest and
stimulation.
Future studies using MR spectroscopy combined with fMRI are needed to resolve these issues and will
be very useful for investigations of the physiologic basis of fMRI signals.
What is the Underlying Neuronal Activity that Drives the fMRI Signals?
With these uncertainties in mind we come back to the question of what neuronal processes are mainly
reflected by CBF and the BOLD signal. Synaptic activity may be excitatory or inhibitory, and this
synaptic activity may or may not produce spiking activity. Different fMRI signals that differ between
2 of 18 12-04-2010 00:26
brain areas or between subjects might be due to different types of neuronal activity or differences in
neurovascular coupling (or both). This clarification is important for two reasons. First, as pointed out
earlier, the hemodynamic response is tightly coupled to oxygen metabolism. The neurovascular
coupling translates very different neuronal and astrocytic processes into one dimension of CBF
response, most likely reflecting just the total oxygen metabolism of these processes. Thus, different
activities of neurons and astrocytes can lead to the same CBF change. Second, the location of the
hemodynamic response may differ from the location of the neuronal activity not only because the
hemodynamic response is controlled in a coarse manner from feeding arterioles, but also because
local processing and input from another brain area may elicit different responses (see below).
Spiking versus Synaptic Activity

The current debate about precisely which aspect of neural activity drives the CBF change is usually
framed in terms of synaptic activity versus spiking activity. Synaptic activity includes neurotransmitter
release and recycling, post-synaptic potential changes, and ion fluxes initiated by the neurotransmitter,
while spiking activity refers to the generation and propagation of action potentials. Rees et al and
Heeger et al have demonstrated a linear relationship between spiking activity in monkeys and BOLD
signal in a similar cortical area in humans; in the former study 9 spikes per second per neuron in V5
and in the latter study 0.4 spikes per second per neuron in V1 was found to correspond to 1% BOLD
signal change.173,174 One of the difficulties in interpreting these studies is that the spiking and
post-synaptic activity are usually strongly correlated.
Two studies have successfully dissociated synaptic and spiking activity. In one study, Lauritzen and
colleagues stimulated either the parallel or the climbing fibers in the rat cerebellum, which have either
inhibitory or excitatory effect on the Purkinje cells in (for a review see reference 175). With this method
they were able to minimize the spiking activity while synaptic activity continued, and this correlated with
a CBF increase due to stimulation despite zero spiking activity. A second study comparing BOLD
signal with high-frequency electrical activity or multi-unit activity (thought to reflect spiking activity) and
the mean local field potential (thought to reflect synaptic activity) found a slightly higher correlation
between BOLD signal and local field potential (LFP).176 Recently, by injecting serotonin, the same
177
group succeeded in suppressing the multi-unit activity while leaving the LFP unchanged. The BOLD
response also remained unchanged, enforcing the strong connection between LFP and BOLD signal.
However, the relationship between electrical activity of neuronal ensembles and CBF might be even
more complicated. In a recent study from Caesar et al, using topical application of GABAA (muscimol)
receptor agonist in the cerebellum of the rat, the baseline CBF did not change whereas the spiking
activity and the CBF response to climbing fiber stimulation were reduced.178 However, the agonist did
not affect the LFP. One conclusion might be that the correlation of LFP and CBF is valid for excitation
circuits but not for inhibition circuits (see "Does Inhibition Produce a BOLD Response?").
In summary, the experimental results favor synaptic activity rather than spiking activity as the prime
driver of CBF changes. This is reinforced by theoretical estimates that for the primate brain 70% or
more of the required energy metabolism is due to synaptic activity (see Fig. 9-2). These findings might
have important implications in interpreting CBF and BOLD changes regarding the spatial information
encoded by these signals.
Input signals to one area or in one neuron are mainly represented by synaptic activity, whereas spiking
activity represents the output signals.176 Therefore the current suggestion is that CBF represents the
input to an area rather than the output. This would have the important implication that BOLD and CBF
changes in one area can be caused remotely by input signals of another area and hence that these
changes may spatially misrepresent the regions with increased spiking activity. However, local
processing in one area evokes both spiking and synaptic activity because of the many local
connections, so it is reasonable to assume that most of the neuronal signals causing CBF changes are
within the area measured. Thus, mapping CBF changes to infer neural activity changes should only
miss activated areas that are spiking with minimal synaptic activity. In addition, subthreshold activity in
one area has a broader spatial extent than spiking activity, such as LFP, and thus CBF should have a
3 of 18 12-04-2010 00:26
broader spatial distribution than the spiking activity.

page 279
page 280
Simultaneous Measurements of Electroencephalography and Event-

Related Field Potentials with fMRI
Event-related field potentials (ERP) and electroencephalography (EEG) measured with fMRI have
been utilized to investigate noninvasively the relationship between CBF and underlying neuronal activity.
ERP and LFP have a close relationship because both are assumed to be a weighted sum of
post-synaptic currents. In contrast, EEG oscillations reflect mainly synchronized neuronal activity.
Arthurs and colleagues have found a linear correlation between somatosensory evoked potentials and
BOLD response in humans.179 Using two-pulse stimulation with different inter-stimulus intervals up to
100 ms, Ogawa et al showed in rats that neuronal interaction measured as suppression of the ERP
180
magnitude to the second stimulus is reflected by the BOLD signal. By the way, this experiment
showed that although the hemodynamic response evolves in seconds, the magnitude of the
hemodynamic response could encode events and interactions on a millisecond time scale. In a recent
study intradural ERP in patients and BOLD signal in healthy volunteers during the same visual
stimulation were measured.181 ERP and BOLD signal did not correspond coherently for all areas.
Furthermore, additional analysis revealed no consistent correlation between the EEG and BOLD
signal. In summary, the connection between ERP and BOLD signal clearly needs further investigation.
In the near future, the main problems in combining both measures (e.g., a non-magnetic EEG device,
EEG time course artifacts caused by switching MR gradients, etc.) will be solved and this will lead to
routinely combined fMRI-EEG measurements in healthy volunteers and in disease populations. In the
last few years many groups have made combined measurements with fMRI and EEG. As an example
from the increasing literature, several groups have correlated BOLD baseline fluctuations (resting
either with eyes open or eyes closed) with α-power in EEG and have found negative correlations in the
visual cortex and positive correlations in the thalamus.182 This finding is supported by the hypothesis
that the generator of the α-rhythm is the thalamus creating a subsequent local deactivation in the visual
cortex.
However, in our opinion the α-power (or a complex composite of EEG oscillations) cannot be regarded
as a general predictor of the CBF time course. Synchronized activity of the pyramidal cells reflected in
the EEG oscillations does not have to be more (or less) energy demanding than non-synchronized
activity and, thus, can cause similar CBF and BOLD response. That is, in other areas other EEG
oscillations are expected to correlate with CBF changes. Therefore the positive finding of these studies
should be considered as a special case of high correspondence of CBF and EEG. However, to
understand the neural basis of the fMRI signals it is not only the EEG signals that correlate with the
BOLD signal that are important but also the EEG signals that do not correlate. Clearly more work is
needed to explore the relation between changes in the fMRI signals and the underlying
electrophysiologic signals, which can be measured directly by invasive means or indirectly
noninvasively with EEG and ERP.
Does Inhibition Produce a BOLD Response?

In the cortex the number of excitatory neurons is approximately five times larger than the number of
inhibitory neurons.183 Release of the main excitatory transmitter glutamate triggers increased
184
metabolism in astrocytes, which in turn has been shown to correlate to neuronal oxidative
28
metabolism. In addition, the coupling between CBF and CMRO2 has been shown to be the same
during increased and decreased activity within the range of physiologic manipulations.185 Thus, the
association of excitatory synaptic activity with increased energy metabolism is clear. Does the main
inhibitory neurotransmitter, GABA, produce the same increase (or even a decrease) in metabolism as
glutamate?
4 of 18 12-04-2010 00:26
Excitation and inhibition share common energy consuming processes in synapses, such as clearing
neurotransmitter from the synaptic cleft and repackaging neurotransmitter in vesicles. For both types
of activity ionic gradients must be restored, although the energy requirements may be different
because the ions are different. Excitatory activity opens sodium channels, while inhibitory activity opens
chloride and/or potassium channels. The sodium distribution across the cell membrane is the farthest
from equilibrium, and so the thermodynamic energy cost may be higher for pumping sodium than
chloride or potassium. Therefore, it is reasonable to assume that inhibition also is energy demanding,
but it is not clear whether it is as demanding as excitatory activity. That is, like excitation, inhibition
should cause an increase in CBF, which in fact was found in the hippocampus186 and in rat
187
cerebellum. However, in cell cultures no increase in metabolism caused by GABA release was found
188
in astrocytes. In support of this finding, no change in CBF during inhibition was observed using a
go/no-go event-related paradigm and combined transcranial magnetic stimulation and fMRI
189
measures. Interpreting these results in a conclusive way is difficult, because inhibition might also
cause a decrease in CBF by suppressing the excitatory activity in the subsequent neuronal circuit: for
example, a decrease in CBF is often found during sensorimotor stimulation in the ipsilateral cortex area
due to transcallosal inhibition (see, e.g., reference 190). In summary, the net effect of inhibition might
be an increase, no change, or a decrease of CBF (for a review see reference 191).
What is the Significance of the Transients of the BOLD Signal?

page 280
page 281
Add to lightbox
Figure 9-20 A typical BOLD response for a block design stimulus. Transients are an occasional brief
initial dip at the beginning, an initial overshoot, and a prolonged post-stimulus undershoot.
The evoked hemodynamic response has some typical characteristics. Notable features of experimental
measurements of the BOLD response are the transients, an occasional brief initial dip at the beginning
or the more common prolonged post-stimulus undershoot.20,26 The positive BOLD response reaches
its maximum after approximately 5 to 8 seconds. For a block design an initial overshoot is observed if
a post-stimulus undershoot is present in the response to an event. A typical BOLD response is shown
in Figure 9-20. The post-stimulus undershoot often appears as an apparent lowering of the baseline
after the first stimulus block, when the undershoot has not fully resolved before the next stimulus block
begins. In general, to clearly distinguish the undershoot, the rest block should be longer than the
stimulus block.
In the balloon model,20 such transients have two distinct sources. The initial dip is modeled as a slight
delay (approximately 1 s) of the CBF response compared to the CMRO2 response. The post-stimulus
undershoot arises in the model because CBV is slower to recover than CBF and CMRO2. Then if the
oxygen extraction fraction returns to baseline at the end of the stimulus, but the venous blood volume
remains elevated, total deoxyhemoglobin will be higher than baseline, reducing the BOLD signal.
In the original version of the balloon model20 it was noted that an initial dip could result from a rapid
5 of 18 12-04-2010 00:26
rise in blood volume. A number of optical studies found that the initial dip period corresponded to an
increase of deoxyhemoglobin but without a dip in deoxyhemoglobin, suggesting a combined change in
CBV and E.69,192 A more recent study, however, found a corresponding initial decrease in
oxyhemoglobin in conjunction with an increase in deoxyhemoglobin, more clearly suggesting a change
73
in E. Some recent data show that in fact the tissue oxygenation decreases shortly after beginning of
the stimulation.193 However, whether this tissue deoxygenation is translated into vascular
deoxygenation has still to be proven. Although more data are clearly needed to resolve the
experimental inconsistencies, recent experimental work suggests the change in E as the source of the
initial dip. Despite its elusive nature,26 it has excited great interest because it appears to be better
localized than the primary BOLD effect (see below), and because it may represent a transient
uncoupling of CBF and oxygen metabolism.
Post-Stimulus Undershoot
The most frequently observed transient is a post-stimulus undershoot that often lasts for tens of
seconds. A possible explanation for this phenomenon is that the CBV change resolves more slowly
than the CBF change at the end of the stimulus. This effect has been seen in a study in a rat model, 21
and two similar theoretical models have been proposed to explain the effect (the delayed compliance
21 20
model and the balloon model ). Both of these models are based on the distensibility of the venous
vessels, and the long time constant for CBV to adjust is then treated as a biomechanical property of
the vessels. Note that this view involves some subtlety with regard to the relationship between CBV
and CBF. In order to increase CBF, the vascular resistance must be decreased by dilating the
arterioles. This active volume change on the arterial side, the direct cause of the CBF change, is
thought to be a small fraction of the total CBV change. The venous change, however, is thought to be a
passive response to the increased CBF and corresponding pressure changes. However, much of this
picture is still speculative, although the resulting model curves often describe experimental BOLD data
quite well. A recent finding supports this basic picture: using a sequence sensitive to blood volume
changes (called VASO), Lu and colleagues reported that the early change in VASO signal is parallel to
the increase in CBF, consistent with the idea that the early increase in CBV is caused by the arterioles,
which in turn leads to the CBF increase.134 However, the VASO signal after the stimulus ended
showed a much slower recovery to baseline than CBF, consistent with the idea that the late part of the
VASO signal is dominated by venous blood volume changes.
Other explanations of the post-stimulus undershoot are still under discussion: 1. CMRO2 recovers to its
baseline value more slowly than CBF,136 hence oxygen is extracted even after CBF recovered to
baseline producing transient hypo-oxygenation; 2. CBF drops below baseline after the stimulus ending,
presumably due to neuronal inhibition and a corresponding decrease of neural activity in the
post-stimulus period.177
Only the first explanation is not compatible with a consistent coupling of CBF and CMRO2 and requires
a transient uncoupling. However, as pointed out above, experimental data suggest that the
post-stimulus undershoot is a result of the temporal mismatch of CBF and CBV and enhanced by a
CBF undershoot. However, the temporal dynamics of the post-stimulus undershoot and its dependence
on the baseline CBF value are not well understood and are still under investigation.
Nonlinearity of the BOLD Response

page 281
page 282
Although the standard statistical analysis assumes a linear response, such that the net response to
several stimuli is simply the sum of the individual responses to a single stimulus, the hemodynamic
response is often nonlinear. The BOLD response typically exhibits a temporal nonlinearity, such that an
appropriately shifted and added response to a brief stimulus over-predicts the true response to an
extended stimulus.85,194-199 This temporal nonlinearity is most pronounced when the brief stimulus is
less than about 4 s and the extended stimulus is longer than 6 s. Comparing short and long duration
stimuli that are both longer than approximately 4 s, the temporal nonlinearity is reduced.
6 of 18 12-04-2010 00:26
A nonlinearity such as this could arise in several ways. In the step from the stimulus to the evoked
neural activity, adaptation can produce an initial sharp rise in activity that decreases to a lower plateau
level, even with a constant stimulus. In addition, the step from neural activity to CBF response could be
nonlinear, for example through a ceiling effect on CBF change.
Friston et al suggested that a large part of the nonlinearity of the BOLD response arises from the
transformation of a CBF response to the BOLD signal change.85 In a subsequent study they showed
that the Volterra kernel characterization of experimentally observed nonlinearities could be accounted
for with the balloon model plus a linear transformation up to the CBF response, again supporting the
idea that the primary nonlinearity is in the transformation from the CBF to the BOLD response. 200 The
two sources of nonlinearity (neural or vascular) can be distinguished experimentally by whether the
nonlinearity is present in just the BOLD response or in both the BOLD and CBF responses. For
example, Miller et al199 found that both visual and motor cortices exhibited a nonlinear BOLD response,
but only the visual cortex showed a nonlinear CBF response.
This source of nonlinearity in the step from CBF to BOLD can be explained as a BOLD ceiling effect.
Even an infinite CBF change could still produce only a finite BOLD response, corresponding to
removing all deoxyhemoglobin from the voxel. This effect produces an over-prediction of the amplitude
of a long duration stimulus from a short duration stimulus when the flow change due to the short
stimulus is substantially weaker than the flow change due to the longer stimulus. That is, the BOLD
ceiling effect should introduce a nonlinear response when the shorter stimulus is narrower than the
width of the CBF response, which is approximately 4 s, in good agreement with experimental data.
In addition to the nonlinearity of the amplitude, there are nonlinearities of the timing of the response,
201 202
such as latency or response width. A recent proposal by Behzadi et al and Baird and Warach is a
promising approach for accounting for these effects. Instead of treating CBF as a linear convolution
with the neural activity, it is assumed that neural activity releases a vasoactive agent that alters CBF,
as in the model described in reference 200. The difference is that the agent is modeled as acting on
the compliance of the vessel. The compliance in turn is treated as a combination of the smooth muscle
tension, which can be controlled by the agent, and a fixed elastic component that becomes a more
dominant factor in determining compliance when the vessel is expanded. In this way, the same
concentration of the agent will have a greater effect on CBF when the flow is lower, and potentially a
different effect on the nonlinearity of the response.
The Physiologic Baseline Strongly Affects the BOLD Signal

The fact that the BOLD signal depends on a combination of changes in CBF, CBV, and CMRO2, and
also on the baseline physiologic state, makes it difficult to interpret the magnitude of the BOLD signal
change unambiguously without further experimental information. For this reason, the BOLD effect has
been used primarily as a mapping tool, based on detecting signal changes, rather than as a probe of
the underlying physiology based on a detailed analysis of the BOLD response.
However, combined measurements of BOLD and CBF can be used to model the effects of the
physiologic baseline on the BOLD response. Experiments have found that when baseline CBF is
increased by breathing CO2 or administering acetazolamide , the BOLD response is reduced. For
49
example, Brown et al found that acetazolamide raised baseline CBF by 20% and the BOLD
response in the motor cortex with finger tapping was reduced by 35%, but the CBF change (∆CBF)
with activation stayed the same.
With CO2 inhalation CBF increases but it is thought that CMRO2 remains the same. This means that at
this new baseline the oxygen extraction fraction E must be smaller than it was at the previous baseline,
and this implies that there is less deoxyhemoglobin present at the new baseline. Thus, even if the
fractional changes of the physiologic quantities were the same, the BOLD signal would be reduced.
However, in addition, the experimental data suggest that ∆CBF remains constant despite the raised
baseline CBF, so the fractional CBF change is smaller at the elevated baseline, and this will further
reduce the BOLD response.
7 of 18 12-04-2010 00:26
As a numerical example, consider an activation that produces a 30% change in CBF and a 10%
change in CMRO2 from the initial baseline state, giving a BOLD signal change of 1.36% (calculated
with Equation 9-10 and A = 0.1). If the subject now breathes CO2 that produces a 20% increase of
baseline CBF with no change in CMRO2, the raw BOLD signal baseline will increase by 1.82%. If the
same activation on top of this new baseline state produces the same absolute changes in CBF and
CMRO2, we can calculate the net effect by first considering that if this net physiologic change had
occurred in the original baseline state, the relevant changes would have been a 50% change in CBF
and a 10% change in CMRO2, which would have given a BOLD signal change of 2.61% from the
original baseline. Subtracting the BOLD baseline shift of the new state, the BOLD signal change from
the new baseline is 0.79%. This is a 42% reduction of the BOLD response to the same activation due
just to the 20% increase of the CBF baseline, and is in reasonable agreement with the experimental
data.
page 282
page 283
The implication of this sensitivity of the BOLD signal to the baseline state is that, potentially, many
factors could alter the baseline state of a patient group (e.g., anxiety or vasoactive medications) that
could make their BOLD responses significantly different from a healthy population even if the neural
responses in the two groups were identical. For this reason, the ∆CBF measured with ASL techniques
may prove to be a much more robust approach for quantitative fMRI studies.
While these considerations show how changes in the baseline CBF can affect the BOLD response, we
have not dealt with the issue of what determines the baseline CBF. A recent theory, which requires
further development, is that CBF is regulated to maintain a constant ratio of O 2 to CO2 at the
mitochondria, in order to preserve the thermodynamic free energy available from oxidative metabolism
203
of glucose . The O2 concentration at the mitochondria ([O2]) is determined by E, so to increase the
diffusive flux of O2 from capillaries to mitochondria, while maintaining [O2], requires E to decrease with
activation. In addition, increasing CO2 in the blood increases CO2 in the tissue as well, degrading the
[O2]/[CO2] ratio, and this is again restored by decreasing E. Thus, the model predicts that CBF should
increase with CO2, and that an additional increase is required to meet the needs of increased CMRO2.
As commonly used, the terms activation and deactivation are relative terms, so the choice of an
appropriate task to define the neural activity of the baseline condition plays a critical role in an fMRI
experiment.204 The uncritical use of the notion of baseline may lead to erroneous interpretation of fMRI
data. Essentially, the brain is always intrinsically active as the high value of "resting" metabolism
shows. The brain has a weight of approximately 2% of the weight of the whole body, but accounts for
20% of the whole metabolism. Strictly speaking, in stimulation experiments, two different active states
are compared with each other and one is referred arbitrarily to be the baseline state. This is especially
problematic for complex brain functions and thus the baseline state has to be chosen carefully.
Do BOLD Correlations Reveal Long-Range Patterns of Connectivity?

As pointed out earlier (see "Design and Analysis of BOLD-fMRI Experiments" and "Artifacts and
Noise"), the BOLD signal varies due to physiologic fluctuations. Beside heartbeat and respiration
110
related changes the so-called "slow oscillations" (approximately 0.1 Hz) are observed. It is currently
debated whether these oscillations are vascular or neuronal in origin. The vascular hypothesis refers to
brain blood pressure autoregulation, with 0.1 Hz as the mean frequency of the autoregulation, so that
the vascular system acts as a frequency filter at 0.1 Hz. Thus, according to this view, correlations of
these oscillations reveal only vascular symmetry.
However, a growing number of studies are now using BOLD baseline fluctuations in order to find
205
modular neuronal networks in the brain. The basis for this is the assumption that neuronal
connectivity is reflected by BOLD signal fluctuations. Neuronal connected areas have simultaneous
fluctuations in baseline activity and thus should create temporally correlated fluctuations in BOLD
signal. However, although this is certainly true, until now no estimation exists of the magnitude and the
frequency of these BOLD fluctuations. Therefore, results of this approach should be interpreted
8 of 18 12-04-2010 00:26
cautiously.
Spatial and Temporal Resolution

The transformation of neuronal signals to vascular dynamics may lead to information loss:
1. Qualitatively: Different dimensions of neuronal processes like inhibition and excitation or

presynaptic and post-synaptic activity are transformed into vascular changes. Clever
experimental designs are needed in order to recover this information and distinguish these
processes from measured vascular changes.
2. Time domain: The vascular response is on the order of one second, whereas the neuronal
changes are on the order of milliseconds. Thus, neuronal changes are averaged and temporally
blurred in the vascular domain. However, as noted above, Ogawa et al have shown that using
two-pulse stimulation of 10 ms duration the amplitude of the vascular changes can encode
neuronal changes which are separated by only a few milliseconds.180 This idea needs further
elaboration and more experiments are required to elucidate the temporal limits of fMRI. In
particular, interaction analysis and nonlinearities seem to provide a versatile tool to improve the
temporal limits of fMRI.
3. Spatial domain: It is not clear what the minimum spatial scale is for the vascular changes to
occur. In the olfactory cortex it was shown by Chaigneau and colleagues in 2003 with
two-photon optical imaging that capillaries 100 μm remote from the activated area are not
showing any stimulus-correlated change.206 In a recent review, Iadecola pointed out that this
finding is an argument against diffusion-controlled changes in CBF and an argument for direct
control of CBF by neurons. This direct control can be employed by interneurons, which integrate
the neuronal activity in local neuronal circuits.207
Harrison et al showed that vascular imaging signals correspond well with the underlying vascular
structure as shown in Figure 9-21.208 However, the translation of the neuronal activity to CBF response
introduces spatial blurring. It is often assumed that CBF changes are actively controlled by the
arterioles. The feeding region of an arteriole is on the order of a millimeter,207 i.e., changes of neuronal
activity within different neuronal populations in the "receptive field" of the same arteriole cannot be
spatially separated by fMRI signals. Thus, fMRI signals reflect activity of a population of neurons, and
a subpopulation of these neurons cannot be spatially resolved by fMRI. In addition, due to intravascular
signaling, arterial blood vessels upstream from the location of the neuronal activity also dilate. This
effect is called retrograde vasodilation,207 introducing further spatial blurring.
page 283
page 284
9 of 18 12-04-2010 00:26
Add to lightbox
Figure 9-21 Vascular structure of a chinchilla measured with corrosion casts and region of changes in
the imaging signal (top right, inset). Vascular imaging signals correspond well with the underlying
structure of the vasculature. (Adapted from Harrison RV, et al: Blood capillary distribution correlates
with hemodynamic-based functional imaging in cerebral cortex. Cereb Cortex 12:225-233, 2002, by
permission of Oxford University Press.)
Beyond the CBF response the "initial dip" (see "What is the Significance of the Transients of the BOLD
Signal?") has been suggested as a way to improve the spatial resolution of imaging signals. The dip is
thought to be produced by initial oxygen consumption, and so might be more precisely located than the
ensuing CBF and BOLD changes. Kim and colleagues have measured cortical columns in the visual
cortex of cats with a functional distance of the columns of approximately 1 mm. 209 However,
Logothetis, who has shown that the "early fMRI signal" was also found in the sagittal sinus, challenged
the interpretation of this study.210 Studies in monkeys176,211 found correlation between neuronal activity
and BOLD in parenchyma, but this decreased close to large vessels. For typical fMRI in humans with
imaging resolution 3 mm × 3 mm × 3 mm, the inclusion of small draining veins in the imaging voxel does
not unduly skew the representation of neuronal activity. As resolution is increased, steps need to be
taken to reduce the influence of large vessels. At higher fields, the use of spin-echo acquisition123 or
diffusion weighting during a BOLD experiment reduces the signal contribution from large veins.
In a different study, Kim et al measured the spatial correlation of single- and multi-unit activity with
BOLD signal and found a linear relationship between these signals when the voxel size is greater than
2
approximately 3 × 3 mm . Because most studies utilize voxel size greater than this value, the BOLD
signal reflects accurately the underlying neuronal signal. 212 Finally, a recent study using ASL
213
techniques was able to resolve cortical columns, suggesting fairly tight spatial control of CBF. In
sum, the limits of the spatial resolution of the CBF and BOLD response are not well understood and
need further investigation.
REFERENCES
1. Pauling L, Coryell CD: The magnetic properties and structure of hemoglobin, oxyhemoglobin, and
carbonmonoxyhemoglobin. Proc Natl Acad Sci USA 22:210-216, 1936. Medline Similar articles
3. Kwong KK, Belliveau JW, Chester DA, et al: Dynamic magnetic resonance imaging of human brain activity during primary
sensory stimulation. Proc Natl Acad Sci USA 89:5675-5679, 1992. Medline Similar articles
10 of 18 12-04-2010 00:26
4. Ogawa S, Lee TM, Nayak AS, Glynn P: Oxygenation-sensitive contrast in magnetic resonance image of rodent brain at high
magnetic fields. Magn Reson Med 14:68-78, 1990. Medline Similar articles
5. Ogawa S, Tank DW, Menon R, et al: Intrinsic signal changes accompanying sensory stimulation: functional brain mapping
with magnetic resonance imaging. Proc Natl Acad Sci USA 89:5951-5955, 1992. Medline Similar articles
6. Frahm J, Bruhn H, Merboldt KD, et al: Dynamic MR imaging of human brain oxygenation during rest and photic stimulation. J
7. Bandettini PA, Wong EC, Hinks RS, et al: Time course EPI of human brain function during task activation. Magn Reson Med
8. Turner R, Le Bihan D, Moonen CT, et al: Echo-planar time course MRI of cat brain oxygenation changes. Magn Reson Med
27:159-166, 1991.
9. Turner R, Jezzard P, Wen H, et al: Functional mapping of the human visual cortex at 4 and 1.5 tesla using deoxygenation
contrast EPI. Magn Reson Med 29:277-279, 1993.
10. Constable RT, McCarthy G, Allison T, et al: Functional brain imaging at 1.5 T using conventional gradient echo MR imaging
techniques. Magn Reson Imaging 11:451-459, 1993.
11. Sereno MI, Dale AM, Reppas JB, et al: Functional MRI reveals borders of multiple visual areas in humans. Science 268:
889-893, 1995.
12. Engel SA, Rumelhart DE, Wandell BA, et al: fMRI of human visual cortex. Nature 369:525, 1994. Medline
Similar articles
13. Erecinska M, Silver IA: Ions and energy in mammalian brain. Prog Neurobiol 43:37-71, 1994. Medline
Similar articles
14. Ames A III: CNS energy metabolism as related to function. Brain Res Brain Res Rev 34:42-68, 2000. Medline
Similar articles
15. Attwell D, Laughlin SB: An energy budget for signaling in the grey matter of the brain. J Cereb Blood Flow Metab
16. Attwell D, Iadecola C: The neural basis of functional brain imaging signals. Trends Neurosci 25:621-625, 2002. Medline
Similar articles
17. Ido Y, Chang K, Woolsey TA, Williamson JR, et al: NADH: sensor of blood flow need in brain, muscle and other tissues.
FASEB J 15:1419-1421, 2001. Medline Similar articles
18. Shulman RG, Hyder F, Rothman DL: Lactate efflux and the neuroenergetic basis of brain function. NMR Biomed
19. Grubb RL, Raichle MF, Eichling JO, Ter-Pogossian MM: The effects of changes in PaCO2 on cerebral blood volume, blood
flow, and vascular mean transit time. Stroke 5:630-639, 1974. Medline Similar articles
20. Buxton RB, Wong EC, Frank LR: Dynamics of blood flow and oxygenation changes during brain activation: the balloon
model. Magn Reson Med 39:855-864, 1998. Medline Similar articles
21. Mandeville JB, Marota JJ, Ayata C, et al: Evidence of a cerebrovascular post-arteriole Windkessel with delayed
compliance. J Cereb Blood Flow Metab 19:679-689, 1999. Medline Similar articles
22. Fox PT, Mintum MA, Raichle ME, et al: Mapping the visual cortex with positron emission tomography. Nature 323:806-809,
23. Mandeville JB, Marota JJ, Ayata C, et al: MRI measurement of the temporal evolution of relative CMRO2 during rat forepaw
stimulation. Magn Reson Med 42:944-951, 1999. Medline Similar articles
24. Lee SP, Duong TQ, Yang G, et al: Relative changes of cerebral arterial and venous blood volumes during increased
cerebral blood flow: implications for BOLD fMRI. Magn Reson Med 45:791-800, 2001. Medline Similar articles
25. Hyder F, Rothman DL, Shulman RG: Total neuroenergetics support localized brain activity: implications for the
interpretation of fMRI. Proc Natl Acad Sci USA, 99:10771-10776, 2002.
26. Buxton RB: The elusive initial dip. Neuroimage 13:953-958, 2001. Medline Similar articles
27. Villringer A, Dirnagl U: Coupling of brain activity and cerebral blood flow: basis of functional neuroimaging. Cerebrovasc
Brain Metab Rev 7:240-276, 1995. Medline Similar articles
28. Rothman DL, Behar KL, Hyder F, Shulman RG: In vivo NMR studies of the glutamate neurotransmitter flux and
neuroenergetics: Implications for brain function. Annu Rev Physiol 65:401-427, 2003. Medline Similar articles
29. Magistretti PJ, Pellerin L, Rothman DL, Shulman RG: Energy on demand. Science 283:496-497, 1999. Medline
Similar articles
page 284
page 285
30. Zonta M, Angulo MC, Gobbo S, et al: Neuron-to-astrocyte signaling is central to the dynamic control of brain
microcirculation. Nat Neurosci 6:43-50, 2003. Medline Similar articles
31. Fields RD, Stevens-Graham B: New insights into neuron-glia communication. Science 298:556-562, 2002. Medline
Similar articles
32. Erecinska M, Silver IA: Metabolism and role of glutamate in mammalian brain. Prog Neurobiol 35:245-296, 1990.
11 of 18 12-04-2010 00:26

33. Nedergaard M, Ransom B, Goldman SA: New roles for astrocytes: redefining the functional architecture of the brain.
Trends Neurosci 26:523-530, 2003. Medline Similar articles
34. Chih CP, Lipton P, Roberts EL Jr: Do active cerebral neurons really use lactate rather than glucose? Trends Neurosci
35. Fox PT, Raichle ME: Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during
somatosensory stimulation in human subjects. Proc Natl Acad Sci USA 83:1140-1144, 1986. Medline Similar articles
36. Buxton RB, Frank LR: A model for the coupling between cerebral blood flow and oxygen metabolism during neural
stimulation. J Cereb Blood Flow Metab 17:64-72, 1997. Medline Similar articles
37. Buxton RB: Coupling between CBF and CMRO2 during neuronal activity. In Tomita M, Kanno I, Hamel E (eds): Brain
Activation and CBF Control. Amsterdam: Elsevier Science, 2002, pp 23-32.
38. Gjedde A, Poulsen PH, Ostergaard L: On the oxygenation of hemoglobin in the human brain. Adv Exp Med Biol 471:67-81,
39. Gjedde A: Cerebral blood flow change in arterial hypoxemia is consistent with negligible oxygen tension in brain
mitochondria. Neuroimage 17:1876-1881, 2002. Medline Similar articles
40. Hyder F, Shulman RG, Rothman DL: A model for the regulation of cerebral oxygen delivery. J Appl Physiol 85:554-564,
41. Seitz RJ, Roland PE: Vibratory stimulation increases and decreases the regional cerebral blood flow and oxidative
metabolism: a positron emission tomography (PET) study. Acta Neurol Scand 86:60-67, 1992. Medline Similar articles
42. Marrett S, Gjedde A: Changes of blood flow and oxygen consumption in visual cortex of living humans. Adv Exp Med Biol
43. Davis TL, Kwong KK, Weisskoff RM, Rosen BR: Calibrated functional MRI: mapping the dynamics of oxidative
metabolism. Proc Natl Acad Sci USA 95:1834-1839, 1998. Medline Similar articles
44. Hoge RD, Atkinson J, Gill B, et al: Linear coupling between cerebral blood flow and oxygen consumption in activated
human cortex. Proc Natl Acad Sci USA 96:9403-9408, 1999. Medline Similar articles
45. Kim SG, Rostrup E, Larsson HB, et al: Determination of relative CMRO2 from CBF and BOLD changes: significant
increase of oxygen consumption rate during visual stimulation. Magn Reson Med 41:1152-1161, 1999. Medline
Similar articles
46. Kastrup A, Kruger G, Neuman-Haefelin T, et al: Changes of cerebral blood flow, oxygenation, and oxidative metabolism
during graded motor activation. Neuroimage 15:74-82, 2002. Medline Similar articles
47. Springett R, Wylezinska M, Cady EB, et al: Oxygen dependency of cerebral oxidative phosphorylation in newborn piglets. J
Cereb Blood Flow Metab 20:280-289, 2000. Medline Similar articles
48. Kastrup A, Kruger G, Glover GH, et al: Assessment of cerebral oxidative metabolism with breath holding and fMRI. Magn
49. Brown GG, Eyler Zorrilla LT, Georgy B, et al: BOLD and perfusion response to finger-thumb apposition after acetazolamide
administration: Differential relationship to global perfusion. J Cereb Blood Flow Metab 23:829-837, 2003. Medline
Similar articles
50. Hoge RD, Atkinson J, Gill B, et al: Investigation of BOLD signal dependence on cerebral blood flow and oxygen
consumption: the deoxyhemoglobin dilution model. Magn Reson Med 42:849-863, 1999. Medline Similar articles
51. Kennan RP, Zhong J, Gore JC: Intravascular susceptibility contrast mechanisms in tissues. Magn Reson Med 31:9-21,
52. Fisel CR, Ackerman JL, Buxton RB, et al: MR contrast due to microscopically heterogeneous magnetic susceptibility:
numerical simulations and applications to cerebral physiology. Magn Reson Med 17:336-347, 1991. Medline
Similar articles
53. Hardy PA, Henkleman RM: On the transverse relaxation rate enhancement induced by diffusion of spins through
inhomogeneous fields. Magn Reson Med 17:348-356, 1991. Medline Similar articles
54. Boxerman JL, Hamberg LM, Rosen BR, Weisskoff RM: MR contrast due to intravascular magnetic susceptibility
perturbations. Magn Reson Med 34:555-566, 1995. Medline Similar articles
55. Yablonsky DA, Haacke EM: Theory of NMR signal behavior in magnetically inhomogenous tissues: the static dephasing
regime. Magn Reson Med 32:749-763, 1994. Medline Similar articles
56. Weisskoff RM: Basic theoretical models of BOLD signal change. In Bandettini PA, Moonen CTW (eds): Functional MRI.
Berlin: Springer, 1999.
57. Boxerman JL, Bandettini PA, Kwong KK, et al: The intravascular contribution to fMRI signal change: Monte Carlo modeling
and diffusion-weighted studies in vivo. Magn Reson Med 34:4-10, 1995. Medline Similar articles
58. Oja JME, Gillen J, Kauppinen RA, et al: Venous blood effects in spin-echo fMRI of human brain. Magn Reson Med
59. van Zijl PC, Eleff SM, Ulatowski JA, et al: Quantitative assessment of blood flow, blood volume and blood oxygenation
effects in functional magnetic resonance imaging. Nat Med 4:159-167, 1998. Medline Similar articles
12 of 18 12-04-2010 00:26
60. Ogawa S, Menon RS, Tank DW, et al: Functional brain mapping by blood oxygenation level-dependent contrast magnetic
resonance imaging: a comparison of signal characteristics with a biophysical model. Biophys J 64:803-812, 1993. Medline
Similar articles
61. Gati JS, Menon RS, Ugurbil K, Rutt BK: Experimental determination of the BOLD field strength dependence in vessels and
tissue. Magn Reson Med 38:296-302, 1997. Medline Similar articles
62. Mandeville JB, Marota JJ, Kosofsky BE, et al: Dynamic functional imaging of relative cerebral blood volume during rat
forepaw stimulation. Magn Reson Med 39:615-624, 1998. Medline Similar articles
63. Menon RS, Ogawa S, Hu X, et al: BOLD based functional MRI at 4 tesla includes a capillary bed contribution: echo-planar
imaging correlates with previous optical imaging using intrinsic signals. Magn Reson Med 33:453-459, 1995.
64. Hu X, Le TH, Ugurbil K: Evaluation of the early response in fMRI in individual subjects using short stimulus duration. Magn
65. Yacoub E, Hu X: Detection of the early negative response in fMRI at 1.5T. Magn Reson Med 41:1088-1092, 1999.
66. Yacoub E, Le TH, Ugurbil K, Hu X: Further evaluation of the initial negative response in functional magnetic resonance
67. Yacoub E, Hu X: Detection of the early decrease in fMRI signal in the motor area. Magn Reson Med 45:184-190, 2001.
68. Ernst T, Hennig J: Observation of a fast response in functional MR. Magn Reson Med 32:146-149, 1994. Medline
Similar articles
69. Malonek D, Grinvald A: Interactions between electrical activity and cortical microcirculation revealed by imaging
spectroscopy: implications for functional brain mapping. Science 272:551-554, 1996. Medline Similar articles
70. Vanzetta I, Grinvald A: Evidence and lack of evidence for the initial dip in the anesthetized rat: implications for human
functional brain imaging. Neuroimage 13:959-967, 2001. Medline Similar articles
71. Mayhew J, Johnston D, Martindale J, et al: Increased oxygen consumption following activation of brain: theoretical footnotes
using spectroscopic data from barrel cortex. Neuroimage 13:975-987, 2001. Medline Similar articles
72. Lindauer U, Royl G, Leithner C, et al: No evidence for early decrease in blood oxygenation in rat whisker cortex in response
to functional activation. Neuroimage 13:988-1001, 2001. Medline Similar articles
73. Devor A, Dunn AK, Andermann ML, et al: Coupling of total hemoglobin concentration, oxygenation, and neural activity in rat
somatosensory cortex. Neuron 39:353-359, 2003. Medline Similar articles
74. Friston KJ, et al: Statistical parametric maps in functional imaging: A general linear approach. Hum Brain Mapping
2:189-210, 1995.
75. D'Esposito M, Deouell LY, Gazzaley A: Alterations in the BOLD fMRI signal with ageing and disease: a challenge for
neuroimaging. Nat Rev Neurosci 4:863-872, 2003. Medline Similar articles
76. Bandettini PA, Jesmanowicz A, Wong EC, et al: Processing strategies for time-course data sets in functional MRI of the
human brain. Magn Reson Med 30:161-173, 1993. Medline Similar articles
77. Jezzard P: Physiological noise: Strategies for correction. In Moonen CTW, Bandettini PA (eds): Functional MRI. Berlin:
Springer-Verlag, 1999, pp 173-181.
78. Friston KJ, Jezzard P, Turner R: Analysis of functional MRI time-series. Hum Brain Mapping 1:153-171, 1994.
79. Buckner RL, Bandettini PA, O'Craven KM, et al: Detection of cortical activation during averaged single trials of a cognitive
task using functional magnetic resonance imaging. Proc Natl Acad Sci USA 93:14878-14883, 1996. Medline
Similar articles
80. Buckner RL, Goodman J, Burock M, et al: Functional-anatomic correlates of object priming in humans revealed by rapid
presentation event-related fMRI. Neuron 20:285-296, 1998. Medline Similar articles
81. Burock MA, Buckner RL, Woldorff MG, et al: Randomized event-related experimental designs allow for extremely rapid
presentation rates using functional MRI. NeuroReport 9:3735-3739, 1998. Medline Similar articles
82. Dale AM: Optimal experimental design for event-related fMRI. Hum Brain Mapping 8:109-114, 1999.
83. Dale AM, Buckner RL: Selective averaging of rapidly presented individual trials using fMRI. Hum Brain Mapping 5:329-340,
1997.
84. Friston KJ, Fletcher P, Josephs O, et al: Event-related fMRI: characterizing differential responses. Neuroimage 7:30-40,
85. Friston KJ, Josephs O, Rees G, Turner R: Non-linear event related responses in fMRI. Magn Reson Med 39:41-52, 1998.
86. Friston KJ, Zarahn E, Josephs O, et al: Stochastic designs in event-related fMRI. Neuroimage 10:607-619, 1999.
87. Josephs O, Turner R, Friston KJ: Event related fMRI. Hum Brain Mapping 5:243-248, 1997.
88. Zarahn E, Aguirre G, D'Esposito M: A trial based experimental design for fMRI. Neuroimage 5:179-197, 1997. Medline
Similar articles
89. Liu TT, Frank LR, Wong EC, Buxton RB: Detection power, estimation efficiency, and predictability in event-related fMRI.
Neuroimage 13:759-773, 2001. Medline Similar articles
13 of 18 12-04-2010 00:26
90. Buracas GT, Boynton GM: Efficient design of event-related fMRI experiments using M-sequences. Neuroimage
91. Liu TT, Frank LR: Efficiency, power, and entropy in event-related FMRI with multiple trial types. Part I: theory. Neuroimage
92. Liu TT: Efficiency, power, and entropy in event-related fMRI with multiple trial types. Part II: design of experiments.
93. Van Kylen J, Jesmanowicz A, Wong EC, Hyde JS: Material influenced susceptibility induced phase shifts. Presented at the
third meeting of the International Society for Magnetic Resonance in Medicine: p 762, May 1995, Nice, France.
page 285
page 286
94. Tyszka JM, Mamelak AN: Quantification of B0 homogeneity variation with head pitch by registered three-dimensional field
mapping. J Magn Reson 159:213-218, 2002. Medline Similar articles
95. Wilson JL, Jenkinson M, Jezzard P: Protocol to determine the optimal intraoral passive shim for minimisation of
susceptibility artifact in human inferior frontal cortex. Neuroimage 19:1802-1811, 2003. Medline Similar articles
96. Wilson JL, Jezzard P: Utilization of an intra-oral diamagnetic passive shim in functional MRI of the inferior frontal cortex.
97. Jezzard P, Balaban RS: Correction for geometric distortion in echo planar images from B 0 field distortions. Magn Reson
98. Reber PJ, Wong EC, Buxton RB, Frank LR: Correction of off resonance-related distortion in echo-planar imaging using
EPI-based field maps. Magn Reson Med 39:328-330, 1998. Medline Similar articles
99. Woods RP, Grafton ST, Holmes CJ, et al: Automated image registration: I. General methods and intrasubject, intramodality
validation. J Comput Assist Tomogr 22:139-152, 1998. Medline Similar articles
100. Woods RP, Grafton ST, Watson JD, et al: Automated image registration: II. Intersubject validation of linear and nonlinear
models. J Comput Assist Tomogr 22:153-165, 1998. Medline Similar articles
101. Weisskoff RM, Baker J, Belliveau J, et al: Paper presented at the 12th meeting of the Society for Magnetic Resonance in
Medicine: p 7, May 1993, New York.
102. Jezzard P, Lebihan D, Cuenod C, et al: Paper presented at the 12th meeting of the Society for Magnetic Resonance in
Medicine: p 1392, May 1993, New York.
103. Purdon PL, Weisskoff RM: Effect of temporal autocorrelation due to physiological noise and stimulus paradigm on
voxel-level false-positive rates in fMRI. Hum Brain Mapping 6:239-249, 1998.
104. Peltier SJ, Noll DC: Systematic noise compensation for simultaneous multislice acquisition using rosette trajectories
(SMART). Magn Reson Med 41:1073-1076, 1999.
105. Stenger VA, Peltier S, Boada FE, Noll DC: 3D spiral cardiac/respiratory ordered fMRI data acquisition at 3 Tesla. Magn
Reson Med 41:983-991, 1999.
107. Hu X, Kim S-G: Reduction of signal fluctuations in functional MRI using navigator echoes. Magn Reson Med 31:495-503,
108. Hu X, Le TH, Parrish T, Erhard P: Retrospective estimation and correction of physiological fluctuation in functional MRI.
109. Britt RH, Rossi GT: Quantitative analysis of methods for reducing physiological brain pulsations. J Neurosci Methods
110. Obrig H, Neufang M, Wenzel R, et al: Spontaneous low frequency oscillations of cerebral hemodynamics and metabolism
in human adults. Neuroimage 12:623-639, 2000. Medline Similar articles
111. Smith AM, Lewis BK, Ruttimann UE, et al: Investigation of low frequency drift in fMRI signal. Neuroimage 9:526-533,
112. Guimaraes AR, Melcher JR, Talavage TM, et al: Imaging subcortical auditory activity in humans. Hum Brain Mapping
6:33-41, 1998.
113. Biswal B, DeYoe AE, Hyde JS: Reduction of physiological fluctuations in fMRI using digital filters. Magn Reson Med
114. Frank LR, Buxton RB, Wong EC: Estimation of respiration-induced noise fluctuations from undersampled multislice fMRI
data. Magn Reson Med 45:635-644, 2001. Medline Similar articles
115. Mitra PP, Pesaran B: Analysis of dynamic brain imaging data. Biophys J 76:691-708, 1999. Medline
Similar articles
116. Glover GH, Li TQ, Ress D: Image-based method for retrospective correction of physiological motion effects in fMRI:
RETROICOR. Magn Reson Med 44:162-167, 2000. Medline Similar articles
117. Andersen AH, Gash DM, Avison MJ: Principal component analysis of the dynamic response measured by fMRI: a
generalized linear systems framework. Magn Reson Imaging 17:795-815, 1999. Medline Similar articles
14 of 18 12-04-2010 00:26
118. McKeown MJ, Hansen LK, Sejnowsk TJ: Independent component analysis of functional MRI: what is signal and what is
noise? Curr Opin Neurobiol 13:620-629, 2003. Medline Similar articles
119. Weisskoff RM: Simple measurement of scanner stability for functional NMR imaging of activation in the brain. Magn
120. Frackowiak RSJ, Lenzi GL, Jones T, Heather JD: Quantitative measurement of regional cerebral blood flow and oxygen
metabolism in man using 15O and positron emission tomography: theory, procedure, and normal values. J Comput Assist
Tomogr 4:727-736, 1980.
121. Detre JA, Leigh JS, Williams DS, Koretsky AP: Perfusion imaging. Magn Reson Med 23:37-45, 1992. Medline
Similar articles
122. Wong EC, Buxton RB, Frank LR: Implementation of quantitative perfusion imaging techniques for functional brain
mapping using pulsed arterial spin labeling. NMR Biomed 10:237-249, 1997. Medline Similar articles
123. Lee SP, Silva AC, Kim SG: Comparison of diffusion-weighted high-resolution CBF and spin-echo BOLD fMRI at 9.4 T.
Magn Reson Med 47:736-741, 2002.
124. Buxton RB, Frank LR, Wong EC, et al: A general kinetic model for quantitative perfusion imaging with arterial spin
labeling. Magn Reson Med 40:383-396, 1998. Medline Similar articles
125. Buxton RB: Introduction to Functional Magnetic Resonance Imaging: Principles and Techniques. Cambridge: Cambridge
University Press, 2002.
126. Hoge R, et al: Flow-metabolism regulation during brain activation and respiratory manipulations. In Tomita M, Kanno I,
Hamel E (eds): Brain Activation and CBF Control. Amsterdam: Elsevier Science BV, 2002, pp 33-38.
127. Belliveau JW, Kennedy DN, Mackinstry RC, et al: Functional mapping of the human visual cortex by magnetic resonance
imaging. Science 254:716-719, 1991. Medline Similar articles
128. Lia TQ, Guang Chen Z, Ostergaard L, et al: Quantification of cerebral blood flow by bolus tracking and artery spin tagging
methods. Magn Reson Imaging 18:503-512, 2000. Medline Similar articles
129. Dubowitz DJ, Bernheim KA, Chen DY, et al: Enhancing fMRI contrast in awake-behaving primates using intravascular
magnetite dextran nanoparticles. Neuroreport 12:2335-2340, 2001. Medline Similar articles
130. Leite FP, Tsao D, Vanduffel W, et al: Repeated fMRI using iron oxide contrast agent in awake, behaving macaques at 3
Tesla. Neuroimage 16:283-294, 2002.
131. Vanduffel W, Fize D, Mandeville JB, et al: Visual motion processing investigated using contrast agent-enhanced fMRI in
awake behaving monkeys. Neuron 32:565-577, 2001. Medline Similar articles
132. Liu TT, Luh W-M, Wong EC, et al: A method for dynamic measurement of blood volume with compensation for T2
changes. In Proceedings of International Society for Magnetic Resonance in Medicine, May 2000, Denver.
133. Wong EC, Buxton RB, Frank LR: A method for dynamic imaging of blood volume. In Fifth Meeting, International Society for
Magnetic Resonance in Medicine, May 1997, Vancouver.
134. Lu H, Golay X, Pekar JJ, et al: Functional magnetic resonance imaging based on changes in vascular space occupancy.
1 2
135. Wong EC, Liu TT, Luh WM, et al: T and T selective method for improved SNR in CSF-attenuated imaging: (T2FLAIR).
136. Frahm J, Kruger G, Merboldt KD, et al: Dynamic uncoupling and recoupling of perfusion and oxidative metabolism during
focal activation in man. Magn Reson Med 35:143-148, 1996. Medline Similar articles
137. Prichard J, Rothman D, Novotny E, et al: Lactate rise detected by 1H NMR in human visual cortex during physiologic
stimulation. Proc Natl Acad Sci USA 88:5829-5831, 1991.
138. Hu Y, Wilson GS: A temporary local energy pool coupled to neuronal activity: fluctuations of extracellular lactate levels in
rat brain monitored with rapid-response enzyme-based sensor. J Neurochem 69:1484-1490, 1997. Medline
Similar articles
139. Kuhr WG, Korf J: Extracellular lactic acid as an indicator of brain metabolism: continuous on-line measurement in
conscious, freely moving rats with intrastriatal dialysis. J Cereb Blood Flow Metab 8:130-137, 1988. Medline
Similar articles
140. Madsen PL, Cruz NF, Sokoloff L, Dienel GA: Cerebral oxygen/glucose ration is low during sensory stimulation and rises
above normal during recovery: excess glucose consumption during stimulation is not accounted for by lactate efflux from or
accumulation in brain tissue. J Cereb Blood Flow Metab 19:393-400, 1999. Medline Similar articles
141. Kuwabara T, Watanabe H, Tsuji S, Yuasa T: Lactate rise in the basal ganglia accompanying finger movements: a
localized 1H-MRS study. Brain Res 670:326-328, 1995.
142. Richards TL, Gates GA, Gardner JC, et al: Functional MR spectroscopy of the auditory cortex in healthy subjects and
patients with sudden hearing loss. Am J Neuroradiol 18:611-620, 1997. Medline Similar articles
143. Richards TL, Duger SR, Corina D, et al: Dyslexic children have abnormal brain lactate response to reading-related
language tasks. Am J Neuroradiol 20:1393-1398, 1999. Medline Similar articles
144. Richards TL, Corina D, Serafini S, et al: Effects of a phonologically driven treatment for dyslexia on lactate levels
measured by proton MR spectroscopic imaging. Am J Neuroradiol 21:916-922, 2000. Medline Similar articles
15 of 18 12-04-2010 00:26
145. Mangia S, Garreffa G, Bianciardi M, et al: The aerobic brain: lactate decrease at the onset of neural activity.
Neuroscience 118:7-10, 2003. Medline Similar articles
146. Janz C, Speck O, Hennig J: Time-resolved measurements of brain activation after a short visual stimulus: new results on
the physiological mechanisms of the cortical response. NMR Biomed 10:222-229, 1997. Medline Similar articles
147. Serafini S, Steury K, Richards T, et al: Comparison of fMRI and PEPSI during language processing in children. Magn
148. Le Bihan D: Looking into the functional architecture of the brain with diffusion MRI. Nat Rev Neurosci 4:469-480, 2003.
149. Moseley ME, Mintorovitch J, Cohen Y, et al: Early detection of ischemic injury: comparison of spectroscopy, diffusion-,
T2-, and magnetic susceptibility-weighted MRI in cats. Acta Neurochir Suppl (Wien) 51:207-209, 1990.
150. Hossmann KA, Hoehn-Berlage M: Diffusion and perfusion MR imaging of cerebral ischemia. Cerebrovasc Brain Metab
Rev 7:187-217, 1995. Medline Similar articles
151. Sotak CH: The role of diffusion tensor imaging in the evaluation of ischemic brain injury-a review. NMR Biomed
152. Mori S, van Zijl PC: Fiber tracking: principles and strategies-a technical review. NMR Biomed 15:468-480, 2002.
153. Zhong J, Petroff OA, Prichard JW, Gore JC: Changes in water diffusion and relaxation properties of rat cerebrum during
status epilepticus. Magn Reson Med 30:241-246, 1993. Medline Similar articles
154. Cohen LB: Changes in neuron structure during action potential propagation and synaptic transmission. Physiol Rev
155. Andrew RD, MacVicar BA: Imaging cell volume changes and neuronal excitation in the hippocampal slice. Neuroscience
page 286
page 287
156. Darquie A, Poline JB, Poupon C, et al: Transient decrease in water diffusion observed in human occipital cortex during
visual stimulation. Proc Natl Acad Sci USA 98:9391-9395, 2001. Medline Similar articles
157. Saleem KS, Pauls JM, Augath M, et al: Magnetic resonance imaging of neuronal connections in the macaque monkey.
Neuron 34:685-700, 2002. Medline Similar articles
158. Pautler RG, Koretsky AP: Tracing odor-induced activation in the olfactory bulbs of mice using manganese-enhanced
magnetic resonance imaging. Neuroimage 16:441-448, 2002. Medline Similar articles
159. Pautler RG, Mongeau R, Jacobs RE: In vivo trans-synaptic tract tracing from the murine striatum and amygdala utilizing
manganese enhanced MRI (MEMRI). Magn Reson Med 50:33-39, 2003. Medline Similar articles
160. Pautler RG, Silva AC, Koretsky AP: In vivo neuronal tract tracing using manganese-enhanced magnetic resonance
161. Van der Linden A, Verhoye M, van Meir V, et al: In vivo manganese-enhanced magnetic resonance imaging reveals
connections and functional properties of the songbird vocal control system. Neuroscience 112:467-474, 2002. Medline
Similar articles
162. Kida I, Hyder F, Behar KL: Inhibition of voltage-dependent sodium channels suppresses the functional magnetic
resonance imaging response to forepaw somatosensory activation in the rodent. J Cereb Blood Flow Metab 21:585-591,
163. Gruetter R: In vivo 13C NMR studies of compartmentalized cerebral carbohydrate metabolism. Neurochem Int
41:143-154, 2002.
164. Rothman DL: Studies of metabolic compartmentation and glucose transport using in vivo MRS. NMR Biomed
165. Duelli R, Kuschinsky W: Brain glucose transporters: relationship to local energy demand. News Physiol Sci 16:71-76,
166. Harder DR, Zhang C, Gebremedhin D: Astrocytes function in matching blood flow to metabolic activity. News Physiol Sci
167. Zheng D, LaMantia AS, Purves D: Specialized vascularization of the primate visual cortex. J Neurosci 11:2622-2629,
167a. Raichle ME, MacLeod AM, Snyder AZ, et al: A default mode of brain function. Proc Natl Acad Sci USA 98:676-682,
168. Fox PT, Raichle ME, Mintun MA, Dence C: Nonoxidative glucose consumption during focal physiologic neural activity.
Science 241:462-464, 1988. Medline Similar articles
169. Sappey-Marinier D, Calabrese G, Fein G, et al: Effect of photic stimulation on human visual cortex lactate and phosphates
using 1H and 31P magnetic resonance spectroscopy. J Cereb Blood Flow Metab 12:584-592, 1992.
170. Shulman RG, Hyder F, Rothman DL: Cerebral energetics and the glycogen shunt: neurochemical basis of functional
imaging. Proc Natl Acad Sci USA 98:6417-6422, 2001. Medline Similar articles
16 of 18 12-04-2010 00:26
171. Hertz L, Fillenz M: Does the "mystery of the extra glucose" during CNS activation reflect glutamate synthesis? Neurochem
Int 34:71-75, 1999. Medline Similar articles
172. Chih CP, Roberts EL Jr: Energy substrates for neurons during neural activity: a critical review of the astrocyte-neuron
lactate shuttle hypothesis. J Cereb Blood Flow Metab 23:1263-1281, 2003. Medline Similar articles
173. Rees G, Friston K, Koch C: A direct quantitative relationship between the functional properties of human and macaque
V5. Nat Neurosci 3:716-723, 2000. Medline Similar articles
174. Heeger DJ, Huk AC, Geisler WS, Albrecht DG: Spikes versus BOLD: what does neuroimaging tell us about neuronal
activity? Nat Neurosci 3:631-633, 2000. Medline Similar articles
175. Lauritzen M, Gold L: Brain function and neurophysiological correlates of signals used in functional neuroimaging. J
Neurosci 23:3972-3980, 2003. Medline Similar articles
176. Logothetis NK, Pauls J, Augath M, et al: Neurophysiological investigation of the basis of the fMRI signal. Nature
177. Logothetis NK: The underpinnings of the BOLD functional magnetic resonance imaging signal. J Neurosci 23:3963-3971,
178. Caesar K, Thomsen K, Lauritzen M: Dissociation of spikes, synaptic activity, and activity-dependent increments in rat
cerebellar blood flow by tonic synaptic inhibition. Proc Natl Acad Sci USA 100:16000-16005, 2003. Medline
Similar articles
179. Arthurs OJ, Williams EJ, Carpenter TA, et al: Linear coupling between functional magnetic resonance imaging and
evoked potential amplitude in human somatosensory cortex. Neuroscience 101:803-806, 2000. Medline
Similar articles
180. Ogawa S, Lee TM, Stepnoski R, et al: An approach to probe some neural systems interaction by functional MRI at neural
time scale down to milliseconds. Proc Natl Acad Sci USA 97:11026-11031, 2000. Medline Similar articles
181. Huettel SA, McKeown MJ, Song AW, et al: Linking hemodynamic and electrophysiological measures of brain activity:
evidence from functional MRI and intracranial field potentials. Cereb Cortex 14:165-173, 2004. Medline Similar articles
182. Goldman RI, Stern JM, Engel J, Cohen MS: Simultaneous EEG and fMRI of the alpha rhythm. Neuroreport 13:2487-2492,
183. Beaulieu C, Colonnier M: A comparison of the number of neurons in individual laminae of cortical areas 17, 18 and
posteromedial suprasylvian (PMLS) area in the cat. Brain Res 339:166-170, 1985.
184. Magistretti PJ, Pellerin L: Cellular bases of brain energy metabolism and their relevance to functional brain imaging:
evidence for a primary role of astrocytes. Cerebral Cortex 6:50-61, 1996. Medline Similar articles
185. Uludag K, Dubowitz DJ, Yoder EJ, et al: Coupling of cerebral blood flow and oxygen metabolism during physiological
activation and deactivation measured with fMRI. Neuroimage 23:148-155, 2004. Medline Similar articles
186. Ackermann RF, Finch DM, Babb TL, Engel J: Increased glucose metabolism during long-duration recurrent inhibition of
hippocampal pyramidal cells. J Neurosci 4:251-264, 1984. Medline Similar articles
187. Mathiesen C, Caesar K, Lauritzen M, et al: Modification of activity-dependent increases of cerebral blood flow by
excitatory synaptic activity and spikes in rat cerebellar cortex. J Physiol 512:555-566, 1998. Medline Similar articles
188. Chatton JY, Pellerin L, Magistretti PJ: GABA uptake into astrocytes is not associated with significant metabolic cost:
implications for brain imaging of inhibitory transmission. Proc Natl Acad Sci USA 100:12456-12461, 2003. Medline
Similar articles
189. Waldvogel D, van Gelderen P, Muellbacher W, et al: The relative metabolic demand of inhibition and excitation. Nature
190. Allison JD, Meador KJ, Loring DW, et al: Functional MRI cerebral activation and deactivation during finger movement.
Neurology 54:135-142, 2000. Medline Similar articles
191. Wobst P, Villringer A: Inhibition. In Tomita M, Kanno I, Hamel E (eds): Brain Activation and CBF Control. Amsterdam:
Elsevier Science, 2002, pp 23-32.
192. Jones M, Berwick J, Johnston D, Mayhew J: Concurrent optical imaging spectroscopy and laser-Doppler flowmetry: the
relationship between blood flow, oxygenation, and volume in rodent barrel cortex. Neuroimage 13:1002-1015, 2001.
193. Thompson JK, Peterson MR, Freeman RD: Single-neuron activity and tissue oxygenation in the cerebral cortex. Science
194. Boynton GM, Engel SA, Glover GH, Heeger DJ: Linear systems analysis of functional magnetic resonance imaging in
human V1. J Neurosci 16:4207-4221, 1996. Medline Similar articles
195. Robson MW, Dorosz JL, Gore JC: Measurements of the temporal fMRI response of the human auditory cortex to trains of
tones. Neuroimage 7:185-198, 1998. Medline Similar articles
196. Glover GH: Deconvolution of impulse response in event-related fMRI. Neuroimage 9:416-429, 1999. Medline
Similar articles
197. Vasquez AL, Noll DC: Nonlinear aspects of the BOLD response in functional MRI. Neuroimage 7:108-118, 1998.
198. Birn RM, Saad ZS, Bandettini PA: Spatial heterogeneity of the nonlinear dynamics in the FMRI BOLD response.
17 of 18 12-04-2010 00:26
Neuroimage 14: 817-826, 2001.

199. Miller KL, Luh WM, Liu TT, et al: Nonlinear temporal dynamics of the cerebral blood flow response. Hum Brain Mapping
13:1-12, 2001.
200. Friston KJ, Mechelli A, Turner R, Price CJ: Nonlinear responses in fMRI: the Balloon model, Volterra kernels, and other
hemodynamics. Neuroimage 12:466-477, 2000. Medline Similar articles
201. Behzadi Y, Restom K, Liu TT: Modeling the effect of baseline arteriolar compliance on BOLD dynamics. Paper presented
at the 12th Scientific Meeting of the International Society for Magnetic Resonance in Medicine, May 2004, Kyoto, Japan.
202. Baird AE, Warach S: Magnetic resonance imaging of acute stroke. J Cereb Blood Flow Metab 18:583-609, 1998.
203. Obata T, Liu TT, Miller KL, et al: Discrepancies between BOLD and flow dynamics in primary and supplementary motor
areas: application of the balloon model to the interpretation of BOLD transients. Neuroimage 21:144-153, 2004. Medline
Similar articles
204. Stark CE, Squire LR: When zero is not zero: the problem of ambiguous baseline conditions in fMRI. Proc Natl Acad Sci
USA 98:12760-12766, 2001. Medline Similar articles
205. Biswal B, Yetkin FZ, Haughton VM, Hyde JS: Functional connectivity in the motor cortex of resting human brain using echo
planar MRI. Magn Reson Med 34:537-541, 1995. Medline Similar articles
206. Chaigneau E, Oheim M, Audinat E, Charpak S: Two-photon imaging of capillary blood flow in olfactory bulb glomeruli.
Proc Natl Acad Sci USA 100:13081-13086, 2003. Medline Similar articles
207. Iadecola C: Neurovascular regulation in the normal brain and in Alzheimer's disease. Nat Rev Neurosci 5:347-360, 2004.
208. Harrison RV, Harel N, Pansar J, Mount RJ: Blood capillary distribution correlates with hemodynamic-based functional
imaging in cerebral cortex. Cereb Cortex 12:225-233, 2002. Medline Similar articles
209. Kim DS, Duong TQ, Kim SG: High-resolution mapping of iso-orientation columns by fMRI. Nat Neurosci 3:164-169, 2000.
210. Logothetis N: Can current fMRI techniques reveal the micro-architecture of cortex? [letter; comment]. Nat Neurosci
211. Disbrow EA, Slutsky DA, Roberts TP, Krubitzer LA: Functional MRI at 1.5 tesla: a comparison of the blood oxygenation
level-dependent signal and electrophysiology. Proc Natl Acad Sci USA 97:9718-9723, 2000.
212. Kim DS, Ronen I, Olman C, et al: Spatial relationship between neuronal activity and BOLD functional MRI. Neuroimage
213. Duong TQ, Yacoub E, Adriany G, et al: Microvascular BOLD contribution at 4 and 7 T in the human brain: gradient-echo
and spin-echo fMRI with suppression of blood effects. Magn Reson Med 49:1019-1027, 2003.
© 2010 Elsevier
18 of 18 12-04-2010 00:26
IFFUSION EIGHTED AGNETIC ESONANCE MAGING

Roland Bammer
Chunlei Liu
Julia Po
Michael E. Moseley
SYNOPSIS
During the past decade, diffusion-weighted imaging (DWI), which provides quantitative measures of the
molecular motion of water in three dimensional space, has shown great promise in diagnosing acute
stroke. Promising reports are appearing regarding the application of DWI to other regions in the human
body, such as the spine, abdomen, and breast. This chapter provides an introduction into the basics of
diffusion-weighted magnetic resonance imaging (MRI). The potential of various MR sequences in
concert with diffusion preparation are discussed with respect to acquisition speed, spatial resolution,
and sensitivity to bulk physiologic motion. More advanced diffusion measurement techniques, such as
motion correction and navigation and eddy current effects and corrections, as well as the impact of
parallel imaging on DWI, are also discussed.
© 2010 Elsevier
1 of 1 12-04-2010 00:27
INTRODUCTION
Within the last decade, the capability of DWI has undergone continuous improvement to probe random
microscopic motion of water protons on a per pixel basis using pulsed magnetic field gradients. Such
DWI techniques have advanced far beyond the experimental arena towards routine clinical applications
to detect acute cerebral ischemia and are also the subject of research to investigate other diseases of
the CNS, such as multiple sclerosis, dyslexia, schizophrenia, or trauma.1-6 Recently, several reports on
the use of DWI for other areas of the human body have been published, some with very encouraging
results.
Diffusion-weighted imaging in acute stroke has become popular since changes in proton self-diffusion
are an early indicator of alterations in cellular homeostasis.1 Early detection of these alterations can
dramatically impact treatment decisions and the therapeutic outcome for stroke victims. 2 This is mainly
because DWI is able to detect acute ischemic lesions within the "window of opportunity" for advanced
stroke therapies and promises to help reduce the extent and severity of ischemic damage in acute
stroke victims.
page 288
page 289
Generally, pathologic processes tend to alter the magnitude of structural organization either by
destruction or regeneration of membranous elements or by a change in cellularity (e.g., scarring,
inflammatory or neoplastic infiltration). Shifts in the number of water protons between tissue
compartments due to changes in permeability, osmolarity, or active transportation may occur in
parallel. These aspects will also have an impact on the extent of proton mobility, or diffusivity, which
can be followed by DWI. Therefore, the probing of diffusional changes in human tissue opens a totally
new approach in describing related morphologic damage that may be associated with focal or diffuse
abnormalities seen on conventional MRI.4,5 By following the cascade of changes in diffusivity, there
may be a means to better understand the different processes of tissue destruction and repair
mechanisms that can take place in several pathological processes.
Besides important clinical information, DWI and diffusion tensor imaging (DTI)7,8 have also made
valuable contributions in basic neuroscience and improved the understanding of certain
pathophysiologic processes. For the first time in vivo, DTI allows one to investigate the orientation of
major white matter fiber tracts and gross anatomic brain connectivity9 in a repeatable and
nondestructive manner, and offers, to some extent, a quantitative measure for the loss of structural
coherence and eventually the destruction of myelin sheaths.
Here, the anisotropic diffusion behavior can be associated with the orientation of fibers. Within fibrous
tissues, such as the white matter of the brain, proton self-diffusion is directional10: water molecules
diffuse at least twice as fast in the direction parallel to fibers relative to the perpendicular direction.
This information can be utilized to study the orientation of anisotropic tissues by traversing a continuous
7,8
path of greatest diffusivity from an initial set of seedpoints. In general, these techniques are known
9
as MR fiber tracking or tractography and will be addressed elsewhere in more detail. The ability to
outline axonal fiber bundles in neuronal networks is important in order to understand normal and
pathological processes affecting brain functions. For example, complex cognitive and motor-oriented
processes that involve different functional areas of the brain are mediated by such neural networks.
The study of neural association networks is, therefore, essential in understanding how different
functionally active areas interact, and also how the human brain reacts to trauma or pathology. In
11
concert with functional magnetic resonance imaging (fMRI), the availability of a noninvasive technique
that outlines fascicles could enhance the understanding of the spatiotemporal interaction of normal
brain function and adaptive processes such as brain plasticity. Finally, DTI may play an important role
in treatment planning for neurosurgical interventions or dose sculpting in radiation therapy by adding
this technique to the diagnostic battery available to the oncologist.
In the following section we will start with a brief introduction into the basic mechanisms of diffusion and
1 of 2 12-04-2010 00:28
diffusion-weighted MR experiments.
© 2010 Elsevier
2 of 2 12-04-2010 00:28
MATHEMATICAL-PHYSICAL PRINCIPLES OF DIFFUSION-WEIGHTED

IMAGING
Diffusion is the process by which matter is transported from one part of a system to another as a
result of random translational molecular motions driven by internal kinetic energy. Translational diffusion
is the most fundamental form of transport and is responsible for all chemical reactions. If it were
possible to watch individual molecules of dye (effectively this can be done by replacing the molecules
with particles small enough to share the molecular motions but just large enough to be visible under the
microscope), it would be found that the motion of each molecule is random. That is, in a dilute solution
each molecule of dye behaves independently of the others, and each is constantly undergoing collision
with solvent molecules having no preferred direction of motion. The motion of a single molecule (i.e.,
Brownian motion*) can thus be described in terms of the familiar "random walk" picture,12,13 and,
although it is possible to calculate the mean-square distance traveled in a given interval of time, it is not
possible to say in what direction a given molecule will move during that time interval.
By the use of labeled protons instead of dye, one can observe the consequences of the random
movement of these spins (self-diffusion). For a single static spin the cumulative phase shift in the
presence of a magnetic field gradient is given by:
where the first term represents the phase accrual due to the static B0-field, and the second term is due
to the effects of a magnetic field gradient. The phase term of the latter is proportional to the strength
of the field gradient, the duration of the gradient, and the spatial location of the spin. From Equation
10-1 it is apparent that a magnetic field gradient can be used to label the position of a spin by means
of minute differences in the Larmor frequency. This sets the basis for measuring diffusion.
*The phenomenon is named after the English botanist Robert Brown, who in 1827 was the first to establish the fact that the motion
is not the motion of living organisms.
page 289
page 290
Add to lightbox
Figure 10-1 Diffusion-weighted spin-echo sequence. Diffusion-weighting gradients of strength GDiff,
duration δ, and spacing ∆ are applied during each TE/2 period. The diffusion-weighted echo is
sampled at the time t = TE when the spin-echo is formed. The diffusion attenuation is only dependent
on the parameters GDiff, ∆, and δ but does not depend on t1. (GM, readout direction; GP, phase-
encode direction; GS, section select direction; RF, radio frequency; TE, echo time.)
The most common approach to render MRI sensitive to diffusion is to use a spin-echo pulse sequence,
1 of 2 12-04-2010 00:29
in which equal rectangular gradient pulses are played out before and after the 180-degree refocusing
pulse (Fig. 10-1). For a moving individual spin, the degree of phase accrual due to the applied field
gradient is proportional to the spin displacement in the direction of the gradient. At the time when the
spin-echo is formed, the total phase shift of one particular spin is, therefore:
In the absence of motion, the phase shifts due to the two gradient pulses will cancel, but if all spins are
moving coherently in the field then all spins will acquire identical phase according to Equation 10-2.
However, in the case of diffusion, the displacement function x(t) is random and the phase shifts
accumulated by individual nuclei will differ. It is beyond the scope of this chapter to show that all these
random phase shifts of individual spins interfere destructively and lead ultimately to a signal attenuation
14
without accumulating a net phase shift. In this context, the corresponding echo attenuation can be
expressed as:
and depends upon the apparent diffusion coefficient (D), which is sometimes abbreviated as ADC, and
the diffusion-sensitizing factor or b-value, which can be calculated for a spin-echo sequence with
rectangular diffusion-encoding gradients as follows15:
Here, δ is the duration of one diffusion-encoding gradient lobe, ∆ is the time interval between the
leading edges of the gradient lobes, G is the strength of the diffusion encoding magnetic field gradient,
and γ is the gyromagnetic ratio. By using the characteristic signal equation of diffusion (Eq. 10-3), the
diffusion coefficient can be calculated from two measurements with different diffusion attenuations (i.e.,
b0 and b1) according to:
© 2010 Elsevier
2 of 2 12-04-2010 00:29
BASIC PULSE SEQUENCES FOR DIFFUSION-WEIGHTED IMAGING

Spin-Echo- and Stimulated Echo-Based Diffusion-Weighted Imaging
In the previous section, the basic principles of diffusion were explained by means of a diffusion-
weighted spin-echo sequence. However, stimulated echoes can also be used to achieve diffusion-
weighted contrast. Merboldt et al16 placed one diffusion-encoding gradient between the first and the
second 90-degree pulse and another one after the third 90-degree radiofrequency (RF) pulse of a
stimulated echo sequence (Fig. 10-2) to generate diffusion-weighted images. Notice that a stimulated
echo can be formed by three RF pulses (these RF pulses need not necessarily be 90-degree pulses to
form a stimulated echo): spins are tipped into the transverse plane by the first RF pulse and lose their
phase coherence. The second RF pulse tips half of the transverse magnetization (i.e., all vector
components perpendicular to the phase of the second RF pulse) from the transverse plane along the
z-axis. While the magnetization is "stored" along the z-axis (mixing time, TM), it is affected solely by
the much slower T1 relaxation process. Ultimately, after the third RF pulse, which tips spins back into
the transverse plane, and after the second precession interval (TE/2), a diffusion-weighted stimulated
echo is formed at time TE + TM.
The diffusion-weighted stimulated echo sequence is of particular interest for tissues with short T2
relaxation times (e.g., liver) and can also be combined with different readout strategies, such as
echo-planar imaging (EPI) or spiral imaging. The diffusion weighting is mainly determined by the
T1-sensitive TM interval and, therefore, allows one to choose short echo times with reasonable
diffusion weighting. However one should keep in mind that stimulated echoes theoretically provide only
half the signal compared with spin-echoes, whereby the corresponding signal equation is:
Steady-State Free Precession Diffusion-Weighted Imaging

page 290
page 291
Add to lightbox
Figure 10-2 Diffusion-weighted stimulated echo sequence. Diffusion-weighting gradients of strength
GDiff are played out after the first and the third RF pulse. The second RF tips half of the spins back
along the z-axis. During the mixing time (TM) these spins are affected only by the much slower T1
relaxation time. During TM, GCrush spoils transverse magnetization away that could interfere with the
diffusion-weighted signal when the stimulated echo is formed. (GM, readout direction; GP, phase-
encode direction; GS, section select direction; RF, radio frequency; TE, echo time.)
1 of 2 12-04-2010 00:29
If a train of equidistant RF pulses with flip angle α and a relaxation time shorter than T2 (TR<T2) is
applied, a condition of steady-state free precession (SSFP) will develop; because of the very short
TR, SSFP imaging allows rapid image formation. SSFP imaging has long been known for its high
sensitivity to flow and diffusion in the presence of magnetic field gradients (Fig. 10-3) and has,
therefore, received some attention from several research groups.13,17-22 In contrast to spin and
stimulated echoes, the signal formation in SSFP is a complex interplay between numerous spin and
23-25
stimulated echoes that may be formed through a multitude of so-called coherence pathways,
limited only by the natural decay times T1 and T2. This complex signal formation renders the
quantification of diffusion very difficult, and the diffusion attenuation (b-factor) may vary from tissue to
tissue since the b-factor for this sequence is also determined by parameters like relaxation times and
B1-uniformity. Unlike in the cases of spin-echo- and stimulated echo-based DWI, one has to deal not
only with "T2-shine-through" effects but also with the fact that b-values are weighted by the underlying
relaxation times and other confounding factors.
© 2010 Elsevier
2 of 2 12-04-2010 00:29
DIFFUSION TENSOR IMAGING

For a better understanding of the remainder of this chapter we will briefly review the concept of
diffusion tensor imaging (DTI). A more detailed description of this matter can be found elsewhere in
this book (see Chapter 11, Diffusion Tensor Imaging).
Add to lightbox
Figure 10-3 One cycle of a diffusion-weighted steady-state free precession (SSFP) sequence. The
ECHO-component is generally very sensitive to diffusion. All imaging gradients are completely
balanced in this example sequence, and the total effective gradient is equal in each cycle to establish
a steady state. To diminish the bulk motion sensitivity, bipolar diffusion gradients can be used to
reduce the first gradient moment and also (spiral) navigator echoes can be implemented. (GM,
readout direction; GP, phase-encode direction; GS, section select direction; RF, radio frequency; TR,
relaxation time.)
In media with anisotropic Gaussian diffusion properties, it has been shown that the displacement front
of a diffusing substance can be modeled as an ellipsoid.12 Hence, in the absence of noise, each
diffusion-weighted measurement reveals, for every voxel during a defined observation interval, the
root-mean-square displacement of the substance from the origin to a point on the ellipsoid surface
along the direction of the diffusion sensitizing gradient.7,8 Acquisition of MR data by using various
gradient orientations samples a set of points on the ellipsoid surface to define its size, shape, and
orientation to within the limits of sampling error. The mathematical construct used to characterize
anisotropic Gaussian diffusion is a second-order diffusion tensor (D).7,8,12
page 291
page 292
To determine all six unknown elements of the diffusion tensor, independent measurements with
diffusion gradients are applied sequentially along at least six noncollinear directions. For each gradient
direction, a diffusion-weighted image M(bk) can be obtained, where the measured signal intensity can
be generally described by the characteristic signal equation:
1 of 3 12-04-2010 00:30
with
T
Here, Gk(t) = [Gk,x(t), Gk,y(t), Gk,z(t)] is a vector containing the waveforms of all gradients applied
along each principle gradient axis for the measurement along the k-th direction. Although the gradient
directions have to be noncollinear, there is no definite convention regarding their absolute orientations.
The quantities most inherent to the diffusion tensor are the three principal diffusivities or eigenvalues of
D (λ1, λ2, and λ3) which are the principal diffusion coefficients measured along the three intrinsic
coordinate directions that constitute the local fiber frame of reference in each voxel. Each of these
eigenvalues can be linked to a principal direction-the so-called eigenvector-(v1, v2, v3) that is also
intrinsic to the tissue and whose orientation relative to the gradient frame of reference can vary from
voxel to voxel. By sorting their magnitude (i.e., λ1>λ2>λ3), maps for each of the three eigenvalues can
be generated. A second-rank tensor (3×3 matrix) can be diagonalized, such that only the three
eigenvalues remain along the diagonal:
where v1, v2, and v3 are the so-called eigenvectors of the tensor D.
Different scalar measures to characterize the properties of the diffusion tensor have been proposed;
their significance lies in being rotationally invariant and independent of the sorting of the eigenvalues.26
Rotational invariance states that such measures are independent of the orientation of the anisotropic
structure with respect to the frame of reference. The isotropic component of the tensor is
characterized by the so-called "trace of the diffusion tensor" TR(D):
In other words, the trace is proportional to the orientation-averaged apparent diffusivity. In acute
stroke imaging, for example, the trace of the diffusion tensor is generally used to increase lesion
conspicuity by removing unwanted anisotropic structures that may arise in white matter tracts. Besides
this scalar measure of isotropic diffusion, several scalar measures exist to characterize the amount of
diffusion anisotropy and express the degree to which the diffusion tensor deviates from isotropy. The
most frequently used measure is the fractional anisotropy FA(D):
Although diffusion anisotropy exists in unmyelinated nerves (as shown in studies of garfish olfactory
27 28
nerves and neonates ), it is widely assumed that the myelin sheath surrounding a nerve fiber acts as
2 of 3 12-04-2010 00:30
the main barrier to water diffusion. Recently, a new method has been introduced that allows one to
trace different fiber bundles in a noninvasive fashion and is called diffusion tensor MRI (DT-MRI)
tractography or fibertracking. Fibertracking relies on the assumption that the eigenvector associated
with the largest eigenvalue is aligned with the direction of the fiber bundle due to the highly directional
diffusivity in nerve fibers. This new technique opens a completely new arena of examinations since it
allows one to trace different fiber tracts multiple times and from various locations in a nondestructive
fashion.
© 2010 Elsevier
3 of 3 12-04-2010 00:30
DIFFUSION-WEIGHTED IMAGING BEYOND TENSORS

It has long been recognized that the Gaussian model for diffusion can be inappropriate within a
complex substrate such as human tissue. One way in which this model fails is in the presence of
multiple fiber directions within a single imaging voxel. Therefore, applying a single fiber diffusion tensor
model (by using a diffusion ellipsoid) has to be reconsidered because one is trying to fit the
characteristic equation of an ellipsoid to data that will not support this kind of model. A novel approach
to this problem is to measure the diffusion coefficient at high angular resolution29 in order to more
accurately detect variations in diffusion along different directions, and to decompose the resulting
three-dimensional ADC representation into a set of orthogonal three-dimensional functions, known as
30
spherical harmonics. The resulting shape (i.e., diffusion coefficient surface) from such a
measurement is indicative of the underlying fiber structure.
page 292
page 293
Multiexponential and Multicompartmental Diffusion

It has been already mentioned that the physical basis of clinical DWI is the random thermal motion of
tissue water. Conventionally, this motional distribution profile is assumed to be Gaussian. As a result,
the echo-attenuation curve can be modeled as an exponentially decaying function of the b-value. In
either the isotropic or anisotropic case, the signal characterized by Equations 10-3 and 10-4 is a single
exponential function. In biological tissues, molecular diffusion may be hindered by its interaction with
barriers, such as cell membranes and organelles. This interaction influences the distribution profile of
the molecular displacement, and, therefore, affects the echo-attenuation curve. Here, it is worthwhile
to point out the difference between anisotropic diffusion and restricted or hindered diffusion. With the
diffusion observation times currently available on clinical scanners, one sees hindered diffusion, which
is (not entirely correctly) interpreted as anisotropic diffusion tensor. With diminishing observation times,
diffusion in biological tissue becomes more and more isotropic, and hence in this substrate diffusion
anisotropy is a function of the diffusion observation time (i.e., the duration and separation of the
diffusion weighting gradients).
In recent years, more and more experiments in biological tissues have demonstrated non-exponentially
attenuated echo amplitude with increasing b-values. Thus, new models have been proposed to
interpret this finding, but many issues remain yet to be solved.
In an in vitro study by Assaf et al,31 diffusion measurements were performed on water and N-acetyl
aspartate (NAA) molecules in excised brain tissue using b-values up to 28.3 × 104 and 35.8 × 104
2
s/mm for water and NAA, respectively. They observed that signal attenuation of water and NAA due
to diffusion over the entire range of b-values examined was not mono-exponential. In a study by
32
Mulkern et al performed on a clinical scanner, the signal decay from brain tissue in vivo was
measured for b-values up to 6000 s/mm2. The signal decay with variable b-value showed non-mono-
exponential behavior and was modeled by a bi-exponential function.
Further studies revealed that the non-mono-exponential behavior is also highly influenced by the
orientation of fibers with respect to the diffusion gradient. When measuring diffusion parallel to the long
axis of the neuronal fibers, the fraction of the slow diffusing component decreases significantly
compared with that measured perpendicular to the axis. In fact, we observed mono-exponential
33
behavior in the corpus callosum when the diffusion encoding gradient was parallel to the fibers (Fig.
10-4), even though the signal was clearly bi-exponential when measured with the gradient
perpendicular to the fibers (Fig. 10-5).
1 of 5 12-04-2010 00:31
Add to lightbox
page 293
page 294
Figure 10-4 Representative set of isotropic diffusion-weighted diffusion scans. The b-values are as
2 2 2 2 2 2
follows: A, 0 s/mm ; B, 250 s/mm ; C, 500 s/mm ; D, 1000 s/mm ; E, 1500 s/mm ; F, 2000 s/mm ;
2 2 2 2 2 2
G, 2500 s/mm ; H, 3000 s/mm ; I, 3500 s/mm ; J, 4000 s/mm ; K, 4500 s/mm ; L, 5000 s/mm . The
increasing signal hyperintensity with increasing b-value is due to the lower isotropic diffusion
coefficient of white matter in contrast to gray matter. Although appearing similar to fractional
anisotropy maps, this contrast mechanism is not due to diffusion anisotropy. Hence, it also shows
signal hyperintensities in areas of merging fibers, such as seen in the white matter of the frontal lobe.
Notice that with increasing b-values the distortions from eddy currents are getting ever stronger. In
addition, the image contrast changes dramatically with increasing b-values. Thus, the image
2
co-registration is getting increasingly difficult for b-values greater than 1500 to 2000 s/mm .
Add to lightbox
Figure 10-5 Decay curves for regions of interest in the splenium and genu corpus callosum, where the
diffusion gradients are perpendicular or parallel to the fibers' orientation. The markers indicate the
experimental data. Solid lines are corresponding fitted curves. The top two curves show
measurements perpendicular to the fibers, whereas the bottom two curves show measurements with
diffusion encoding parallel to the fibers.
2 of 5 12-04-2010 00:31
Much of the analysis of the multi-exponential echo-attenuation curves is based on the following
non-exchange (or slow-exchange) two-compartment model:
where Dfast and Dslow are the fast and slow apparent diffusion coefficients, respectively, and ffast is the
fraction of the fast diffusing component. For anisotropic media, multicomponent ADC analysis can be
extended to a multicomponent tensor analysis. Alternatively, the exchange condition can be weakened
and an additional parameter can be introduced, which turns the model-depending upon this
parameter-more and more into a fast-exchange model. In a manner similar to the two-compartment
model given by Equation 10-12, each compartment can be characterized by its own diffusion tensor
instead of a diffusion coefficient:
Here, D1 and D2 represent the diffusion tensor of the first and second compartments, respectively.
More compartments can be added into this model with the penalty of the increasing number of free
parameters and the associated problem that the parameter estimation can be trapped in a local
extremum of the cost function.
In most cases, the slow-exchange model can phenomenologically describe the multi-exponential
behavior well. However, the physiological origin and underlying mechanism of multi-exponential
echo-attenuation curves in MR diffusion experiments with biological tissue are still subjects of much
debate. Some studies have attributed this behavior to compartmentation of water pools; others have
suggested that restricted diffusion in one compartment can also lead to a non-mono-exponential
behavior of the MR signal. Human tissues are heterogeneous systems containing micrometer-scale
compartments separated by impermeable or semi-permeable membranes. These compartments differ
in size and shape. Where the exchange of water between compartments is slow on the MR timescale
(i.e. the diffusion encoding time), many physical MR parameters will show a superposition of values.
For example, it is often assumed that the existence of multi-exponential transverse relaxation curves is
an indication of tissue compartmentation. Similarly, tissue compartmentation can be a natural
explanation for multi-exponential echo attenuation curves. Peled et al34 have correlated water diffusion
and T2 relaxation to compartmentation in frog sciatic nerve. In this study, three T2 values were
identified: water diffusion was found to be unrestricted in the component of the signal with intermediate
T2 times (~100 ms). It was suggested that this was compatible with extracellular space. Furthermore,
restricted diffusion was observed in the component with the longest T2 times, supporting the
assignment of this component to water in an intracellular space.
Although the slow-exchange model can phenomenologically describe the bi-exponential behavior well,
the physical significance of the parameters is still unclear, especially the signal fraction of each
component. It has been suggested that the fast water population represents extracellular space and
the slow water population, intracellular space. The main difficulty with this assignment is that the fast
and slow signal fractions are essentially the opposite of the known volume fractions for the two
spaces. For example, some groups have reported a fast component fraction of 70%, yet extracellular
space contains approximately 20% of total brain water. Furthermore, the fractions have been found to
be highly sensitive to the orientation of the neuronal fibers with respect to the diffusion gradient.
Because of this difficulty, some investigators have suggested an alternative explanation for the multi-
exponential behavior. There is evidence that both fast and slow water ADC components can arise from
one water pool. Sehy et al35 have identified and characterized both fast and slow-diffusing water
populations in the intracellular space of the Xenopus oocyte. The fast and slow intracellular
components have diffusion coefficients of 1060 × 10-6 mm2/s and 160 × 10-6 mm2/s, respectively. The
3 of 5 12-04-2010 00:31
fast component represents about 89% of the total water signal. Recently, Sukastanskii et al36 have
demonstrated in theory that the presence of restrictive barriers in a one-compartment model can lead
to a quasi-two-compartment behavior of the MR signal.
Unfortunately, no consensus has been reached regarding the origins of the multicomponent ADC, thus
far. In general, multi-exponential decay can be reasonably expected when there is a statistical
distribution of any relevant physical property within a voxel. Nevertheless, many difficulties still remain
in identifying appropriate models to describe and interpret experimental data obtained in
heterogeneous tissues.
The q-Space Approach

An alternative approach to analyze multi-exponential signal attenuation is to use q-space analysis.37,38
In q-space representation, the MR signal measured in diffusion experiments is described by using the
probability density function (PDF) of the molecular random spin displacement. The vector q is defined
as γδG/2π. The echo amplitude that can be obtained from a q-space experiment is expressed as:
where E∆(q) is the echo amplitude as a function of q; R is the average displacement vector over the
time period ∆; and P(R,∆) is the PDF of the random spin displacement.
page 294
page 295
Add to lightbox
Figure 10-6 Magnetic resonance diffusive diffraction. Simulation of the echo attenuation, E(q, ∆), for
spins trapped between two parallel plates with separation 2L, and for perfectly reflecting walls. Four
2
different diffusion times are displayed in multiples of (2L) /D, and ∆ are respectively 0.1, 0.2, 0.5, and
0.7 (dotted, dashed, dashed-dotted, and solid dotted). Also shown is the theoretical curve for the
infinite diffusion time diffraction pattern, E(q, ∞) (solid).
From Equation 10-14 it is evident that E∆(q) is the Fourier transform of the PDF of the random spin
displacement. Hence, the inverse Fourier relationship between E∆(q) and P(R,∆) can be used to obtain
the PDF of the displacement (Fig. 10-6). This is accomplished by sampling the q space and
subsequently performing a discrete Fourier transformation with respect to q. To reconstruct a
high-resolution three-dimensional PDF is still a difficult task because of the long scan time required to
sample the three-dimensional q space. Most current q-space techniques have been limited to
one-dimensional projections of the PDF or low resolution PDF. An advantage of the q-space approach
is that no specific diffusion model needs to be assumed.
Various statistical measures can be obtained from the reconstructed displacement profile (i.e., the
4 of 5 12-04-2010 00:31
PDF), such as mean displacement and kurtosis (i.e., the standardized fourth statistical moment of the
PDF). Multiple diffusion components can be separated by fitting the PDF with a combination of multiple
Gaussian functions. In a study on isolated bovine optic nerve and rat brain (in vitro), Assaf et al39 were
able to identify two diffusing components from the displacement profiles. By measuring the mean
displacement, the slow-decaying component was found to be restricted to a diffusing distance of
approximately 2 μm.
Model of Non-Gaussian Diffusion

The diffusion tensor model is associated with the assumption of Gaussian diffusion. The observation of
the multi-exponential behavior of echo attenuation in diffusion experiments has made this assumption
questionable. Biological tissues in general and brain white matter in particular are heterogeneous
structures containing various impermeable and semi-permeable membranes. These physical barriers
hinder molecular diffusion, which results in a constrained distribution profile of the displacement. This
effect is most pronounced at regions where white matter fibers intersect. However, it cannot be
appreciated with the naked eye to what extent this factors into apparently normally oriented structures.
The statistical property of a Gaussian random vector can be fully characterized by its covariance
matrix (second order statistics). For a zero mean Gaussian random vector, only the second order
cumulant is non-zero. Given its second order cumulant, the PDF of a Gaussian random vector is
determined. Following Einstein's derivation of the relationship between the variance of particle's
random displacement and the diffusion coefficient, a linear relationship between the second order
cumulant and the diffusion tensor can be obtained. Therefore, by measuring the diffusion tensor, a
Gaussian diffusion process can be characterized. However, in general, higher-order cumulants are
nontrivial for a non-Gaussian random vector, and the PDF can not be determined by using only the
second-order cumulant. To fully characterize a non-Gaussian diffusion process that can exist in
biological tissues, it becomes necessary to measure its higher order statistics.
Recently, a generalized diffusion tensor imaging (GDTI) method was proposed to measure the higher
order statistics of a non-Gaussian diffusion process. The GDTI formalism can be derived by
generalizing Fick's law to a higher order partial differential equation (PDE) and solving the Bloch
equation with diffusion terms. A linear relationship exists between higher order cumulants and higher
order tensor (HOT) coefficients of the PDE. The resultant signal equation is expressed in an infinite
series of the higher order cumulants. For the special case of Gaussian diffusion, the signal equation
can be reduced to the conventional signal equation as given in Equation 10-3. The GDTI method
extends Torrey's approach of deriving the signal equation for a MR diffusion experiment to the
non-Gaussian diffusion case.
By combining a series of diffusion MR measurement with various diffusion encodings, higher order
cumulants of a non-Gaussian diffusion process can be estimated by solving the GDTI signal equation.
Just as diffusion tensor is used to characterize Gaussian diffusion in DTI, higher order tensor
coefficients serve to characterize non-Gaussian diffusion in GDTI. Moreover, these HOT coefficients
can be used to reconstruct non-Gaussian PDFs. The 3D PDF can be visualized by plotting its
isosurface. The non-Gaussianality can be characterized by computing its skewness map that can be
visualized similarly. One great advantage of the GDTI formalism is that it provides a series of
parameters with well-defined physical meanings to characterize the property of a diffusion process,
and this is achieved without assuming any specific diffusion model.
© 2010 Elsevier
5 of 5 12-04-2010 00:31
DIFFUSION-WEIGHTED IMAGING IN STROKE

Diffusion-weighted MRI is recognized to be an important imaging modality because of its ability to
detect central nervous injury within minutes of its onset, whereas other conventional imaging techniques
(e.g., MRI and CT) still fail to detect a lesion for at least a few hours1,40,41 (Fig. 10-7).
It is thought that the loss of neural homeostasis and cell membrane function within ischemic cells leads
to increased cell membrane permeability immediately after the onset of ischemia. In this context,
experiments in cats have shown a pronounced influx of water and Na+ cations decreasing the interstitial
space from 18.9 to 8.5 volume percent while the net water content remains constant. 42 These results
are in accordance with findings in impedance measurements where a reduction of extracellular space
from 24% to 12% was estimated.43 Therefore, the secondary shift of water from the extracellular
space to the intracellular compartment and its associated effects are postulated to be some of the
underlying cellular events that lead to the marked signal change on DWI.44 Correspondingly, changes
on DWI and ADC maps were reported as early as 45 minutes after onset of stroke in animal studies1
and within the first hour in humans,45 and remain positive for several hours to days.
Add to lightbox
Figure 10-7 Acute ischemic stroke of an 18-year-old male patient. Top row, Approximately 1.5 to 2
hours after onset of clinical symptoms, conventional X-ray computed tomography (CT) shows no clear
signs of infarction. Middle row, Series of diffusion-weighted images obtained immediately after CT
examination allows exact delineation of injured tissue. Bottom row, Follow-up CT performed after
several days clearly confirms findings of diffusion-weighted imaging (DWI) examination. Generally, the
outcome in older stroke victims is better due to more progressed brain atrophy. The DWI examination
was performed with a navigated diffusion-weighted interleaved echo-planar imaging sequence.
The precise mechanism underlying acute changes in tissue water diffusion following cerebral ischemia
still remains controversial. To explain this complex situation, one must consider various tissue
parameters, such as the intra- and extracellular diffusion constants, the exchange time, the
extracellular tortuosity, and the intracellular diffusion restriction. Consequently, several
pathomechanisms have been proposed to be causative for rapid ADC decline, namely: 1. displacement
1 of 2 12-04-2010 00:32
of extracellular highly diffusible water into the intracellular compartment that is supposedly
characterized by markedly reduced diffusivity due to a multitude of factors (e.g., presence of
plasmalemma and organellar membranes, the cytoskeletal network, the high concentration of
proteins)41,46-49; 2. changes in cell-membrane permeability and reduction of transmembrane water
50-52
movement ; 3. reduced interstitial space, making it more difficult for extracellular water to diffuse
around the cells (i.e., increased tortuosity); in the extreme case some of it will be trapped in small
pockets in between the swollen cells53-56; 4. reduction of intracellular diffusivity owing to an increase in
47,57-59
the viscosity of the intracellular milieu ; 5. decrease of energy-dependent intracellular
58
circulation.
The true underlying cellular pathomechanism(s) responsible for ADC depletion has not yet been
convincingly shown and some findings are even contradictory. Nevertheless, it might be anticipated that
the previously described pathomechanisms are not mutually exclusive and so it seems reasonable that
a combination of some of these effects would be causative for residual ADC reduction in acute
ischemic injury.
© 2010 Elsevier
2 of 2 12-04-2010 00:32
DIFFUSION IN THE PRESENCE OF PHYSIOLOGIC MOTION

page 296
page 297
Add to lightbox
Figure 10-8 A, Axial diffusion-weighted image with navigator-echo correction and pulse-triggering. B,
Without any correction, motion in the presence of strong diffusion-weighting gradients results in severe
image degradation. C, Diffusion-weighted imaging with pulse-triggering only. While motion artifacts
are less severe, the overall quality is still very poor.
The greatest technical challenge associated with DWI is to overcome the effects of macroscopic
motion, while retaining its sensitivity to the microscopic motion. To create diffusion-weighted contrast,
DWI pulse sequences must be sensitive to molecular motion on the order of several micrometers and
consequently are also sensitive to bulk tissue motion. Even small (submilimeter) displacements during
the diffusion-encoding phase will cause large phase changes in the resultant echo signal. Because bulk
motion (pulsatile motion, peristalsis, or involuntary patient motion) on this level is likely to be different
60
during each echo acquisition, each echo will be perturbed differently from one excitation to the next.
In other words, this additionally accrued spin phase will be overlayed on the phase information that is
impressed by regular phase-encoding. For this reason, patient motion during the diffusion-sensitizing
period produces phase errors that manifest as ghosting (Fig. 10-8A and B). The simplest way to
minimize excursions of the head when imaging the brain is to securely support the head with foam pads
and Velcro straps, but these are generally uncomfortable, not fully efficient, and cannot be applied to
examine, for example, the solid organs of the abdomen.
Investigators have measured brain parenchymal motions and CSF pulsations using different MRI
61
methods. Greitz et al attributed the major sources of physiologic brain motion to: 1. systolic
compression of the ventricles, and 2. caudal movement of the midline structures and brainstem. Rapid
motion of the brain is detectable within the first 200 to 250 ms after the electrocardiogram R-wave
trigger, while smaller amplitude motion is noted from 250 to 300 ms to the next R-wave, with maximum
brain motion being detectable in the brainstem, the cervical spinal cord, and the periventricular
regions.62 To reduce the artifacts arising from pulsatile motion, cardiac gating or pulse triggering was
proposed. Although gating reduces the residual velocities considerably, the motion-induced artifacts
cannot be completely eliminated (Fig. 10-8C). However, even with single-shot methods, cardiac
triggering and an appropriate gate delay help to reduce the point-to-point scatter in diffusion
measurements.
In a first approximation, the net result of bulk motion in the presence of the diffusion-encoding gradients
can be modeled by a zeroth-order phase term and a net shift of the recorded signal away from the
origin of the k-space.60 In general, these shifts are different for each phase-encoding step; if more
than one echo per excitation is recorded (e.g., with interleaved EPI or spiral imaging), the measured
data on that particular k-space trajectory is equally affected by this perturbation (Fig. 10-9).
Besides signal averaging and cardiac triggering, several attempts have been made to overcome the
problem of motion in multishot DWI sequences. The use of bipolar gradient pulses is one such attempt.
By placing a pair of bipolar gradients around the 180-degree pulse and enabling gradient-moment
1 of 4 12-04-2010 00:32
nulling, the sequences compensate for velocity (and higher order motion). However, this strategy will
also reduce the sensitivity to diffusion dramatically. Alternatively, removing the conventional phase-
encoding is another option to overcome motion-induced phase artifacts. This can be accomplished
either by means of line scan diffusion imaging63,64 or by applying k-space trajectories, such as are
used for radial65 or spiral66 imaging techniques, which are less vulnerable to motion. Another very
promising method to circumvent ghosting artifacts is to use an additional non-phase-encoded echo
67-71
immediately prior to or after the imaging echo. This additional echo is termed "navigator echo"
(Fig. 10-10).
The main intention behind the original navigator echo correction idea is that both echoes are diffusion-
weighted and contain the phase terms from bulk motion, but only the imaging echo contains the
additional phase-encoding to form the image. Therefore, the effect of motion can be retrospectively
corrected in the imaging echo. In the past, different forms of navigator corrections were devised: 1.
zero-order phase correction; 2. zero- and first-order phase correction in the read-out direction; 3. zero-
and first-order phase correction in the read-out and phase-encode directions; and 4. navigator images.
In this context, the navigator image is the most accurate navigation scheme. Generally, an
oversampled k-space center serves as the navigator image from which the k-space shifts are
determined and the original image k-space data is regridded accordingly.68,69,71 The determination of
the phase correction terms can be done either in the k-space domain68,69,71 or more efficiently in the
72
image domain. Recently k-space trajectories and novel correction schemes have been proposed that
22
inherently acquire a navigator image and provide a more effective retrospective correction,
respectively.21 A careful investigation of this improved correction scheme, which is based on the idea
of autofocusing, demonstrates that this processing technique is essentially equal to and can be
accomplished by the conjugate gradient iteration utilized for the Generalized SENSitivity Encoding
(GSENSE) method. Here, only the coil sensitivity estimates need to be replaced by the phase
estimates measured for each interleave.
page 297
page 298
Add to lightbox
2 of 4 12-04-2010 00:32
Figure 10-9 Left column, k-space trajectory of an interleaved echo-planar imaging (IEPI) acquisition
(top) and the corresponding image (bottom). Right column, IEPI k-space trajectory (top). Each profile
belonging to one particular interleaf is shifted in k-space by a certain amount along kRO and kPE.
Since the motion in the strong diffusion-encoding gradient field is random, the k-space shifts of each
interleaf are unpredictable. Without correction the reconstructed image is heavily distorted (bottom).
Add to lightbox
page 298
page 299
Figure 10-10 Pulse sequence for diffusion-weighted imaging with an additional navigator echo
acquisition prior to the acquisition of the regular image echo. The navigator echo (without phase-
encoding) is recorded at the time TEDiff when the spin-echo is formed by the first two RF pulses. At
the time TE the magnetization is refocused again by a second 180° RF pulse and the diffusion-
weighted image echo is read out. Note that the time TEDiff is more or less the limiting factor for the
diffusion-weighting whereas the composite echo time TE + TEDiff is responsible for T2-weighting of
the image. The dotted gradient waveform in the phase-encoding direction (GP) depicts a possible
second navigator echo perpendicular to the first one to monitor k-space shifts in the phase-encoding
direction. (GM, readout direction; GS, section select direction; RF, radio frequency; TE, echo time.)
Add to lightbox
3 of 4 12-04-2010 00:32
Add to lightbox
Figure 10-11 A, Single-shot diffusion-weighted spin-echo EPI pulse sequence. The entire k-space is
filled with a single EPI-readout train. Since the EPI train occupies a lot of the time available in the
second TE/2 interval (tEPI/2), the maximum duration of the diffusion gradient (GDiff) lobes is mostly
determined by the time TE/2-tEPI/2. Small additional bipolar gradients can be included after the first
diffusion gradients to slightly increase the diffusion-weighting. B, Corresponding single-shot sequence
for parallel imaging. The EPI readout is much shorter and allows longer diffusion encoding gradients
(GDiff). Thus, for a given b-value the echo time can be much shorter and can compensate in part for the
reduced number of profiles. Notice that the tEPI/2-period can be shortened further by means of partial
Fourier imaging techniques. (GM, readout direction; GP, phase-encode direction; GS, section select
direction; RF, radio frequency; TE, echo time.)
Despite all these readout strategies, the most common form of data acquisition is still a single-shot
73,74
read out, in particular single-shot EPI. This is because the acquired phase error due to bulk motion
is equal in each phase-encoding step and does not affect the image reconstruction, and nowadays the
hardware of most state-of-the-art clinical scanners is easily capable of producing EPI scans of
diagnostic quality. Moreover, recent developments in advanced DWI techniques require the acquisition
of an excessive number of diffusion-weighted scans along different directions or with different
amplitudes. From this perspective, the capabilities for rapid image formation are also very important to
keep clinical DWI within an acceptable time frame.
© 2010 Elsevier
4 of 4 12-04-2010 00:32
PULSE SEQUENCES
Diffusion-Weighted Single-Shot Echo-Planar Imaging
An EPI pulse sequence samples all the data points necessary for reconstruction of an image after the
application of a single RF excitation pulse (or a 90- to 180-degree RF-pulse combination) (Fig.
10-11A). With EPI, an image can be acquired within a fraction of a second and physiologic motion can
be minimized. For DWI, the general advantage of single-shot EPI is the fact that each profile in
k-space acquires the same motion-induced phase error and k-space shift. Thus, in a magnitude
reconstructed image these phase error terms are of no consequence. However, the maximum
attainable spatial resolution of EPI can be markedly limited by T2* decay during the long period of data
acquisition. In addition, EPI has only a very small bandwidth per pixel along the phase-encoding
75
direction. Hence, EPI is very susceptible to off-resonance effects, such as main field inhomogeneity,
local susceptibility gradients, and chemical shift, which all may lead to severe image degradation.
Typical regions that are affected by such artifacts are brain regions in the vicinity of air cavities, such
as around the posterior fossa, and the nasal sinuses. The skull base, the infratentorial portion of the
brain, the cervicothoracic junction, and abdominal regions close to the bowels.
page 299
page 300
Add to lightbox
Figure 10-12 Comparison between conventional diffusion-weighted single-shot EPI (top) and
sensitivity-encoding (SENSE) EPI readout (bottom) in a stroke patient with hemorrhagic
transformation. By means of the faster k-space traversal with SENSE the chemical shift artifact (solid
arrow) can be strongly reduced. Moreover, magnetic susceptibility artifacts or artifacts from
B0-inhomogneities (open arrows) can also be markedly diminished.
To reduce the artifacts caused by the low bandwidth per pixel, diffusion-weighted single-shot EPI can
76-81
be combined with recently introduced parallel imaging strategies, such as sensitivity encoding
79
(SENSE). SENSE is a valuable complement to regular gradient encoding in MRI. It uses some of the
spatial information contained in the individual elements of an RF coil array to more efficiently traverse
k-space. SENSE can, therefore, serve to accelerate virtually any conventional MRI technique without
interfering with the numerous different contrast mechanisms used in MRI, such as diffusion weighting.
It has been demonstrated that SENSE can help to improve single-shot EPI and fast-spin-echo (FSE)
scans by reducing artifacts, as well as by improving spatial resolution.78,82,83 Single-shot SENSE-EPI
(Fig. 10-11B) reduces the train of gradient-echoes in the EPI-readout in combination with a faster
k-space traversal per unit time. The resultant increased bandwidth per pixel in the phase-encoding
1 of 9 12-04-2010 00:33
direction and the shortened EPI train thus improve image quality (Fig. 10-12). The net penalty of the
accelerated k-space traversal is aliasing. With SENSE, however, image acquisition takes place with an
RF coil array where the component coil images and corresponding coil sensitivity estimates of each
coil can be used to fully unfold the desired diffusion-weighted image. In theory, the SENSE technique
allows for an acceleration of the k-space traversal by the same number as there are component coils
available. Therefore, the susceptibility to typical EPI-artifacts and image blurring can be markedly
diminished. Since the EPI-readout duration is also shortened and hence the T2* filtering is diminished,
another improvement with SENSE-EPI is resolution enhancement. With a properly adapted diffusion-
weighted single-shot EPI pulse sequence, the shortened EPI train improves spatial resolution to a
range previously possible only with multishot techniques.
Diffusion-Weighted Multi-Shot Echo-Planar Imaging

In interleaved EPI (IEPI), the EPI trajectory is split into multiple interleaves (Nint), where in k-space the
adjacent profiles of one interleaf are Nint-times farther apart than in single-shot EPI. Therefore, each
interleaf transverses the k-space along the phase-encoding direction Nint-times faster than does
regular EPI.68-71 In IEPI, off-resonance artifacts and image blurring can thus be decreased by taking
advantage of the increased k-space velocity. Because Nint excitations are required to fill up the entire
k-space, the problems with random phase shifts from bulk motion are also reintroduced. Using
navigator echoes, however, the motion-induced view-to-view phase alterations between each k
line/interleaf can be monitored and retrospectively corrected for motion. To further reduce image
degradation, additional plethysmographic RR-gating should be used to compensate for pulsatile brain
movement.
Overall, the IEPI method provides a good trade-off between measurement speed, maximum gradient
strength, and the aforementioned sensitivity to characteristic EPI-related artifacts (Fig. 10-13). With
regard to readout speed, IEPI is less demanding and thus IEPI can be installed on clinical MRI
systems equipped with less powerful gradients. However for low numbers of interleaves, SENSE-EPI
is superior to IEPI since no navigators are required.
Nevertheless, great improvements in phase navigation techniques have been made recently by using
an autofocusing approach.21 This approach has significantly increased the robustness of navigator
correction and allows for the acquisition of diffusion-weighted images at high definition.
Diffusion-Weighted Line Scan Imaging

Line scan imaging63,64 is especially attractive for DWI because no phase-encoding steps are
necessary, so additional phase terms due to bulk motion during the diffusion-weighting period are
irrelevant. In diffusion-weighted line scan imaging (LSDI), each column is formed by the intersection of
two planes selected by two slice-selective RF pulses, each with a different angulation relative to the
desired image plane. A standard frequency-encoding readout is then performed along the selected
column to obtain the spatial information along this column. By shifting the resultant cross-sectional area
(defined by the intersecting pulses) within the image plane, one can scan the image column by column.
Assembling adjacent lines ultimately yields the final image. The width of the selected column
determines the in-plane resolution, while the height of the cross-section defines the slice thickness of
the image.
page 300
page 301
2 of 9 12-04-2010 00:33
Add to lightbox
3 of 9 12-04-2010 00:33
Add to lightbox
page 301
page 302
Figure 10-13 A, Coronal diffusion-weighted single-shot EPI scans in a patient suffering from epilepsy.
Note the strong artifacts from off-resonant spins (i.e., susceptibility artifacts) close to the skull base.
Signal pile-up or cancellation and geometric warping are apparent. B, Corresponding diffusion-
weighted scans in the same patient with interleaved EPI readout (192 × 256, Nint = 13). Besides much
better spatial resolution, the mesotemporal lobe appears with smaller artifacts.
Add to lightbox
4 of 9 12-04-2010 00:33
Add to lightbox
Figure 10-14 A, Line scan diffusion imaging (LSDI) sequence. Only spins in the shaded cross-
sections form a diffusion-weighted spin-echo since it is only in these areas that spins have seen both
excitation (dotted lines) and refocusing (solid lines) RF pulses. B, LSDI in two female patients
suffering from acute benign compression fractures. Both image sets show, from left to right, scans
without and with diffusion-weighting, and the corresponding map of the isotropic diffusion coefficient.
Diffusion-weighted line scan imaging has demonstrated its usefulness for imaging with a small field-
of-view (FOV), such as subcortical U-fibers,84 and the spine85-87 (Fig. 10-14) and has great potential in
areas that are not accessible by EPI due to large susceptibility artifacts. Unfortunately, LSDI is partly
limited by the relatively longer scan times and inherently lower signal-to-noise ratio (SNR). It follows
88
that for some applications, such as DWI of the spinal cord, IEPI might still be the better choice,
although both methods have never been directly compared. In contrast to conventional 2D Fourier
methods, the SNR penalty of line scan imaging is immediately evident, considering that only a single
line contributes to each spin-echo, while the whole slice contributes to the spin ensemble in 2D spin
warp imaging.
Diffusion-Weighted Imaging with Fast Spin-Echo

In the past, FSE imaging has been successfully used in daily clinical practice as a fast and robust
image readout strategy with negligible artifacts. Thus, FSE promises to be an attractive alternative to
gradient-echo based methods such as single- or multishot EPI for image acquisition in DWI.
Unfortunately, RF power deposition within a subject can be considerably high if a series of compact
180-degree pulses is applied. This is especially the case in multislice imaging or when B0 is high, and
thus echo trains with lower flip angles are recommended.
Due to variable flip angle distribution throughout the slice in slice-selective imaging and due to specific
absorption rate (SAR) constraints, the RF pulses of an FSE train usually deviate from ideal 180-degree
pulses. This has the net effect that each RF pulse splits the magnetization into different components
and much of the signal power migrates from the primary echoes into higher order and stimulated
echoes. Ultimately, when the net spin phase is not in phase with the phase of the RF pulses of the FSE
readout-Carr-Purcell-Meiboom-Gill (CPMG) condition-the different coherences may interact
destructively. The violation of the CPMG condition happens frequently in the presence of motion during
the diffusion-weighting preparation phase that precedes the FSE readout. Specifically, if the spins have
accrued an arbitrary phase due to bulk motion, the part of the residual magnetization that is in phase
with subsequent FSE RF-pulses interferes constructively and the magnetization may asymptotically
reach a steady state. Conversely, the signal component that has a phase that is perpendicular to the
phase of the subsequent FSE readout pulses interacts destructively and may lead to amplitude decay
and phase fluctuations. Overall, the sum of the in-phase and the quadrature signal component leads to
unintended amplitude changes and phase fluctuations that are competing with regular phase-encoding.
It has long been known that the FSE sequence is very sensitive to the phase of image echoes so that
any deviation of the CPMG-phase condition leads to severe ghosting. 89 Therefore, the major difficulty
in using FSE DWI in vivo is the presence of such phase errors, and proper remedies to overcome such
artifacts are required.
5 of 9 12-04-2010 00:33
page 302
page 303
One approach for diffusion-weighted FSE imaging is the displaced ultrafast low-angle rare (U-FLARE)
technique,90 which relies on the separation of different echo families so that they do not interfere. This
technique can be modified by imbalanced readout gradients in conjunction with diffusion-weighted
half-Fourier single-shot turbo spin-echo (HASTE) acquisition in order to split and subsequently sample
the two echo families separately.91 The resultant two images can then be combined by adding them in
magnitude mode. Although one can avoid artifacts caused by the violation of the CPMG-phase
condition, these methods are somewhat limited due to 1. a prolonged echospacing; 2. an SNR penalty
due to the required higher bandwidth to sample two echoes between adjacent refocusing pulses; and
3. the requirement of dummy echoes to provide an equilibrium between both echo families. A further
refinement of the diffusion-weighted displaced U-FLARE method was proposed by Alsop et al92 and
McKinnon et al93 that utilizes only half the echo signal. More recently, Le Roux et al94,95 presented a
method for diffusion-weighted FSE that relies on a non-CPMG FSE approach that employs a modified
quadratic phase law for the phase setting of the refocusing pulses. The latter is also the basis for the
PROPELLER FSE DWI method, which is addressed in the next section.
Diffusion-Weighted Imaging with Radial k-Space Acquisition

It has been found that radial acquisition of Fourier data combined with filtered back-projection
reconstruction can yield diffusion-weighted images with reasonable quality. 96 More recently, radial
scanning methods that sample more than one radial line in k-space per excitation and, therefore,
increase the acquisition speed were presented.65
The relative insensitivity of radial scanning methods to motion in DWI can be attributed to the fact that
the phase errors associated with motion during the diffusion-weighting can be removed in the image
reconstruction process by employing a magnitude-only filtered back-projection reconstruction. But
notice also that the k-space shifts orthogonal to the direction of the trajectory cannot be resolved by
taking the magnitude.
Add to lightbox
6 of 9 12-04-2010 00:33
Add to lightbox
page 303
page 304
Figure 10-15 A, Example of a k-space sampling trajectory for "periodically rotated overlapping parallel
lines with enhanced reconstruction" (PROPELLER) MRI. The bold lines indicate the measured area of
k-space (called a "blade") by one echo train in a fast spin-echo (FSE) experiment. In subsequent
relaxation times, the blade is rotated in order to measure the remaining parts of k-space, while
resampling the center of k-space each time. B, Diffusion-weighted images using single-shot EPI (left
column) and PROPELLER (right column) at the same slice locations for a patient with temporal lobe
infarcts (top row) and a patient with an infarct in the pons (bottom row). (A and B, courtesy of J. Pipe,
Barrows Neurological Institute, Phoenix, AZ).
As mentioned in the previous section, FSE methods for DWI have far less B0-related artifacts (e.g.,
from metal, sinuses, and eddy currents) than have EPI-based scanning techniques. Nevertheless, FSE
DWI is challenged by the fact that the diffusion-weighted signal will generally not meet the CPMG
condition, leading to unstable echoes. 88,91,92 A very elegant method to combine FSE with advances in
radial scanning has recently been published by Pipe et al.72 This work is based on the "Periodically
Rotated Overlapping ParallEL Lines with Enhanced Reconstruction" (PROPELLER) method, which has
inherent 2D navigator information in each FSE echo train. PROPELLER FSE DWI uses FSE data
collection, which provides far greater immunity to geometric distortion than that obtained with EPI
sequences. It also mitigates signal instability, present in all diffusion-weighted FSE methods, by varying
the phase of the refocusing pulses using a scheme previously reported. 97 Since the sequence is radial
in nature, errors are expressed in a rather benign fashion, similar to the aforementioned projection
reconstruction methods. The resulting k-space coverage for data acquisition is illustrated in Figure
10-15A. For each TR, the collected echoes are phase-encoded to form a sufficiently sampled data
strip (blade), which goes through the center of k-space. The width of the blade in k-space is equal to
the echo train length (ETL) divided by the field of view (FOV). In subsequent TRs this blade is rotated,
so that together the blades measure a circular region of k-space formed by their union, and separately
they all measure a smaller, circular region in the center of k-space formed by their intersection (see
Fig. 10-15A). The basic idea of PROPELLER is that this oversampled region in the center of k-space
can be compared between blades to correct for inconsistencies prior to combining the data. For
diffusion applications, the primary inconsistency will be image-space phase differences. After data
correction, the blades are weighted to correct for sampling density variation as well as uncorrected
error, then gridded onto a Cartesian array,98 and Fourier transformed to form the image (Fig. 10-15B).
7 of 9 12-04-2010 00:33
Some limitations of the original PROPELLER method are 1. its incompatibility with array coils; 2.
limited blade width for navigator correction; and 3. limited FSE readout flip angle reduction to maintain
stable echoes. The latter is of particular importance for SAR reduction at 3T as well as when one
considers the stronger B1 inhomogeneities at higher fields. Generating a pair of orthogonal blades for
odd and even echoes can remedy the limitation for phased array coils. The width of the blade for a
certain ETL can be improved either by using SENSE encoding technology (i.e., increasing the k-space
increment at the cost of aliasing) or by introducing additional gradient-echoes, such as are used for
GRAdient and Spin-Echo (GRASE) techniques.
Diffusion-Weighted Spiral Imaging

Spiral scanning (Fig. 10-16A) provides an alternative scheme to collect diffusion-weighted k-space
data66,99 and has so far received much attention for fast MRI in cardiac and functional MRI.100,101
Spiral imaging has a number of potential advantages over other fast imaging sequences. First, spiral
imaging is less demanding for the gradient system because the trajectory is continually turning. This is
accomplished by slowly changing from one gradient amplifier to the other. Thus, the need for rapid
switching gradient amplifier polarity, such as is needed in EPI, is reduced. Moreover, both gradient
axes contribute equal amounts of power to the k-space sweep, rather than having one axis contribute
as in conventional 2D acquisition schemes. Spiral imaging is also relatively insensitive to flow artifacts
and ghosting. This is simply because the center of k-space is collected at the start of the scan and so
has zero transverse moments of all orders (the moments of gradients in MRI determine how sensitive
a particular pulse sequence is to motion). Another inherent advantage in the case of interleaved spirals
is that the k-space origin is redundantly sampled by the spiral trajectories, providing self-navigating
capabilities. The T2/T2* decay during a spiral scan leads to circular symmetric blurring instead of
severe blurring along a particular direction. This can be understood by the fact that in spiral scanning
the k-space velocity in the angular direction is much higher than in the radial direction.
Add to lightbox
8 of 9 12-04-2010 00:33
Add to lightbox
page 304
page 305
Figure 10-16 A, Interleaved spiral scanning trajectory with four arms starting under incremental angles
of 90°. Since the spiral trajectories oversample the center of the k-space and always start from the
center of the k-space, spiral scanning has inherent 2D-navigator capabilities. B, Self-navigated
interleaved variable density spiral diffusion tensor imaging (DTI, 16 interleaves). Diffusion-weighted
images with diffusion encoding along six different directions (a-f); map of the trace of the diffusion
tensor (g); isotropically diffusion-weighted image (h); fractional anisotropy map (i).
To date, however, the widespread use of spiral scanning in routine clinical practice has been limited by
the lack of time-varying gradient-waveform capabilities and the lack of appropriate image
reconstruction resources. Further improvement of this technique can be achieved by means of variable-
density spiral scans. This method is currently under investigation in our laboratory (Fig. 10-16B). With
variable-density spirals,102 the first portion of the trajectory oversamples the center of k-space to
resolve the motion-induced k-space shifts and thus provides inherent 2D navigation capabilities.
© 2010 Elsevier
9 of 9 12-04-2010 00:33
EDDY CURRENTS AND CORRECTION SCHEMES
Add to lightbox
Figure 10-17 Effect of eddy current induced image warping on diffusion tensor maps. A, Fractional
anisotropy map derived from EPI-based diffusion tensor imaging (DTI) scans. The phase-encode
direction was along the left-to-right direction. Due to the different geometric distortions that are caused
by playing out diffusion encoding gradients along different directions, the diffusion tensor scans are
slightly misregistered and produce erroneous diffusion tensor estimates. B, Fractional anisotropy map
after an unwarping algorithm was applied. The misregistration was almost completely eliminated and
gray/white matter areas can be much better appreciated. Also the hyperintense artifact surrounding
the brain could be removed.
Eddy currents are introduced by switching strong diffusion gradient pulses. Besides motion, they
comprise a major source for image misregistration in DWI. Misregistration in the diffusion-weighted
images can generate anomalous contrast, evident as blurring and displacement. Quantitative
measurements are also erroneous due to the comparison of dissimilar materials in the same pixel. The
image misregistration is most apparent at (inhomogeneous) tissue interfaces, such as at the edges of
the brain. Likewise, the comparison of pixels in air with shifted tissue pixels between differing gradient
directions produces large amounts of contrast that is unrelated to tissue diffusion anisotropy; this
creates, for example, the typical pseudo-contours in FA maps of the brain (Fig. 10-17A). This spurious
contrast can also be seen at the gray/white matter interfaces (see Fig. 10-17A) or in the breasts at
interfaces between lesions and regular glandular tissue. Overall, diffusion contrast and quantitative
measurements clearly improve with better image registration (Fig. 10-17B).
According to fundamental physical principles, the rapid switching of magnetic fields is always linked to
an electric field, which generates eddy currents in surrounding low-resistance conductors (e.g., in the
Dewar, RF coils, and shielding). These in turn generate secondary magnetic fields that spoil the
magnetic field homogeneity and can have serious effects on the image quality of magnetic resonance
diffusion experiments. The induced eddy currents are proportional to the field gradient switching rate
(dB/dt) and have associated magnetic fields, which not only distort the gradient profiles around the
sample, but also can persist for tens of milliseconds up to seconds after the gradient pulse has been
turned off. For MRI in general, the eddy current induced spatiotemporal field distortions result in a
slower rise and settling time of the gradient waveform. For DWI, the problems are twofold: first, the
effective diffusion attenuation is affected by eddy current induced gradients. Second, and more
significant, the residual eddy current background gradients lead to perturbation of the prescribed
k-space trajectory and to image distortions in echo-planar or spiral imaging.
In general, the spatial distribution of the eddy current fields depends upon their conductive pathways; in
some cases, they may result primarily in an overall time-varying field shift (e.g., the conducting path
forms a loop around the magnet bore) or in gradient fields. Here, a given eddy current can be
produced from its respective gradient term (self term), or it can arise from one of the two other
gradient axes terms (cross term). Generally, these eddy currents are modeled by time-varying
1 of 3 12-04-2010 00:33
functions; specifically, these are exponentials with a wide range of decaying time constants; 103,104 the
105-107
effective eddy current will be a superposition of multiple exponential decay terms. The
time-varying eddy current effects introduce magnetic field inhomogeneities that are comprised of 1. a
spatially invariant term [epsi ]0(t) (i.e., B0 eddy current); 2. linear magnetic field gradient terms [epsi
]x,y,z(t); and 3. higher order eddy current terms that are usually neglected.
For a rectangular gradient waveform, the rise and fall portions of the diffusion gradient introduce eddy
currents that are equal but of opposite sign. In the case of relatively short rectangular pulses (e.g.,
imaging gradients), these eddy currents will tend to cancel each other out and will lead to minimal
residual gradients. On the other hand, if the trapezoid is relatively long in duration (e.g., for diffusion
encoding gradients) the currents from the rising and falling ramps of the gradient waveform will not be
canceled out completely, and considerably large residual eddy currents will remain. For a Stejskal-
Tanner diffusion gradient pair (i.e., the classical diffusion-encoding gradient pair), the residual eddy
current is approximately double that of a simple gradient pulse. Conversely, the eddy currents
generated by a pair of diffusion gradients with opposite polarity (bipolar gradients) will tend to cancel
each other out,108 but their encoding efficiency is less than 25% of the original unipolar gradients.
page 305
page 306
If the eddy currents are long enough, they extend into the acquisition window and the observed signal
may be distorted. One possible solution might be the use of short diffusion gradient pulses in
combination with a stimulated echo sequence. For appropriate diffusion attenuation, the mixing time
(i.e., the period between the second and third RF pulse, where the magnetization is aligned along the z
axis and is not subject to T2 decay) can be of sufficient length to allow half of the echo time long
enough to let the eddy current effects dissipate. This approach is especially attractive for high b-value
experiments, since the signal decay and eddy-current-related distortions are more pronounced for
these types of experiments (see Fig. 10-4). Single-shot diffusion-weighted echo-planar and spiral
imaging suffer greatly from geometric distortions that are introduced by residual eddy currents.
Despite effective countermeasures, such as gradient current pre-emphasis and active gradient
shielding, residual eddy currents are still sufficient to distort MRI scans. The signal resulting from
deviations on an arbitrary k-space trajectory can be written as follows:
Since the bandwidth per pixel along the phase-encoding is very low in EPI, magnetic field deviations
produced by eddy currents have concurrent effects to the regular phase-encoding in EPI. The net
effect of such field fluctuations can be approximated as shear, shift, and scaling along the phase-
encoding direction. In contrast, the dominant effect of eddy-current-induced gradients for diffusion-
weighted spiral imaging is no longer along one dimension and, therefore, more difficult to correct. The
resultant effects from eddy currents are time-variant deviations from the carrier frequency and
deviations from the originally designed k-space trajectory. The former causes blurring, whereas the
latter causes geometric distortions and depends on the axis of the diffusion encoding.
Using the aforementioned distortion model, several post-processing methods have been suggested to
correct for these geometric distortions in EPI. To our knowledge, similar corrections are not known for
non-Cartesian trajectories thus far. Our implementation for EPI data is very similar to published
algorithms and employs numerical methods to automatically find optimal warping parameters to
perfectly remove these distortions: it is assumed that shear and scale are symmetric around the center
of the image and that all effects are only effective along the phase-encoding direction.
2 of 3 12-04-2010 00:33
By analyzing the pixels along the phase-encoding direction, the shear, scale, and shift can be
transferred into a simple shift and scale operation. Thus, the problem is cast into multiple (for each
position along the readout dimension), easily-solvable one-dimensional optimization tasks with two
unknowns. This approach was originally introduced by Haselgrove et al109 using an exhaustive search
for proper shift and scale parameters that optimize the degree of coregistration between a template
and a distorted profile. We use a modified version of this approach that employs a fast converging
numerical optimization algorithm, which allows one to identify the warping parameters that maximize
110,111
our cost function. Here, the optimizer is iteratively searching the warping parameter space; the
algorithm transforms the profile according to these parameters, calculates the similarity measure/cost
function (e.g., cross-correlation, normalized mutual information), and adapts the parameters to
converge to a maximum of the cost function. Further processing is comprised of the following: 1.
calculating the average image scale from the estimated scale factors for each one-dimensional profile;
and 2. weighted linear least squares fitting to obtain the shift parameters over the readout coordinate.
For the latter, the slope of this linear fit yields the shear parameter, whereas the intercept yields the
shift parameter. Using the three parameters (shear, scale, and shift), the value of each pixel from the
distorted image is warped to the corrected position. Since the spatial transformation is usually no
longer an integer value, this method requires that the continuous data be regridded back onto an
equidistant grid. This entire registration step can be performed in an iterative fashion until the
registration quality is sufficient.
Recently, a very effective, sequence-based method for suppression of eddy currents has been
introduced for EPI112 and spiral imaging.66 Here, a conventional spin-echo diffusion-weighted sequence
is modified by introducing another refocusing pulse within the given TE interval. Two bipolar field
gradients of length δ1+δ2 and δ3+δ4 are used, with the RF refocusing pulses dividing each bipolar
gradient pair (δ2+δ3 = TE/2 and δ1+δ4 = TE/2-tRO) (Fig. 10-18). The time tRO is the time required for
the imaging gradients that shortens the maximum duration of the diffusion encoding gradients. Since
the gradient pulses are shorter and have opposite polarity, the build-up of potent eddy currents is
strongly reduced. Conversely, for the diffusing spins the bipolar gradient pair together with the
refocusing RF pulse act as one long diffusion encoding gradient with efficient encoding power.
Compared with other pulse-sequence based correction schemes, this sequence is the first to nullify
residual eddy-current-induced fields with almost no loss of scanning efficiency. Depending on the RF
pulse implementation, only an extra 3 to 8 ms have to be spent for the second refocusing pulse.
© 2010 Elsevier
3 of 3 12-04-2010 00:33
NONUNIFORMITIES IN DIFFUSION ENCODING

page 306
page 307
Add to lightbox
Figure 10-18 Twice-refocused dual spin-echo diffusion-weighted pulse sequence. The RF pulses
(excite and refocus), diffusion gradients GDiff of lengths δ1, δ2, δ3, and δ4, and an EPI readout are
shown. The sequence allows any diffusion gradient lengths such that the time between the two
refocusing pulses is TE/2, and the dephasing and rephasing due to the diffusion gradients are equal.
Since the gradient pulses are shorter and have opposite polarity, the build-up of potent eddy currents
is strongly reduced. Conversely, for the diffusing spins the bipolar gradient pair together with the
refocusing RF pulse act as one long diffusion encoding gradient with efficient encoding power. With
knowledge of the principle eddy current decay time constant, diffusion gradient lengths can be
calculated so that eddy current buildup is nulled prior to readout.
High-performance gradient systems use gradient coils that are designed for modest fields of view in
order to limit dB/dt, which can cause nerve stimulation. As a result, magnetic field gradients can be
significantly nonuniform. Such nonuniformity is well known to cause spatial image warping in MR
images. On clinical scanners, there are methods routinely used to retrospectively correct for these
geometric distortions. However, little has been reported about the influence of gradient distortion on
DWI. These gradient nonuniformities over a clinical whole-body unit can be quite significant and usually
depend on the ratio between the size and design of the gradient coil and the FOV chosen. On a state-
of-the-art clinical scanner, typical errors in absolute diffusion measurements range from approximately
-10% to +20% for a FOV of 25 cm, but they can vary depending on the vendor. The presence of such
gradient nonuniformities is unfortunate because nonuniformity ultimately causes the accuracy of
absolute diffusion information to vary spatially. Additionally, it can be shown that spatial deviations from
the desired gradients can significantly perturb the interpretation of diffusion tensor information,
specifically the orientation of the eigenvectors (see Chapter 11, Diffusion Tensor Imaging).
However, the knowledge of the mismatch between the desired and the actual gradient field allows for
a straightforward correction by using an adapted calculation of the diffusion information. The mismatch
can be determined either by phantom measurements or, more conveniently, by utilizing gradient
perturbation models (i.e., determining the deviation from the true gradient linearity; Fig. 10-19), such as
is routinely used to correct for nonuniformity-induced image warping. A more detailed analysis of
theses artifacts and correction schemes is beyond the scope of this chapter and can be found
elsewhere.113
© 2010 Elsevier
1 of 1 12-04-2010 00:34
DIFFUSION-WEIGHTED IMAGING OUTSIDE THE NEUROCRANIUM

There is a large body of publications about applications of DWI and DTI in the brain. Some of them are reviewed in
other chapters of this book. Here, we review clinical applications of DWI/DTI for other parts of the human body that
have not received much attention thus far, but may bear promising potential for the future.
Breast
Motivation
page 307
page 308
Add to lightbox
Figure 10-19 Calculated diffusion tensor elements (D xx, Dyy, Dzz), isotropic diffusion coefficient <D>, and fractional
anisotropy (FA) measured in a spherical phantom (Ø ~ 25 cm) before (first and third rows) and after (second and
fourth rows) correction of the gradient uniformity. The images were acquired at the isocenter z = 0 cm (top two
rows), and at the plane z = 6 cm (bottom two rows). Notice the strong signal fluctuations within the object before the
gradient uniformity correction. At z = 6 cm some residual distortions are still visible due to incomplete unwarping of
eddy-current-induced misregistration.
Although the sensitivity of breast MRI for invasive breast cancer has been repeatedly reported to be close to
114,115
100%, the specificity limits its clinical applicability because of concern for false-positive lesions. Disparities in
specificity values arise from differences in technique, patient population, and interpretation criteria, among other
factors. Several modalities contribute to lesion characterization. High spatial resolution MR with contrast
116,117
enhancement is used to describe architectural features of breast lesions. Combinations of features predict
116
the likelihood of malignancy with 84% accuracy. Dynamic contrast-enhanced T1-weighted imaging gives
information on tissue vascularity and vascular permeability. The kinetic profile of lesion enhancement can help in
lesion characterization, because most malignancies enhance more quickly and to a greater extent than do benign
118-120
lesions. T2-weighted imaging further distinguishes certain benign from malignant lesions: in T2-weighted turbo
spin-echo pulse sequences, fibroadenomas (benign abnormality) are hyperintense, while malignancies tend to be
iso- or hypointense.121 However, many lesions do not fall neatly into classical appearances of benign or malignant
tissue based on these imaging techniques, and call for biopsy or continued follow-up in the absence of other
alternatives. In particular, while cysts may be easily distinguished from invasive carcinomas, there remains
considerable overlap in the imaging appearances of ductal carcinoma in situ (DCIS), infiltrating lobular carcinoma,
and fibrocystic change; and between invasive carcinoma, benign intraductal papilloma, and fibroadenoma.
Development of further imaging characterization is, therefore, necessary.
Diffusion-weighted imaging may potentially provide additional information to improve the specificity of breast MR for
malignant entities. The ADC calculated at a location in tissue is a measure of the mobility and tortuosity of water
122
protons and, hence, indirectly of tissue cellularity. In breast tissue, the apparent diffusivity can be affected by cell
density, organization, and membrane permeability, as well as by the extracellular matrix and tissue microcirculation.
Since these characteristics are altered in tumors, it can be speculated that DWI will improve characterization of
breast lesions by supplying an additional parameter to a lesion profile.
1 of 11 12-04-2010 00:35
Beyond lesion characterization, breast DWI may also be applied to monitor the efficacy of therapy regimens for
breast carcinoma. As tumor cells undergo necrosis and apoptosis, there is a corresponding increase in water
mobility. Lower cellularity may contribute to the increase in diffusivity measured, as well as cell membrane lysis and
permeability from toxic effects. Along these lines, DWI has been used in brain tumors to look for early signs of
response to chemotherapy.123,124 In the setting of breast cancer, DWI is also being used to look at changes
125,126
induced by radiation or chemotherapy to human breast carcinoma tissue grafted onto mice. However, no
study has been conducted yet to investigate the efficacy of DWI for treatment monitoring in representative sample
of patients treated for breast cancer.
Previous Studies in Lesion Characterization
Table 10-1. Apparent Diffusion Coefficient Values from Breast Diffusion-Weighted Imaging
studies (10-3mm2/s)*
b Values Fat
2
Group (mm /s) Suppression Cyst Benign Malignant
128 0, 1000 2.35 ±0.08 1.57 ±0.23
Guo et al yes 0.97 ±0.20 (30)
(4) (17)
Kinoshita et al129 0, 700
no
1.495 ±0.181 1.216 ±0.189
(6) (16)
Sinha et al 0-289.7 2.65 ±0.30
130 yes 2.01 ±0.46 (6) 1.60 ±0.36 (17)
(orthogonal) (6)
Sinha et al 1074 (max) 2.79 ±0.34
130 yes 1.35 ±0.10 (2) 1.01 ±0.17 (2)
(tetrahedral) (2)
page 308
page 309
*Data are mean ± SD. Sample sizes are in parentheses.
Several studies have demonstrated that mean apparent diffusivities in normal glandular breast tissue and benign
127-129 127
lesions are higher than in breast malignancies (Table 10-1). Guo et al performed DWI with single-shot EPI
and followed up on lesions with pathological analysis. They found that lower ADCs in malignant lesions correspond
with higher tumor cellularity functioning as a barrier to water movement. Diffusion-weighted imaging is most
effective in distinguishing fibroadenoma from invasive malignancy. The higher ADCs of fibroadenomas may be
attributed to myxoid change of the stroma, which allows for greater water mobility. In this context, a potential limit
to the specificity of DWI occurs when cellularity does not correlate with malignancy. For example, a malignant
lesion such as scirrhous adenocarcinoma has low cell density, and hence a high ADC, and it can be misclassified as
benign by DWI. Likewise a papilloma, a benign entity with high cellularity, may be falsely identified as malignant on
the basis of its low ADC.127 Hence, further cross-sectional studies on a large number of breast tumor patients are
warranted before the true potential of breast DWI can be revealed with respect to its diagnostic specificity.
Kinoshita et al128 evaluated diffusion-weighted half-Fourier single-shot turbo spin-echo (HASTE) as a diagnostic
tool. They imaged a group of patients with mammographic abnormalities or palpable masses with this fast
spin-echo based sequence. Their sequence is much less affected by chemical shift or susceptibility artifacts than is
EPI. They found lower mean ADCs in invasive ductal carcinomas than in fibroadenomas. However, HASTE
128
produces images with lower SNRs and more spatial blurring and is, therefore, a less sensitive tool. Moreover,
the long FSE readout is more sensitive to long T2 lesions, such as fibroadenomas. Conversely, short T2 lesions will
demonstrate very little SNR and, hence, the ADC measurements can be perturbed.
Sinha et al129 experimented with two different diffusion-encoding schemes: in one, they applied low b-value (b = 0
2
to 234.3 s/mm ) diffusion gradients in three orthogonal directions to image normal volunteers and patients with
2
known lesions. In the other, they applied high b-value (maximum b = 1074 s/mm ) diffusion gradients in four
tetrahedral directions to image a small group of patients. In both groups, they found higher mean ADCs in the
benign lesions than in the malignancies. At low b-values, the authors found that the ADC values generated reflect
perfusion effects. Hence, they suggest the capacity of DWI to detect tumor vasculature in a background of normal
129
tissue vasculature.
Due to several factors, including the variance in imaging protocols, the actual ADC values range widely between
studies for both benign and malignant lesions. However, there have consistently been statistically significant
differences between ADCs of benign and malignant lesions within each study. In order for there to be comparability
2 of 11 12-04-2010 00:35
between studies, there needs to be standardization of such variables as patient population sampled, b-values used,
number of directions the gradients are applied, and whether or not fat suppression has been used.
Technical Points
Diffusion-weighted imaging of the breast presents particular challenges: The shape of the breast distorts the
background magnetic field as the breast protrudes into airspace and acts as an air-tissue interface, causing a
dramatic local change in magnetic susceptibility. For this reason, the region of the nipple is particularly susceptible
to geometric distortions, especially in single-shot EPI scans.
The lipid and glandular content of the breast can demonstrate a mixture of free and lipid-bound protons within
voxels of interest, which introduces problems that are not common in DWI of the brain and spinal cord. However,
measurements of ADC can be biased by selection of a region of interest (ROI) because of the differential content
of microscopic fatty infiltration in glandular tissue. For example, Englander et al calculated a range of ADCs in
123
different patients for ROIs that all appeared to be homogeneously fibroglandular, which-besides biological
variability-could also be indicative of a variable fat content in each patient's breast tissue. Fat suppression is,
therefore, a prerequisite and should be performed in breast DWI in order to address the following: 1. signal from
fat causes off-resonance artifacts in single-shot EPI; 2. lipid-bound protons are very immobile with very low ADC
and, hence, fat is a confounder in the diffusion-weighted proton signal observed from a voxel. In other words, if the
fat signal is not suppressed, the apparent ADC is a composite measurement that is comprised of fractional
contributions from fat and tissue. This notion might sound evident and straightforward; however, in the past this fact
has often been neglected and has led to confusion in the interpretation of ADC measurements of non-neural tissue
and can-at least to some extent-explain the wide variation of values that have been reported. Nevertheless, in the
presence of fatty tissue, whether or not lipid-bound protons contribute to the overall MR signal, diffusion is
restricted simply due to fact that the fat molecules impose some hindrance on the diffusing protons.
As with all MRI measurements, the voxel size in breast DWI is limited by the requirements of the sequence for
sufficient SNR, because low SNR causes limited precision but also reduced accuracy in ADC measurements. The
SNR is also affected by the b-value chosen, because signal is attenuated as diffusion gradients are increased in
strength (greater b-values). The choice of slice thickness, in-plane resolution, and b-value should produce enough
signal to get a good SNR. As a rule of thumb, the SNR of the diffusion-weighted signal should not become lower
than 3 or 4 to avoid ADC underestimation from signal biasing that occurs in low-SNR MR magnitude data. Likewise,
129
ADC measurements can be biased by flow effects in the low b-value range and the reference b-value should be
2
slightly elevated (b ≈ 50 to 100 s/mm ) to spoil the contribution from moving spins.
Concerning the Radiologist

Partridge et al found the ADCs of fibroglandular tissue were lowest during the second week of the menstrual cycle
130
and highest during the fourth week. Although the fluctuations were small (coefficient for variation of 5.5%), they
130
indicate that the cycle week should be taken into account by the clinician evaluating a breast lesion.
page 309
page 310
Because contrast material may produce confounding signal, DWI scans should be done before contrast
administration. Contrast material leaks into the interstitial space and causes T2* decay, such that there is a chance
for loss of signal or production of geometric distortions. We hypothesize that the leakage of contrast can cause
local susceptibility changes that could eventually alter the diffusion weighting to an unknown extent.
Each diffusion-weighted breast MRI pulse sequence has its unique set of problems. The shortcomings of diffusion-
weighted EPI include blurriness and susceptibility artifacts due to magnetic field distortions. Diffusion-weighted
79
single-shot echo-planar imaging (EPI) has been used together with sensitivity encoding (SENSE) to reduce
82
artifacts and blurring and to give better spatial resolution. Our preliminary work in using SENSE with diffusion-
weighted breast MRI has produced images with a resolution of approximately 2 mm, allowing good identification of
breast lesions when correlated with T1- and T2-weighted images (Fig. 10-20). Due to the quick image formation
and the little additional effort-altogether approximately 2 minutes per breast-single-shot EPI in combination with
SENSE appears to be a useful breast DWI screening method.
3 of 11 12-04-2010 00:35
Add to lightbox
Add to lightbox
Figure 10-20 Diffusion-weighted echo-planar imaging (EPI) of the breast using the twice-refocused dual spin-echo
sequence. A, Three slices from a patient with cystic abnormalities in the breast. From left to right: Unweighted (b0)
SE-EPI image; isotropic diffusion-weighted image (b ≈ 500 s/mm2); and calculated map of mean diffusivity (trace).
Notice that the cystic lesions have higher diffusion coefficients than does the surrounding glandular tissue. B, Two
slices from a patient with an invasive breast cancer. The lesions are highlighted by arrows. From left to right:
Unweighted (b0) SE-EPI image, isotropic diffusion-weighted image (b~500 s/mm(2)), and calculated map of mean
diffusivity (trace). In addition, corresponding water-only-excited T2w-FSE scans of these lesions are shown
(rightmost column). The top two rows are acquired with conventional single-shot EPI, whereas the bottom two rows
are acquired using SENSE acquisition principles (reduction factor = 2.0). It is apparent from this image how much
reduction in geometric distortion can be achieved. Moreover, image blurring, especially in the diffusion-weighted
images, was dramatically improved. Notice that the diffusion coefficient of the tumor lesions is significantly lower
than that of the surrounding glandular tissue.
Currently, in addition to the standard breast MRI protocol, at the Stanford University Medical Center, we do
diffusion-weighted EPI with SENSE after suitable coil sensitivity calibration. Rather than performing DWI of a single
slice of tissue selected for an ROI, we image the entire breast with a multielement phased array coil. At readout
then the radiologist may select ROIs at any location in the breast to determine the ADCs. We use the twice-
refocused approach to reduce geometric distortions due to eddy currents. Patients are imaged prone with a 1.5-T
scanner (GE Signa Short-Bore Twin) using a phased-array eight element breast surface coil. The FOV is typically
24 cm × 24 cm but can be enlarged dependent upon requirement, slice thickness is 8 mm, and matrix size is 128 ×
2
128. We apply diffusion gradients in six directions with a b-value of 600 s/mm . Prior to ADC calculation, DWI
scans are corrected for distortions from eddy-current-induced warpings and magnitude averaged. Finally the mean
4 of 11 12-04-2010 00:35
or isotropic diffusion coefficient is calculated by averaging over all six directions.

page 310
page 311
It remains to be shown whether or not DWI provides additional information for the diagnostic work-up of patients
suffering from breast cancer. Further, more detailed studies are warranted, particularly in imaging different
carcinomas. It is also remains to be shown whether DWI and T2-weighted images provide the clinician with unique
sets of complementary information about breast tissue, as both relate to the tissue free water content. Our initial
findings demonstrate the effectiveness of DWI in separating malignant from benign lesions. However, these results
are preliminary and no sound conclusions can yet be drawn until further study is done.
Abdomen
Motivation
Diffusion-weighted imaging may be useful in imaging the abdominal organs for purposes of lesion detection and
131-134 133,135-139
characterization, as well as in the evaluation of diffuse parenchymal diseases. Application within the
abdomen, however, has been hindered by the presence of bulk physiologic motion-such as respiration, peristalsis,
and blood flow-which is orders of magnitude greater in amplitude than that of diffusion, but ultrafast MRI strategies
allow the effects of diffusion within the abdomen to be measured by essentially freezing bulk physiologic motion.
With recent developments in gradient hardware, both the diffusion-encoding time for a given b-value and EPI
readout time can now be substantially reduced. This is of particular interest when imaging tissues with short T2*
times and in areas prone to off-resonance.
Previous Studies
-3 2
Table 10-2. Apparent Diffusion Coefficients of the Abdominal Organs (×10 mm /s)*
Group Kidney Cortex Medulla Pancreas Liver Spleen
Namimoto et
1.63 ±0.34 0.69 ±0.31 0.78 ±0.35
al134
Ichikawa et 1.94
133 5.76 ±0.06 2.28 ±1.23 1.44 ±0.05
al ±0.19
Namimoto et 2.55 2.84
138 2.28 ±0.07
al ±0.62 ±0.72
Kim et al131 1.79/2.14 1.55/1.16 0.85/0.99
±0.49/0.44 ±0.37/0.42 ±0.25/0.54
141 2.89 2.18
Ries et al
±0.28 ±0.36
Chan145 1.98 ±0.37
147 1.64
Murtz et al 0.96 ±0.09 0.60 ±0.04
±0.09
*Data are mean ± SD.
Since the implementation of EPI imaging sequences into clinical MRI systems, there have been multiple preliminary
139
investigations of their application to abdominal DWI. In 1994, Muller et al reported that significantly different
ADCs were obtainable within normal liver, spleen, and muscle as well as within various focal and diffuse
abnormalities of the liver. This was followed by several reports of the application of DWI in the
liver,131,132,134,136,137,140 kidney,138,141-144 and various other organs (Table 10-2). However, image quality suffered
from the limitations of available hardware and pulse sequences at the time that these studies were performed. As
146
was addressed by Okada et al, the suppression of signal from vascular flow by even small diffusion gradients is
also very helpful in delineating lesions from the otherwise bright vessels. Typical ADC measurements for the solid
organ that were found in one of our previous studies are shown in Table 10-2. In this particular study, SNR ranged
from 27.0 in liver to 44.1 in kidneys for the diffusion-weighted images, and from 19.6 in liver to 39.0 in kidneys in
reference images. Apparent diffusion coefficients obtained in the renal medulla, renal cortex, liver, spleen, and
-6 -6 -6 -6
pancreas were (2091 ± 55) × 10 , (2580 ± 53) × 10 , (1697 ± 52) × 10 , (1047 ± 82) × 10 , and (2605 ± 168) ×
-6 2
10 mm /s, respectively (mean ± SE). The calculated diffusion anisotropy indices obtained in the kidney in one of
our studies suggests weakly anisotropic diffusion in the renal medulla (0.129) but not cortex (0.067), possibly as a
result of the radial orientation of blood vessels and collecting structures within the pyramids. This finding
5 of 11 12-04-2010 00:35
141-143
corroborates the results of other investigators and supports the contention that rotationally invariant diffusion
values should be measured in the kidneys. Although the anisotropy index that was used in our study is a less robust
method of determining anisotropy than are true diffusion tensor measurements, with current methods it is unrealistic
to generate sufficient SNR in 10 to 13 slices within a single breath-hold to accurately calculate the full tensor.
Diffusion-weighted imaging may be also able to provide functional information. Recently, we assessed the ability of
DWI to reflect and lateralize renal dysfunction by comparing renal ADC with serum creatinine values. We found that
there was a significant decrease in renal ADC in patients with renal insufficiency, and a linear correlation between
renal ADC and serum creatinine in patients without unilateral disease. There was a striking difference in ADCs
between normal and compromised kidneys in patients with unilateral renal disease. In patients with unilateral
disease, ADCs of the affected kidneys were even lower than those of kidneys with bilateral disease. Excluding the
patients with unilateral disease, the average ADCs obtained in the renal parenchyma of patients with normal
-6 2
creatinine (≤1.6 mg/dL) were (2437± 111) × 10 mm /s, whereas average ADCs obtained in the renal parenchyma
of patients with elevated serum creatinine levels was (2171 ± 214) × 10-6 mm2s. We believe that this technique has
great potential for the evaluation of patients with renal disease, particularly those with unilateral disease, such as
renal artery stenosis and congenital hydronephrosis, which may not be reflected in cruder measures of renal
function, such as serum creatinine.
page 311
page 312
Technical Considerations
Interleaving slices allows multiple sections with multiple numbers of excitation (NEX) to be obtained within a single
breath-hold. Due to its rapid acquisition capabilities, one of the greatest benefits of single-shot EPI, combined with
DWI, is that the phase-perturbations produced by motion are equal in each phase-encoding step and, therefore, do
not lead to ghosting artifacts. As mentioned earlier, T2* relaxation during the EPI readout can lead to considerable
image blurring, especially in tissues like the liver. A rapid k-space traversal can be achieved with powerful gradients
or by interleaving the EPI trajectory. Interleaved (multishot) EPI has been shown to be a potential alternative to
improve image quality, but it requires phase navigation. Such navigation strategies have been helpful in reducing
ghosting artifacts in the brain, but in the abdomen, where there is more extensive motion, the effectiveness of
navigator echoes is likely to be limited.
At our institution, we routinely use a diffusion-weighted single-shot EPI sequence. In order to minimize image
blurring and off-resonance effects, the rapidity of the k-space readout is maximized and TE is minimized by a
combination of factors including the use of ramp sampling, a rectangular FOV, partial k-space acquisition, a system
with high performance magnetic field gradients (40 mT/m), and phased-array capabilities. If available, further speed
and image quality improvement can be achieved by including SENSE technology. Typical scanning parameters are
the following: FOV = 30 × 30 cm to 40 × 40 cm; TR = 2600 to 3200 ms; TE = 45.3 to 55.1 ms, matrix 128 × 128;
2
bandwidth ±125 kHz, and b-values ranging between 300 and 600 s/mm . A very short echo time is particularly
important in the abdomen, where organs such as the liver have very short T2 relaxation times that can clearly limit
the available SNR. In order to achieve maximum SNR, we usually use a diffusion encoding gradient combination that
provides us with the shortest possible TE. Here, a protocol is used that plays out diffusion gradients along all three
axes simultaneously (tetrahedral encoding), providing a square root of three stronger gradients than single axis
encoding. In order to suppress the effects of perfusion on the diffusion-weighted images, a small diffusion-weighting
2
(b ≈ 50 s/mm ) can be used on the reference scans in order to dephase inflowing spins. Furthermore, by acquiring
multiple signal averages for each diffusion-weighted image, pulsatility effects cancel out and the SNR can be
increased. This strategy permitted the acquisition of 10 to 13 images in a single breath-hold. To further minimize the
influence from pulsatile motion, pulse-triggering in conjunction with single-shot diffusion-weighted EPI can be
147
used. This approach limits the rate of data acquisition, ultimately yielding fewer signal averages and far fewer
image slices during a single breath-hold, which are compromises in both SNR and anatomic coverage.
6 of 11 12-04-2010 00:35
Add to lightbox
Figure 10-21 Isotropic diffusion-weighted images with b = 50 s/mm2 (top row) and b = 350 s/mm2 (middle row) and
corresponding map of the diffusion coefficient (bottom row). While the healthy left kidney (curved arrow)
demonstrates normal signal characteristics, the right kidney shows markedly increased signal in the diffusion-
weighted scan (straight arrow) and considerably reduced diffusivity in the cortex, which is most likely due to
ischemia. Conversely, the medulla appears hypointense in the maps of the diffusion coefficient and the tumor mass
itself is relatively heterogenous in signal characteristics.
While early work has shown potential for DWI in the evaluation of both focal and diffuse diseases of the liver and
kidneys, further work in this area is necessary to clarify its true clinical utility and indications. Given the sensitivity of
DWI to cerebral ischemia, one may surmise that DWI may be useful for detecting ischemia in abdominal organs,
such as the kidneys. In this context, we recently found a dramatically altered diffusion condition in the cortex of one
kidney of a patient with an obstructing renal cell carcinoma. Magnetic resonance angiography and subsequent
histopathological examination clearly revealed vascular compromise in the diseased kidney. The DWI scans and
corresponding coefficient maps clearly revealed markedly decreased diffusion in the obstructed kidney (Fig. 10-21).
Additionally, ADC is a physical tissue property that could help in the characterization of focal lesions or diffuse
parenchymal disease by MRI, just as T1, T2, and enhancement characteristics are currently used. We believe that
the ease and rapidity of this technique, coupled with the high image quality and quantitative diffusion data, make it
desirable for clinical use as a simple add-on to abdominal MRI examinations, ultimately contributing to our greater
collective understanding of the role of DWI in the abdomen.
Spine and Spinal Column

Diffusion-Weighted Imaging in the Spinal Column
page 312
page 313
Because of its high vascularity, the spine is a common target site for metastatic spread. 148 Hence, vertebral
metastases are frequently observed in patients with cancer (30-70%) and, besides pain, they may lead to various
secondary problems, which range from fractures to compression of the dural sac by epidural masses. During
recent years, MRI has evolved into one of the preferred methods for detecting vertebral metastasis. This is mainly
because MRI rapidly detects normal bone marrow replacement with more hypercellular tissue or with other
149,150
pathologies that cause a higher water content. However, the specificity of conventional MRI is often poor and
does not allow for discrimination of whether the cause of an acute vertebral compression fracture is osteoporosis
or metastasis.151,152 Occasionally, morphological signs-such as complete replacement of vertebral marrow,
involvement of the posterior elements, and epidural or paraspinal masses-can be used to improve the diagnostic
accuracy, but they may be equivocal.
Results from recent studies raise hope that DWI might be able to differentiate benign from malignant acute
vertebral fractures.20,153-156 Specifically, it has been reasoned that proton diffusivity is elevated in osteoporotic
20
fractures because of bone marrow edema, whereas metastatic lesions might change diffusivity only moderately
or even decrease it. In this context, Zhou et al156 postulated that the high cellularity of metastatic lesions, especially
that of actively growing tumors, would increase their intracellular volume fraction relative to the interstitial space.
Because the water diffusion coefficient is approximately 10 times lower in the intracellular space than that in the
7 of 11 12-04-2010 00:35
extracellular space, this should lower the ADC values of metastatic vertebral infiltration.
Baur et al20 reported that osteoporotic fractures (n = 22) generally showed hypo- or isointense signal on diffusion-
weighted steady-state free precession (SSFP) sequences relative to adjacent unaffected bone marrow and found
that metastatic fractures (n = 17) appeared hyperintense on SSFP DWI. In a different study, they found signal
hyperintensities on DWI scans in all 25 malignant vertebral fractures, while 35 out of 38 benign fractures appeared
153
hypo- or isointense ; however, 3 out of the 38 benign lesions were false positives, showing hyperintense signal on
DWI scans. Similar specificities of DWI regarding the separation of benign and pathologic vertebral fractures were
157 158
reported by Tasaly et al and Matoba et al, although they were in slightly different patient populations.
Conversely, in a consecutive series of fifteen patients with proven malignant vertebral lesions, Castillo et al159 found
that these lesions appeared either hypo- (n = 5), hyper- (n = 3), or isointense (n = 2) and concluded that SSFP
160
DWI does not offer any advantage over conventional MRI. In a recent review article, Baur et al clarified that the
patients enrolled in the Castillo study are not the primary target group for DWI in spine (i.e., patients with sclerotic
metastases and previously treated metastases). Baur and coworkers suggested the following inclusion criteria: 1.
unknown reason for the vertebral collapse; 2. lack of sclerotic metastases; and 3. no prior therapy.
Unfortunately, all the aforementioned studies have been performed without absolute quantification of the diffusion
coefficient. Hence, the diagnosis was based solely on the signal intensities of DWI examinations and the diffusion
information was obscured by confounding effects on the MR contrast, such as T2-shine-through or T2* effects.
Another problem that led to some controversy about the diagnostic validity of SSFP diffusion sequences emerged
from the fact that SSFP diffusion sequences lack the ability to absolutely quantify diffusion. This is mainly because
the final echo formation of SSFP is comprised of different echo components and the magnitude of these different
contributions depends heavily on the underlying relaxation and sequence parameters (e.g., TR, TE, and flip angle).
Further reports from (semi)-quantitative diffusion measurements have since been reported: Spuentrup et al155
analyzed 35 lesions (18 acute osteoporotic or traumatic fractures; 17 untreated malignant infiltrations) and found
significant differences between malignant lesions with and without fractures in DWI scans that were normalized to
unweighted images. Significant relative signal loss was also found for osteoporotic and traumatic fractures,
156
whereas only small changes were found in metastatic lesions. Truly quantitative measurements by Zhou et al
have demonstrated that acute benign vertebral fractures (n = 12) have higher diffusivity than malignant lesions (n =
15). However, they also showed that both groups overlap considerably. DWI of the benign fractures demonstrated
hypo- (n = 1), iso- (n = 5), and hyperintense lesions (n = 6). The isotropic diffusion coefficients of the group ranged
-6 -6 2
from 240 × 10 to 350 × 10 mm /s and were slightly lower than values of normal vertebral bodies (270-330 ×
-6 2
10 mm /s) in the same subjects. On the other hand, DWI of the metastatic vertebral lesions demonstrated hypo-
(n = 4), iso- (n = 1), hyperintense (n = 9), and heterogeneous (n = 1) lesions. The isotropic diffusion coefficients of
-6 -6 2
this group ranged from 130 × 10 to 200 × 10 mm /s and were, therefore, considerably lower than values of
normal vertebral bodies in the same subjects (270-330 × 10-6 mm2/s). Zhou et al156 concluded from this that DWI
intensity values alone were highly unspecific but that measurement of the diffusion coefficient improved diagnostic
161
specificity. Recently, Herneth et al quantified ADC values in 22 patients with vertebral compression fractures and
also showed that vertebral metastases had ADC values that were significantly lower than normal vertebrae (690
-6 2 -6 2
±240 × 10 mm /s versus 1660 ±380 × 10 mm /s). Likewise, pathologic compression fractures had significantly
-6 2 -6 2
lower ADC than had benign compression fractures (710 ±270 × 10 mm /s versus 1610 ±370 × 10 mm /s). The
156
absolute ADC numbers, however, differed significantly from the results observed by Zhou et al. One possible
161
explanation for this effect is that Herneth et al used fat suppression for their IEPI acquisition whereas the other
group did not. The consequences for ADC measurements depending upon the usage of fat suppression has already
been addressed.
page 313
page 314
8 of 11 12-04-2010 00:35
Add to lightbox
Figure 10-22 Spinal cord ischemia. A, T2-weighted sagittal MRI 30 hours after symptom onset shows a
hyperintense lesion with edematous expansion (arrows) in the central portion of the lumbar cord and the conus
medullaris. B, Diffusion-weighted (b = 1000 s/mm2) MRI demonstrates strong hyperintensity in this region. The
corresponding apparent diffusion coefficient was markedly reduced (490 × 10-6 mm2/s) confirming the ischemic
pathology. Corresponding T2-weighted MRI (C) showed only faint signal hyperintensities (arrow). Infarction of the
spinal cord can be also clearly seen on axial DWI scans (E and F: arrows) in another subject with acute spinal cord
ischemia. Similarly, the diffusion was reduced in the ischemic lesions. Axial gradient-echo scans (D) did not show
any signal changes at all. (A-C, images courtesy A. Gass, University of Mannheim, Germany; D-E, images
courtesy F. Fazekas, University of Graz, Austria.)
Besides the apparent potential of DWI to increase diagnostic specificity, a recent study has suggested that DWI
can also improve the monitoring of therapeutic efficacy for metastatic vertebral spread. A change in ADC of
162
infiltrated areas may reflect therapeutic effects that are otherwise difficult to establish in an objective manner. In
162
this context, Byun et al demonstrated that DWI shows decreasing signal intensity of metastatic disease of the
vertebral bone marrow (23 out of 24 patients) with successful therapy. Unfortunately, these results were compared
against only one patient with treatment failure. Thus, DWI as an adjunct in monitoring treatment response appears
to be promising, but studies of larger patient groups and more quantitative results are certainly necessary for more
sound conclusions.
At our institution, we perform LSDI on all patients that undergo vertebroplasty and biopsy. The relevant protocol
parameters are as follows: FOV = 67.5 × 270 mm, acquisition matrix = 128 × 256, slice thickness = 6 mm, skip =
0, TE = 39.2 ms, TRsweep = 1900 ms, TRline-to-line = 149 ms (shortest line-to-line TR), number of excitations = 6,
number of slices = 3, receiver bandwidth = ±15.63 kHz, tetrahedral encoding, and b ≈ 650 s/mm2. The total
acquisition time is about 6.5 minutes (Fig. 10-22).
Diffusion-Weighted Imaging in the Spinal Cord

As in the brain, DWI of the spinal cord could contribute both to the diagnostic process and to the understanding of
pathophysiologic processes in many disorders of the spinal cord. Damage to the spinal cord may be caused by a
wide range of pathologies, and can result in profound functional disability. Diffusion-weighted imaging may improve
our understanding of the nature and evolution of structural damage in various types of spinal cord disease. Possible
applications include improved detection of ischemic lesions, clarification of the relationship between clinical disability
and structural damage to the cord, and monitoring of neuroprotective or anti-inflammatory therapies.
page 314
page 315
Quantitative diffusion measurements in healthy volunteers demonstrate diffusion coefficients in the spinal cord
9 of 11 12-04-2010 00:35
88,163-166
comparable to those of the brain and also indicate diffusion anisotropy. We found that the ratio between
diffusion along the fibers and across them differed by at least a factor of two.164 This directional preference allows
one to visualize major fiber bundles, such as the corticospinal tract or transverse bundles in the pontine region. In
the clinical setting, several aspects challenge for the radiologist: First, the small size of the spinal cord and the
adjacent CSF space makes it difficult at times to exactly quantify diffusion and to distinguish between gray and
white matter. Second, partial volume effects have to be considered and can affect measurements. Here especially,
the rapid transition from CSF space to the spinal cord can cause ringing artifacts which, if uncorrected, can falsify
DWI measurements.165 Third, anisotropic diffusion is characterized most accurately by DTI.165 However, due to the
aforementioned problems, DTI of the spinal cord appears much more difficult than conventional DWI, and one still
has to be cautious when interpreting new results.
Diffusion-weighted imaging may be equally well suited to detecting spinal cord infarction as it is for the
neurocranium. Conventional MRI can be sensitive to intramedullary abnormalities following spinal cord ischemia only
after days. Even then it is often hard to discriminate such changes from those caused by other etiologies such as
myelitis. Such information, however, would be invaluable in planning more aggressive therapies. So far, only a few
164,167,168
reports have been published about reduced diffusion in spinal cord infarction. Unfortunately, all of the
studies have investigated no more than two patients. This reflects existing difficulties in applying DWI to the spinal
cord in the routine clinical setting.
If possible, a navigated interleaved EPI or another robust technique that provides high spatial definition should be
used to perform a DWI examination. Because of its robustness against motion, a phased-array compatible
PROPELLER DWI sequence may be one of these methods. Also, LSDI has been proven to be very well suited to
image the spine. The long extension of the spine along one dimension makes it an ideal object to be imaged with
LSDI, since only a few lines are required to cover the entire width of the spinal cord. Likewise, navigated diffusion-
weighted interleaved EPI in concert with pulse triggering has demonstrated its usefulness to image the spinal cord
with high definition and at higher speed than have the aforementioned methods. A typical navigated diffusion-
weighted interleaved EPI sequence protocol is as follows: FOV = 28 cm, TR = 2 RR intervals, pulse triggering, TE
= minimum, matrix 192 × 256, EPI train length 13 to 15, fat suppression, tetrahedral encoding, NEX = 2, slice
2
thickness = 4 mm with 0.4 mm gap, and b ≈ 600-800 s/mm . Alternatively, a single-shot EPI protocol with SENSE
169
reduction can be utilized to image the spinal cord.
Besides its application in ischemia, DWI appears promising for trauma patients; while DWI of human spinal cord
injury is still in its infancy, it has already been performed successfully for a longer period in animals. The initial cord
injury and axonal transection cause only part of the functional deficits that ultimately occur. Increased functional loss
is related to "secondary injury", an immune response which can persist for several days and results in increased
170
lesional size, swelling, and ultimately the additional degeneration of axonal fiber tracts. The exact stage of
traumatic injury is often difficult to characterize by conventional MRI, whose ultimate goal is to test for the functional
integrity of the axons within the white matter tracts of the spinal cord. Wallerian degeneration above and below the
site of injury is known to be indicative of axonal loss, but on conventional MRI it occurs only with advanced
progression of tissue damage and is not differentiable from edema; diffusion-related parameters obtained from
DWI might be better able to define the type and extent of spinal cord injury earlier than conventional MRI, as
different pathophysiologies may affect diffusion properties differently. For example, in so-called weight drop
experiments in rodents, it has been shown that diffusion anisotropy is a very sensitive marker for damage to the
spinal cord. Here, reduced anisotropy loss or even increased anisotropy was observed if neuroprotective therapy
171
was applied. The latter was explained by the possible enhancement of myelin survival and the diminished
occurrence of cysts, which have increased isotropic diffusivity. Spinal trauma can be complicated further if a
condition known as syringomyelia develops; syringomyelia represents a cystic cavitation within the center of the
spinal cord. In animal models, it has been shown that ADC changes could already be seen on ADC maps after 1
week, which allows for a much earlier and more aggressive treatment than does conventional MRI, which was first
positive 4 weeks after the injury.172
Diffusion-weighted imaging may develop into an important diagnostic adjunct for multiple sclerosis (MS). Focal
lesions within the myelin tend to have great clinical consequences173 because the densely packed fibers traffic to
and from the extremities. Remote fiber changes, resulting from multifocal supratentorial diseases or from motor
neuron disorders, concentrate in the spinal cord. This has led to speculations that disability may correlate best with
174
spinal lesions. We have found that these lesions usually present increased rates of diffusivity. This has been
175
confirmed in other reports by Clark et al who, in their study, found significantly elevated isotropic diffusion
-6 2 -6 2
relative to controls (1180 ±120 × 10 mm /s [n = 4] versus 910 ±50 × 10 mm /s [n = 3]).
In amyotrophic lateral sclerosis (ALS), a chronic degenerative disorder of unknown etiology that affects the first
and second motor neurons, conventional MRI demonstrates only poor specificity. Diffusion-weighted imaging
10 of 11 12-04-2010 00:35
promises to improve diagnostic capabilities for ALS because of its unique sensitivity for the dissolution of neuronal
176
tracts. Recently, Ellis et al showed that isotropic diffusion of the internal capsule was significantly increased,
whereas fractional anisotropy was significantly decreased in patients with ALS relative to healthy controls. Since
ALS affects the whole corticospinal tract, similar changes in diffusion properties might also be found in the spinal
cord.
page 315
page 316
Degenerative spondylosis may cause significant narrowing of the spinal canal and ultimately can lead to
compression of the spinal cord. Protruded or herniated discs especially can also compromise the spinal cord.
Clinically, such patients usually present with chronic progressive signs of myelopathy. Early diagnosis is of
164
paramount importance to prevent severe disability. In our studies, we found that spondylotic myelopathy
presented with reduced ADC values, whereas the surrounding cord demonstrated elevated diffusivity. The former is
presumably due either to cord compression or to vascular compromise, while the latter is due to surrounding
165
edema. Ries et al recently studied a patient suffering from severe spinal stenosis with cord compression. In
contrast to conventional MRI, the spatial extent of diffusion abnormalities was much larger. When interpreting DWI
scans of patients with spondylotic myelopathy, one should consider the extent of susceptibility differences (caused
by protruded discs or osteophytes that have moved toward the anterior aspect of the cord) that could lead to false
positive readings.
Because of the altered cellular matrix of neoplastic tissue, DWI may add important information to the staging of
tumors, or at least it may help to differentiate different types of neoplasm. Encouraging results have been shown in
the brain, where researchers have reported the ability of DWI to differentiate between cerebral tumor types177;
similar results might be anticipated for the spine. In high grade, heterogeneous tumors such as glioblastoma
multiforme or high grade astrocytomas, however, ADC values can vary over a large range, 178,179 and a general
differentiation based on ADC could be difficult. Fiber tracking, on the other hand, could shed light onto whether a
tumor invades or displaces fiber tracts. This may have consequences for the planning of surgical intervention.
Functional Diffusion-Weighted Imaging in Brain Mapping

The role of functional magnetic resonance imaging (fMRI) has become increasingly important in neuroimaging, for
example in the presurgical mapping of gray matter.180 Contrast mechanisms based on the blood oxygenation level,
volume, and flow changes have been used to noninvasively detect brain activation secondary to the neuronal
activity. Recently, several efforts have focused on alternative contrast mechanisms that may offer shorter temporal
delays and more direct spatial localization in brain mapping. These efforts include detection of Lorentzian forces
from neural activation from phase-sensitive images, use of diffusion weighting to sensitize the image to changes in
incoherent displacements, and mapping small changes in the ADC from neural activity.
Phantom studies now suggest the possibility of detecting the destructive phase addition from the spatially
incoherent, yet temporally synchronized, displacements caused by the Lorentz force experienced during electrical
181
conduction within a strong magnetic field.
Another approach has employed heavy diffusion weighting to remove the vascular signal and sensitize the minute
182
and incoherent displacement in order to detect fast dynamic signal changes synchronized to a task. Li and Song
observed fast functional signal changes upon performance of a task by using a heavy diffusion-weighting protocol,
which possessed a different time course from the corresponding BOLD response, suggesting a different origin from
the common BOLD signal sources. The authors conclude that the characteristics of the signal changes suggest that
they may be more directly linked to the neuronal activity temporally and spatially than the BOLD response. Song et
al have recently expanded this approach to include the mapping of the diffusion tensor to detect synchronized fMRI
183
signal changes during brain activation.
Le Bihan and colleagues observed changes in the apparent diffusion coefficient of water in the visual cortex of the
human brain during activation by a flickering checkerboard task activation paradigm. The ADC decrease was less
than 1% but was significant and reproducible, and closely followed the time course of the activation paradigm. The
observed ADC findings were ascribed to a transient swelling of cortical cells that occurred with neural activation. 184
These preliminary results from a variety of new imaging techniques suggest a new approach to produce images of
brain activation with MRI from signals directly associated with neuronal activation, and not through changes in local
blood flow. More importantly, the methods may in the near future provide noninvasive maps of the neural activation
pathways in the CNS.
© 2010 Elsevier
11 of 11 12-04-2010 00:35
CONCLUSION AND FUTURE OUTLOOK

Diffusion-weighted imaging has seen rapid growth and development, quickly escalating it from an
experimental tool to an established clinical methodology whose primary use has been for the evaluation
of acute cerebral ischemia. Much like T1 and T2 relaxation, diffusivity can be thought of as an intrinsic
tissue property. Thus, DWI may also be useful in imaging extracranial organs, such as the solid organs
within the abdomen, or abnormalities in the musculoskeletal system. The ability to determine diffusion
coefficients in vivo has great potential for furthering our understanding of normal and abnormal
physiology, as well as for characterizing focal and diffuse disease within the human body. Historically,
bulk physiologic motion has hampered the application of DWI to a wide variety of clinical questions.
However, developments in MR hardware, pulse sequences, and computer science have led to a robust
tool with reasonable image quality and outstanding diagnostic potential. With the most recent technical
advances, further clinical trails are now warranted to prove the ability of DWI to facilitate the diagnostic
work-up of other diseases.
Acknowledgements
The authors acknowledge their support from the National Institutes of Health (1R01EB002771,
1R01NS35959), the Center of Advanced MR Technology at Stanford (P41RR09784), and the Lucas
Foundation. We are also grateful to Drs Chow, Chang, Daniel, Do, and Atlas (Stanford University) and
Drs Fazekas and Augustin (Karl-Franzens University of Graz).
REFERENCES
1. Moseley ME, Cohen Y, Mintorovitch J, et al: Early detection of regional cerebral ischemia in cats: Comparison of diffusion-
and T2-weighted MRI and spectroscopy. Magn Reson Med 14:330-346, 1990. Medline Similar articles
2. Lansberg MG, Norbash AM, Marks MP, et al: Advantages of adding diffusion-weighted magnetic resonance imaging to
conventional magnetic resonance imaging for evaluating acute stroke. Arch Neurol 57:1311-1316, 2000. Medline
Similar articles
3. Nakahara M, Ericson K, Bellander BM: Diffusion-weighted MR and apparent diffusion coefficient in the evaluation of severe
brain injury. Acta Radiol 42:365-369, 2001. Medline Similar articles
page 316
page 317
4. Filippi M, Cercignani M, Inglese M, et al: Diffusion tensor magnetic resonance imaging in multiple sclerosis. Neurology
5. Bammer R, Fazekas F: Diffusion imaging in multiple sclerosis. Neuroimaging Clin N Am. 12:71-106, 2002. Medline
Similar articles
6. Klingberg T, Hedehus M, Temple E, et al: Microstructure of temporo-parietal white matter as a basis for reading ability:
Evidence from diffusion tensor magnetic resonance imaging. Neuron 25:493-500, 2000. Medline Similar articles
7. Basser PJ, Mattiello J, LeBihan D: Estimation of the effective self-diffusion tensor from the nmr spin echo. J Magn Reson B
8. Basser PJ: Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. NMR
Biomed 8:333-344, 1995. Medline Similar articles
9. Mori S, Crain BJ, Chacko VP, van Zijl PC: Three-dimensional tracking of axonal projections in the brain by magnetic
resonance imaging. Ann Neurol 45:265-269, 1999. Medline Similar articles
10. Le Bihan D: Clinical perfusion and diffusion imaging. New York: Raven Press, 1993.
11. Ogawa S, Lee TM, Nayak AS, Glynn P: Oxygenation-sensitive contrast in magnetic resonance image of rodent brain at
high magnetic fields. Magn Reson Med 14:68-78, 1990. Medline Similar articles
12. Crank J: The mathematics of diffusion. New York: Oxford University Press, 1956.
13. Gudbjartsson H, Patz S: Simultaneous calculation of flow and diffusion sensitivity in steady-state free precession imaging.
14. Carr HY, Purcell EM: Effects of diffusion on free precession in nuclear magnetic resonance experiments. Phys Rev
94:630-638, 1954.
15. Stejskal EO, Tanner JE: Spin diffusion measurements: Spin echoes in the presence of a time-dependent field gradient. J
Chem Phys 42:288-292, 1965.
16. Merboldt KD, Hanicke W, Bruhn H, et al: Diffusion imaging of the human brain in vivo using high-speed steam MRI. Magn
17. Le Bihan D, Turner R, MacFall JR: Effects of intravoxel incoherent motions (IVIM) in steady-state free precession (SSFP)
imaging: Application to molecular diffusion imaging. Magn Reson Med 10:324-337, 1989. Medline Similar articles
1 of 7 12-04-2010 00:36
18. Wu E, Buxton R: Effects of diffusion on the steady-state magnetization with pulsed field gradients. J Magn Reson
90:243-253, 1990.
19. Buxton RB: The diffusion sensitivity of fast steady-state free precession imaging. Magn Reson Med 29:235-243, 1993.
20. Baur A, Stabler A, Bruning R, et al: Diffusion-weighted MR imaging of bone marrow: Differentiation of benign versus
pathologic compression fractures. Radiology 207:349-356, 1998. Medline Similar articles
21. Miller KL, Pauly JM: Nonlinear navigated motion correction for diffusion imaging. 10th Annual Meeting of ISMRM, Honolulu,
HI, 2002. p1110.
22. Miller KL, Meyer CM, Pauly JM: Self-navigated spirals for high resolution steady-state diffusion imaging. 10th Annual
Meeting of ISMRM, Honolulu, HI, 2002, p 1110.
23. Hennig J: Echoes-how to generate, recognize, use or avoid them in MR-imaging sequences: Part 1. Concepts Magn
Reson 3:125-143, 1991.
24. Hennig J: Echoes-how to generate, recognize, use or avoid them in MR-imaging sequences: Part 2. Concepts Magn
Reson 3:179-192, 1991.
25. Scheffler K: A pictorial description of steady-states in rapid magnetic resonance imaging. Concepts Magn Reson
11:291-304, 1999.
26. Basser PJ, Pierpaoli C: Microstructural and physiological features of tissues elucidated by quantitative diffusion tensor
MRI. J Magn Reson B 111:209-219, 1996. Medline Similar articles
27. Beaulieu C, Fenrich FR, Allen PS: Multicomponent water proton transverse relaxation and T2-discriminated water diffusion
in myelinated and nonmyelinated nerve. Magn Reson Imaging 16:1201-1210, 1998. Medline Similar articles
28. Wimberger DM, Roberts TP, Barkovich AJ, et al: Identification of "premyelination" by diffusion-weighted MRI. J Comput
29. Frank LR: Anisotropy in high angular resolution diffusion-weighted MRI. Magn Reson Med 45:935-939, 2001. Medline
Similar articles
30. Frank LR: Characterization of anisotropy in high angular resolution diffusion-weighted MRI. Magn Reson Med
31. Fisher E, Cothren R, Tkach J: Knowledge-based 3D segmentation of MR images for quantitative MS lesion tracking. Proc
SPIE Medical Imaging 3034:19-25, 1997.
32. Mulkern RV, Gudbjartsson H, Westin CF, et al: Multi-component apparent diffusion coefficients in human brain. NMR
33. Liu C, Bammer R, Moseley ME: Multi-exponential analysis of gray and white matter structures in human brain. 10th annual
meeting of ISMRM, Honolulu, HI, 2002, p 1151.
34. Peled S, Cory DG, Raymond SA, et al: Water diffusion, t(2), and compartmentation in frog sciatic nerve. Magn Reson Med
35. Sehy JV, Ackerman JJ, Neil JJ: Evidence that both fast and slow water ADC components arise from intracellular space.
36. Sukstanskii AL, Yablonkiy DA: Effects of restricted diffusion on MR signal formation. J Magn Reson 157:92-105, 2002.
37. Callaghan PT, Coy A, MacGowan D, et al: Diffraction-like effects in NMR diffusion studies of fluids in porous solids. Nature
351:467-469, 1991.
38. Callaghan PT: Principles of nuclear magnetic resonance microscopy, Oxford: Clarendon Press, 1993.
39. Assaf Y, Cohen Y: Structural information in neuronal tissue as revealed by q-space diffusion NMR spectroscopy of
metabolites in bovine optic nerve. NMR Biomed 12:335-344, 1999. Medline Similar articles
40. Mintorovitch J, Moseley ME, Chileuitt L, et al: Comparison of diffusion- and T2-weighted MRI for the early detection of
cerebral ischemia and reperfusion in rats. Magn Reson Med 18:39-50, 1991. Medline Similar articles
41. Moseley ME, Kucharczyk J, Mintorovitch J, et al: Diffusion-weighted MR imaging of acute stroke: Correlation with
T2-weighted and magnetic susceptibility-enhanced MR imaging in cats. Am J Neuroradiol 11:423-429, 1990. Medline
Similar articles
42. Hossmann KA: Development and resolution of ischemic brain swelling. In Pappius HM, Feindel W (eds): Dynamics of brain
edema. Heidelberg, Germany: Springer Verlag, 1976.
43. Verheul HB, Balazs R, Berkelbach van der Sprenkel JW, et al: Comparison of diffusion-weighted MRI with changes in cell
volume in a rat model of brain injury. NMR Biomed 7:96-100, 1994. Medline Similar articles
44. Le Bihan D: Molecular diffusion nuclear magnetic resonance imaging. Magn Reson Q 7:1-30, 1991. Medline
Similar articles
46. van Gelderen P, de Vleeschouwer MH, DesPres D, et al: Water diffusion and acute stroke. Magn Reson Med 31:154-163,
2 of 7 12-04-2010 00:36
47. Zhong J, Petroff OA, Prichard JW, Gore JC: Changes in water diffusion and relaxation properties of rat cerebrum during
status epilepticus. Magn Reson Med 30:241-246, 1993. Medline Similar articles
48. Benveniste H, Hedlund LW, Johnson GA: Mechanism of detection of acute cerebral ischemia in rats by diffusion-weighted
magnetic resonance microscopy. Stroke 23:746-754, 1992. Medline Similar articles
49. Anderson AW, Zhong J, Petroff OA, et al: Effects of osmotically driven cell volume changes on diffusion-weighted imaging
of the rat optic nerve. Magn Reson Med 35:162-167, 1996. Medline Similar articles
50. Helpern JA, Ordidge RJ, Knight RA: The effect of cell membrane water permeability on the apparent diffusion coefficient of
water. 11th Annual Meeting of SMRM, 1992, p 1201.
51. Waldeck AR, Nouri-Sorkhabi MH, Sullivan DR, Kuchel PW: Effects of cholesterol on transmembrane water diffusion in
human erythrocytes measured using pulsed field gradient. NMR Biophys Chem 55:197-208, 1995.
52. Pfeuffer J, Bröer S, Bröer A, et al: Water channels in glial cells detected by diffusion-weighted 1H NMR spectroscopy. 5th
Annual Meeting of ISMRM, Vancouver, Canada, 1997, p 508.
53. Sykova E, Svoboda J, Polak J, Chvatal A: Extracellular volume fraction and diffusion characteristics during progressive
ischemia and terminal anoxia in the spinal cord of the rat. J Cereb Blood Flow Metab 14:301-311, 1994. Medline
Similar articles
54. Norris DG, Niendorf T, Leibfritz D: Health and infarcted brain tissues studied at short diffusion times: The origins of
apparent restriction and the reduction in apparent diffusion coefficient. NMR Biomed 7:304-310, 1994. Medline
Similar articles
55. Latour LL, Svoboda K, Mitra PP, Sotak CH: Time-dependent diffusion of water in biological model system. Proc Natl Acad
Sci 91:1229-1233, 1994. Medline Similar articles
56. van der Toorn A, Sykova E, Dijkhuizen RM, et al: Dynamic changes in water ADC, energy metabolism, extracellular space
volume, and tortuosity in neonatal rat brain during global ischemia. Magn Reson Med 36:52-60, 1996. Medline
Similar articles
57. Wick M, Nagatomo Y, Prielmeier F, Frahm J: Alteration of intracellular metabolite diffusion in rat brain in vivo during
ischemia and reperfusion. Stroke 26:1930-1933 [discussion 1934], 1995. Medline Similar articles
58. Duong TQ, Ackerman JJ, Ying HS, Neil JJ: Evaluation of extra- and intracellular apparent diffusion in normal and globally
ischemic rat brain via 19F NMR. Magn Reson Med 40:1-13, 1998.
59. Neil JJ, Duong TQ, Ackerman JJ: Evaluation of intracellular diffusion in normal and globally-ischemic rat brain via 133Cs
NMR. Magn Reson Med 35:329-335, 1996.
60. Anderson AW, Gore JC: Analysis and correction of motion artifacts in diffusion weighted imaging. Magn Reson Med
61. Greitz D, Wirestam R, Franck A, et al: Pulsatile brain movement and associated hydrodynamics studied by magnetic
resonance phase imaging. The Monro-Kellie doctrine revisited. Neuroradiology 34:370-380, 1992. Medline
Similar articles
62. Poncelet BP, Wedeen VJ, Weisskoff RM, Cohen MS: Brain parenchyma motion: Measurement with cine echo-planar MR
63. Gudbjartsson H, Maier SE, Jolesz FA: Double line scan diffusion imaging. Magn Reson Med 38:101-109, 1997.
64. Gudbjartsson H, Maier SE, Mulkern RV, et al: Line scan diffusion imaging. Magn Reson Med 36:509-519, 1996.
65. Trouard TP, Theilmann RJ, Altbach MI, Gmitro AF: High-resolution diffusion imaging with DIFRAD-FSE (diffusion-weighted
radial acquisition of data with fast spin-echo) MRI. Magn Reson Med 42:11-18, 1999. Medline Similar articles
66. Bammer R, Glover GH, Moseley ME: Diffusion tensor spiral imaging. 10th Annual Meeting of the International Society of
Magnetic Resonance in Medicine, Honolulu, HI, 2002, p 1111.
67. de Crespigny AJ, Marks MP, Enzmann DR, Moseley ME: Navigated diffusion imaging of normal and ischemic human
brain. Magn Reson Med 33:720-728, 1995. Medline Similar articles
68. Butts K, de Crespigny A, Pauly JM, Moseley M: Diffusion-weighted interleaved echo-planar imaging with a pair of
orthogonal navigator echoes. Magn Reson Med 35:763-770, 1996. Medline Similar articles
page 317
page 318
69. Butts K, Pauly J, de Crespigny A, Moseley M: Isotropic diffusion-weighted and spiral-navigated interleaved epi for routine
imaging of acute stroke. Magn Reson Med 38:741-749, 1997. Medline Similar articles
70. Bammer R, Stollberger R, Augustin M, et al: Diffusion-weighted imaging with navigated interleaved echo-planar imaging
and a conventional gradient system. Radiology 211:799-806, 1999. Medline Similar articles
71. Atkinson D, Porter DA, Hill DL, et al: Sampling and reconstruction effects due to motion in diffusion-weighted interleaved
echo planar imaging. Magn Reson Med 44:101-109, 2000. Medline Similar articles
72. Pipe JG, Farthing VG, Forbes KP: Multishot diffusion-weighted FSE using propeller MRI. Magn Reson Med 47:42-52,
73. Mansfield P: Multi-planar image formation using NMR spin echoes. J Phys C 10:L55-L58, 1977.
3 of 7 12-04-2010 00:36
74. Turner R, Le Bihan D, Maier J, et al: Echo-planar imaging of intravoxel incoherent motion. Radiology 177:407-414, 1990.
75. Farzaneh F, Riederer SJ, Pelc NJ: Analysis of T2 limitations and off-resonance effects on spatial resolution and artifacts in
echo-planar imaging. Magn Reson Med 14:123-139, 1990. Medline Similar articles
76. Griswold MA, Jakob PM, Heidemann RM, et al: Generalized autocalibrating partially parallel acquisitions (GRAPPA).
77. Griswold MA, Jakob PM, Nittka M, et al: Partially parallel imaging with localized sensitivities (PILS). Magn Reson Med
78. Griswold MA, Jakob PM, Chen Q, et al: Resolution enhancement in single-shot imaging using simultaneous acquisition of
spatial harmonics (SMASH). Magn Reson Med. 41:1236-1245, 1999. Medline Similar articles
79. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P: SENSE: Sensitivity encoding for fast MRI. Magn Reson Med
80. Pruessmann KP, Weiger M, Bornert P, Boesiger P: Advances in sensitivity encoding with arbitrary k-space trajectories.
81. Sodickson DK, Manning WJ: Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency
82. Bammer R, Keeling SL, Augustin M, et al: Improved diffusion-weighted single-shot echo-planar imaging (EPI) in stroke
using sensitivity encoding (SENSE). Magn Reson Med 46:548-554, 2001. Medline Similar articles
83. Bammer R, Auer M, Keeling SL, et al: Diffusion tensor imaging using single-shot sense-EPI. Magn Reson Med
84. Mamata H, Mamata Y, Westin CF, et al: High-resolution line scan diffusion tensor MR imaging of white matter fiber tract
anatomy. Am J Neuroradiol 23:67-75, 2002. Medline Similar articles
85. Murphy BP, Zientara GP, Huppi PS, et al: Line scan diffusion tensor MRI of the cervical spinal cord in preterm infants. J
86. Robertson RL, Maier SE, Mulkern RV, et al: MR line-scan diffusion imaging of the spinal cord in children. Am J Neuroradiol
87. Bammer R, Herneth AM, Maier SE, et al: Line scan diffusion imaging of the spine. Am J Neuroradiol 24:5-12, 2003.
88. Bammer R, Augustin M, Prokesch RW, et al: Diffusion-weighted imaging of the spinal cord: Interleaved echo-planar
imaging is superior to fast spin-echo. J Magn Reson Imaging 15:364-373, 2002. Medline Similar articles
89. Wan X, Parker DL, Lee JN, et al: Reduction of phase error ghosting artifacts in thin slice fast spin-echo imaging. Magn
90. Norris DG, Bornert P, Reese T, Leibfritz D: On the application of ultra-fast rare experiments. Magn Reson Med 27:142-164,
91. Schick F: Splice: Sub-second diffusion-sensitive MR imaging using a modified fast spin-echo acquisition mode. Magn
92. Alsop DC: Phase insensitive preparation of single-shot rare: Application to diffusion imaging in humans. Magn Reson Med
38:527-533, 1997.
93. McKinnon GC, Zhou XJ, Leeds NE: Phase corrected complex averaging for diffusion-weighted spine imaging. Annual
Meeting of ISMRM, Denver, CO, 2000, p 802.
94. Le Roux P, McKinnon GC: Non CPMG fast spin-echo with full signal. Annual Meeting of ISMRM, Sydney, Australia, 1998, p
574.
95. Le Roux P: Non-CPMG fast spin-echo in practice. Annual Meeting of ISMRM, Denver, CO, 2000, p 1530.
96. Gmitro AF, Alexander AL: Use of a projection reconstruction method to decrease motion sensitivity in diffusion-weighted
97. Shaka A, Rucker S, Pines A: Iterative Carr-Purcell trains. J Magn Reson 77:606-611, 1988.
98. Jackson JI, Meyer CH, Nishimura DH, Macovski A: Selection of a convolution function for Fourier inversion using gridding.
IEEE Trans Med Imag 10:473-478, 1991.
99. Li TQ, Takahashi AM, Hindmarsh T, Moseley ME: ADC mapping by means of a single-shot spiral MRI technique with
application in acute cerebral ischemia. Magn Reson Med 41:143-147, 1999. Medline Similar articles
100. Meyer CH, Hu BS, Nishimura DG, Macovski A: Fast spiral coronary artery imaging. Magn Reson Med 28:202-213, 1992.
101. Glover GH, Lai S: Self-navigated spiral fMRI: Interleaved versus single-shot. Magn Reson Med 39:361-368, 1998.
102. Liao JR, Pauly JM, Brosnan TJ, Pelc NJ: Reduction of motion artifacts in cine MRI using variable-density spiral
trajectories. Magn Reson Med 37:569-575, 1997. Medline Similar articles
103. van Vaals JJ, Bergman AH: Optimization of eddy current compensation. J Magn Reson 90:52-70, 1990.
4 of 7 12-04-2010 00:36
104. Jehenson P, Westphal M, Schuff N: Analytical method for the compensation of eddy current effects induced by pulsed
magnetic field gradients. J Magn Reson 90:264-278, 1990.
105. Ahn CB, Cho ZH: Analysis of eddy currents in nuclear magnetic resonance imaging. Magn Reson Med 17:149-163, 1991.
106. Ahn CB, Cho ZH: Analysis of the eddy-current induced artifacts and the temporal compensation in nuclear magnetic
resonance imaging. IEEE Trans Med Imag 10:47-52, 1991.
107. Barker BR, Archer BT, Erdman WA, Peshock RM: MRI gradient waveform model for automated sequence calibration.
Med Phys 19:1483-1489, 1992. Medline Similar articles
108. Trudeau TD, Dixon WT, Hawkins J: The effect of inhomogeneous sample susceptibility on measured diffusion anisotropy
using NMR imaging. J Magn Reson B 108:22-30, 1995. Medline Similar articles
109. Haselgrove JC, Moore JR: Correction for distortion of echo-planar images used to calculate the apparent diffusion
coefficient. Magn Reson Med 36:960-964, 1996. Medline Similar articles
110. de Crespigny AJ, Moseley ME: Eddy current induced image warping in diffusion-weighted EPI. 6th Annual Meeting of
ISMRM, Sydney, Australia, 1998, p 661.
111. Bammer R, Auer M: Correction of eddy-current induced image warping in diffusion-weighted single-shot EPI using
constrained non-rigid mutual information image registration. 9th Annual Meeting of ISMRM, Glasgow, Scotland, 2001, p 508.
112. Reese TG, Heid O, Weisskoff RM, Wedeen VJ: Reduction of eddy-current-induced distortion in diffusion MRI using a
twice-refocused spin echo. Magn Reson Med 49:177-182, 2003. Medline Similar articles
113. Bammer R, Markl M, Barnett A, et al: Analysis and generalized correction of the effect of spatial gradient field distortions
in diffusion-weighted imaging. Magn Reson Med 50:560-569, 2003. Medline Similar articles
114. Orel SG, Schnall MD: MR imaging of the breast for the detection, diagnosis, and staging of breast cancer. Radiology
115. Hulka CA, Smith BL, Sgroi DC, et al: Benign and malignant breast lesions: Differentiation with echo-planar MR imaging.
116. Nunes LW, Schnall MD, Orel SG: Update of breast MR imaging architectural interpretation model. Radiology
117. Nunes LW, Schnall MD, Orel SG, et al: Breast MR imaging: Interpretation model. Radiology 202:833-841, 1997.
118. Heywang-Koebrunner SH: Contrast-enhanced magnetic resonance imaging of the breast. Invest Radiol 29:94-104, 1994.
119. Kaiser WA, Zeitler E: MR imaging of the breast: Fast imaging sequences with and without Gd-DTPA. Preliminary
observations. Radiology 170:681-686, 1989. Medline Similar articles
120. Stomper PC, Winston JS, Herman S, et al: Angiogenesis and dynamic MR imaging gadolinium enhancement of
malignant and benign breast lesions. Breast Cancer Res Treat 45:39-46, 1997. Medline Similar articles
121. Kuhl CK, Klaschik S, Mielcarek P, et al: Do T2-weighted pulse sequences help with the differential diagnosis of
enhancing lesions in dynamic breast MRI? J Magn Reson Imaging 9:187-196, 1999. Medline Similar articles
122. Le Bihan D, Turner R, Douek P, Patronas N: Diffusion MR imaging: Clinical applications. Am J Roentgenol 159:591-599,
1992.
123. Chenevert TL, Stegman LD, Taylor JM, et al: Diffusion magnetic resonance imaging: An early surrogate marker of
therapeutic efficacy in brain tumors. J Natl Cancer Inst 92:2029-2036, 2000. Medline Similar articles
124. Mardor Y, Pfeffer R, Spiegelmann R, et al: Early detection of response to radiation therapy in patients with brain
malignancies using conventional and high b-value diffusion-weighted magnetic resonance imaging. J Clin Oncol
125. Mardor Y, Kaplan O, Sterin M, et al: Noninvasive real-time monitoring of intracellular cancer cell metabolism and response
to lonidamine treatment using diffusion weighted proton magnetic resonance spectroscopy. Cancer Res 60:5179-5186, 2000.
126. Chinnaiyan AM, Prasad U, Shankar S, et al: Combined effect of tumor necrosis factor-related apoptosis-inducing ligand
and ionizing radiation in breast cancer therapy. Proc Natl Acad Sci USA 97:1754-1759, 2000. Medline Similar articles
127. Guo Y, Cai YQ, Cai ZL, et al: Differentiation of clinically benign and malignant breast lesions using diffusion-weighted
imaging. J Magn Reson Imaging 16:172-178, 2002. Medline Similar articles
128. Kinoshita T, Yashiro N, Ihara N, et al: Diffusion-weighted half-Fourier single-shot turbo spin echo imaging in breast tumors:
Differentiation of invasive ductal carcinoma from fibroadenoma. J Comput Assist Tomogr 26:1042-1046, 2002. Medline
Similar articles
129. Sinha S, Lucas-Quesada FA, Sinha U, et al: In vivo diffusion-weighted MRI of the breast: Potential for lesion
characterization. J Magn Reson Imaging 15:693-704, 2002. Medline Similar articles
130. Partridge SC, McKinnon GC, Henry RG, Hylton NM: Menstrual cycle variation of apparent diffusion coefficients measured
in the normal breast using MRI. J Magn Reson Imaging 14:433-438, 2001. Medline Similar articles
131. Kim T, Murakami T, Takahashi S, et al: Diffusion-weighted single-shot echoplanar MR imaging for liver disease. Am J
Roentgenol 173:393-398, 1999.
5 of 7 12-04-2010 00:36
132. Ichikawa T, Haradome H, Hachiya J, et al: Diffusion-weighted MR imaging with a single-shot echoplanar sequence:
Detection and characterization of focal hepatic lesions. Am J Roentgenol 170:397-402, 1998.
page 318
page 319
133. Ichikawa T, Haradome H, Hachiya J, et al: Diffusion-weighted MR imaging with single-shot echo-planar imaging in the
upper abdomen: Preliminary clinical experience in 61 patients. Abdom Imaging 24:456-461, 1999.
134. Namimoto T, Yamashita Y, Sumi S, et al: Focal liver masses: Characterization with diffusion-weighted echo-planar MR
135. Abragam A: Principles of nuclear magnetism. New York: Oxford University Press, 1961.
136. Amano Y, Kumazaki T, Ishihara M: Single-shot diffusion-weighted echo-planar imaging of normal and cirrhotic livers using
a phased-array multicoil. Acta Radiol 39:440-442, 1998. Medline Similar articles
137. Stahlberg F, Brockstedt S, Thomsen C, Wirestam R: Single-shot diffusion-weighted echo-planar imaging of normal and
cirrhotic livers using a phased-array multicoil. Acta Radiol 40:339, 1999. Medline Similar articles
138. Namimoto T, Yamashita Y, Mitsuzaki K, et al: Measurement of the apparent diffusion coefficient in diffuse renal disease
by diffusion-weighted echo-planar MR imaging. J Magn Reson Imaging 9:832-837, 1999. Medline Similar articles
139. Muller MF, Prasad P, Siewert B, et al: Abdominal diffusion mapping with use of a whole-body echo-planar system.
140. Moteki T, Ishizaka H, Horikoshi H, Matsumoto M: Differentiation between hemangiomas and hepatocellular carcinomas
with the apparent diffusion coefficient calculated from turboflash MR images. J Magn Reson Imaging 5:187-191, 1995.
141. Ries M, Jones RA, Basseau F, et al: Diffusion tensor MRI of the human kidney. J Magn Reson Imaging 14:42-49, 2001.
142. Fukuda Y, Ohashi I, Hanafusa K, et al: Anisotropic diffusion in kidney: Apparent diffusion coefficient measurements for
clinical use. J Magn Reson Imaging 11:156-160, 2000. Medline Similar articles
143. Muller MF, Prasad PV, Bimmler D, et al: Functional imaging of the kidney by means of measurement of the apparent
diffusion coefficient. Radiology 193:711-715, 1994. Medline Similar articles
144. Siegel CL, Aisen AM, Ellis JH, et al: Feasibility of MR diffusion studies in the kidney. J Magn Reson Imaging 5:617-620,
145. Chan JH, Tsui EY, Luk S, et al: Diffusion-weighted MR imaging of the liver: distinguishing hepatic abscess from cystic or
necrotic tumor. Abdom Imaging 26:161-165, 2001. Medline Similar articles
146. Okada Y, Ohtomo K, Kiryu S, Sasaki Y: Breath-hold T2-weighted MRI of hepatic tumors: Value of echo planar imaging
with diffusion-sensitizing gradient. J Comput Assist Tomogr 22:364-371, 1998. Medline Similar articles
147. Murtz P, Flacke S, Traber F, et al: Abdomen: Diffusion-weighted MR imaging with pulse-triggered single-shot sequences.
148. Malawer MM, Delaney T: Treatment of metastatic cancer to bone. Cancer: Principle and practice of oncology. DeVita VT
Jr, Hellman S, Rosenberg SA (eds). Philadelphia, Pa: Lippincott 4th ed, 1993, pp 2225-2245.
149. Vogler JB 3rd, Murphy WA: Bone marrow imaging. Radiology 168:679-693, 1988.
150. Daffner RH, Lupetin AR, Dash N, et al: MRI in the detection of malignant infiltration of bone marrow. Am J Roentgenol
146:353-358, 1986.
151. An HS, Andreshak TG, Nguyen C, et al: Can we distinguish between benign versus malignant compression fractures of
the spine by magnetic resonance imaging? Spine 20:1776-1782, 1995. Medline Similar articles
152. Yuh WT, Zachar CK, Barloon TJ, et al: Vertebral compression fractures: Distinction between benign and malignant
causes with MR imaging. Radiology 172:215-218, 1989. Medline Similar articles
153. Baur A, Huber A, Ertl-Wagner B, et al: Diagnostic value of increased diffusion weighting of a steady-state free precession
sequence for differentiating acute benign osteoporotic fractures from pathologic vertebral compression fractures. Am J
Neuroradiol 22:366-372, 2001. Medline Similar articles
154. Leeds NE, Kumar AJ, Zhou XJ, McKinnon GC: Magnetic resonance imaging of benign spinal lesions simulating
metastasis: Role of diffusion-weighted imaging. Top Magn Reson Imaging 11:224-234, 2000. Medline Similar articles
155. Spuentrup E, Buecker A, Adam G, et al: Diffusion-weighted MR imaging for differentiation of benign fracture edema and
tumor infiltration of the vertebral body. Am J Roentgenol 176:351-358, 2001.
156. Zhou XJ, Leeds NE, McKinnon GC, Kumar AJ: Characterization of benign and metastatic vertebral compression fractures
with quantitative diffusion MR imaging. Am J Neuroradiol 23:165-170, 2002. Medline Similar articles
157. Tasaly N, Unlu E, Cokal N: Can we differentiate benign versus malignant vertebral fractures with diffusion-weighted MR
imaging? [abstract]. Radiology 217:683, 2000.
158. Matoba M, Tonami H, Yokota H: Role of diffusion-weighted MRI and P31-MRS in differentiating between malignant and
benign vertebral compression fractures. 7th Annual Meeting of the ISMRM, Philadelphia, PA, 1999, p 1038.
159. Castillo M, Arbelaez A, Smith JK, Fisher LL: Diffusion-weighted MR imaging offers no advantage over routine noncontrast
MR imaging in the detection of vertebral metastases. Am J Neuroradiol 21:948-953, 2000. Medline Similar articles
160. Baur A, Dietrich O, Reiser M: Diffusion-weighted imaging of the spinal column. Neuroimaging Clin N Am 12:147-160,
6 of 7 12-04-2010 00:36

161. Herneth AM, Philipp MO, Naude J, et al: Vertebral metastases: Assessment with apparent diffusion coefficient. Radiology
162. Byun WM, Shin SO, Chang Y, et al: Diffusion-weighted MR imaging of metastatic disease of the spine: Assessment of
response to therapy. Am J Neuroradiol 23:906-912, 2002. Medline Similar articles
163. Clark CA, Barker GJ, Tofts PS: Magnetic resonance diffusion imaging of the human cervical spinal cord in vivo. Magn
164. Bammer R, Fazekas F, Augustin M, et al: Diffusion-weighted MR imaging of the spinal cord. Am J Neuroradiol
165. Ries M, Jones RA, Dousset V, Moonen CT: Diffusion tensor MRI of the spinal cord. Magn Reson Med 44:884-892, 2000.
166. Holder CA, Muthupillai R, Mukundan S Jr, et al: Diffusion-weighted MR imaging of the normal human spinal cord in vivo.
Am J Neuroradiol 21:1799-1806, 2000. Medline Similar articles
167. Gass A, Back T, Behrens S, Maras A: MRI of spinal cord infarction. Neurology 54:2195, 2000. Medline
Similar articles
168. Weidauer S, Dettmann E, Krakow K, Lanfermann H: [diffusion-weighted MRI of spinal cord infarction-description of two
cases and review of the literature]. Nervenarzt 73:999-1003, 2002. Medline Similar articles
169. Cercignani M, Horsfield MA, Agosta F, Filippi M: Sensitivity-encoded diffusion tensor MR imaging of the cervical cord.
170. Liu XZ, Xu XM, Hu R, et al: Neuronal and glial apoptosis after traumatic spinal cord injury. J Neurosci 17:5395-5406,
171. Nevo U, Hauben E, Yoles E, et al: Diffusion anisotropy MRI for quantitative assessment of recovery in injured rat spinal
cord. Magn Reson Med 45:1-9, 2001. Medline Similar articles
172. Schwartz ED, Yezierski RP, Pattany PM, et al: Diffusion-weighted MR imaging in a rat model of syringomyelia after
excitotoxic spinal cord injury. Am J Neuroradiol 20:1422-1428, 1999. Medline Similar articles
173. Fazekas F, Kapeller P: Diseases of the spinal cord. In Greenberg JO (ed): Neuroimaging: A Companion to Adams and
Victors Principles of Neurology. New York: McGraw-Hill, 1999, pp 521-542.
174. Nijeholt GJ, van Walderveen MA, Castelijns JA, et al: Brain and spinal cord abnormalities in multiple sclerosis. Correlation
between MRI parameters, clinical subtypes and symptoms. Brain 121:687-697, 1998. Medline Similar articles
175. Clark CA, Werring DJ, Miller DH: Diffusion imaging of the spinal cord in vivo: Estimation of the principal diffusivities and
application to multiple sclerosis. Magn Reson Med 43:133-138, 2000. Medline Similar articles
176. Ellis CM, Simmons A, Jones DK, et al: Diffusion tensor MRI assesses corticospinal tract damage in ALS. Neurology
177. Krabbe K, Gideon P, Wagn P, et al: MR diffusion imaging of human intracranial tumours. Neuroradiology 39:483-489,
178. Le Bihan D, Douek P, Argyropoulou M, et al: Diffusion and perfusion magnetic resonance imaging in brain tumors. Top
179. Muti M, Aprile I, Principi M, et al: Study on the variations of the apparent diffusion coefficient in areas of solid tumor in high
grade gliomas. Magn Reson Imaging 20:635-641, 2002. Medline Similar articles
180. Hendler T, Pianka P, Sigal M, et al: Delineating gray and white matter involvement in brain lesions: three-dimensional
alignment of functional magnetic resonance and diffusion-tensor imaging. J Neurosurg 99:1018-1027, 2003. Medline
Similar articles
181. Bellgowan PS, Saad ZS, Bandettini PA: Understanding neural system dynamics through task modulation and
measurement of functional MRI amplitude, latency, and width. Proc Natl Acad Sci USA 100:1415-1419, 2003. Medline
Similar articles
182. Li T, Song AW: Fast functional brain signal changes detected by diffusion weighted fMRI. Magn Reson Imaging
183. Song AW, Harshbarger T, Li T, et al: Functional activation using apparent diffusion coefficient-dependent contrast allows
better spatial localization to the neuronal activity: evidence using diffusion tensor imaging and fiber tracking. Neuroimage
184. Darquie A, Poline JB, Poupon C, et al: Transient decrease in water diffusion observed in human occipital cortex during
visual stimulation. Proc Natl Acad Sci USA 98:9391-9395, 2001. Medline Similar articles
© 2010 Elsevier
7 of 7 12-04-2010 00:36
IFFUSION ENSOR MAGING UNDAMENTALS

Peter J. Basser
INTRODUCTION
It is remarkable that the MR measurement of the diffusivity of water in neural tissue provides unique
and useful biological and clinical information about tissue composition, the physical properties of its
constituents, tissue microstructure, and architectural organization. Moreover, this information can be
obtained in vivo, noninvasively without contrast agents.
This chapter deals with diffusion-tensor MRI, which provides new information about tissue
microstructure, particularly anisotropic tissues like brain white matter, beyond that provided by
diffusion-weighted MRI and conventional diffusion MRI.
What is diffusion anisotropy? It is simply that the measured apparent diffusion coefficient (ADC) is not
the same in all directions, i.e. it is not directionally uniform (isotropic) as in a jar of water. The
observation of diffusion anisotropy in tissues is usually a clue that the underlying microstructure is
ordered. Characterizing diffusion anisotropy quantitatively can provide information not only about the
alignment direction, but also often about the organization and properties of the ordered elements.
In a biological context, anisotropic diffusion was first observed in NMR experiments with skeletal
1,2
muscle. Interest in the phenomenon, however, was rekindled when it was observed in MRIs of
animal and human brains.3-7
© 2010 Elsevier
1 of 1 14-04-2010 19:28
DIFFUSION AND DIFFUSION-TENSOR MRI

Other chapters in this book explain diffusion MRI, its underpinnings and its applications. There is no
need to repeat this information here. It will be assumed the reader is familiar with the notion of a
diffusion-weighted image (DWI) as well as conventional MRI pulse sequences used to acquire one.
Familiarity with the formula used to calculate an ADC in each voxel from a set of DWIs is also
assumed. The notion of the b-factor or b-value has also been introduced in this context as a measure
of the degree of diffusion weighting.
A reasonable starting point is to compare and contrast diffusion MRI (DI) and diffusion tensor MRI
(DTI), which is sometimes referred to as DT-MRI.8 DI is based upon a one-dimensional Gaussian
model of molecular displacements. The measurement of an ADC is obtained from analyzing the
projection of all molecular displacements along one direction. In tissues, such as brain gray matter,
where the ADC is largely independent of the orientation of the tissue, a single ADC is often sufficient to
characterize the diffusion process at the voxel scale. However, in anisotropic media, such as skeletal
and cardiac muscle1,2,9 and in white matter,5,10-12 where the ADC depends upon the orientation of the
tissue, the 1-D Gaussian model is inadequate to characterize the orientation-dependent water mobility.
page 320
page 321
A more general description of anisotropic free diffusion uses a three-dimensional Gaussian model of
molecular displacements. This model contains a symmetric effective or apparent diffusion tensor of
water, D (e.g. see reference 13) to describe the orientation dependence of diffusion instead of a
scalar diffusion coefficient.
In essence, DTI consists of the measurement of D (and functions of it) from a series of DWIs. D is
determined by using a relationship between the measured echo signal in each voxel and the applied
magnetic field gradient sequence,14-17 which was derived from Stejskal's classic solution to the
modified Bloch-Torrey equation18:
Above, A(b) is the echo magnitude of the diffusion-weighted signal, A(b=0) is the echo magnitude of
the nondiffusion-weighted signal, and bij is a component of the symmetric b-matrix, b. In DI a scalar
b-factor is calculated for each DWI; in DTI a symmetric b-matrix is calculated for each DWI. Whereas
the b-value summarizes the attenuating effect of diffusion and imaging gradients on the MR signal
along one direction,19 the b-matrix summarizes the attenuating effect of all gradient waveforms (i.e., all
14-17
imaging and diffusion gradient sequences) applied in all three directions, x, y, and z. Just as in DI,
each DWI is used with its corresponding b-factor to estimate an ADC using linear regression; in DTI,
each DWI is used with its corresponding b-matrix to estimate D using multivariate linear regression of
Eq. 11-1. Multivariate linear regression is just one of a number of techniques, including nonlinear
regression and singular-value decomposition, that could be used to estimate D from the echo data.
In the limiting case in which the medium is isotropic, it is easy to show that the 3-D Gaussian model
assumed in DTI reduces to the 1-D Gaussian model assumed in DI. To see this, set Dxx = Dyy = Dzz =
D, and Dxy = Dxz = Dyz = 0. Then, Equation 11-1 reduces to:
There are two important differences between the designs of DI and DTI experiments. First, in DTI
diffusion gradients must be applied along at least six noncolinear directions,14 whereas in DI it is
1 of 2 14-04-2010 19:30
sufficient to apply diffusion gradients along only one direction. Second, the notion of "cross-terms"
must be expanded in DTI. Interactions between imaging and diffusion gradients applied in orthogonal
directions, and even between imaging gradients applied in orthogonal directions, can introduce
additional diffusion weighting.15,16 It can be seen from Equation 11-2 that diffusion weighting can be
introduced by imaging (e.g., slice select gradients) applied along the x, y, or z directions. In isotropic
media, however, interactions between gradients applied in orthogonal directions do not produce any
diffusion attenuation, but in anisotropic media these interactions can result in diffusion attenuation.
It is important to note that MRI measurements of the ADC, T1, T2, MT, proton density, phase, and
chemical shift were all preceded by classical NMR measurements of these quantities, which in some
cases were introduced almost a half century before their MRI counterpart. For DTI, however, there
was no classical NMR method developed either to measure the self-diffusion tensor or the effective (or
apparent) diffusion tensor of water (or other species) in the laboratory coordinate frame. So, unlike T1
MRI or DI, implementing DTI first required the developing of a new method to measure D using NMR
means14,20 and then combining it with conventional MRI.8
© 2010 Elsevier
2 of 2 14-04-2010 19:30
GEOMETRIC REPRESENTATION OF TRANSLATIONAL DIFFUSION IN 3-D
Add to lightbox
page 321
page 322
Figure 11-1 The Brownian picture of diffusion is best epitomized by illustrating a possible path taken
by a molecule that is released at point r0 at time t=0 and moves to position r at a later time, t. The
molecule undergoes a "drunk walk" whose random motion is described by a displacement probability
distribution.
The most intuitive way to understand what D means is by considering a Gedanken experiment in which
the Brownian motion of an ensemble of "tagged" water molecules released from the center of a voxel
is followed, as depicted in Figure 11-1. If this experiment is performed in a jar of water, then the rate
of diffusive transport is the same in all directions. Diffusion is said to be isotropic and is completely
specified by a single scalar constant, D, the diffusion coefficient. Thus, diffusion isotropy describes the
case in which the molecular diffusivity is independent of the medium's orientation. For a diffusion time,
∆, the translational displacement distribution is spherically symmetric, and surfaces of constant
probability or water concentration are concentric spheres. When considering the Einstein formula for
21
1-D diffusion :
it is possible to construct a particular sphere whose radius equals the root mean-squared (rms)
displacement of water molecules after diffusion time ∆, a graphical representation of which can be
found in Figure 11-2A.
If the Gedanken experiment is performed in a liquid crystal or system with microscopic aligned rods,
then the rate of diffusive transport will no longer be the same in all directions. Diffusion anisotropy
implies that the translational displacement probability of the diffusing species now depends upon the
medium's orientation. In homogeneous (i.e., spatially uniform) anisotropic media, the voxel-averaged
displacement distribution is given by:
where the column vector r = (x,y,z)T is the net displacement, and |D| is the determinant of D. The
covariance matrix, 2∆ D, characterizes the width of the distribution, which changes with orientation.
Now, when surfaces of constant probability or particle concentration are constructed by setting the
exponent in Equation 11-4 to a constant, then we obtain:
1 of 4 14-04-2010 19:31
that can be rewritten as:
8,13,22
This equation represents a three-dimensional ellipsoid,* called the "diffusion ellipsoid". This is
shown in Figure 11-2B. The six independent parameters (a,b,c,d,e,f) in Equation 11-5b contain all
information required to specify the size, shape, and orientation of this ellipsoid. Not coincidentally the
number of independent elements of D required to specify the form of the three-dimensional
displacement distribution in Equation 11-4 is also six.
The laboratory (x-y-z) coordinate axes can always be rotated so that they are aligned with the local
principal (x'-y'-z') axis of the diffusion ellipsoid in each voxel. In the preferred principal frame, all
off-diagonal elements of D vanish; Equation 11-5A simplifies to:
Above λx', λy', and λz' are the three principal diffusivities (or eigenvalues), corresponding to the three
respective principal directions, [epsi ]x',[epsi ]y', [epsi ]z'; , , and are the root
mean-squared (rms) displacements along these three principal directions at diffusion time ∆,
respectively. The lengths of the major and minor axes of the diffusion ellipsoid represent these three
distances. Note that the direction of the principal (x'-y'-z') axes of the diffusion ellipsoid are generally
not known a priori, and typically do not coincide with the laboratory coordinate axes. It is necessary to
assume that they are different in each voxel.14 An example of an in vivo diffusion ellipsoid image is
Add to lightbox
2 of 4 14-04-2010 19:31
Add to lightbox
*This is true because both D and the matrix of coefficients, are positive definite.
page 322
page 323
Figure 11-2 A, In isotropic media, the rms displacement or diffusion ellipsoid is spherical. B, However,
in anisotropic media, the diffusion ellipsoid can be prolate or oblate, and its three principal directions
are coincident with the eigenvectors of D, [epsi ]1, [epsi ]2, and [epsi ]3.
Add to lightbox
Figure 11-3 The T2-weighted axial image contains an ROI encompassing the lateral ventricles and
corpus callosum. A diffusion ellipsoid image constructed from each voxel within the ROI shows the
size, shape and orientation of the diffusion ellipsoid. CSF is clearly depicted by large spherical
ellipsoids in which diffusion is free and isotropic. Gray matter is depicted by smaller spherical
ellipsoids, indicating isotropic diffusion but lower mean diffusivity than in the CSF. White matter is
depicted by prolate diffusion ellipsoids whose polar axis is aligned with the purported white matter
fiber direction in the corpus callosum. (From Pierpaoli C, Jezzard P, Basser PJ, et al: Diffusion tensor
MR imaging of the human brain. Radiology 201:637-648, 1996.)
In summary, both diagonal and off-diagonal elements of D measured in DTI, maps of which are shown
in Figure 11-4, are essential in specifying the probability distribution in Equation 11-4 as well as in
characterizing the size, shape, and orientation of the diffusion ellipsoid constructed from D in each
voxel.
3 of 4 14-04-2010 19:31
© 2010 Elsevier
4 of 4 14-04-2010 19:31
QUANTITATIVE PARAMETERS PROVIDED BY DTI

Scalar parameters that embody distinct intrinsic features listed earlier in italics can be displayed as an
image. Parameters describing the size and shape of the ellipsoid should be rotationally invariant, i.e.,
independent of the orientations of the anisotropic structure, patient's body within the MR magnet, the
applied diffusion sensitizing gradients, and choice of the laboratory coordinate system in which the
components of D and the magnet field gradients are measured.23,24
Possibly the most important quantitative scalar parameters provided by DTI are the three sorted
eigenvalues, λ1, λ2, and λ3 (i.e., the eigenvalues of D described above, λx',λy', and λz', but sorted
according to size). An example of an eigenvalue image is shown in Figure 11-5. Combinations of these
quantities are used to characterize the size and shape of the diffusion ellipsoid. The orientational
information is encoded by the three eigenvectors of D.
Add to lightbox
Figure 11-4 DTI provides data for each slice within the imaging volume that consists of images of the
three diagonal elements of D (top row), and of the three off-diagonal elements of D (bottom row).
(From Pierpaoli C, Jezzard P, Basser PJ, et al: Diffusion tensor MR imaging of the human brain.
Radiology 201:637-648, 1996.)
© 2010 Elsevier
1 of 1 14-04-2010 19:32
SIZE OF DIFFUSION ELLIPSOID

Quantities described later characterize the size of the diffusion ellipsoid, independent of orientation and
shape:
Physically, Trace(D) is three times the orientationally averaged diffusivity, <D>, which can be obtained
25
by arithmetically averaging the ADC distribution uniformly over all possible directions. Trace(D)
measures an intrinsic property of the tissue, which is independent of fiber orientation, gradient
directions, etc. An example of a Trace(D) image of brain parenchyma is provided in Figure 11-6.
page 323
page 324
Add to lightbox
Figure 11-5 Images or maps of the three sorted eigenvalues, λ1, λ2, and λ3. Different regions of brain
can be distinguished by examining these images. CSF is uniformly bright in these images, indicating
diffusion is isotropic and high there. Grey matter regions are nearly uniform in brightness, while white
matter regions, like the corpus callosum at the center, is bright in the λ1 image but dark in the λ2 and
λ3 images, indicating significant diffusion anisotropy there. (From Pierpaoli C, Jezzard P, Basser PJ,
et al: Diffusion-tensor MR imaging of the human brain. Radiology 201:637-648, 1996.)
What is the clinical value of Trace(D)? Moseley et al. discovered in animals,26-28 and Warach et al.
later showed in humans29,30 that a reduction of the ADC in brain parenchyma sensitively indicates the
onset and severity of a cerebral ischemic event. Moreover, an elevation in ADC was observed in
chronic stroke, often leading to lacunar infarcts. Thus, it seemed possible to follow an acute stroke in
progress, and the subsequent chronic degeneration it caused using the ADC. However, this aim was
hampered by anisotropic diffusion observed in white matter. There, DWI intensity and the ADC depend
on the direction of the diffusion sensitizing gradient with respect to the white matter fiber
orientation.3,11,31 Thus, it was generally not possible to determine whether an observed drop or
recovery in the DWI or ADC was of a physiological origin (brought on by the ischemic event itself) or
arose because of the relative orientation of the white matter tracts and the applied diffusion gradient.
In ischemia monitoring applications, diffusion anisotropy in white matter produced directionally
dependent intensity variations in DWIs and in ADC maps that complicated their interpretation, and was
considered by many to be a confounding artifact.
The solution to this vexing problem was provided by DTI. Unlike the DWI signal intensity and the ADC,
Trace(D) is inherently insensitive to both the direction of the diffusion sensitizing gradient and of the
orientation of white matter in a voxel.23 Thus, displaying Trace(D) or <D> instead of the ADC
eliminates all orientational dependence seen in DWIs and ADC maps, so changes in signal can be
ascribed solely to changes in the tissue's physiologic state.
1 of 2 14-04-2010 19:33
The use of <D> in acute and chronic stroke assessment is probably the most important current clinical
application of DTI, although it is not widely known that Trace(D) and <D> (sometimes referred to as
the "mean ADC") are actually diffusion-tensor-derived parameters. A second reason for the success of
DTI in stroke assessment (and other brain disorders) is the surprising finding that in normal brain
parenchyma Trace(D) is highly uniform.32-35 Remarkably, it has virtually the same value in both white
35 34
and gray matter. Van Gelderen, et al first demonstrated in cats, and Ulug et al in humans that an
image of Trace(D) (or of <D>) defines ischemic regions better than the corresponding DWI or ADC
map.
Add to lightbox
page 324
page 325
Figure 11-6 T2-weighted amplitude images (top row) and Trace(D) maps (bottom row) obtained for
the same slice. Trace(D) images are remarkably uniform in normal brain parenchyma. Hyperintense
regions correspond to CSF or fluid-filled regions. (From Pierpaoli C, Jezzard P, Basser PJ, et al:
Diffusion tensor MR imaging of the human brain. Radiology 201:637-648, 1996.)
Several MR sequences have been proposed to produce a "trace-weighted" or "isotropically weighted"

MRI. Clever schemes were proposed by Wong et al. and Mori et al. that use only two DWIs.36,37 One
way to compute a trace-weighted image is to compute <D> and then to generate an image whose
intensity is a function of it, such as exp(-b <D>) where b is defined in Equation 11-2. In such an image,
ischemic regions would appear bright while normal parenchyma appears dark. A general formalism for
producing trace-weighted images from a set of DWIs can be found in reference 38.
© 2010 Elsevier
2 of 2 14-04-2010 19:33
SHAPE OF DIFFUSION ELLIPSOID
Add to lightbox
Figure 11-7 Whole-brain axial slices showing the distribution of FA. (Courtesy of Derek Jones.)
Characterizing the degree of diffusion anisotropy is tantamount to characterizing the shape of the
three-dimensional diffusion ellipsoid, independent of its orientation and size. Such a measure should
also be rotationally invariant, i.e., independent of the sample's placement or its orientation with respect
to the (laboratory) x-y-z reference frame.8 One way of deriving these parameters is by analyzing the
39,40
anisotropic part of the D or the "diffusion deviation tensor" in each voxel, D :
where above, the familiar mean diffusivity, <D>, is seen, multiplied by the identity tensor, I. In defining
the deviation tensor, the part of D that is isotropic is simply subtracted off. What remains is the
2
anisotropic part of D. The magnitude of the diffusion deviation tensor, Trace(D ), can be shown to be
proportional to the sample variance of the eigenvalues or principal diffusivities39:
This quantity, which is rotationally invariant, is used in several popular diffusion anisotropy measures,
such as the Relative Anisotropy (RA)39:
and the Fractional Anisotropy (FA)39:
Both characterize the degree of "out-of-roundness" of the diffusion ellipsoid. The RA is just the
coefficient of variation of the eigenvalues. The FA measures the fraction of the magnitude of the
diffusion tensor that is anisotropic.39 An example of a whole-brain FA map is provided in Figure 11-7.
Other anisotropy measures have also been proposed, some differing from the FA and RA above by
41
only a scale factor.
Another useful "shape" parameter is the third moment or skewness of the eigenvalues:
1 of 2 14-04-2010 19:35
When the skewness is positive the diffusion ellipsoid is prolate (i.e., cigar shaped-a shape observed in
white matter fibers in the corpus callosum and in the pyramidal tract in monkeys32 and in humans33).
When it is negative, the diffusion ellipsoid is oblate (i.e., pancake shaped-a shape observed in white
33
matter in the subcortical white matter regions in the centrum semiovale in humans ). Clearly, it would
be useful to determine higher moments of the distribution of eigenvalues of D, such as the Kurtosis(λ),
in order to characterize diffusion anisotropy more completely, however, typically noise in DWIs
introduces enough bias in the estimates of the eigenvalues to make these higher-order statistics
inaccurate. Background noise even causes overestimation of the variance of the eigenvalues, and other
measures of diffusion anisotropy due to a phenomenon called "eigenvalue repulsion".32
page 325
page 326
Add to lightbox
Figure 11-8 A T2-weighted image juxtaposed with a direction-encoded color map. This color scheme
overcomes the problem of using line drawings to depict fiber orientation within a voxel, since all fibers
can be perceived, and their orientations determined from the color wheel in the legend. (From Pajevic
S, Pierpaoli C. Color schemes to represent the orientation of anisotropic tissues from diffusion
tensor data: application to white matter fiber tract mapping in the human brain. Magn Reson Med
42:526-540, 1999. Reproduced with permission of Wiley-Liss, Inc., a subsidiary of John Wiley &
Sons, Inc.)
Three novel tensor-derived parameters have been proposed that measure the degree to which D is
line-like, plane-like, and sphere-like (corresponding to diffusion ellipsoids that are prolate, oblate, and
42
spherical, respectively). This tensor decomposition provides a useful and different geometric
interpretation of the diffusion process. These diffusion measures, however, are not rotationally invariant
(as are the three moments of D, the mean, variance, and skewness, described earlier), which could
introduce additional statistical bias in their distributions (see Reference 32).
© 2010 Elsevier
2 of 2 14-04-2010 19:35
ORIENTATION OF DIFFUSION ELLIPSOIDS AND THEIR SPATIAL

DISTRIBUTION
Another important development in DTI has been the introduction of quantities that reveal the distribution
or spatial pattern of diffusion ellipsoids within an imaging volume. Some features involve the spatial rate
of change of tensor-derived quantities within the imaging volume, such as the gradient of <D>, or the
gradient of the magnitude of the deviation tensor described earlier.43 The former provides information
about the location of boundaries between cerebrospinal fluid (CSF) and tissue parenchyma, while the
latter provides information about boundaries between white and gray matter. Typically, it is challenging
to calculate these quantities from noisy, discrete, voxel-averaged DTI data, so some spatial smoothing
43
for example, using a continuous approximation to the diffusion tensor field or a regularization
44
scheme, must be performed beforehand.
Initially, it was proposed that in ordered fibrous tissues, the eigenvector associated with the largest
eigenvalue within a voxel lies parallel to the local fiber direction. 8 Imaging methods that apply this idea
include direction field mapping, in which the local fiber direction (or more correctly, orientation) is
displayed as a vector (or line segment) in each voxel, and fiber-tract color mapping, in which a color,
assigned to a voxel containing anisotropic tissue, is used to signify the local fiber tract direction (see
Figure 11-8).45 An important achievement of color mapping has been the ability to identify
unambiguously all major commissural, association and projection pathways in the human brain46,47 as
© 2010 Elsevier
1 of 1 14-04-2010 19:35
DTI FIBER TRACTOGRAPHY

DTI fiber tractography is a natural extension of diffusion ellipsoid imaging. Re-examine the diffusion
ellipsoid image containing the corpus collosum shown in Figure 11-3. Imagine now that adjacent,
coherently ordered ellipsoids suggestive of a continuous fiber tract can be connected. What is obtained
is an object like a link-sausage that represents a possible fiber tract. DTI fiber tractography48-54 is the
formal name given to this procedure in which coherently ordered fiber tract trajectories within the brain
and other fibrous tissues are mathematically followed. Like color mapping, it is also based on the idea
that in ordered fibrous tissues, the eigenvector associated with the largest eigenvalue within a voxel is
8
parallel to the local fiber direction. A nice review of this emerging field is given in reference 55.
page 326
page 327
1 of 2 14-04-2010 19:37
Add to lightbox
Figure 11-9 This direction-encoded color map depicts the major association, commissural, and
projection pathways in the brain. (From Pajevic S, Pierpaoli C: Color schemes to represent the
orientation of anisotropic tissues from diffusion tensor data: application to white matter fiber tract
mapping in the human brain. Magn Reson Med 42:526-540, 1999. Reproduced with permission of
Since its introduction, several different schemes have been proposed to follow fiber tracts. In the
so-called "streamline" approaches, fiber-tract trajectories are generated from the local fiber tract
direction field in much the same way fluid streamlines are generated from a fluid velocity field. Starting
from a "seed point", fibers are launched in both directions until some stopping or "termination" criteria
are satisfied, such as the FA drops below its level in background noise. A fiber tract consists of all
points along such a continuous trajectory. One disadvantage of this computational strategy is that
errors can accumulate during the tract following process. In general, it cannot be assured that the
trajectories represent actual or even probable pathways of nerve fibers.
page 327
page 328
The second strategy is probabilistic in nature.56-62 A seed point is assumed to be connected to all
points within the imaging volume but the most probable connections are those that minimize some
function, such as the strain energy of the pathway.56 This approach holds great promise in that many
paths are explored, but only those that are frequently traversed are assigned a high likelihood. Of
course, a disadvantage of this approach is that it is not known which physical constraints nature uses
to construct nerve pathways. The development of probabilistic approaches that do not rely on such
functional minimization appear to be quite promising,63 but further validation studies still should be
performed.
A hybrid tractography method, recently proposed by Jones and Pierpaoli,64 uses the streamline in
conjunction with an empirical statistical scheme such as bootstrapping. In this way, many possible
paths can be generated from the same DWI data set and the variability of these paths can be
assessed. This is done without making any assumptions about the source of artifacts in the DWIs. A
disadvantage of this approach is that it generally requires acquiring more DWIs than is necessary for a
typical DTI study.
It is important to note that a number of well-documented artifacts in DTI fiber tractography arise
because a discrete, coarsely sampled, noisy, voxel-averaged direction field data,49 are used to
65
reconstruct topologically complex nerve pathways. These artifacts can produce "phantom"
connections between different brain regions that do not exist anatomically, or can result in missing
anatomical connections that do exist. There are usually a number of thresholds and free parameters
that can be set in existing tractography codes whose adjustment can alter the findings. Therefore,
great care must be exercised both in obtaining "anatomical connectivity" obtained using DTI fiber
tractography.50 Clearly, some skepticism is warranted when interpreting such data-less in the coherent
primary pathways, but increasingly in finer white matter structures.
Artful computer graphics-inspired schemes have been proposed to visualize fiber tract trajectories. 66
Zhang et al. have used stream tubes to visualize fiber tract trajectories and stream surfaces to
visualize regions in which fibers cross within a plane. 67,68 An example of a stream tube representation
of fiber tracts in the brain is given in Figure 11-10. These imaginative strategies will find increased use
in displaying features of high-dimensional data inherent in DTI.
© 2010 Elsevier
2 of 2 14-04-2010 19:37
DWI ARTIFACTS
Add to lightbox
Figure 11-10 An example of computed fiber tracts in the brain. Here, fiber tracts are represented as
"stream tubes," which are generated by following the direction of maximum diffusivity starting from
"seedpoints" located in white matter ROIs. Novel computer visualization schemes such as this one will
play an increasingly important role in representing the high-dimensional data contained in the diffusion
tensor MRI data set. (Courtesy of Derek Jones.)
As seen earlier, DWI data is used to estimate D. Clearly any random or systematic artifacts in DWIs
will adversely affect the estimate of D. Some of these artifacts have been discussed elsewhere in this
book in Chapter 10 on DWI; however, many cause specific anomalies in DTI data, so they are
described here as well. DWI artifacts are presented in order of importance as follows: Subject motion
during the DWI acquisition can cause an artifactual redistribution of signal intensities within DWIs. Rigid
body motion, such as rotation and translation are the easiest to correct for, since they involve applying
a uniform phase offset to an entire image. This problem has been addressed by incorporating
navigator echoes in the DWI pulse sequence.69,70 More pernicious, however, is nonrigid body motion
caused by eye movements, pulsations of CSF etc. At present, these artifacts cannot be entirely
eliminated in DWIs, but they can be mitigated by the use of fast echo-planar DWI sequences and
cardiac gating. Subject motion can also occur during the acquisition of a series of DWIs, like those
used to estimate D. Recently a proposed remedy is to warp the set of DWIs to a common template
using a nonrigid transformation that maximizes a measure of similarity among them.71
page 328
page 329
The large, rapidly switched magnetic field produced by the gradient coils during the diffusion sequence
induces eddy currents in the conductive structures within the MRI scanner. These, in turn produce an
additional unwanted, slowly decaying magnetic field. Two undesirable effects result: the field gradient
at the sample differs from the prescribed field gradient (thus causing in a deviation from the prescribed
b-matrix), and the slowly decaying field causes geometrical distortion of the DWI, which depends on
the gradient strength and direction. These artifacts can adversely affect DTI studies because D is
generally calculated in each voxel from a multiplicity of DWIs assuming that the prescribed gradients
are the same as the gradients actually being applied to the tissue, and that each voxel contains the
1 of 2 14-04-2010 19:37
same tissue from scan to scan. Thus, uncompensated image distortion caused by eddy currents can
lead to systematic errors in these estimated diffusion parameters. Unfortunately, single-shot diffusion-
weighted echo planar image (EPI) acquisitions are quite susceptible to eddy-current artifacts so
correction schemes have to be used. One widely used strategy is to warp each DWI to a common
template; another is to try to use a model of the effects of the eddy current on the phase map to
correct it.72 Nonrigid transformations of DWI data have recently been proposed to remedy this problem
71
as well.
Large discontinuities in the bulk magnetic susceptibility, such as those that occur at tissue-air
interfaces produce local magnetic field gradients which are notorious for their contribution to image
distortion, particularly during EPI. In addition to the image distortion, susceptibility variations within the
brain adversely affect DWIs because the additional local gradients act like diffusion gradients, causing
the b-matrix to be spatially varying. Fortunately, this effect appears to be limited to the volume of the
brain adjacent to the sinuses.73
While at low levels of diffusion weighting the logarithm of the signal attenuation decreases linearly with
increasing b-value, background noise causes the DWI intensity eventually to approach a baseline
"noise floor" as the degree of diffusion weighting increases. Noise introduces an error in the estimated
ADC or D when using Equations 1 or 2 in this regime. Background noise leads primarily to a bias in
variance-based anisotropy measures of D, (such as the RA and FA) as noise always introduces
variability in the eigenvalues that makes isotropic media appear anisotropic, and anisotropic media
appear more anisotropic.32 Noise also biases the mean and variance of the corresponding
eigenvectors of D.74 Unfortunately, the ability to characterize the deleterious effects of this artifact
exceeds current understanding of how to remedy it,32,75-77 although some recent work on the effects of
background noise on DWIs has provided some new insights into how noise affects DTI measurements
and how it can be ameliorated.78
The following problems can produce artifacts as well. Improper refocusing of rf pulses leads to added
signal loss. Background gradients can be present because of improper shimming, which lead to
additional signal attenuation if not properly compensated. This problem can often be remedied by
measuring the background gradients directly79 and incorporating them explicitly in the formula relating
the signal intensity and the ADC or diffusion tensor. Flipping the signs of the diffusion gradients on
alternate averages has been suggested as an aid in removing most cross-terms. Clearly, cross-terms
arising between imaging gradients, albeit generally small, would not be corrected using this method.
Moreover, averaging should be done on the logarithm of the intensity rather than the intensity,
otherwise gradient cancellation will not occur. Gradient nonlinearity and miscalibration can lead to
errors in the calculation of the ADC or D from a set of DWIs. The diffusion attenuation at different
gradient strengths must be known for these formulae to provide meaningful results. If these gradients
are not well calibrated, or they are not linear, signal attenuation observed in the image will be
misattributed to diffusion processes in the sample. If the gradient in the x, y and z directions are
coupled to one another (i.e., there is cross-talk) because of misalignment, then gradients applied in
logical directions may have components in other directions. Such cross-talk could lead to problems,
80,81
particularly in DTI.
© 2010 Elsevier
2 of 2 14-04-2010 19:37
ISSUES IN INFERRING TISSUE MICROSTRUCTURE FROM NMR SIGNAL

Inferring the microstructure and underlying architectural organization of tissue using diffusion imaging
data has been a long sought-after goal; however, its attainment is complicated by several factors.
First, the homogeneity of tissue within each voxel cannot be assumed. Numerous microscopic
compartments exist in brain parenchyma. Typically gray matter, white matter, and CSF can occupy the
same voxel. The number of these distinct tissue types and their distribution within the voxel are
generally not known a priori. At a microstructural level, gray and white matter are heterogeneous,
having a distribution of macromolecular structures with a range of sizes, shapes, composition, and
physical properties (such as T2, D). Differences in relaxation parameters can lead to different rates of
echo attenuation in each compartment, making it more difficult to explain the cause of signal loss within
a voxel. There are also irregular boundaries between macromolecular and microscopic-scale
compartments. Water diffusion may be restricted in some compartments and hindered in others. Some
water will be associated with certain macromolecules while some will be free to diffuse. Another
unknown is whether there is water exchange between compartments, which can also affect the
relaxation rates of the spin system. Water movement within and between compartments is still not well
understood, although tractable models exist to describe it. Owing to differences in blood flow and
thermal conductivity, temperature cannot even be assumed to be uniform throughout a tissue sample.
82-84
It is well known that temperature affects the measured diffusivity, typically by about 1.5% for each
84
degree (C).
As an aside, the list given earlier in part explains why the underlying cause of diffusion anisotropy has
not been fully elucidated in brain parenchyma, although at low b-values typical of DTI, most
investigators ascribe it to ordered, heterogeneous structures, such as large oriented extracellular and
intracellular macromolecules, super-macromolecular structures, organelles, and membranes. In the
central nervous system, diffusion anisotropy is not simply caused by myelin in white matter, since it has
been shown in several studies that even before myelin is deposited, diffusion anisotropy can be
measured using MRI.85-89 Thus, despite the fact that increases in myelin are temporally correlated with
increases in diffusion anisotropy, structures other than the myelin sheath must also be contributing to
diffusion anisotropy. This is important, because there is a common misconception that the degree of
diffusion anisotropy can be used as a quantitative measure or "stain" of myelin content, when no such
simple relationship exists.
page 329
page 330
Recently, with the advent of stronger magnetic field gradients, several groups have reported
90
nonmonoexponential decay of the MR signal intensity as a function of b-value, e.g., and have
attempted to infer from it properties of distinct tissue compartments. Putting aside the complexities of
obtaining stable estimates of discrete exponentials (i.e., diffusion relaxography), numerous
microstructural and architectural configurations could produce the same multi-exponential relaxation
data. For example, Peled et al. showed that a system of impermeable tubes with a distribution of
diameters consistent with those found in histological brain slices could give rise to multi-exponential
decay of the signal.91 Similar behavior is expected when there is a statistical distribution of any
relevant physical property or microstructural dimension within a voxel. It is unlikely that a particular
exponential in a multi-exponential model of diffusion attenuation can ever meaningfully be assigned to a
particular and distinct tissue compartment. Clearly, without invoking additional a priori information about
tissue structure, tissue composition, the physical properties of the different compartments, and their
spatial distribution, determining tissue microstructural and architectural features from the NMR signal
remains a challenging ill-posed inverse problem.
© 2010 Elsevier
1 of 1 14-04-2010 19:38
BEYOND DTI
DTI completely characterizes the Gaussian contribution of the water diffusion process within tissues. 92
Still, there is increasing evidence of restricted diffusion caused by water being trapped within regions
bounded by impermeable barriers, which is described by a nonGaussian displacement profile, features
of which several new methods have been proposed to characterize. These include diffusion spectrum
93 94
imaging (DSI), a variant of Callaghan's 3-D q-space MRI, high angular resolution diffusion imaging
(HARDI),95,96 persistent angular structure MRI (PAS-MRI),97 generalized diffusion-tensor MRI
98,99 100 101
(GDTI), Q-ball MRI, composite hindered and restricted model of diffusion (CHARMED) MRI
and multi-tensor MRI.102,103
First, as a general requirement, all these methods must either subsume DTI or reduce to it in the limit
of low b (or low q), where the Gaussian diffusion model applies.92 Moreover, their experimental
designs typically entail acquiring DWIs with a greater number of directions and/or a larger range of
gradient strengths than are typically used in DTI. In principle, while one may obtain more information, it
is generally at the price of additional acquisition time. In applications such as acute stroke monitoring,
this price is unjustifiably high. In this case, it may be sufficient to produce a mean ADC or trace map,
or even a DWI. In a careful neurological assessment of fiber tract organization in the brain, more
information than DTI provides may be warranted. Clearly, the clinical or biological application
determines the appropriate displacement imaging method used.
Another important point is that the methods mentioned earlier (including DTI) are all predicated on the
assumption that each voxel in each DWI always contains the same tissue mass. While this assumption
is not necessarily satisfied in DTI (although methods have been developed to correct this artifact), it is
much less likely to be satisfied as the number of DWIs increases or as the b-value increases. Other
forms of distortion (particularly eddy current distortion) can become severe at high b-values although
104
some strategies have been proposed to ameliorate these. At high b-values, however, bulk and
small-scale motion are difficult to correct, since often there are few landmarks to identify in strongly
diffusion-weighted images. Typically, since most of the signal attenuation is used to estimate D from
the signal decay, little is left over to characterize the nonGaussian contribution. In this regime, noise
becomes increasingly prominent in DWIs. The noise either has to be reduced or fitted in order not to
introduce artifacts in the processed data. Recent work in the effects of noise in high b-value DWI is
given in reference 78. Collectively these issues present serious technical challenges to all high b or high
q methods, which will have to be ameliorated.
© 2010 Elsevier
1 of 1 14-04-2010 19:38
LONGITUDINAL AND MULTI-SITE STUDIES

In performing multi-site or longitudinal DTI studies, several additional issues arise. The most basic is
how to compare high-dimensional D data obtained from different subjects or from the same subject at
different times. This tensor data contains both scalar and vector information. Applying warping
transformations developed only for scalar images, such as those implemented in statistical parametric
mapping (SPM), produces nonsensical results when applied to DTI data without taking appropriate
precautions to preserve the features of the tissue structure.105 Current understanding of admissible
transformations that can be applied to warp and register diffusion-tensor field data is still limited. A
second issue is the proliferation of measures derived from D to characterize different features of
isotropic and anisotropic diffusion. Consistent definitions of quantities such as the orientationally
averaged diffusivity, RA, FA, etc. should be used. It is advisable to use the same imaging acquisition
hardware, reconstruction software, and post-processing routines to help control for unnecessary
variability. All sites should use a well-characterized phantom, even if it is isotropic, to ensure that no
serious systematic artifacts exist, and that features of the DWIs are stable in time and among
platforms.
page 330
page 331
Statistical analysis of DTI data is complicated by several factors. Although in an ideal DTI experiment,
D has been shown to be distributed according to a multivariate normal distribution, 106 and Trace(D)
has been shown to be distributed according to a univariate normal distribution, 107 the parametric
distribution of many other tensor-derived parameters is unknown. Then, nonparametric statistical
hypothesis testing methods should be used to determine whether an observed difference between
different regions of interest (ROIs) is statistically significant. Empirical methods like the
bootstrap-which allow determination of the distribution of a statistical parameter empirically without
knowledge of the form of its distribution a priori-show great promise in this application,106 particularly
now that a large number of single-shot DWIs can be acquired during a single scanning session, making
this method practicable. Statistical properties of measured DTI parameters can then be compared on
a voxel-by-voxel basis. Moreover, the bootstrap method can be used to assure data quality.
© 2010 Elsevier
1 of 1 14-04-2010 19:39
CONCLUSION
DTI provides a means to probe tissue structure at different levels of hierarchical organization. While
experimental diffusion times are associated with water molecule displacements on the order of
microns, these molecular motions are then averaged within a millimeter scale voxel, and then
subsequently assembled into an image volume on a meter scale that encompasses tissues and
organs. This integration permits us to study and elucidate complex structural features of tissue
spanning length scales from the macromolecular to the macroscopic. Various DTI-derived parameters
described earlier, such as maps of the eigenvalues of the diffusion tensor, its trace, measures of the
degree of diffusion anisotropy and organization, estimates of fiber direction and white matter tract
trajectory provide a unique multi-scale description of tissue architecture and organization.
Acknowledgment
Thanks go to Liz Salak, who edited this manuscript.
REFERENCES
1. Cleveland GG, Chang DC, Hazlewood CF, et al: Nuclear magnetic resonance measurement of skeletal muscle: anisotrophy
of the diffusion coefficient of the intracellular water. Biophys J 16:1043-1053, 1976. Medline Similar articles
2. Tanner JE: Self diffusion of water in frog muscle. Biophys J 28:107-116, 1979. Medline Similar articles
3. Chenevert TL, Brunberg JA, Pipe JG: Anisotropic diffusion in human white matter: demonstration with MR techniques in vivo.
4. Douek P, Turner R, Pekar J, et al: MR color mapping of myelin fiber orientation. J Comput Assist Tomogr 15:923-929, 1991.
5. Moseley ME, Kucharczyk J, Asgari HS, et al: Anisotropy in diffusion-weighted MRI. Magn Reson Med 19:321-326, 1991.
6. Sakuma H, Nomura Y, Takeda K, et al: Adult and neonatal human brain: diffusional anisotropy and myelination with diffusion-
weighted MR imaging. Radiology 180:229-233, 1991. Medline Similar articles
7. Moonen CT, Pekar J, de Vleeschouwer MH, et al: Restricted and anisotropic displacement of water in healthy cat brain and
in stroke studied by NMR diffusion imaging. Magn Reson Med 19:327-332, 1991. Medline Similar articles
8. Basser PJ, Mattiello J, Le Bihan D: MR diffusion tensor spectroscopy and imaging. Biophys J 66:259-267, 1994.
9. Garrido L, Wedeen VJ, Kwong KK, et al: Anisotropy of water diffusion in the myocardium of the rat. Circ Res 74:789-793,
10. Turner R, Le Bihan D, Maier J, et al: Echo-planar imaging of intravoxel incoherent motion. Radiology 177:407-414, 1990.
11. Moseley ME, Cohen Y, Kucharczyk J, et al: Diffusion-weighted MR imaging of anisotropic water diffusion in cat central
nervous system. Radiology 176:439-445, 1990. Medline Similar articles
12. Henkelman RM, Stanisz GJ, Kim JK, et al: Anisotropy of NMR properties of tissues: Anisotropy of NMR properties of
tissues. Magn Reson Med 32:592-601, 1994. Medline Similar articles
13. Crank J: The mathematics of diffusion, 2nd ed. Oxford: Oxford University Press, 1975.
14. Basser PJ, Mattiello J, Le Bihan D: Estimation of the effective self-diffusion tensor from the NMR spin echo. J Magn Reson
B 103:247-254, 1994. Medline Similar articles
15. Mattiello J, Basser PJ, Le Bihan D: Analytical expression for the b matrix in NMR diffusion imaging and spectroscopy. J
Magn Reson A 108:131-141, 1994.
16. Mattiello J, Basser PJ, Le Bihan D: The b matrix in diffusion tensor echo-planar imaging. Magn Reson Med 37, 292-300,
1997.
17. Mattiello J, Basser PJ, Le Bihan D: Analytical calculation of the b matrix in diffusion imaging. In Bihan DL (ed): Diffusion
and Perfusion Magnetic Resonance Imaging. New York: Raven Press, 77-90, 1995.
18. Stejskal EO, Tanner JE: Spin diffusion measurements: spin echoes in the presence of time-dependent field gradient. J
Chem Phys 42:288-292, 1965.
19. Neeman M, Freyer JP, Sillerud LO: Pulsed-gradient spin-echo studies in NMR imaging. Effects of the imaging gradients on
the determination of diffusion coefficients. J Magn Reson 90:303-312, 1990.
20. Basser PJ, Mattiello J, Le Bihan D: In 11th Annual Meeting of the SMRM, Berlin. 1:1222, 1992.
21. Einstein A: Investigations on the Theory of the Brownian Movement. New York: Dover Publications, 1926.
22. Stejskal EO: Use of spin echoes in a pulsed magnetic field gradient to study restricted diffusion and flow. J Chem Phys
43:3597-3603, 1965.
1 of 4 14-04-2010 19:40
23. Basser PJ, Le Bihan D: In 11th Annual Meeting of the SMRM, Berlin. 1:1221, 1992.
24. Basser PJ, Mattiello J, Pierpaoli C, Le Bihan D: In IEEE 16th Annual International Meeting of the IEEE EMBS, Baltimore.
445, 1994.
25. Kärger J, Pfeifer H, Heink W. In Waugh J (ed): Advances in Magnetic Resonance. New York: Academic Press, 12: 1-89,
1988.
26. Moseley ME, Kucharczyk J, Mintorovitch J, et al: Diffusion-weighted MR imaging of acute stroke: correlation with
T2-weighted and magnetic susceptibility-enhanced MR imaging in cats. Am J Neuroradiol 11:423-429, 1990. Medline
Similar articles
27. Moseley ME, Mintorovitch J, Cohen Y, et al: Early detection of ischemic injury: comparison of spectroscopy, diffusion-, T2-,
and magnetic susceptibility-weighted MRI in cats. Acta Neurochir Suppl (Wien) 51:207-209, 1990.
28. Moseley ME, Cohen Y, Mintorovitch J, et al: Early detection of regional cerebral ischemia in cats: comparison of diffusion-
and T2-weighted MRI and spectroscopy. Magn Reson Med 14:330-346, 1990. Medline Similar articles
29. Warach S, Gaa J, Siewert B, et al: Acute human stroke studied by whole brain echo planar diffusion-weighted magnetic
resonance imaging. Ann Neurol 37:231-241, 1995. Medline Similar articles
31. Doran M, Hajnal JV, Van Bruggen N, et al: Normal and abnormal white matter tracts shown by MR imaging using
directional diffusion weighted sequences. J Comput Assist Tomogr 14:865-873, 1990. Medline Similar articles
32. Pierpaoli C, Basser PJ: Toward a quantitative assessment of diffusion anisotropy. Magn Reson Med 36:893-906, 1996.
33. Pierpaoli C, Jezzard P, Basser PJ, et al: Diffusion tensor MR imaging of the human brain. Radiology 201:637-648, 1996.
34. Ulug AM, Beauchamp N, Bryan RN, van Zijl PCM: Absolute quantitation of diffusion constants in human stroke. Stroke
35. van Gelderen P, de Vleeschouwer MHM, DesPres D, et al: Water diffusion and acute stroke. Magn Reson Med
36. Wong EC, Cox RW, Song AW: Optimized isotropic diffusion weighting. Magn Reson Med 34:139-143, 1995. Medline
Similar articles
37. Mori S, van Zijl PC: Diffusion weighting by the trace of the diffusion tensor within a single scan. Magn Reson Med 33:41-52,
38. Basser PJ, Jones DK: Diffusion-tensor MRI: theory, experimental design and data analysis - a technical review. NMR
39. Basser PJ: Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. NMR
40. Basser PJ, Pierpaoli C: Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor
MRI. J Magn Reson B 111:209-219, 1996. Medline Similar articles
41. Shimony JS, McKinstry RC, Akbudak E, et al: Quantitative diffusion-tensor anisotropy brain MR imaging: normative human
data and anatomic analysis. Radiology 212:770-784, 1999. Medline Similar articles
42. Westin CF, Maier S, Khidir B, et al: In MICCAI'99; Cambridge: Springer-Verlag: LNCS-1679:441-452, 1999.
43. Pajevic S, Aldroubi A, Basser PJ: A continuous tensor field approximation of discrete DT-MRI data for extracting
microstructural and architectural features of tissue. J Magn Reson 154:85-100, 2002. Medline Similar articles
44. Poupon C, Clark CA, Frouin V, et al: Regularization of diffusion-based direction maps for the tracking of brain white matter
fascicles. Neuroimage 12:184-195, 2000. Medline Similar articles
45. Pajevic S, Pierpaoli C: Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data:
application to white matter fiber tract mapping in the human brain. Magn Reson Med 42:526-540, 1999. Medline
Similar articles
46. Makris N, Worth AJ, Sorensen AG, et al: Morphometry of in vivo human white matter association pathways with diffusion-
weighted magnetic resonance imaging. Ann Neurol 42:951-962, 1997. Medline Similar articles
47. Pajevic S, Pierpaoli C: Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data:
application to white matter fiber tract mapping in the human brain. Magn Reson Med 43:921, 2000. Medline
Similar articles
page 331
page 332
48. Wedeen VJ, Davis TL, Weisskoff RM, et al: In Proceedings of the First International Conference for Functional Mapping of
the Human Brain, Paris, 1:36, 1995.
49. Basser PJ: In 6th Annual Meeting of the ISMRM, Sydney, 2:1226, 1998.
50. Basser PJ, Pajevic S, Pierpaoli C, et al: In vivo fiber tractography using DT-MRI data. Magn Reson Med 44:625-632, 2000.
2 of 4 14-04-2010 19:40
51. Conturo TE, Lori NF, Cull TS, et al: Tracking neuronal fiber pathways in the living human brain. Proc Natl Acad Sci U S A
52. Mori S, Crain BJ, Chacko VP, et al: Three-dimensional tracking of axonal projections in the brain by magnetic resonance
imaging. Ann Neurol 45:265-269, 1999. Medline Similar articles
53. Poupon C, Clark CA, Frouin V, et al: In 8th Annual Meeting of the ISMRM: Philadelphia, 325, 1999.
54. Jones DK, Simmons A, Williams SC, et al: Non-invasive assessment of axonal fiber connectivity in the human brain via
diffusion tensor MRI. Magn Reson Med 42:37-41, 1999. Medline Similar articles
55. Bammer R, Acar B, Moseley ME: In vivo MR tractography using diffusion imaging. Eur J Radiol 45:223-234, 2003.
56. Tuch DS, Belliveau JW, Wedeen V: In Eighth Annual Meeting of the ISMRM, ISMRM, Berkeley, CA. 791, 2000.
57. Parker GJ, Haroon HA, Wheeler-Kingshott CA: A framework for a streamline-based probabilistic index of connectivity
(PICo) using a structural interpretation of MRI diffusion measurements. J Magn Reson Imaging 18:242-254, 2003. Medline
Similar articles
58. Koch M, Glauche V, Finsterbusch J, et al: Estimation of anatomical, connectivity from diffusion tensor data. NeuroImage
13:S176, 2001.
59. Koch MA, Norris DG, Hund-Georgiadis M: An investigation of functional and anatomical connectivity using magnetic
resonance imaging. Neuroimage 16:241-250, 2002. Medline Similar articles
60. Batchelor PG, Hill DLG, Calamante F, Atkinson D: In 17th International Conference, IMPI'01; Information Processing in
Medical Imaging 121-133, 2001.
61. Parker GJ, Wheeler-Kingshott CA, Barker GJ: Estimating distributed anatomical connectivity using fast marching methods
and diffusion tensor imaging. IEEE Trans Med Imaging 21:505-512, 2002. Medline Similar articles
62. Behrens TEJ, Jenkinson M, Brady JM, Smith SM: In ISMRM: Honolulu. 1142, 2002.
63. Behrens TE, Woolrich MW, Jenkinson M, et al: Characterization and propagation of uncertainty in diffusion-weighted MR
64. Jones DK, Pierpaoli C: In 12th Annual ISMRM: Kyoto, Japan. 2004.
65. Pierpaoli C, Barnett A, Pajevic S, et al: Water diffusion changes in Wallerian degeneration and their dependence on white
matter architecture. Neuroimage 13:1174-1185, 2001. Medline Similar articles
66. Catani M, Howard RJ, Pajevic S, Jones DK: Virtual in vivo interactive dissection of white matter fasciculi in the human
brain. Neuroimage 17:77-94, 2002. Medline Similar articles
67. Zhang S, Curry C, Morris DS, Laidlaw DH: In Proceedings of the IEEE Visualization Conference: Utah, 2000
68. Zhang S, Bastin ME, Laidlaw DH, et al: Visualization and analysis of white matter structural asymmetry in diffusion tensor
MRI data. Magn Reson Med 51:140-147, 2004. Medline Similar articles
69. Anderson AW, Gore JC: Analysis and correction of motion artifacts in diffusion weighted imaging. Magn Reson Med
70. Ordidge RJ, Helpern JA, Qing ZX, et al: Correction of motional artifacts in diffusion-weighted MR images using navigator
echoes. Magn Reson Imaging 12:455-460, 1994. Medline Similar articles
71. Rohde GK, Barnett AS, Basser PJ, et al: Comprehensive approach for correction of motion and distortion in diffusion-
weighted MRI. Magn Reson Med 51:103-114, 2004. Medline Similar articles
72. Jezzard P, Barnett AS, Pierpaoli C: Characterization of and correction for eddy current artifacts in echo planar diffusion
73. Clark CA, Barker GJ, Tofts PS: Improved reduction of motion artifacts in diffusion imaging using navigator echoes and
velocity compensation. J Magn Reson 141:52-61, 1999. Medline Similar articles
74. Basser PJ, Pajevic S: Statistical artifacts in diffusion tensor MRI (DT-MRI) caused by background noise. Magn Reson Med
75. Martin KM, Papadakis NG, Huang CL, et al: The reduction of the sorting bias in the eigenvalues of the diffusion tensor.
76. Basser PJ, Pajevic S: Statistical artifacts in diffusion tensor MRI (DT-MRI) caused by background noise. Magn Reson Med
77. Basser PJ, Pajevic S: In 7th Annual ISMRM: Philadelphia, 1788, 1999.
78. Jones DK, Basser PJ: Squashing peanuts and smashing pumpkins. How noise distorts diffusion-weighted MR data. Magn
79. Jara H, Wehrli FW: Determination of background gradients with diffusion MR imaging. J Magn Reson Imaging 4:787-797,
80. Basser PJ: In Third Meeting of the SMR: Nice, France, 308, 1995.
81. Basser PJ, Mattiello J, Le Bihan D: Method and system for measuring the diffusion tensor and for diffusion tensor imaging.
USPTO 5:529-531, 1996.
82. Le Bihan D, Delannoy J, Levin RL: Temperature mapping with MR imaging of molecular diffusion: application to
3 of 4 14-04-2010 19:40
hyperthermia. Radiology 171:853-857, 1989. Medline Similar articles

83. Mills R: Sef-diffusion in normal and heavy water in the range 1-45°. Journal of Physical Chemistry 77:685-688, 1973.
84. Simpson JH, Carr HY: Diffusion and nuclear spin relaxation in water. Physical Review 111:1201-1202, 1958.
85. Neil JJ, Shiran SI, McKinstry RC, et al: Normal brain in human newborns: apparant diffusion coefficient and diffusion
anistropy measured by using diffusion tensor MR imaging. Radiology 209:57-66, 1998. Medline Similar articles
86. Beaulieu C, Allen PS: Water diffusion in the giant axon of the squid: implications for diffusion-weighted MRI of the nervous
system. Magn Reson Med 32:579-583, 1994. Medline Similar articles
87. Beaulieu C, Allen PS: Determinants of anisotropic water diffusion in nerves. Magn Reson Med 31:394-400, 1994.
88. Wimberger DM, Roberts TP, Barkovich AJ, et al: Identification of "premyelination" by diffusion-weighted MRI. J Comput
89. Beaulieu C: The basis of anisotropic water diffusion in the nervous system - a technical review. NMR Biomed 15:435-455,
90. Mulkern RV, Gudbjartsson H, Westin CF, et al: Multi-component apparent diffusion coefficients in human brain. NMR
91. Peled S, Cory DG, Raymond SA, et al: Water diffusion, T(2), and compartmentation in frog sciatic nerve. Magn Reson Med
92. Basser PJ: Relationships between diffusion tensor and q-space MRI. Magn Reson Med 47:392-397, 2002. Medline
Similar articles
93. Wedeen VJ, Reese TG, Tuch DS, et al: In 8th Annual Meeting of the ISMRM: Denver, 82, 2000.
94. Callaghan PT. Principles of nuclear magnetic resonance microscopy. Oxford: Oxford University Press, 1991.
95. Frank LR: Characterization of anisotropy in high angular resolution diffusion-weighted MRI. Magn Reson Med
96. Frank LR: Anisotropy in high angular resolution diffusion-weighted MRI. Magn Reson Med 45:935-939, 2001. Medline
Similar articles
97. Jansons KM, Alexander DC: Persistent angular structure: new insights from diffusion magnetic resonance imaging data.
Inverse Probl 19:1031-1046, 2003.
98. Ozarslan E, Mareci TH: Generalized diffusion tensor imaging and analytical relationships between diffusion tensor imaging
and high angular resolution diffusion imaging. Magn Reson Med 50:955-965, 2003. Medline Similar articles
99. Liu CL, Bammer R, Moseley ME: Generalised diffusion tensor imaging (GDTI): a method for characterising and imaging
diffusion anistropy caused by non-Gaussian diffusion. Israel J Chem 43:145-154, 2003.
100. Tuch DS, Reese TG, Wiegell MR, et al: Diffusion MRI of complex neural architecture. Neuron 40:885-895, 2003.
101. Assaf Y, Freidlin RZ, Rohde GK, et al: New modeling and experimental framework to characterize hindered and restricted
water diffusion in brain white matter. Magn Reson Med 52:965-978, 2004. Medline Similar articles
102. Inglis BA, Bossart EL, Buckley DL, et al: Visualization of neural tissue water compartments using biexponential diffusion
tensor MRI. Magn Reson Med 45:580-587, 2001. Medline Similar articles
103. Clark CA, Le Bihan D: Water diffusion compartmentation and anisotropy at high b values in the human brain. Magn Reson
104. Reese TG, Heid O, Weisskoff RM, et al: Reduction of eddy-current-induced distortion in diffusion MRI using a twice-
refocused spin echo. Magn Reson Med 49: 177-182, 2003.
105. Alexander DC, Pierpaoli C, Basser PJ, et al: Spatial transformations of diffusion tensor magnetic resonance images.
IEEE Trans Med Imaging 20:1131-1139, 2001. Medline Similar articles
106. Pajevic S, Basser PJ: Parametric and non-parametric statistical analysis of DT-MRI data. J Magn Reson 161:1-14, 2003.
107. Basser PJ, Pajevic S: Dealing with uncertainty in DT-MRI data. Israel J Chem 43:129-144, 2003.
© 2010 Elsevier
4 of 4 14-04-2010 19:40
ERFUSION MAGING OF THE RAIN
David Alsop
INTRODUCTION
Disorders of blood flow underlie most of the major causes of death in the developed world. Insufficient
blood flow, or ischemia, is the primary mechanism of damage in myocardial infarction, pulmonary
embolism, and stroke. The development of a blood-supplying vasculature, or angiogenesis, is the key
enabling step in the growth of most cancerous tumors.1 For these and other pathologies, imaging of
blood flow can provide valuable information for guiding diagnosis and monitoring therapies.
Perfusion, or tissue-specific blood flow, is an established physiologic measure of the volume of blood
flowing through the microvasculature of a mass of tissue in a certain period of time. Often expressed in
ml/100 g/min, perfusion is a quantitative measure which can be determined with many different
experimental techniques. This means that perfusion, unlike T1 or T2, has a quantitative meaning
outside MRI and values should be comparable across measurement modalities. Often, the theoretical
equations used for determining perfusion can be translated from one measurement technique to
another.
The primary unifying factor in noninvasive measures of blood flow is the use of a tracer. We can further
divide blood flow measures based on the characteristics of these tracers. Tracers which easily pass
through blood vessel walls are referred to as diffusible tracers. These include H2O, or H2O15 as used
in positron emission tomography (PET), and Xe as used in computed tomography-based perfusion
measures and some nuclear medicine measures. Tracers which remain within the vessel walls are
known as intravascular tracers. Their use for perfusion imaging in the brain was pioneered with
iodinated contrast in computed tomography (CT)2 and then later Gd-DTPA and other magnetic
resonance (MR) contrast agents. Unfortunately, these agents are partially diffusible in most tissues of
the body and separating perfusion from vessel permeability effects can be a challenge. A final class of
agents are microsphere type tracers. The prototypical microspheres are slightly larger than capillaries
and hence they get stuck within the microvasculature indefinitely. While widely used for invasive animal
studies, they are considered too risky for human studies. Clinical agents such as the SPECT agent
Tc-HMPAO mimic microspheres by remaining in the target tissue for a long time but they achieve this
by diffusing into tissue and then undergoing a chemical transformation within the tissue that
discourages diffusing back out.3 A final distinction between agents is stability. Radioactively labeled
agents decay with time while many MR and CT contrast agents are chemically stable. Decay must be
considered when perfusion is quantified.
page 333
page 334
A variety of different tracers have been employed for perfusion imaging with MRI. Perfusion can be
17,4 2 5,6 6-8
measured in principle with O H 2O, a number of compounds with large quantities of F, and with
9,10 13 11
hyperpolarized Xe gas or C -containing compounds. These tracers have the advantage that the
normal concentration of these agents in tissue is almost zero. Since the signal from these other nuclei
can be well separated from the water proton signal usually observed in clinical MRI, the only signal
observed is due to perfusion effects. However, these agents generally produce weak signals, causing
the resulting images to be noisy with low spatial resolution, and approval of the agents for clinical use
has not been actively pursued. The main options for clinical and clinical research applications of
perfusion MRI employ the strong signal from water protons in tissue. 12
Perfusion imaging with the proton signal can be performed either by dynamic imaging following the
bolus injection of an MR contrast agent or by labeling of the water in the inflowing arterial blood. The
effects of contrast agents in the vessels or tissue can be quite strong and consequently contrast
agent-based methods have been widely explored. When the contrast agent is observed using T2- or
1 of 2 14-04-2010 19:43
T2*-weighted imaging, the signal change is due to the magnetic susceptibility of the contrast agent and
the technique has become known as dynamic susceptibility contrast (DSC) imaging.13-15 This approach
16,17
is mostly employed in brain. When the contrast agent is observed using T1-weighted imaging, the
signal change is due to the T1 shortening effects of the contrast agent and, since shorter T1 is usually
brighter on T1-weighted imaging, the technique is often referred to as dynamic contrast enhancement
(DCE) imaging. DCE MRI is usually employed outside the brain. Imaging without contrast agent is
possible by selectively affecting the MRI signal from water in the inflowing arterial blood.18,19 This
arterial spin labeling (ASL) technique produces a smaller signal change than the contrast agent
methods but the lack of an injection, the use of diffusible water as the tracer, and a number of other
methodologic advantages make this approach an interesting option for a number of research and
clinical applications.
In this chapter, the methods and challenges of these techniques will be described with only a few hints
of their applications. Their use in particular pathologies and their role in clinical evaluation will be
discussed in the chapters covering the relevant diseases.
© 2010 Elsevier
2 of 2 14-04-2010 19:43
PERFUSION IMAGING WITH CONTRAST AGENTS

Magnetic Contrast Agents in Neural Tissues
Because the blood-brain barrier tightly blocks access to paramagnetic ions, such as Gd, Dy and Mn, and safe
doses of these agents require these ions to be contained within much larger molecules, or chelates, there are no
approved MR contrast agents that can cross the intact blood-brain barrier. Thus within brain tissue, contrast
agents remain within the vasculature and are intravascular agents. Only in enhancing tissues, such as neoplasms,
do the agents actually enter the brain. Let's first examine the effects of an intravascular tracer and return to the
special case of blood-brain barrier disruption below.
Intravascular Tracers
Since the brain blood volume is between 3% and 6%, the possible signal changes resulting from alterations to the
intravascular signal are relatively small. Fortunately, the presence of magnetic agents within the vasculature can
have effects beyond the vessels.20-25
Magnetic contrast agents can act on spins outside the vessels in two ways. First, spins can move between small
blood vessels in the microvasculature and the tissue. Water molecules are in constant and rapid thermal motion.
Frequent collisions with other water molecules, membranes, and large macromolecules prevent the molecules from
moving very far. The random diffusive motion of these molecules does lead to slow movement. On the time scale
of 100 ms, water can move in and out of a nearby capillary. Consequently, a larger volume than just the vascular
volume can experience the effects of the contrast agent. The second, and more significant, way an intravascular
contrast agent can affect tissue signal is by altering the magnetic fields in the nearby tissue. A strong magnet can
have magnetic effects even at a considerable distance. Likewise, the presence of magnetic contrast agent in the
blood vessels will cause nonuniform magnetic fields in the nearby tissue (Fig. 12-1). These nonuniform fields will
cause a spread in the frequencies of tissue spins because frequency is proportional to magnetic field strength.
After excitation, the spins will start to get out of phase and the total signal will decrease. This decay of the signal is
known as T2* decay and the primary effect of the nonuniform fields is to decrease T2*.
20,23
Somewhat surprisingly, spin-echo T2 will also decrease when contrast agent is in the vessels. While
spin-echoes are supposed to eliminate the effects of nonuniform magnetic fields, spin-echo refocusing is only
successful when no motion is present. The contrast agent in small vessels causes large magnetic field gradients in
the tissue which act like strong diffusion-sensitizing gradients (please refer to Chapter 10 on diffusion imaging for
more details). When a long TE image is acquired, these intrinsic gradients are "on" for the whole time, causing
diffusion attenuation in the image. Note that T2 as measured by sequences which rapidly refocus the spins for fast
imaging, such as fast spin-echo or turbo spin-echo, will not be as greatly changed because of the closely spaced
refocusing pulses used in these sequences. This reduced sensitivity to T2 changes caused by an intravascular
contrast agent is closely related to the reduced sensitivity to hemorrhage.
The effect of contrast agent in large vessels on T2 of the surrounding tissue is very weak because the gradients
produced depend on the size of the vessel. Consequently, it has been argued that T2-based DSC studies show
less prominent large vessel effects than T2* studies, for which the effect of the agent is independent of the
vessel's size. Large vessels are still very prominent on spin-echo DSC studies, however, probably because of
intravascular T2 changes resulting from diffusion of water between red blood cells, which do not contain contrast
agent, and the plasma, which does. While it would seem that the diffusion rate in the tissue, which can be strongly
affected in acute stroke, would have a big effect on the T2 with contrast agent present, clinical studies and
20
theoretical investigations indicate that the effect of diffusion changes is small.
page 334
page 335
1 of 13 14-04-2010 19:46
Add to lightbox
Add to lightbox
Figure 12-1 The effect of an intravascular contrast agent in the brain vasculature. The agent cannot cross the
blood-brain barrier into the tissue or into the red blood cells (A) but the magnetic field from the contrast agent
extends into the tissue (B) and makes a nonuniform magnetic field in the tissue.
13,14,26
It is generally assumed that T2 and T2* depend upon contrast agent concentration according to the following
relationship:
where C is the concentration of the contrast agent in the blood, CBV is the fraction of the voxel volume occupied
by blood vessels and R or R* is the relaxivity of the agent. The T2 and T2* effects of a contrast agent are almost
entirely determined by its magnetic susceptibility, i.e., its tendency to enhance magnetic fields, so the molecular
chemistry of the agent is important only to the extent that it affects susceptibility. Unfortunately, the relaxivity of the
agent is not certain. A theory for the T2* effects of contrast in vessels has been developed22 that permits a
method to estimate R* and which supports the idea that R* is not strongly dependent on tissue type as long as the
2 of 13 14-04-2010 19:46
blood volume is predominantly in vessels large enough that diffusion is not important, usually larger than capillaries,
and that they are randomly oriented. There has, however, been a report demonstrating that R* is significantly
smaller in tissues with complex microvascular structure, such as kidney and spleen, than in other tissues. 27 The
theoretical and experimental basis for a tissue-independent relaxivity for spin-echo studies, R, is much weaker. In
20,23
theoretical models, R is highly dependent on the fraction of the blood volume in small vessels. In addition, the
28,29
variability of the ratio of R* to R has been used as an approach to measure the average vessel size in tissues.
In normal brain tissues, however, blood volume images generated with spin-echo appear to be qualitatively similar
to those obtained with gradient-echo, except in regions where a large artery or vein is present.
In addition to T2 and T2* changes, high intravascular concentrations of T1 contrast agents, such as Gd-DTPA, do
cause changes on T1-weighted images. The T1 change in the blood volume causes an increase of signal on
T1-weighted imaging because the blood signal is increased. In addition, the diffusion and exchange of water from
the microvasculature to the tissue leads to additional T1 enhancement. Still, the T1 change is quite small in normal
tissue and depends upon vascular volume, water permeability, contrast concentration, and diffusion in a complex
fashion. Recent work has suggested that low contrast dose and high T1 weighting can make this technique
viable.30 While perfusion imaging with T1 contrast has been proposed for use in the normal brain,31 methodologic
challenges make it still an area of development rather than application.
page 335
page 336
In tumors and in tissue following injury, the blood-brain barrier may not prevent contrast agents from entering the
tissue from the microvasculature. Still, because of the large size of the agents, the permeability of the
microvasculature plays an important role in determining the amount of contrast that enters the tissue. Once within
the tissue, the contrast agent usually resides only in the extracellular space but because of rapid diffusion of water
across cell membranes, the effect of the agent on T1 is similar to what it would be if uniformly distributed in tissue.
The T1 of the tissue as a function of concentration is given by:
The strong T1 effect of the contrast agent when it enters tissue overshadows any intravascular effect on T1 and is
responsible for the strong enhancement of some pathologies on T1-weighted, contrast-enhanced scans.
The Relationship Between Contrast Dynamics and Perfusion

Before we examine the MR-specific techniques in detail, it is helpful to consider the dynamics of an injected tracer
and its relationship to perfusion and other hemodynamic parameters. The discussion is simpler if we limit ourselves
to either fully diffusible agents, which might be the case for Gd-DTPA in very porous microvasculature, and
intravascular agents, as would be the case for Gd-DTPA in the brain with an uncompromised blood-brain barrier.
Add to lightbox
Figure 12-2 If a narrow bolus could be injected right into the feeding arterioles of each tissue voxel, the
concentration in each voxel would first jump to a level proportional to the flow in each voxel and then decay away
with the mean transit time.
Let's consider injecting an amount of contrast very rapidly into the arteriole that feeds a particular cube of tissue
(Fig. 12-2). Right after the injection, if we inject quickly enough, the total concentration of contrast in the cube will
be equal to the amount of contrast injected divided by the volume of the cube. If we inject from further up the
arterial tree, the contrast will be divided across a number of cubes of tissue. Since the contrast is well mixed with
3 of 13 14-04-2010 19:46
the blood and the blood flows to different cubes in proportion to the blood flow, the contrast agent will be divided
proportional to blood flow also. Thus the contrast concentration immediately after injection is proportional to the
flow to the tissue. With time, this initial concentration then decays away because of outflow of the contrast agent
into the veins. The average time it takes for the contrast agent to leave is known as the mean transit time (MTT).
The MTT and blood flow are related by the accessible volume fraction (AVF):
The accessible volume fraction is the fraction of the volume in each voxel that the contrast agent can get into. For
fully diffusible agents, the AVF is essentially 100%. For an intravascular tracer, the AVF is the fraction of the voxel
volume occupied by blood plasma. (Typically intravascular tracers also do not enter blood cells, hence just the
plasma volume is the AVF.) The accessible volume fraction is also equal to the area under the concentration
versus time curve.
In principle, then, measuring relative blood flow should be very easy. The concentration at the beginning of the
tissue concentration curve is proportional to blood flow and other interesting parameters may be obtained from the
2,32,33
concentration decay curve. In practice, measuring perfusion is not so easy because practical intravenous
bolus injections produce quite broad arterial concentrations in the brain tissue. Usually some early-arriving contrast
begins to leave the tissue through the veins before the end of the inflowing bolus. This means there is no time
where the concentration is simply proportional to the flow. Instead, one must observe the concentration over time
in the tissue as well as the concentration in a feeding artery and perform mathematical operations to infer the flow.
Typically the arterial concentration is measured in a large feeding vessel using the same images acquired to
measure tissue signal. A related problem is that it takes additional time for the contrast to move from the large
feeding vessel to the tissue of interest. The delay of the contrast agent arrival and, more importantly, any
34
broadening of the bolus that occurs as a result can complicate the interpretation of the flow.
An example of a typical arterial bolus width for a dynamic contrast study and the corresponding concentration
curve in healthy tissue is shown in Figure 12-3. Because the bolus width is comparable to or greater than the
healthy tissue transit time, the tissue concentration curve looks very similar to the arterial bolus shape. From this
example, it is clear that knowledge of the arterial input bolus is essential to inferring flow from the concentration
measurements over time.
Dynamic Susceptibility Contrast Imaging with T2 or T2* Contrast

page 336
page 337
Based on the considerations of the previous sections, a standard approach to DSC imaging in the brain has
evolved using T2- or T2*-sensitive imaging at high temporal resolution, about every 1.5-2 seconds during and after
the bolus injection of a dose of clinically approved paramagnetic contrast agent. The imaging is typically performed
with echo-planar imaging because it can produce T2- or T2*-sensitive images from multiple slices within a second,
though a number of recent advances in fast imaging have reopened the consideration of different imaging
sequences. The intravenous bolus is rapidly injected, often with a power injector, and typically a single clinical dose
of Gd-DTPA (0.1 mM/kg) or related contrast agent is used for T2* imaging and often a double dose (0.2 mM/kg)
to produce a comparable change for T2-weighted imaging. The bolus is injected rapidly to minimize the broadening
of the bolus but because of the broadening that occurs from systemic transit, the speed of bolus injection is often
not the limiting factor on the bolus width in the cerebral arteries.35
One of the greatest challenges of DSC MRI is that the transit time through the tissue is so fast. The transit time
through the vascular system in the brain is of the order of 4 seconds, a relatively short time. A similar time scale
for transit is observed with iodinated contrast in CT. Consider instead the longer time scales in other types of
perfusion measurements. When a diffusible tracer that enters the tissue is used, such as Xe used in stable Xe
36 15 37
CT or H2 used in PET flow imaging, the typical time to pass through the tissue is about 20 times longer
because of the much larger accessible volume. Tc-HMPAO, a SPECT contrast agent used for flow sensitivity,
enters the tissue and is chemically altered so it stays in the tissue even longer. Consequently, an MRI DSC
perfusion imaging study requires a much tighter bolus and much higher temporal resolution imaging to measure
perfusion than a study with a diffusible tracer. On the other hand, the entire MRI perfusion experiment can be
performed in a much shorter time than one with stable diffusible tracers.
4 of 13 14-04-2010 19:46
Add to lightbox
Figure 12-3 A typical bolus shape after intravenous injection, blue, and the corresponding tissue concentration
curve, orange, from a region of interest. Because the bolus is broader than the normal tissue mean transit time, the
tissue concentration is only slightly broader than the injected bolus.
Typical signal intensities from a T2*-sensitive DSC study in a patient presenting with acute stroke symptoms are
shown in Figures 12-4 and 12-5. The signal intensity drops in the tissue with the arrival of contrast because of the
shortened T2* in the tissue when the vascular concentration is high. On the side contralateral to the stroke, the
signal intensity drops rapidly and then recovers fairly quickly because the normal flow in this region washes the
contrast back out. Depending on the contrast agent used, the sequence parameters, and the volume of the blood
vessels in the tissue, the signal intensity can drop from a few percent to more than a factor of 2. Eventually the
signal rises again after the bolus washes through the tissue but usually before the signal can recover to 0, the
bolus returns after circulating through the heart and lungs again. This second bolus is strongly attenuated because
of further spreading of the bolus as well as absorption into the permeable tissues outside the brain. Usually a third
recirculation is hard to detect but the image intensity never fully returns to the original intensity because the arterial
38
concentration of contrast agent will remain significant for many minutes after the injection. With the new class of
39
blood pool agents, the recirculation may need to be better understood because the arterial concentration will
remain higher than with agents that are absorbed elsewhere in the body. In the stroke region, the drop of the
signal is weaker and more prolonged, indicating abnormal hemodynamics in this region.
Add to lightbox
page 337
page 338
Figure 12-4 The signal from regions of interest in the unaffected (blue line) and affected (green line) hemisphere of
a patient presenting with acute stroke symptoms during a bolus injection. The affected hemisphere signal
decreases and recovers more slowly than the contralateral side.
5 of 13 14-04-2010 19:46
Add to lightbox
Figure 12-5 T2*-weighted echo-planar images acquired at 2-s intervals during the bolus injection of Gd-DTPA in
the same patient as Fig. 12-4. The asymmetry in the arrival of the bolus in the MCA territories is visually apparent.
The curves in Figure 12-4 show information on perfusion, blood volume, and transit time through the tissue.
Extracting this information, in either quantitative or qualitative form, from the image time series is challenging. A
number of simple approaches as well as complex algorithms have been proposed. The approaches are most
clearly divided by the objectives. Qualitative techniques are usually designed to provide useful diagnostic contrast
with a minimum of image post-processing steps. Quantitative techniques are targeted at obtaining absolute or at
least relative measures of blood flow, blood volume or tissue mean transit time. Since quantification methods may
increase processing time, noise and motion sensitivity, it is not necessarily true that they will provide greater
40
diagnostic accuracy. However, it seems likely that separating intrinsic tissue parameters such as perfusion from
extrinsic variables such as bolus width, contrast dose, and upstream vascular transit delays will lead to increased
41
diagnostic accuracy.
Qualitative DSC Analysis Methods

A great deal can be learned from the time series of echo-planar images merely by viewing the images sequentially
or in a cine loop with the grayscale level adjusted for high contrast. Often the limited post-processing software
available on the scanner makes this the only option for quick interpretation of the images. The drop in signal in the
arteries and then the tissue can be readily seen and asymmetries in arrival time, departure time, and degree of
attenuation can be clinically interpreted. This approach is especially useful if the subject moves during the bolus
arrival. While more sophisticated approaches require that the same region of tissue be in the same position from
timepoint to timepoint, qualitative viewing of the images can still be performed. Motion between images is a
significant problem for DSC MRI but individual images acquired with snapshot echo-planar imaging are rarely
contaminated by motion. Images from key timepoints, such as when the bolus first arrives somewhere in the brain,
can be useful diagnostic indicators if motion makes image-to-image comparison challenging.
page 338
page 339
6 of 13 14-04-2010 19:46
Add to lightbox
Figure 12-6 Subtraction of the T2* intensity from a baseline image highlights the effects of the injected bolus on the
signal intensity. Detail in the inflow and outflow patterns is more readily appreciated on such subtraction images.
Assuming that patient motion is not excessive, more sophisticated comparison of multiple images over time can be
performed. The simplest such comparison is just subtraction. If each of the images is subtracted from the first
image, or an average of the first several images acquired before contrast arrival, the effects of the contrast agent
inflow are highlighted (Fig. 12-6). With this subtraction, only changes in the signal intensity induced by the contrast
agent appear in the image and the inflowing bolus is more readily visually detected. As with the unsubtracted time
series, viewing of the images over time can provide valuable diagnostic information. Early arrival and rapid
clearance are generally related to higher flow while very late arrival suggests low flow and potentially vascular
disease. While the subtraction has made the effects of contrast bolus arrival easy to visualize, interpretation of the
images still requires viewing of a series of images. Much of the information can be derived from just a few images,
such as the image when contrast fully enters normal tissue, image number 4 in Figure 12-5, and a later image to
show later arrival in some regions, image number 8. However, unusual bolus width or timing or unusual
hemodynamics can complicate the interpretation of this smaller set of images. For this reason, numeric
calculations are often performed to provide better summaries of the data.
42
A very popular summary of bolus passage data is the time to peak concentration (TTP) map. This is a map that
provides a clean measure of the arrival time of the bolus in the tissue and can be a clear indicator of some
abnormalities. When the concentration is at a maximum in a region of tissue, the signal intensity is at a minimum.
The TTP is calculated one pixel at a time by searching through all the signal intensities as a function of time for that
pixel. The minimum signal intensity is found and the time corresponding to that minimum intensity is stored as the
TTP. There are some problems with TTP maps. First, the limited temporal resolution of most acquisitions means
the TTP values are measured in discrete values of 1-2 seconds. Since a typical range of TTP for normal tissue is
less than 5 seconds, the TTP map cannot detect subtle changes. This problem can be largely eliminated by
interpolating the data between the obtained timepoints or by fitting the signal intensity to a model. A second
problem with TTP maps is ambiguity in the TTP in regions where the concentration is very low such that the
minimum signal intensity is mostly a result of noise. This can occur in regions with very low flow, such as in the
core of a stroke or in ventricles where there is no perfused tissue. An example of a TTP map derived from the
data in Figure 12-5 is shown in Figure 12-7.
Quantitative Analysis of DSC MRI

page 339
page 340
7 of 13 14-04-2010 19:46
Add to lightbox
Figure 12-7 Calculated images obtained from the time series of T2* images acquired in the acute stroke patient
shown in Fig. 12-5. The time to peak image (left) highlights the affected territory. The relative blood volume image
(second from left) shows a slight elevation of blood volume in the same region. Mean transit time (third from left)
and relative blood flow images (right) show the defect apparent in the MCA territory as well. These last two
quantitative images, calculated with SVD deconvolution, may be more specific than the time-to-peak map.
While the TTP map is a popular measure of hemodynamic abnormality, it is not a quantitative measure of
physiology and it may be nonspecific since vascular delay can be mistaken for reduced perfusion. 41 Several
quantitative physiologic measures can be extracted from the signal as a function of time, though more complex
mathematical manipulations are required. The first step in quantitative analysis of the signal over time is to convert
the signal changes into concentration changes. Using the assumed linear relationship between contrast agent
concentration and changes in relaxation rate, one can calculate the concentration using the following equation:
where S is the signal at each timepoint, S0 is the signal before the contrast arrives, ln is the natural logarithm, and
R is the relaxivity of the contrast agent. The relaxivity is not easy to define because it depends, in principle, on
microvascular properties, diffusion coefficient and other unknown variables but it is common to assume that it is
constant across the brain. Challenges in determining the relaxivity are a major reason why absolute quantification
of blood flow is rarely performed. The conversion of the signal intensities in Figure 12-4 to relative concentrations
is shown in Figure 12-8.
Add to lightbox
Figure 12-8 Relative concentrations in regions of interest in contralateral (blue line) and ipsilateral (orange line)
temporal cortex in an acute stroke patient are calculated from the signal intensities in those regions (Fig. 12-4)
using the logarithm of the ratio to the baseline image.
Once the relative concentration of contrast is calculated with the above equation, the calculation of relative
cerebral blood volume (rCBV) can be performed. For intravascular tracers, the relative plasma volume is the
accessible volume fraction, which can be obtained by measuring the area under the concentration versus time
curve. Assuming constant hematocrit, rCBV is proportional to the plasma volume. Since the area under the curve is
unaffected by the width of the bolus, the area can be easily calculated as the sum of the concentration over time
as long as enough time is allowed for the entire bolus to enter and exit everywhere in the brain. If too little time is
43
allowed, those regions with delayed arrival or prolonged MTT will show artificially low blood volume. While the
arterial input concentration does not return to 0 for a long time after the bolus injection, the low concentration
8 of 13 14-04-2010 19:46
present after the second pass is usually not sufficient to bias the blood volume measurement.
To generate maps of CBV, the area under the concentration curve, also known as the discrete integral of the
curve, is calculated pixel by pixel. Calculation of the CBV in this manner is very susceptible to motion-related noise.
Suppose that, after the bolus injection, the patient moves slightly, causing the intensity in the pixel to decrease
relative to baseline. The calculation of the area will be dramatically increased by the small shift, because it keeps
adding area for as long as data are obtained. For these reasons, the calculation of the area is usually stopped at a
time shortly after the bolus peak, though this will cause underestimation of regions with very long delays or MTT. A
relative CBV map derived from the data in Figure 12-5 is shown in Figure 12-7.
page 340
page 341
Add to lightbox
Figure 12-9 Input arterial concentrations measured from voxels highlighted in arterial phase images (dark blue
line) and voxels selected using a short first-moment criterion (light blue line). The input function obtained from the
anatomically defined voxels is broader than the input function obtained from short first-moment voxels. Physical
simulations indicate that non-linear effects in the vessels make purely anatomic selection of arterial voxels for input
arterial concentration measurement inaccurate.
Quantification of perfusion, or cerebral blood flow (CBF), and MTT requires measurement of the arterial bolus
concentration where it enters the voxel. Arterial concentration is usually measured by MRI techniques rather than
invasive measurement and is most often performed using the T2- or T2*-weighted echo-planar images used for
measuring the tissue concentration. Unfortunately, the effect of contrast in larger cerebral arteries on the intensity
44
of an echo-planar image is complex. The T2 relaxation of blood with paramagnetic contrast is not linear with
concentration and the phase shifts and chemical shift induced by the paramagnetic agent, combined with the partial
volume effects caused by the low spatial resolution, cause very strange nonlinear effects which depend on angle of
the vessel, details of the echo-planar sequence, and the size of the vessel. For these reasons, specialized
acquisition sequences for the measurement of the arterial input have been proposed. Proposed methods include
very short TE, high-resolution sequences for the measurement of the amplitude or phase of the signal from a
44,45
vessel. These approaches usually also require the measurement of the angle of the vessel relative to the
magnetic field and modeling of the contamination from tissue within the voxel. While these are important areas of
development, currently the vast majority of studies are performed with only the T2- or T2*-sensitive echo-planar
images used for imaging the tissue. An approach for measuring arterial concentration from such images is clearly
required.
To assess methods for measuring input functions from echo-planar data, we have performed simulations of the
physical phenomena which play a role in the signal changes observed near arterial vessels on echo-planar
46
imaging. These simulations, and comparisons with in vivo data, suggest that those voxels with large blood volume
and the shortest first moment of the concentration curve tend to mirror the true arterial concentration (Fig. 12-9).
Presumably similar quantitative and qualitative curve shape criteria can also be used to choose the best voxels for
26
measuring input function. These criteria select the voxels in or near large vessels which are the least distorted by
the nonlinear effects which occur in echo-planar imaging of DSC. Our results indicate, however, that selection of
arterial voxels, or voxels near arterial vessels, based on anatomic criteria without consideration of the shape of the
curves is a poor approach to measuring arterial input function. Studies reporting good results using manually
selected voxels near arteries could be relying on subjective quality assessment by the user. 14,47 These simulations
provide some theoretical support for the use of arterial bolus measurements measured from echo-planar images
9 of 13 14-04-2010 19:46
for quantitative DSC MRI but it remains clear that this approach is not perfect and improved approaches to input
function measurement are desirable.
Even if the arterial concentration is accurately measured in an artery in the base of the brain, this concentration
34
need not faithfully represent the concentration in the small arteries feeding tissue throughout the brain. In transit
from the large artery to the tissue, the bolus can be delayed and broadened, especially in vascular disease and
stroke where the flow can arrive by multiple, circuitous collateral paths. We have proposed using a local measure
48
of the arterial concentration to accurately represent the concentration feeding the tissue voxels. In this approach,
the voxels with large blood volume and the shortest first moment in a region of the brain are used to define the
input function within that region. This approach shows promise in reducing the effects of delay and broadening of
the bolus during transit from the base of the brain and may improve the predictive value of DSC MRI for tissue
outcome. Examples of local input functions are shown in Figures 12-10 and 12-11.
Because of the nonlinear phenomena affecting the signal in and around arterial vessels, it is unlikely that one can
measure the absolute concentration in an artery. Instead, techniques such as those above for measuring the bolus
focus on the relative concentration over time. With this information it is possible to calculate the absolute MTT but
calculated blood volume and blood flow measurements are relative; this means the ratio of values in different brain
regions is probably accurate but the absolute value of flow or volume in any region is not known. Several
approaches to recovering absolute flow measures have been proposed. One group has proposed that the flow
and volume measurements obtained using a standardized protocol are correct except for a scale factor, which
49
they measured by comparison to PET measures. It is uncertain how robust this scale factor approach is to
different measurement approaches, bolus shapes, and the presence of pathology. Others have proposed
performing a measurement of whole-brain blood flow using phase contrast angiography in the cerebral vessels as
50
a method to obtain an individual scale factor for each image. Such scale factor approaches are desirable
because they can simultaneously correct for uncertainties in arterial input concentrations and the relaxivity, R or R*,
of the contrast agent in the tissue.
page 341
page 342
Add to lightbox
Figure 12-10 A local input function derived from a T2*-weighted time series in a healthy control subject. The
derived local arterial concentration at 2-s intervals is shown as a function time in one axial slice. The bolus arrives
later and is broader in the posterior circulation.
10 of 13 14-04-2010 19:46
Add to lightbox
page 342
page 343
Figure 12-11 A local input function derived from a T2*-weighted time series in a patient with acute right MCA
stroke. The derived local arterial concentration at 2-s intervals is shown as a function time in one axial slice. The
arrival of the contrast in the affected territory is delayed and broadened relative to the contralateral side. Use of the
local input function may help eliminate errors in the determination of blood flow and mean transit time in such
studies.
Many of the problems with measuring arterial concentration stem from using the same images for tissue and
arterial measurement. If additional images are obtained, a number of significant improvements are possible.51 One
of the simplest schemes is a dual echo experiment, which obtains each of the slices twice after excitation. The
shorter TE image is less attenuated by contrast agent concentration, so it can be more sensitive for measuring
arterial contrast agent concentration, while the longer TE image is ideal for tissue concentration measurement.
This dual echo approach has the added advantage that the effects of T1 can be separated from the effects of
52-54
T2*. The ratio of the signal in the short and long TE images depends only upon T2*. Another scheme is to
measure the frequency change which occurs in a vessel containing contrast agent.45 The frequency change is very
linear with contrast concentration so many of the problems of T2 and T2* intensity models are avoided. However,
frequency is usually measured by observing the phase of the voxel on a longer TE image. Flowing blood can also
have phase shifts from motion in the presence of gradients and these phase shifts could cause excess noise or
systematic errors. Phase measurements can also be complicated by the presence of some tissue mixed with the
44
vessel in the voxel. More complicated modeling is required to extract an input function in this case.
Quantifying Perfusion Using the Measured Arterial Bolus

The purpose of measuring the arterial concentration curve is to remove its effects from the tissue concentration
curve. Ideally one would like to produce corrected tissue concentration curves that approximate how they would
have appeared if the bolus were much less than 1 s wide. With such curves, the simple approach to quantifying
hemodynamic parameters described earlier could be used to calculate the blood flow, blood volume, and MTT.
The mathematical process for removing the effect of the arterial bolus to produce corrected tissue concentration
curves is known as deconvolution. There are a number of numeric methods for deconvolution which have their
disadvantages and advantages but they all share one basic property. That is, no method can perfectly deblur the
tissue concentration curve and the more it attempts to be perfect, the more sensitive it is to the presence of noise.
Because all deconvolution methods share this imperfection, they usually provide some method to select the
trade-off between better removal of blurring and reduced noise. In Fourier deconvolution,26,33 which uses the same
fast Fourier transform (FFT) employed in most MR image reconstruction, a filter must be introduced to avoid
ridiculously high levels of noise. The more filtering is applied, the lower the noise level but more blurring will be left
in the result. Another popular deconvolution method,33 singular value decomposition (SVD), also uses a threshold
to eliminate excessive sensitivity to noise. Though the mathematics of SVD are less familiar to the MRI community
11 of 13 14-04-2010 19:46
than FFT, the effect of increasing the threshold is very similar to increasing the filtering used for Fourier
55,56
deconvolution.
One problem with both these methods is that the deconvolved tissue curves are prone to ringing artifacts related to
the Gibbs artifact in MRI. Recently modified deconvolution methods have been proposed that reduce the ringing
and may produce more accurate deconvolved tissue curves. 57,58 Both SVD and Fourier deconvolution are relatively
fast algorithms, which will not cause excessive processing times.
An alternative to deconvolution is to assume a model for the tissue clearance curve with just a few free parameters
33
to determine its width and amplitude. A number of model clearance curves with different parameters are then
mathematically smoothed by the arterial input function and compared to the actual tissue concentration curve. This
approach can be highly advantageous if the model is a good indicator of the tissue response but can perform very
poorly if not. The specification of a model makes it easier to avoid systematic errors, which can occur in the
deconvolution procedure, such as overestimation of short mean transit times, because of the filtering applied.
Delay, Dispersion, and Mean Transit Time

The Achilles heel of the DSC technique is its inability to separate effects of flow between the region of interest
(ROI) used for arterial concentration measurement to the tissue and the effects of flow through the tissue. In fact,
all the preceding discussion has assumed that blood and contrast agent move nearly instantaneously from the
arterial sampling location to the tissue. If there is a delay in the arrival of the contrast agent, then the deconvolved
tissue concentration curve will simply be shifted in time. Theoretically, the signal intensity at the first timepoint of
the concentration time curve was to be used for flow measurement. Because of the possibility of delay, the
47
maximum of the signal intensity curve must be used instead. Recently the vulnerability of a popular
34 59,60
implementation of the SVD method to delays has been emphasized but solutions have been proposed.
Delayed arrival is thus less of a problem for perfusion measurement than additional blurring of the bolus during
transit to the tissue. Such blurring, sometimes called dispersion, is an almost unavoidable companion of delay
since all the blood in a vessel does not move at the same speed. In addition, if the delay is a result of collateral
flow, there may be multiple paths for the supply of blood with different delays for each. Improved understanding of
dispersion in arterial networks might make possible a first-order correction for dispersion based on the arrival
delay and properties of the tissue clearance curve. The local input function approach described above is another
potential solution, though its robustness and accuracy must still be tested. Despite these promising new
approaches, in current practice, perfusion measurements in regions with delayed and dispersed arrival will tend to
underestimate true perfusion by an uncertain factor. Because delayed arrival will increase time to peak, TTP
images will delineate regions of delayed arrival even if perfusion and MTT in that region are otherwise normal.
Large Vessel Effects on Hemodynamic Maps

page 343
page 344
The theory for the calculation of blood volume, blood flow, and mean transit time from the tissue concentration
curves is flawed in one respect. There is no distinction made between blood flowing through the capillary networks
and blood flowing through the voxel in larger arteries or veins. Thus regions with larger vessels have large blood
volumes, short mean transit times, and very high blood flow values when calculated from this theory. This problem
would be even worse were it not for the nonlinear effects of contrast agents in the vessels described above. Since
spin-echo imaging emphasizes smaller vessels, the effect of large vessels on the hemodynamic maps is reduced,
though not absent. The filtering process used in deconvolution can also attenuate the artificially high blood flow in
regions with large arteries. Nevertheless, large vessel contribution to blood flow maps is a major systematic error
if measurement of tissue perfusion is desired.
Partially Permeable Contrast Agents

The discussion so far has assumed that the contrast agent is purely intravascular. While this is the case for most
cerebral tissues, permeability can be quite important in regions where the blood-brain barrier has been
compromised or in tumors with newly generated, leaky vasculature. When contrast agent leaks into the tissue, it
generally has relatively small effects on T2 or T2* but it can have quite dramatic effects on T1. If the permeability
is so high that most of the contrast agent enters the tissue during the first passage of the bolus, then the T2 and
T2* effects of the contrast will be greatly reduced because the contrast agent is much more uniformly distributed
throughout the tissue. A further complicating factor is the shortening of T1 that can occur with contrast leakage.
Since the T2*-sensitive images are usually somewhat T1 weighted as well, the signal intensity increases with
contrast in the tissue but decreases with contrast in the vessels.
If permeability effects are considered undesirable, then there are several strategies that can be employed. One is
12 of 13 14-04-2010 19:46
61
to use a contrast agent, such as Dy-DTPA, which has very little effect on T1. Attenuation of T2* or T2 is not
avoided but the complicating factor of T1 enhancement is. A related strategy is to give a smaller dose of contrast
agent prior to the study. This dose will often shorten the tissue T1 sufficiently that further T1 changes have no
effect on the weakly T1-weighted images. The recent development of a number of larger contrast agents designed
to stay within the vasculature for an extended period offers a simpler option: choose an agent for which
permeability is small enough to be unimportant during the first pass.
Measurement of Perfusion Using T1 Contrast

In principle, T1-weighted images can be used to measure blood flow and related parameters. While blood volume
and other hemodynamic parameters in normal cerebral tissues have been reported using T1-weighted imaging,
generally the T1 signal change is considered far too weak to be reliably used. Instead, T1-weighted imaging is
reserved for imaging of partially permeable tissues. If permeability is high enough relative to flow, one can treat the
agent as fully permeable and use the mathematical model described earlier. In this case, the accessible volume
fraction becomes the extracellular volume fraction, not the vascular volume fraction, but the analysis is the same. In
the opposite limit, where the permeability limits the leakage of contrast agent, the accessible volume fraction is the
same but the peak of the contrast clearance curve is not an indication of flow but of permeability surface
62
product. In practice, it is not known which of the two processes, perfusion or permeability, determines the tissue
T1 contrast kinetics so uncertainty usually remains.
The use of T1 and T2 or T2* measurement may make simultaneous measurement of perfusion and tissue
permeability possible with a single bolus. While this technique is still an area of investigation, the basic approach is
as follows. A series of dual echo images with short TR are acquired during the passage of a bolus through the
brain. T2 or T2* is measured from the ratio of the images acquired at the same time but different TE. The T2*
effects can then be removed from the shorter TE image leaving just T1 changes. Since contrast agent
predominantly affects T2* when it is in the vessels and contrast agent predominantly affects T1 when it is in the
tissue, a two-compartment model including tissue and blood concentration and permeability to the contrast agent
can be used to analyze the tissue signal curves. While the algorithms used to extract perfusion and contrast agent
permeability from the data may be more sensitive to noise and the assumptions of the behavior of contrast agent
in small vessels, this approach can in principle quantify both quantities.52,54 This technique is of great interest for
imaging of brain tumors since the blood-brain barrier is usually compromised.
© 2010 Elsevier
13 of 13 14-04-2010 19:46
ARTERIAL SPIN LABELING

Because DSC imaging produces a large signal change and DSC data can be readily acquired on a scanner
equipped with fast imaging, it has been extensively applied to clinical research. While these studies have shown
promise, they have also emphasized current limitations of the technique including uncertainty in input function
measurement, difficulties in deriving absolute measurements, and potential insensitivity to very short mean transit
times and high flows. The need to perform a timed injection also adds to the complexity, cost, and reliability of the
DSC approach.
18,19
Arterial spin labeling (ASL) is an alternative approach to measuring blood flow with MRI that does not require
the injection of a contrast agent. ASL can be thought of as a natural extension of magnetic resonance angiography
(MRA). In MRA, spatially selective excitation pulses or flow encoding gradient pulses are used to create a
difference between the signal in flowing blood and the signal in the surrounding tissue. Imaging is performed during
or very quickly after the selective pulses are applied so the flowing blood signal is still within the vessels. ASL
methods are quite similar except that after the pulses which differentiate between static tissue and flowing blood,
time is allowed for the blood to move out of the vessels and into the perfused tissue. The signal change in tissue
caused by the selective labeling of the arterial blood is closely related to the blood flow into that tissue.
page 344
page 345
Arterial spin labeling is also closely related to blood flow imaging with H215O PET.37 Both techniques use labeling of
water to measure blood flow and thus some prior experience with PET can be used in the understanding of the
ASL MRI methods. There are several major differences between the two methods. First, the labeling of water in
ASL is achieved noninvasively and precisely in the feeding arteries. Second, the decay rate of the labeled blood
15
signal with ASL is of the order of 1 s rather than minutes as for the H2 O. Finally, with ASL the background static
tissue also contributes a large signal in MRI that is not present in the PET scan.
Add to lightbox
Figure 12-12 An example ASL study. An image is acquired from the brain approximately 1 s after a slab inversion
is applied either superior to the slice, the control image, or inferior to the slice, the labeled image. The inferior
inversion labels blood in the feeding arteries which flows into the tissue. Subtraction of the labeled image from the
control produces a perfusion-sensitive image.
While the various approaches to ASL will be discussed in more detail below, one of the simplest approaches is
63
presented in Figure 12-12. This technique, first proposed by Edelman et al, uses a slab selective inversion prior
to imaging to label the blood in the arteries within that slab. Time is allowed after the inversion for the labeled
1 of 13 14-04-2010 19:56
arterial blood to enter the imaged slice. Since the signal from inverted magnetization is negative, the inflowing
labeled blood decreases the signal in the slice. Because the decrease is very small, the image must be subtracted
from another image without the inversion pulse. The difference between the control image and the labeled image
reflects blood flow into the slice. This approach to ASL exhibits most of the features of all ASL experiments: the
selective labeling, the wait for arrival of labeled blood in the tissue, and the subtraction from a control image. This
subtracted image is already proportional to blood flow and only a calibration factor dependent on the T1 of the
tissue and the properties of the bolus of labeled water is required in order to create a quantitative blood flow map.
Labeling Strategies for Arterial Spin Labeling

There are quite a few different variants of ASL that have been proposed or used for in vivo studies. Most of these
are very similar and differ only in subtle aspects of implementation but there are a few broad categories of
approach. The ASL approach described in Figure 12-12 is an example of spatially selective ASL. Spatially
selective ASL is similar to time-of-flight angiography in that a spatial separation between the imaged region and
inflowing arterial blood is required to cause signal change in the image. The vast majority of ASL studies have been
performed using spatially selective ASL and it is by far the most well-developed technique. It is possible, however,
to label flowing blood using bipolar gradients as in phase contrast angiography. Velocity-selective ASL64,65 can be
used to label blood within the slice or slab under study and thus may permit labeling even closer to the capillary
bed than spatially selective ASL. Because this approach is still in the early stages of development, it will not be
further described here.
page 345
page 346
The vast majority of ASL approaches also rely upon labeling the magnetization along the main magnetic field
direction or Mz. The advantage of labeling Mz is that changes in Mz return back to the equilibrium on the time scale
of T1. Labeling the transverse magnetization, Mx or My, has been proposed66 but since the transverse
magnetization decays with T2, typically 5-10 times faster than T1 decay, the signal produced would be much
smaller. While the spin-locking approach can be used to effectively lengthen T2, it typically requires power too high
for use in humans.
Among spatially selective ASL approaches, the clearest distinction is between continuous ASL and pulsed ASL. In
continuous ASL,18 spins are continuously inverted at a specific location before they enter the tissue. Thus spins
which enter the tissue first are labeled first and spins which enter later are labeled later. In contrast, the pulsed
67,68
labeling techniques apply a single perturbing pulse to the spins outside the slice and then wait as the spins
enter the tissue. The approach of Figure 12-12 is an example of a pulsed ASL approach because the
slab-selective inversion is performed at a single timepoint. Because spins which enter the tissue later are perturbed
earlier than in the continuous technique, there is additional T1 decay of the label and generally the signal change
69,70
with pulsed techniques is smaller than with continuous techniques. Nevertheless, pulsed techniques employ
more traditional RF and gradient strategies, are less subject to certain types of systematic error, and can produce
excellent images. For simplicity the pulsed techniques will be discussed first.
In pulsed techniques, an RF pulse is applied at a time approximately 1 s before imaging. Almost always an
inversion RF pulse is used because it creates the largest change in the arterial magnetization and, by analogy to
inversion recovery, the time before imaging is referred to as TI. Two images are obtained, one with the region
containing the inflowing artery inverted and one where it is not inverted. The primary differences between all the
pulsed labeling strategies (Fig. 12-13) is what inversions are applied to other tissue, including the imaged region,
and what is done to insure that the differences between the two RF preparations do not cause direct effects on the
71
image which are independent of perfusion. Such direct effects can result from imperfect inversion slice profiles
63,72
and differences in the magnetization transfer effects of the two preparation sequences.
2 of 13 14-04-2010 19:56
Add to lightbox
page 346
page 347
Figure 12-13 Comparison of the two most popular pulsed labeled strategies. The EPISTAR technique (top)
employs slab-selective inversions to label inferior blood while leaving the imaged region unaffected. The FAIR
technique (bottom) uses an inversion slab centered on the slice of interest so the imaged slice is always affected.
For labeling, the inversion slab is made very wide, typically non-selective, to label inferior arterial blood.
Arguably the simplest pulsed inversion strategy is the application of a slab-selective inversion pulse inferior to the
imaged brain region to label arterial spins and the application of a superior inversion slab during the control image
preparation. The superior inversion insures equal magnetization transfer effects from the two preparation pulses.
67
This technique was first proposed by Edelman et al and is widely known as the EPISTAR technique. The most
clear-cut alternative strategy employs a nonselective inversion pulse for the labeling and a slab selective inversion
containing the entire imaged region for a control. Though the effect on inflowing spins inferior to the slice is the
same as in EPISTAR, the spins in the imaged slab are inverted in both cases, rather than left unaffected. This
68 73
inversion technique was first proposed by Kwong et al and was termed FAIR by Kim. Because any slight
difference in the flip angle of the two pulses will have a big effect upon the measured signal, special RF pulses with
low sensitivity to variations in RF amplitude and very sharp slice profiles must be employed.74 A number of
variations of these basic labeling techniques have been proposed, many of which may provide small improvements
to the results but variations in implementation and methods of assessment overshadow these small differences.
3 of 13 14-04-2010 19:56
Add to lightbox
Figure 12-14 The principle of flow-driven adiabatic inversion. In the rotating frame of the applied RF, the RF
magnetic field, B1, and the difference in the field along the main magnetic field direction induced by gradients,
∆B0, combine to create an "effective field" around which the blood magnetization, M, precesses or rotates. As the
blood flows along the gradient, ∆B0 switches from positive to negative. The effective field rotates from positive to
negative and the blood magnetization follows this rotation. The net effect is a continuous inversion of blood as it
flows past a particular plane.
75
Continuous ASL usually employs a special RF labeling scheme known as flow-driven adiabatic inversion. One
can approximate continuous saturation19 of spins as they flow past a certain plane by applying a series of thin
saturation pulses repeatedly, but trying the same strategy with inversion pulses is problematic; some spins may
remain within the saturation slab for two or more inversions and get doubly inverted. Because inversion provides a
stronger signal change in ASL, it is considered highly desirable to use a continuous inversion technique. Flow-driven
adiabatic inversion is a simple technique because it requires only turning on a constant gradient and RF; however,
the concept of flow-driven adiabatic inversion is foreign to most MRI practitioners who typically use only pulsed RF.
The mechanism and explanation for flow-driven adiabatic inversion are described in Figure 12-14. For the purposes
of perfusion imaging, however, it is only necessary to understand that the technique defines a plane such that all
spins flowing through the plane get inverted at the time they flow through (Fig. 12-15).
While the pulsed ASL experiment is essentially a decaying bolus experiment where the labeled blood transiently
enters the tissue, the continuous ASL experiment can be operated as a steady-state experiment.18 If the labeling is
left on for a long time, a balance between the inflow of newly labeled spins and the T1 decay of the labeled spins
already in the tissue is reached. This makes quantification of perfusion relatively simple. Of course, the adiabatic
labeling can be turned on for shorter periods and still achieve the inversion.
4 of 13 14-04-2010 19:56
Add to lightbox
page 347
page 348
Figure 12-15 The geometry of a continuous ASL study. A plane is selected to continuously invert blood
magnetization as it flows across the plane. The labeled blood then flows on into the distal vessels and the tissue.
The greatest attraction of continuous ASL is that it can achieve a signal approximately 2-3 times larger than with
pulsed ASL.69 To achieve this large a signal change, however, the labeling must be left on for many seconds. The
decreased number of averages achievable per unit time with this large signal change partially offsets the signal
advantage.76 Still, continuous ASL is generally more sensitive than pulsed techniques and has some advantages in
the details of implementation. The greatest disadvantage of the continuous labeling approach is the unusual RF
requirements of the labeling. First, many scanners are designed specifically for brief, widely spaced pulses of RF
at high power. Flow-driven adiabatic inversion requires weaker RF for an extended period of time so
implementation on some scanners may be difficult. The long duration of the labeling can also lead to high RF power
deposition in the subject that can approach the FDA limits. Nevertheless, continuous labeling can be performed on
many mainstream clinical scanners at power levels well within government safety limits. 70
Calculating Blood Flow from Arterial Spin Labeling Measurements

Unlike DSC, where a series of images must be mathematically analyzed to separate blood flow, blood volume, and
arterial transit effects, the simple subtraction image obtained from ASL can be a good reflection of relative blood
flow without any calculations. However, because differences in tissue T1 and other factors can lead to different
sensitivity to blood flow across the image and quantitative measurements of blood flow can be useful for some
purposes, it can be useful to convert the difference images into actual blood flow maps in physiologic units. This
conversion requires some consideration of how inflow impacts tissue MR signal.
The Modified Bloch Equations

In the absence of flow and RF pulses, the Z direction magnetization behaves according to the Bloch equation:
The derivative on the left of the equation just means the rate of change of Mz with time. In words, this equation
0
means that any difference of Mz from Mz tends to decay away with the time constant T1. When blood flow is
present there is another way for the magnetization to increase or decrease. Spins come in from the arteries and
add whatever their magnetization is to the voxel and spins flow out through the veins and take away their
magnetization. The flow modified Bloch equation18 is:
Here f is the blood flow, or perfusion, ρb is the density of brain tissue, Ma is the magnetization in the inflowing
arteries, and Mv is the magnetization in the outflowing veins. Unlike magnetic contrast agents, which are large
5 of 13 14-04-2010 19:56
molecules, water penetrates the blood-brain barrier rather easily.77 Though some water may not pass into the
tissue at high flow rates, at first approximation one can assume complete free diffusion into the tissue. That means
at the capillary level, Mv equals Mz. Mv cannot be changed separately from Mz.
The ASL signal represents a competition between T1 decay and inflow of labeled magnetization. f, the flow or
perfusion, is the rate of inflow and 1/T1 is the rate of decay, so the fractional change in the image intensity due to
ASL tends to be approximately F × T1. For a T1 of 1 s and a flow of 60 ml/100 g/min, equal to 1 ml/100 g/s =
0.01/s, the ratio is 1%. Typically ASL signal changes in human gray matter are of this order and white matter
changes are considerably smaller. Compared to DSC imaging, the signal change in ASL is quite small and this
small signal discouraged the wider use of the technique. However, with modern scanners equipped with fast
imaging hardware to minimize motion-related noise, images of the perfusion-related signal can readily be
measured, albeit at lower spatial resolution than typically used for clinical MRI.
ASL perfusion measurement differs from DSC imaging in several important ways. First, the tracer, water spins,
77
can readily diffuse into tissue. This makes it impossible to measure venous blood volume with ASL but it also
makes ASL relatively insensitive to the vessel permeability issues that affect DSC perfusion imaging when the
blood-brain barrier is damaged. Another important difference is that the tracer is rapidly decaying. Finally, the
tracer is "injected" into the arteries right near the tissue, so deconvolution of an intravenous bolus is unnecessary.
Typically the arterial input function is assumed based upon the labeling applied, whether pulsed or continuous.
Because the tracer is rapidly decaying, it is important to know the decay rate, T1. T1 of blood can be taken from
78,79
literature values or an attempt can be made to measure it in vivo. Likewise the tissue T1 can be measured,
since it may be different from the T1 of blood. Including the effects of the different T1s complicates the
discussion70 so for now, assume that the T1 of blood and tissue are the same.
page 348
page 349
For pulsed experiments, the effect of the inversion is to create a bolus of labeled spins that is of duration tb. If we
wait long enough for the bolus to enter the tissue, a time TI, before imaging, then the labeled signal will be just the
perfusion times the duration of the bolus times a decay factor for T1 decay over TI.
This equation can be inverted to give f, the perfusion.
If we hadn't waited long enough, then some of the labeled blood destined for the tissue would still be in the feeding
vessels and the perfusion would be underestimated. One of the challenges of pulsed ASL quantification is that t b is
not known because the inversion pulse inverts a thickness of tissue containing arteries, not a duration. A solution to
this problem was proposed by Wong and collaborators.80 They applied a saturation pulse at a certain time after
the inversion to eliminate any labeled blood still remaining in the inverted slab. This makes the tb always be equal to
TI-Tsat.
For continuous ASL experiments the perfusion quantification is slightly different. Since the spins are not all labeled
at the same time, we have to average the T1 decay of the duration of the bolus. For comparison with the pulsed
case, we will call TI the time the labeling is started and tbolus the duration of the bolus; then the solution for f is:
Often the term TI is not used for continuous experiments. Instead w, a post-labeling delay, is defined such that
70
TI=tb+w. As with the pulsed experiment, flow will be underestimated in tissue and vessels will be very
prominently labeled unless one waits long enough for all the bolus to enter the tissue (Fig. 12-16).
If the T1 of tissue and blood are significantly different, then there will be added uncertainty in the quantification of
perfusion. Generally, if one uses the T1 of blood in the above equations and T1 of tissue is shorter, the perfusion
will be underestimated, and if the T1 of tissue is longer, then perfusion will be overestimated. 70 Because there is no
independent measurement of when the blood reaches the tissue, quantification of flow is challenging with greatly
different T1s. Since T1 sensitivity is greater with longer bolus duration, especially with the continuous ASL
technique, shorter boluses should be used in this situation. In clinical applications, the uncertainty in T1 effects
might occur, for example, if high flow was measured in an edematous region where T1 is much longer than blood.
If the label enters the tissue quickly, the measured flow will be very different than if it does not. This might occur,
6 of 13 14-04-2010 19:56
for example, in reperfused ischemic lesions81 or in neoplasms.
With MRI we do not directly measure M, the magnetic moment of the tissue. Instead we measure a signal intensity
in arbitrary units. In order to use the above equations for perfusion quantification, it is necessary to define magnetic
18,69
moments in terms of a reference magnetic moment. Typically one of two reference quantities is used. One
approach is to use the magnetization within each voxel as a reference. Since the signal within a proton density-
weighted image is proportional to M, one simply divides the signal in the ASL difference image by the signal in the
proton density-weighted image:
where now the ASL signal, in scanner units, and the proton density signal in the voxel are used. λ, the blood-brain
partition coefficient, is usually taken from the literature for gray matter and white matter.82 The alternative is to use
a reference intensity somewhere in the image. One choice is to use the intensity of CSF within a ventricle, which
83
has almost the same magnetic moment as pure water, as a reference. This approach requires a uniform
sensitivity profile as is obtained with a good head coil. Because the T1 and T2 of CSF are so different from tissue
and blood, use of shorter TR or longer TE must be explicitly corrected for in the perfusion quantification.
In the above equations, we have assumed perfect inversion of the labeled blood but in practice the inversion does
84
not have to be perfect, especially for the flow-driven adiabatic labeling technique. Typically a correction factor for
imperfect efficiency, α, is included in the quantification equations. The efficiency of a particular labeling approach
can be measured in vivo or estimated with simulations.
Multislice Arterial Spin Labeling Imaging

In principle, acquiring more than one slice with ASL is straightforward but in practice there are a number of
complications. Since ASL benefits from averaging, acquiring one slice at a time would make the imaging of the
whole brain an unacceptably long process. Just as with traditional imaging, interleaving of slice acquisition is the
obvious solution. This approach is challenging for several reasons. First, the ASL label decays with T1. Acquiring
images over a period of time comparable to T1 will cause loss of signal, and signal-to-noise ratio, in the later
images, relative to single-slice acquisition. Second, imaging of one slice may saturate labeled blood in an artery
within that slice that is destined to enter tissue in another slice. Third, it may take longer for blood to enter a thick
slab than to enter a thin one. This longer transit time will necessitate longer waits before imaging and consequently
some signal decay due to T1. Finally, it is harder to control for systematic differences in the direct effects of the
labeling RF on the tissue signal across a thick slab.
The first problem requires that the different slices be acquired as rapidly as possible so that T1 decay of the label
is not a concern. A reasonable number of slices can be acquired with very fast imaging systems, but the 20 or
more slices required for whole brain coverage probably cannot be satisfactorily achieved this way. Imaging time
per preparation sequence can be reduced by using multishot versions of the single-shot sequences currently being
85,86
used for ASL. The use of new parallel imaging strategies should also make possible an increase in slice
coverage, though with a decrease in the signal-to-noise ratio for each slice. Alternatively, 3D acquisition can be
used so that all the slices are acquired at exactly the same time after labeling. Multishot and 3D scans could
increase the noise in a perfusion study because motion-related phase errors can cause ghosts and other artifacts.
The use of background suppression87 can help a great deal with this approach.
page 349
page 350
7 of 13 14-04-2010 19:56
Add to lightbox
Figure 12-16 Examples of continuous ASL images acquired with different delays between labeling and imaging.
The top images have a 100-ms delay and bright vascular signal is apparent. As the delay is increased to 600 ms
(middle row) or 1200 ms (bottom row), the vascular intensity fades and the images become more tissue specific.
Saturation of labeled blood destined for other slices is not an issue as long as image acquisition does not begin
until most of the labeled blood has entered the tissue. Since this timing is required to insure accurate quantification
70
anyway, saturation should not occur. If for some reason a pulse sequence with a very short wait period is used,
then this effect can become very significant.
The longer transit into a thick slab of tissue is an unavoidable result of imaging a thicker region of tissue. If the flow
into the thick slab is so slow that a very long wait has to be used, then the signal-to-noise ratio may suffer
dramatically. In such a case, a series of sequential single-slice scans may give comparable or superior results.
Usually, however, a large volume of tissue is perfused by a relatively large and fast feeding artery, so flow is not
markedly delayed in transit. If there is a stenotic vessel, then delay in the large vessel could be more of a serious
83
issue.
Finally direct labeling effects on the measured signal must be eliminated for accurate results. A number of
strategies to control for direct effects of RF only work perfectly when there is one thin slice. 18 Generally RF can
71 72
affect tissue directly in two ways, either through imperfect slice profiles or magnetization transfer.
All the discussion above has assumed that one could perfectly invert or saturate a square slab of tissue with a very
fast boundary. In practice, there is a transition region between the inversion and no inversion areas. Depending on
the RF pulse used, the transition may be very broad. For a slice near the end of an imaged slab, the transition
region will pass directly through the slice and systematic errors can occur. To overcome these errors, very
carefully engineered inversion pulses have been used. Two adiabatic pulses, the hyperbolic secant pulse and, more
8 of 13 14-04-2010 19:56
recently, the FOCI pulse,74 have become popular for multislice pulsed ASL experiments because of their sharp
slice profiles as well as insensitivity to B1 variations. The FOCI pulse in particular can be made to have very sharp
boundaries and has helped a great deal with pulsed labeling studies.
page 350
page 351
88
Magnetization transfer (MT) is a physical process that can cause attenuation of tissue magnetization by RF at a
considerable offset in frequency from the slice. Even if the slice profile is perfect, the magnetization transfer effects
of two different inversion pulses need not be perfectly matched. Because there is no magnetization transfer in pure
liquids such as water, the systematic error from unmatched magnetization transfer for the labeled and control
image will not be apparent in images of phantoms filled with such liquids. Magnetization transfer is an especially
important issue for adiabatic labeling since the RF is on for a long time. RF must be applied during the preparation
period of the control image in order to avoid a large systematic error. Single-slice studies simply moved the
labeling plane above the slice instead of below the slice. 18 This serves only as a good control for one slice. A
second, small RF coil can be used for the labeling that produces only a weak RF field in the imaged region and
89,90
thus minimizes magnetization transfer effects but this does require a favorable labeling geometry and special
91
hardware. We have shown that sinusoidal modulation of the amplitude of the RF pulse during the control period
can mimic the magnetization transfer effects of the labeling while allowing most of the blood spins to pass
uninverted. Sinusoidal modulation can be thought of as splitting the inversion plane into two separate inversion
planes that doubly invert the inflowing blood. The efficiency of the scheme is not perfect, however, and an
improved control for multislice continuous ASL would be desirable. Talagala et al92 have proposed a strategy that
18
controls for magnetization transfer near the center slice better than the original superior slice strategy, making
possible the imaging of a thin slab near the center slice. MT can also be a significant but smaller problem for
pulsed labeling, especially when adiabatic RF pulses are used. Several groups have suggested strategies involving
double pulses with altered properties that can help to eliminate MT systematic errors. 63,72
Motion Artifacts and Arterial Spin Labeling

While the small signal level of ASL reduces the signal-to-noise ratio of ASL perfusion images, surprisingly good
images can be obtained if motion artifacts can be reduced and eliminated. Though snapshot imaging is typically
used for ASL studies, the subtraction of images acquired at different times still can be sensitive to motion.
Experience with patient studies has shown, however, that good image quality can be obtained with simple
strategies to reduce motion and recent developments suggest even greater robustness to motion is possible.
Typically imaging of perfusion with ASL requires multiple averages to achieve an acceptable signal-to-noise ratio. If
the label and control images are interleaved, i.e., first a control, then a label, then a control, then a label, etc., the
time difference between the two sets of images will be only one TR, even if the averaging continues for many
minutes. This rapid alternation between label and control tends to give improved robustness to motion and other
sources of noise in the scanner, such as drifts in sensitivity. In normal subjects and reasonably co-operative
93
patients, this subtraction strategy is sufficient to achieve good-quality images. Robustness to motion can be
further improved by using post-processing methods to reduce motion. Using one particular strategy,94 we have
been able to drastically reduce residual motion in ASL perfusion images of acute stroke patients and demented
patients, even when speaking during the scan, and more advanced methods such as independent component
analysis95 show even greater promise.
Despite the success of the above strategies, the large background signal and the threat of motion artifacts limit the
scanning strategies and pulse sequences that might be used. An exciting alternative to imaging of the entire
87,96
background signal is to suppress the background signal with slice-selective saturation and inversion pulses. An
example of such a strategy for a pulsed ASL experiment is shown in Figure 12-17. A saturation pulse is applied in
the imaged region followed by a selective or nonselective inversion pulse. Then a series of nonselective pulses is
applied during the TI period of the study. If the timing of the inversion pulses is optimized, reduction of the
background below the gray matter perfusion signal level can be achieved (Fig. 12-18). Examples of 3D whole-brain
perfusion images acquired with background suppressed continuous ASL are shown in Figure 12-19.
9 of 13 14-04-2010 19:56
Add to lightbox
Figure 12-17 A background-suppressed FAIR pulse sequence design. The selective pulse makes the proximal
blood magnetization different from the static tissue magnetization. The nonselective inversion pulses are optimized
to null the static tissue signal but the magnitude of the labeled blood signal is not affected.
Add to lightbox
page 351
page 352
Figure 12-18 The effect of background suppression on the label and control images in an ASL experiment. The
static tissue contribution to the control image (left) and labeled image (middle) is so well suppressed that the less
than 1% signal from the inflowing blood causes a large change in the image appearance. Subtraction of the two
images (right) produces a flow-sensitive image.
10 of 13 14-04-2010 19:56
Add to lightbox
Figure 12-19 Continuous ASL images obtained in a normal volunteer using a 3D background-suppressed, whole-
brain, fast spin-echo sequence. The background suppression makes the use of the higher image quality multishot
sequence practical by reducing the effects of motion.
Applications of Arterial Spin Labeling

page 352
page 353
Blood flow measurement can be useful for a wide range of clinical and research studies. Years of experience using
PET and SPECT methods for blood flow measurement can be readily transferred to the MRI platform. One major
benefit of MRI blood flow measurement is that the functional information can be acquired in spatial registration with
excellent anatomic information obtained without moving the subject. Functional studies are often challenging to
interpret without the anatomic information. With ASL, the blood flow measurement can be obtained as simply
another imaging sequence without the added cost and complexity of a contrast bolus. The absence of a bolus also
means that the measurement can be repeated as frequently as desired. The absolute quantification that ASL
provides also means that values obtained in different scanning sessions or different physiologic states can be
directly compared.
Probably the widest use of ASL has been for the study of brain activation. Increased brain activity causes a
11 of 13 14-04-2010 19:56
localized increase of blood flow, which can be used to map the regions involved in specific tasks.67,73,97 Most MRI
brain activation studies are performed using changes in T2* caused by blood oxygenation changes but ASL has
been receiving increasing attention as a tool for activation imaging. While the within-subject sensitivity of ASL is
lower than oxygenation-sensitive MRI, ASL has better temporal stability and can be used to study brain activity
98,99
changes over long times. ASL signal changes also appear to be more reproducible across subjects, which is
important for comparison of groups. Oxygenation-sensitive studies suffer from signal loss in the base of the frontal
and temporal lobes, which does not occur in ASL studies acquired with the appropriate imaging sequences. ASL
also provides a baseline measure of blood flow so the fractional change in flow can be assessed. Finally, ASL
blood flow measures can be combined with oxygenation-sensitive MRI to measure changes in oxidative metabolism
100
accompanying neural activity.
Ischemia and vascular disease are natural applications for blood flow imaging techniques and ASL has been
applied to acute stroke83,101 and cerebrovascular disease.102 Stroke is a challenging application for ASL both
because the flow can be very low and because the time required for labeled blood to reach the tissue can be very
long. Examples of ASL images from two patients with acute stroke are shown in Figure 12-20. Bright signal in
collateral vessels is apparent in the right MCA territory of one of the cases, indicating delayed arrival of the labeled
blood. Studies performed so far have indicated a great sensitivity of ASL for vascular abnormality but the
specificity for predicting tissue damage is likely poor. Use of much longer wait times for the label to reach the
tissue or alternative labeling strategies, such as velocity-selective ASL, may be needed for this application.
Add to lightbox
Figure 12-20 Echo-planar ASL images acquired in two patients with acute MCA stroke. Low tissue signal is
apparent in the affected territory but in the second patient (bottom images) bright signal is apparent in feeding
vessels. This suggests delayed collateral flow which is underestimated by the ASL study.
The absolute quantification capabilities of ASL have been exploited for several candidate applications in
cerebrovascular disease. Scanning before and after flow augmentation with acetazolamide has been used to
assess cerebrovascular reserve in patients with cerebrovascular stenosis or occlusion. 103 ASL can also be easily
104 105
used to assess the impact of therapies such as carotid endarterectomy and coronary bypass on cerebral
blood flow.
ASL may be of use in the characterization of epilepsy. An initial comparison between ASL and PET glucose
utilization imaging for the lateralization of temporal lobe epilepsy has been reported. Hypoperfusion on ASL MRI
and hypometabolism on PET were in agreement in all but one subject.106 ASL can also be used for ictal imaging
but the utility of this application is limited by the need to scan during the seizure. Epilepsy is an excellent example of
where the combination of anatomic MRI for the assessment of hippocampal sclerosis and blood flow MRI for the
assessment of function is a compelling diagnostic team.
ASL can also complement anatomic studies in the evaluation of dementia and other neurodegenerative diseases.
Initial evaluations of ASL for the assessment of Alzheimer's disease (AD) have been reported that indicate good
sensitivity for detecting hypoperfusion in the parietal and temporal lobes consistent with other methods.107 Early
detection of AD is an important and challenging application and the ability to compare tissue loss and functional
changes in a single scanning session could be an essential advantage of MRI. An example of ASL perfusion
imaging in a patient with mild Alzheimer's is compared to an age-matched control in Figure 12-21. The areas of
most statistically significant perfusion decrease in a group study are shown in Figure 12-22.
Blood flow imaging in tumors has received increased emphasis because anti-angiogenic agents targeting the tumor
12 of 13 14-04-2010 19:56
blood supply have shown exciting promise.108 ASL offers the ability to measure blood flow and compare values
during the course of therapy. A preliminary analysis of ASL blood flow data in renal cell carcinoma metastases
undergoing anti-angiogenic therapy found a strong correlation between blood flow decrease 1 month after therapy
109 110-113
and time to progression. Application of ASL to brain tumors is now being explored. Substantial differences
between blood flow and enhancement patterns have been observed (Fig. 12-23) and the use for characterization
of response and recurrence following therapy is promising.
© 2010 Elsevier
13 of 13 14-04-2010 19:56
CONCLUSION
Both dynamic contrast and arterial spin labeling techniques offer the ability to complement the excellent anatomic
information provided by MRI with tissue-specific functional information. As the acquisition and analysis methods for
these techniques become more seamlessly integrated into commercial scanners, the additional diagnostic
strategies they make possible will become increasingly apparent.
page 353
page 354
Add to lightbox
page 354
page 355
Figure 12-21 Comparison of ASL images acquired in a patient diagnosed with mild Alzheimer's disease (top two
rows) and in an age-matched control subject. Attenuation of signal in the parietal and inferior temporal cortex is
apparent.
1 of 5 14-04-2010 20:10
Add to lightbox
Figure 12-22 Regions of significant decreases in blood flow in Alzheimer's disease compared to normal controls
obtained from a group analysis. This characteristic pattern of flow decrease may be useful for early diagnosis of
the disease.
Add to lightbox
Figure 12-23 Comparison of contrast-enhanced T1-weighted images and ASL blood flow images in a patient with
a brain tumor prior to treatment. Enhancement is very weak in this tumor but the ASL indicates very high perfusion
in the perimeter of the lesion.
REFERENCES
1. Folkman J: New perspectives in clinical oncology from angiogenesis research. Eur J Cancer 32A(14):2534-2539, 1996.
2. Axel L: Cerebral blood flow determination by rapid-sequence computed tomography. Radiology 137:676-686, 1980.
3. Lassen NA, Anderson AR, Friberg L, et al: The retention of [Tc-99m] D,L-Hm-Pao in the human brain after intracarotid bolus injection-a
kinetic analysis. J Cereb Blood Flow Metab 8(6):S13-S22, 1988.
4. Pekar J, Ligeti L, Ruttner Z, et al: Invivo measurement of cerebral oxygen consumption and blood flow using O-17 magnetic resonance
imaging. Magn Reson Med 21(2):313-319, 1991.
5. Detre JA, Subramanian VH, Mitchell MD, et al: Measurement of regional cerebral blood-flow in cat brain using intracarotid (H2o)-H-2 and
H-2 Nmr imaging. Magn Reson Med 14(2):389-395, 1990.
6. Kim SG, Ackerman JJH: Quantification of regional blood flow by monitoring of exogenous tracer via nuclear magnetic resonance
spectroscopy. Magn Reson Med 14(2):266-282, 1990.
7. Barranco D, Sutton LN, Florin S, et al: Use of F-19 Nmr-spectroscopy for measurement of cerebral blood flow-a comparative study using
microspheres. J Cereb Blood Flow Metab 9(6):886-891, 1989.
8. Detre JA, Eskey CJ, Koretsky AP: Measurement of cerebral blood flow in rat brain by f-19-nmr detection of trifluoromethane washout. Magn
Reson Med 15(1):45-57, 1990.
9. Peled S, Jolesz FA, Tseng CH, et al: Determinants of tissue delivery for Xe-129 magnetic resonance in humans. Magn Reson Med
36(3):340-344, 1996.
10. Choquet P, Hyacinth JN, Duhamel G, et al: Method to determine in vivo the relaxation time T1 of hyperpolarized xenon in rat brain. Magn
Reson Med 49(6):1014-1018, 2003.
11. Johansson E, Mansson S, Wirestam R, et al: Cerebral perfusion assessment by bolus tracking using hyperpolarized 13C. Magn Reson
Med 51(3):464-472, 2004.
12. Calamante F, Thomas DL, Pell GS, et al: Measuring cerebral blood flow using magnetic resonance imaging techniques. J Cereb Blood
Flow Metab 19(7):701-735, 1999.
13. Villringer A, Rosen BR, Belliveau JW, et al: Dynamic imaging with lanthanide chelates in normal brain-contrast due to magnetic
susceptibility effects. Magn Reson Med 6(2):164-174, 1988.
14. Rosen BR, Belliveau JW, Buchbinder BR, et al: Contrast agents and cerebral hemodynamics. Magn Reson Med 19(2):285-292, 1991.
15. Edelman RR, Mattle HP, Atkinson DJ, et al: Cerebral blood-flow-assessment with dynamic contrast-enhanced t2-star-weighted MR
imaging at 1.5-T. Radiology 176(1):211-220, 1990.
16. Atkinson DJ, Burstein D, Edelman RR: First-pass cardiac perfusion: evaluation with ultrafast MR imaging. Radiology 174(3 Pt 1):757-76,
2 of 5 14-04-2010 20:10
1990.
17. Rosen BR, Belliveau JW, Vevea JM, et al: Perfusion imaging with NMR contrast agents. Magn Reson Med 14(2):249-265, 1990.
18. Williams DS, Detre JA, Leigh JS, et al: Magnetic resonance imaging of perfusion using spin inversion of arterial water. Proc Natl Acad Sci
USA 89:212-216, 1992. Medline Similar articles
19. Detre JA, Leigh JS, Williams DS, et al: Perfusion imaging. Magn Reson Med 23:37-45, 1992. Medline Similar articles
20. Boxerman JL, Hamberg LM, Rosen BR, et al: MR contrast due to intravascular magnetic susceptibility perturbations. Magn Reson Med
34(4):555-566, 1995.
21. Weisskoff RM, Zuo CS, Boxerman JL, et al: Microscopic susceptibility variation and transverse relaxation-theory and experiment. Magn
Reson Med 31(6):601-610, 1994.
22. Yablonskiy DA, Haacke EM: Theory of NMR signal behavior in magnetically inhomogeneous tissues-the static dephasing regime. Magn
Reson Med 32(6):749-763, 1994.
23. Kiselev VG, Posse S: Analytical model of susceptibility-induced MR signal dephasing: effect of diffusion in a microvascular network. Magn
Reson Med 41(3):499-509, 1999.
24. Fisel CR, Ackerman JL, Buxton RB, et al: MR contrast due to microscopically heterogeneous magnetic susceptibility-numerical
simulations and applications to cerebral physiology. Magn Reson Med 17(2):336-347, 1991.
25. Kennan RP, Zhong JH, Gore JC: Intravascular susceptibility contrast mechanisms in tissues. Magn Reson Med 31(1):9-21, 1994.
26. Rempp KA, Brix G, Wenz F, et al: Quantification of regional cerebral blood flow and volume with dynamic susceptibility contrast-enhanced
MR imaging. Radiology 193(3):637-641, 1994.
27. Johnson KM, Tao JZ, Kennan RP, et al: Intravascular susceptibility agent effects on tissue transverse relaxation rates in vivo. Magn Reson
Med 44(6):909-914, 2000.
28. Dennie J, Mandeville JB, Boxerman JL, et al: NMR imaging of changes in vascular morphology due to tumor angiogenesis. Magn Reson
Med 40(6):793-799, 1998.
29. Tropres I, Grimault S, Vaeth A, et al: Vessel size imaging. Magn Reson Med 45(3):397-408, 2001.
30. Kim YR, Rebro KJ, Schmainda KM: Water exchange and inflow affect the accuracy of T1-GRE blood volume measurements: implications
for the evaluation of tumor angiogenesis. Magn Reson Med 47(6):1110-1120, 2002.
31. Martel AL, Allder SJ, Delay GS, et al: Perfusion MRI of infarcted and noninfarcted brain tissue in stroke: a comparison of conventional
hemodynamic imaging and factor analysis of dynamic studies. Invest Radiol 36(7):378-385, 2001.
32. Weisskoff RM, Chesler D, Boxerman JL, et al: Pitfalls in MR measurement of tissue blood flow with intravascular tracers-which mean
transit time. Magn Reson Med 29(4):553-559, 1993.
33. Ostergaard L, Weisskoff RM, Chesler DA, et al: High resolution measurement of cerebral blood flow using intravascular tracer bolus
passages. 1. Mathematical approach and statistical analysis. Magn Reson Med 36(5):715-725, 1996.
34. Calamante F, Gadian DG, Connely A: Delay and dispersion effects in dynamic susceptibility contrast MRI: simulations using singular value
decomposition. Magn Reson in Med 44:466-473, 2000.
35. van Osch MJ, Vonken EJ, Wu O, et al: Model of the human vasculature for studying the influence of contrast injection speed on cerebral
perfusion MRI. Magn Reson Med 50(3):614-622, 2003.
36. Yonas H, Darhy JM, Marks EC, et al: Cbf measured by Xe-Ct-approach to analysis and normal values. J Cereb Blood Flow Metab
11(5):716-725, 1991.
37. Lammertsma AA, Cunningham MP, Deiber MP, et al: Combination of dynamic and integral methods for generating reproducible functional
CBF images. J Cereb Blood Flow Metab 10:675-686, 1990. Medline Similar articles
38. Levin JM, Kaufman MJ, Russ MH, et al: Sequential dynamic susceptibility contrast MR experiments in human brain-residual contrast agent
effect, steady-state, and hemodynamic perturbation. Magn Reson Med 34(5):655-663, 1995.
39. Brasch RC: New directions in the development of MR imaging contrast-media. Radiology 183(1):1-11, 1992.
page 355
page 356
40. Grandin CB, Duprez TP, Smith AM, et al: Which MR-derived perfusion parameters are the best predictors of infarct growth in hyperacute
stroke? Comparative study between relative and quantitative measurements. Radiology 223(2):361-370, 2002.
41. Perthen JE, Calamante F, Gadian DG, et al: Is quantification of bolus tracking MRI reliable without deconvolution? Magn Reson Med
47(1):61-67, 2002.
42. Tong DC, Yenari M, Obrien MW, et al: Perfusion MRI time to peak (TTP) map compared with diffusion MRI and clinically apparent
neurologic deficit in hyperacute stroke. Stroke 28(1):60, 1997.
43. Perkio J, Aronen HJ, Kangasmaki A, et al: Evaluation of four postprocessing methods for determination of cerebral blood volume and
mean transit time by dynamic susceptibility contrast imaging. Magn Reson Med 47(5):973-981, 2002.
44. van Osch MJ, Vonken EJ, Viergever MA, et al: Measuring the arterial input function with gradient echo sequences. Magn Reson Med
49(6):1067-1076, 2003.
45. Akbudak E, Conturo TE: Arterial input functions from MR phase imaging. Magn Reson Med 36(6):809-815, 1996.
46. Duhamel G, de Bazelaire C, Alsop DC: Input functions from echoplanar hemodynamic studies using dynamic susceptibility contrast: a
numerical model compared to in-vivo results. International Society for Magnetic Resonance in Medicine 11th Scientific Meeting, Toronto,
2003.
47. Ostergaard L, Sorenson AG, Kwong KK, et al: High resolution measurement of cerebral blood flow using intravascular tracer bolus
passages. 2. Experimental comparison and preliminary results. Magn Reson Med 36(5):726-736, 1996.
48. Alsop DC, Wedmid A, Schlaug G: Defining a local input function for perfusion quantification with bolus contrast MRI. International Society
for Magnetic Resonance in Medicine 10th Scientific Meeting, Honolulu, 2002.
49. Ostergaard L, Johannsen P, Host-Poulsen P, et al: Cerebral blood flow measurements by magnetic resonance imaging bolus tracking:
comparison with [(15)O]H2O positron emission tomography in humans. J Cereb Blood Flow Metab 18(9):935-940, 1998.
50. Degaonkar MN, Melhem E, Golay X, et al: A new approach to quantitative cerebral blood flow determination using gadolinium MR
3 of 5 14-04-2010 20:10
perfusion imaging. International Society for Magnetic Resonance in Medicine 11th Scientific Meeting, Toronto, 2004.
51. Perman WH, Gado MH, Larson KB, et al: Simultaneous MR acquisition of arterial and brain signal time curves. Magn Reson Med
28(1):74-83, 1992.
52. Heiland S, Benner T, Debus J, et al: Simultaneous assessment of cerebral hemodynamics and contrast agent uptake in lesions with
disrupted blood-brain-barrier. Magn Reson Imag 17(1):21-27, 1999.
53. Vonken E, van Osch MJP, Bakker CJG, et al: Measurement of cerebral perfusion with dual-echo multi-slice quantitative dynamic
susceptibility contrast MRI. J Magn Reson Imag 10(2):109-111, 1999.
54. Vonken E, van Osch MJP, Bakker CJG, et al: Simultaneous quantitative cerebral perfusion and Gd-DTPA extravasation measurement with
dual-echo dynamic susceptibility contrast MRI. Magn Reson Med 43(6):820-827, 2000.
55. Alsop DC, Schlaug G: The equivalence of SVD and Fourier deconvolution for dynamic susceptibility contrast analysis. International Society
for Magnetic Resonance in Medicine 9th Scientific Meeting, Glasgow, 2001.
56. Liu HL, Pu YL, Liu YJ, et al: Cerebral blood flow measurement by dynamic contrast MRI using singular value decomposition with an
adaptive threshold. Magn Reson Med 42(1):167-172, 1999.
57. Andersen IK, Szymkowiak A, Rasmussen CE, et al: Perfusion quantification using Gaussian process deconvolution. Magn Reson Med
48(2):351-361, 2002.
58. Calamante F, Gadian DG, Connelly A: Quantification of bolus-tracking MRI: improved characterization of the tissue residue function using
Tikhonov regularization. Magn Reson Med 50(6):1237-1247, 2003.
59. Wu O, Ostergaard L, Weisskoff RM, et al: Tracer arrival timing-insensitive technique for estimating flow in MR perfusion-weighted imaging
using singular value decomposition with a block-circulant deconvolution matrix. Magn Reson Med 50(1):164-174, 2003.
60. Smith MR, Lu H, Trochet S, et al: Removing the effect of SVD algorithmic artifacts present in quantitative MR perfusion studies. Magn
Reson Med 51(3):631-634, 2004.
61. Runge VM, Wells JW: Choice of metal ion and formulation concentration for first-pass brain perfusion studies with magnetic resonance
imaging at 1.5 tesla. Invest Radiol 31(7):395-400, 1996.
62. Tofts PS, Brix G, Buckley DL, et al: Estimating kinetic parameters from dynamic contrast-enhanced T-1-weighted MRI of a diffusable
tracer: Standardized quantities and symbols. J Magn Reson Imag 10(3):223-232, 1999.
63. Edelman RR, Chen G: EPISTAR MRI: multislice mapping of cerebral blood flow. Magn Reson Med 40(6):800-805, 1998.
64. Wong E, Liu T, Sidarus K, et al: Velocity selective arterial spin labeling. International Society for Magnetic Resonance Imaging 10th
Scientific Meeting, Honolulu, 2002.
65. Duhamel G, de Bazelaire C, Alsop DC: Evaluation of systematic quantification errors in velocity-selective arterial spin labeling of the brain.
Magn Reson Med 50(1):145-153, 2003.
66. Marro K: FAWSETS: flow-driven arterial water stimulation with elimination of tissue signal. J Magn Reson 124(1):240-244, 1997.
67. Edelman RR, Siewert B, Darby DG, et al: Qualitative mapping of cerebral blood flow and functional localization with echo-planar MR
imaging and signal targeting with alternating radio frequency. Radiology 192:513-520, 1994. Medline Similar articles
68. Kwong KK, Chesler DA, Weisskoff RM, et al: MR perfusion studies with T1-weighted echo planar imaging. Magn Reson Med 34:878-888,
69. Buxton RB, Frank LR, Wong EC, et al: A general kinetic model for quantitative perfusion imaging with arterial spin labeling. Magn Reson
Med 40(3):383-396, 1998.
70. Alsop DC, Detre JA: Reduced transit-time sensitivity in non-invasive magnetic resonance imaging of human cerebral blood flow. J Cereb
Blood Flow Metab 16:1236-1249, 1996. Medline Similar articles
71. Frank LR, Wong E, Buxton R: Slice profile effects in adiabatic inversion: application to multislice perfusion imaging. Magn Reson Med
38(4):558-564, 1997.
72. Golay X, Stuber M, Pruessmann KP, et al: Transfer insensitive labeling technique (TILT): application to multislice functional perfusion
imaging. J Magn Reson Imag 9(3):454-460, 1999.
73. Kim SG: Quantification of relative cerebral blood-flow change by flow- sensitive alternating inversion-recovery (Fair) technique-application
to functional mapping. Magn Reson Med 34(3):293-301, 1995.
74. Yongbi MN, Branch CA, Helpern JA: Perfusion imaging using FOCI RF pulses. Magn Reson Med 40(6):938-943, 1998.
75. Dixon WT, Du LN, Faul DD, et al: Projection angiograms of blood labeled by adiabatic fast passage. Magn Reson Med 3:454-462, 1986.
76. Wong E, Buxton RB, Frank LR: A theoretical and experimental comparison of continuous and pulsed arterial spin labeling techniques for
quantitative perfusion imaging. Magn Reson Med 40(3):348-355, 1998.
77. Raichle ME, Eichling JO, Straatmann MG, et al: Blood-brain barrier permeability of 11C-labeled alcohols and 15O-labeled water. Am J
Physiol 230(2):543-552, 1976.
78. Bryant RG, Marill K, Blackmore C, et al: Magnetic Relaxation in Blood and Blood Clots. Magn Reson Med 13:133-144, 1990. Medline
Similar articles
79. Lu H, Golay X, Pekar JJ, et al: Functional magnetic resonance imaging based on changes in vascular space occupancy. Magn Reson
Med 50(2):263-274, 2003.
80. Wong E, Buxton RB, Frank LR: Quantitative imaging of perfusion using a single subtraction (QUIPSS and QUIPSS II). Magn Reson Med
39(5):702-708, 1998.
81. Wolf RL, Alsop DC, McGurvey ML, et al: Susceptibility contrast and arterial spin labeled perfusion MRI in cerebrovascular disease. J
Neuroimag 13(1):17-27, 2003.
82. Herscovitch P, Raichle ME: What is the correct value for the brain-blood partition coefficient for water? J Cereb Blood Flow Metab
83. Chalela JA, Alsop DC, Gonzales-Atavales JB, et al: Magnetic resonance perfusion imaging in acute ischemic stroke using continuous
arterial spin labeling. Stroke 31(3):680-687, 2000.
84. Maccotta L, Detre JA, Alsop DC: The efficiency of adiabatic inversion for perfusion imaging by arterial spin labeling. NMR Biomed
4 of 5 14-04-2010 20:10
10:216-221, 1997.
85. Sodickson DK, Manning W: Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magn
Reson Med 38(4):591-603, 1997.
86. Pruessmann KP, Weiger M, Scheidegger MB, et al: SENSE: sensitivity encoding for fast MRI. Magn Reson Med 42(5):952-962, 1999.
87. Ye F, Frank JA, Weinberger DR, et al: Noise reduction in 3D perfusion imaging by attenuating the static signal in arterial spin tagging
(ASSIST). Magn Reson Med 44(1):92-100, 2000.
88. Wolff SD, Balaban R: Magnetization transfer contrast (MTC) and tissue water proton relaxation in vivo. Magn Reson Med 10(1):135-144,
1989.
89. Silva AC, Zhang WG, Williams DS, et al: Multislice MRI of rat-brain perfusion during amphetamine stimulation using arterial spin-labeling.
Magn Reson Med 33(2):209-214, 1995.
90. Zaharchuk G, Ledden PJ, Kwong KK, et al: Multislice perfusion and perfusion territory imaging in humans with separate label and image
coils. Magn Reson Med 41(6):1093-1098, 1999.
91. Alsop DC, Detre JA: Multisection cerebral blood flow mr imaging with continuous arterial spin labeling. Radiology 208:410-416, 1998.
92. Talagala S: Multi-slice perfusion MRI using continuous arterial spin labeling: controlling for MT effects with simultaneous proximal and distal
RF. International Society for Magnetic Resonance in Medicine, Berkeley, CA, 1998.
93. Siewert B, Bly BM, Schlaug G, et al: Comparison of the BOLD- and EPISTAR-technique for functional brain imaging by using signal
detection theory. Magn Reson Med 36(2):249-255, 1996.
94. Alsop DC, Detre JA: Reduction of excess noise in fMRI using noise image templates (abstract). International Society for Magnetic
Resonance in Medicine, Berkeley, CA, 1997.
95. Quigley MA, Haughton VM, Carew J, et al: Comparison of independent component analysis and conventional hypothesis-driven analysis
for clinical functional MR image processing. Am J Neuroradiol 23(1):49-58, 2002.
96. Dixon WT, Sardashti M, Castillo M, et al: Multiple inversion recovery reduces static tissue signal in angiograms. Magn Reson Med
97. Kwong KK, Belliveau JW, Chesler DA, et al: Dynamic magnetic resonance imaging of human brain activity during primary sensory
stimulation. Proc Natl Acad Sci (USA) 89:5675-5679, 1992.
page 356
page 357
98. Aguirre GK, Detre JA, Zarahn E, et al: Experimental design and the relative sensitivity of BOLD and perfusion fMRI. Neuroimage
15(3):488-500, 2002.
99. Wang J, Aguirre GK, Kimberg DY, et al: Arterial spin labeling perfusion fMRI with very low task frequency. Magn Reson Med
49(5):796-802, 2003.
100. Davis TL, Kwong KK, Weiskoff RM, et al: Calibrated functional MRI: mapping the dynamics of oxidative metabolism. Proc Natl Acad Sci
USA 95(4):1834-1839, 1998.
101. Siewert B, Schlaug G, Edelman RR, et al: Comparison of EPISTAR and T2*-weighted gadolinium-enhanced perfusion imaging in
patients with acute cerebral ischemia. Neurology 48(3):673-679, 1997.
102. Detre JA, Alsop DL, Vives LR, et al: Noninvasive MRI evaluation of cerebral blood flow in cerebrovascular disease. Neurology
50(3):633-641, 1998.
103. Detre JA, Samuels OB, Alsop DC, et al: Noninvasive magnetic resonance imaging evaluation of cerebral blood flow with acetazolamide
challenge in patients with cerebrovascular stenosis. J Magn Reson Imag 10(5):870-875, 1999.
104. Ances BM, McGarvey ML, Abrahams JM, et al: Continuous arterial spin labeled perfusion magnetic resonance imaging in patients before
and after carotid endarterectomy. J Neuroimag 14(2):133-138, 2004.
105. Floyd TF, McGarvey M, Ochroch EA, et al: Perioperative changes in cerebral blood flow after cardiac surgery: influence of anemia and
aging. Ann Thorac Surg 76(6):2037-2042, 2003.
106. Wolf RL, Alsop DC, Levy-Reis I, et al: Detection of mesial temporal lobe hypoperfusion in patients with temporal lobe epilepsy by use of
arterial spin labeled perfusion MR imaging. Am J Neuroradiol 22(7):1334-1341, 2001.
107. Alsop DC, Detre JA, Grossman M: Assessment of cerebral blood flow in Alzheimer's disease by spin-labeled magnetic resonance
imaging. Ann Neurol 47(1):93-100, 2000.
108. Folkman J: Angiogenesis inhibitors: a new class of drugs. Cancer Biol Ther 2(4 Suppl 1):S127-133, 2003.
109. de Bazelaire C, et al: MRI-assessed changes in tumor blood flow following treatment with PTK/ZK correlate with subsequent tumor
shrinkage or growth in patients with metastatic renal cell carcinoma. Clin Cancer Res 9(16):6139S-6140S, 2003.
110. Schmitt PJ, Kotas M, Tobermann A, et al: Quantitative tissue perfusion measurements in head and neck carcinoma patients before and
during radiation therapy with a non-invasive MR imaging spin-labeling technique. Radiother Oncol 67(1):27-34, 2003.
111. Warmuth C, Gunther M, Zimmer C: Quantification of blood flow in brain tumors: comparison of arterial spin labeling and dynamic
susceptibility-weighted contrast-enhanced MR imaging. Radiology 228(2):523-532, 2003.
112. Weber MA, Gunther M, Lichy MP, et al: Comparison of arterial spin-labeling techniques and dynamic susceptibility-weighted contrast-
enhanced MRI in perfusion imaging of normal brain tissue. Invest Radiol 38(11):712-718, 2003.
113. Weber MA, Thilmann C, Lichy MP, et al: Assessment of irradiated brain metastases by means of arterial spin-labeling and dynamic
susceptibility-weighted contrast-enhanced perfusion mri: initial results. Invest Radiol 39(5):277-287, 2004.
© 2010 Elsevier
5 of 5 14-04-2010 20:10
ONTRAST GENTS
Peter Caravan
Randall B. Lauffer
INTRODUCTION
Approximately 25% to 30% of magnetic resonance imaging (MRI) scans are performed with the
addition of an intravenously administered contrast agent. The contrast agent typically makes diseased
tissue appear brighter (or in some cases darker) than the surrounding tissue. The use of contrast
media in MRI is well established. The first approved contrast agent, Gd-DTPA (gadopentetate
dimeglumine , Magnevist), appeared in 1988 and several other compounds followed. The first
generation of contrast agents consisted of the extracellular fluid (ECF) agents, followed by compounds
designed for liver imaging and compounds given orally for gastrointestinal imaging. There are currently
several compounds in clinical trials designed specifically for imaging blood vessels. At the preclinical
stage there are exciting advances in molecular imaging agents: compounds that detect pH change,
enzymatic activity, specific biomolecules such as fibrin or integrins, and magnetically labeled cells.
The goal of this chapter is to introduce the most important background information and current topics
related to MRI contrast media. These tools should allow the reader to understand the chemical and
physical basis of MRI signal enhancement and to critically evaluate new agents as they are introduced
into the clinical arena. More emphasis is placed on the gadolinium-based agents used for T1-weighted
proton imaging since these represent the majority of contrast agents used clinically. Contrast media
that use other mechanisms or nuclei-T2-weighted proton imaging, magnetization transfer imaging, or
hyperpolarized gases-will be briefly discussed. More detailed accounts of the chemistry and biophysics
of contrast agents have been published.1,2
MRI contrast media are different from imaging agents used in X-ray or scintigraphic imaging in that the
MRI contrast agent is observed indirectly. Barium or iodine absorbs X-rays and the presence of the
contrast media is seen directly on the film. Technetium and thallium emit gamma rays and the
concentration of these isotopes is proportional to the radiation emitted. In most MRI studies water and
lipid hydrogen atoms are being imaged. The MRI contrast agent perturbs in some way the magnetic
properties of these hydrogen atoms and the effect is seen in the image. The indirect effect is less
straightforward to understand. In addition, there are several means by which a contrast agent can alter
the magnet properties of hydrogen atoms, i.e., there are several contrast mechanisms available.
Although this makes matters more complex, it also gives the practitioner several tools to use with MRI
to obtain the desired image.
© 2010 Elsevier
1 of 1 14-04-2010 20:11
HISTORICAL BACKGROUND
page 358
page 359
The use of paramagnetic materials to alter the relaxation properties of water began with the
development of magnetic resonance in the 1940s. Bloch3 used ferric nitrate to enhance the relaxation
rate of water protons. The theory relating to solvent nuclear relaxation rates in the presence of
dissolved paramagnetic metal ions was developed by Bloembergen, Solomon, and others.4-7 Eisinger
and colleagues8 demonstrated that the binding of a paramagnetic metal ion to a macromolecule-in their
case, DNA-enhances the water proton relaxation efficiency of the metal ion through a slowing of the
molecular tumbling time. This phenomenon, called "proton relaxation enhancement" was subsequently
used extensively to study the hydration and structure of metalloenzymes. 9-11 These studies showed
that the magnitude of the effect of a metal ion on relaxation of water protons depended on which metal
ion was used and the chemical environment that the metal ion was in.
The pioneering 1973 work of Lauterbur12 in imaging with MR was extended to human imaging in
13-15 16
1977. Shortly thereafter, Lauterbur and co-workers demonstrated the feasibility of using
paramagnetic materials for tissue discrimination exploiting the differences in proton relaxation times. In
their studies, a salt of manganese (II) was injected into dogs with an occluded coronary artery.
Manganese is known to localize in normal myocardial tissue in preference to infarcted regions. The
longitudinal proton relaxation rates of tissues samples correlated with manganese concentration and
thus normal myocardium could be distinguished from the infarcted region by the difference in relaxation
behavior. Brady, Goldman, and colleagues17,18 confirmed the feasibility of using paramagnetic agents
in imaging studies of excised dog hearts treated similarly with Mn(II). Normal myocardium containing
higher manganese concentrations exhibited greater signal intensity than did the infarcted regions in
MR images; no contrast was present without manganese administration.
The first contrast-enhanced human MRI study was performed by Young et al.19 They used orally
administered ferric chloride to enhance the gastrointestinal tract. The diagnostic potential of
20
paramagnetic agents was first demonstrated in patients by Carr and co-workers. Gadolinium(III)
diethylenetriamine pentetate ([Gd(DTPA)(H2O)]2-) was administered intravenously to patients with
cerebral tumors and provided enhancement of the lesion in the region of cerebral capillary breakdown.
© 2010 Elsevier
1 of 1 14-04-2010 20:13
MAGNETIC PROPERTIES AND NUCLEAR RELAXATION

Contrast agents that function by affecting the relaxation rates of water protons possess electron spins
that are unpaired. Electron spin is analogous to the quantum mechanical property of nuclear spin
possessed by nuclei. For most molecules, electrons are distributed in pairs into various orbitals (an
orbital defines a region in space where an electron has some probability of residing). The two
electrons in a given orbital must have opposite spins (one up, one down) as described by the Pauli
exclusion principle. Because these two spins cancel, there is no net electron spin associated with the
molecule and it is said to be diamagnetic.
Table 13-1. Magnetic Moments of Selected Paramagnetic Metal Ions

Number of Unpaired Magnetic Moment(Bohr
Ion Electrons Magnetons)
Transition metal ions
Copper(II) 1 1.7-2.2
Nickel(II) 2 2.8-4.0
Chromium(III) 3 3.8
Iron(II) 4 5.1-5.5
Manganese (II), 5 5.9
Iron(III)
Lanthanide ions
Praseodymium(III) 2 3.8
Gadolinium(III) 7 8.0
Dysprosium(III) 5 10.6
Holmium(III) 4 10.6
Some substances (usually metal ions) have several orbitals at identical or similar energies. If there are
more orbitals than electron pairs, only one electron is present per orbital and the electrons all have
parallel spins. In this case a net electron spin results and the substance is said to be paramagnetic.
Other paramagnetic substances possess an odd number of electrons; the odd number makes it
impossible for all to pair up and a net electron spin results.
The tumbling of endogenous diamagnetic molecules (water, fats, proteins) creates tiny fluctuating
magnetic fields that induce relaxation. Unpaired electrons generate much larger magnetic fields. As
contrast agents tumble in solution the local magnetic field that they create is very potent at inducing
nuclear relaxation.
The most common paramagnetic species in nature are metal ions that often possess incompletely filled
d or f orbitals. These ions possess between one and seven unpaired electrons and this results in
magnetic moments ranging from 1.7 to 10.6 Bohr magnetons (Table 13-1). The magnetic moment is a
factor that determines the efficiency of nuclear relaxation enhancement and is roughly proportional to
the number of unpaired electrons. For some lanthanide ions other than Gd(III), the orbital angular
momentum has a significant contribution to the magnetic moment in addition to the number of unpaired
electrons. One reason why gadolinium is often used in MRI contrast agents is because of its high
magnetic moment. Other paramagnetic materials include organic free radicals (such as nitroxides) and
molecular oxygen, but these substances have rather low magnetic moments and have received less
attention as MRI contrast agents.
page 359
page 360
1 of 2 14-04-2010 20:18
Table 13-2. Properties of Different Forms of Magnetism

Net Alignment to Relative Magnetic Example of
Type of Magnetism External Field Susceptibility Substance
Diamagnetism Antiparallel -1 Most organic
materials
Paramagnetism Parallel +10 Metal complexes
Superparamagnetism Parallel +5000 Small iron
particles
Ferromagnetism Parallel +25,000 Large iron
particles
Paramagnetism generally involves the magnetism of small isolated ions that only behave as local
magnets in the presence of an external magnetic field. For paramagnetic materials that contain multiple
ions, the total magnetic susceptibility is directly proportional to the number of ions in the material.
Therefore the molar magnetic susceptibility (magnetic susceptibility divided by the number of ions) is
constant. There are other materials that exhibit ferromagnetism and superparamagnetism. For certain
materials such as ferrites (iron oxides) the individual spins of each iron cooperatively, via quantum
mechanical interactions, build up to give the crystal a very large total spin, resulting in a very large
molecular magnetic susceptibility that is a function of the number of spins. Such a material is called
ferromagnetic and its magnetism persists outside the external magnetic field. A weaker form of this is
superparamagnetism: small particles of iron oxides with aligned spins in a magnetic field. Since the
particles are small (submicron), the magnetic susceptibility effect is smaller than for large crystals of
ferrites. Superparamagnets are no longer magnetic outside of the external field. The different types of
magnetism are summarized in Table 13-2 along with their relative effect on magnetic susceptibility.
© 2010 Elsevier
2 of 2 14-04-2010 20:18
DISTINCTION BETWEEN T1 AND T2 AGENTS

All contrast agents shorten both T1 and T2. However, it is useful to classify MRI contrast agents into
two broad groups based on whether the substance increases the transverse relaxation rate (1/T2) by
roughly the same amount that it increases the longitudinal relaxation rate (1/T1) or whether 1/T2 is
altered to a much greater extent. Those in the first category are referred to as T1 agents because, on
a percentage basis, these agents alter 1/T1 of tissue more than 1/T2 owing to the fast endogenous
transverse relaxation in tissue. With most pulse sequences, this dominant T1 lowering effect gives rise
to increases in signal intensity; these are positive contrast agents. The T2 agents largely increase the
1/T2 of tissue selectively and cause a reduction in signal intensity; these are negative contrast agents.
Paramagnetic gadolinium-based contrast agents are examples of T1 agents, while ferromagnetic large
iron oxide particles are examples of T2 agents. Figure 13-1 shows the qualitative effect of such agents
on signal intensity as a function of contrast agent dose or concentration. Gadolinium agents increase
signal intensity until present in such a high concentration that the T2 effect of the agent starts to
decrease signal. Large iron particles have some effect on T1, but the T2 effect is much greater and
these agents cause signal loss at all concentrations. The intermediate case is given by the ultrasmall
particles of iron oxide (USPIO), which can increase signal on T1-weighted images and decrease signal
on T2-weighted images.
© 2010 Elsevier
1 of 1 14-04-2010 20:22
GENERAL REQUIREMENTS FOR MRI CONTRAST AGENTS

MRI contrast agents must be biocompatible pharmaceuticals in addition to nuclear relaxation probes.
Aside from standard pharmaceutical features such as water solubility and shelf stability, the
requirements relevant for metal-ion-based agents can be classified into three general categories.
Relaxivity
The efficiency by which the agent enhances the proton relaxation rates of water is referred to as
relaxivity. This must be sufficiently high to increase significantly the relaxation rates of the target tissue.
The dose of the agent at which such alteration of tissue relaxation rates occurs must, of course, be
nontoxic. MRI can detect increases in 1/T1 as small as 10% to 20%.
Specific In Vivo Distribution

Ideally, to be of diagnostic value, the agent should localize for a period in a target tissue or tissue
compartment in preference to nontarget regions. This is a basic tenet of any agent-based imaging
procedure, in which detection of the agent is usually a simple function of its tissue concentration. For
MRI relaxation agents, this requirement should be qualified: it is sufficient that the relaxation rates of
the target tissue be enhanced in preference to the rates of other tissues. This goal might be attained
by means other than concentration differences. For example, the agent may have a higher relaxivity in
the environment of one tissue versus another.
Add to lightbox
page 360
page 361
Figure 13-1 Dose- or concentration-dependent effect of typical MRI contrast agents on tissue signal
intensity. Gadolinium agents generally increase signal intensity except where present in high
concentrations. Larger iron oxide particles usually lead to signal loss. Ultrasmall iron oxide particles
(USPIO) have higher T1 relaxivity and can cause a signal intensity increase on T1-weighted images or
decrease on T2-weighted images.
In Vivo Stability, Excretability, and Lack of Toxicity

The safety of MRI contrast agents, which of themselves offer no direct therapeutic benefit, is always a
question of appropriate medical concern. The acute and chronic toxicity of an intravenously
administered metal complex is related in part to its stability in vivo and its tissue clearance behavior.
The free metal ions are relatively toxic at doses required for MRI relaxation rate changes, and thus
dissociation of the complex cannot occur to any significant extent. The toxicity of the free ligand also
becomes a factor in the event of dissociation. A diagnostic agent should be excreted within hours to
days of administration.
1 of 2 14-04-2010 20:23
© 2010 Elsevier
2 of 2 14-04-2010 20:23
T1 AGENTS
Gadolinium Chelates Approved for Human Use
Clinical trials for Gd-DTPA (gadopentetate, Magnevist) began in 1983, shortly after the introduction of
clinical MRI, and culminated in 1988 with Gd-DTPA becoming the first approved MRI contrast agent.
There are currently (November, 2004) eight gadolinium-based contrast agents approved for human
use. The first generation of commercial contrast agents consisted of the extracellular fluid (ECF)
agents, followed by compounds designed for liver imaging. Currently there are several blood pool
agents-that is, agents with a longer distribution phase that remain in the blood for longer periods than
the ECF agents-in clinical trials. At the preclinical stage, there are several reports describing molecular
imaging agents.
The majority of contrast media in clinical use are T1 agents based on gadolinium. These will form the
bulk of the discussion. Other metals are capable of providing contrast, and iron and manganese have
been used in commercially approved agents. These compounds will be discussed later in the chapter.
The approved ECF agents are shown in Figure 13-2. These compounds exhibit three similar features:
they all contain gadolinium, they all contain an 8-coordinate ligand binding to gadolinium, and they all
contain a single water molecule coordination site to gadolinium. The multidentate ligand is required for
safety. The ligand provides high thermodynamic stability and kinetic inertness with respect to metal
loss, and enables the contrast agent to be excreted intact. The compounds all exhibit a good to
excellent safety and tolerance record in millions of applications. The gadolinium ion and coordinated
water molecule are essential to providing contrast. The compounds in Figure 13-2 show nearly
identical effects on signal intensity of tissue. Despite differences in chemical structure, all the
compounds have substantially equivalent biodistribution and plasma half-life. The main differences
among the agents are the electrical charge on the molecule, either neutral or anionic, and whether or
not the chelating ligand is cyclic in nature. The relative importance of all of these properties is
discussed in the following sections.
Relaxivity: Theory
Contrast agents shorten T1 and T2. While there are many mechanisms by which this occurs, in many
cases the effect of these mechanisms can be reduced to a single constant, called "relaxivity." The
effect and definition of relaxivity is illustrated in Figures 13-3A and B. In Figure 13-3A, the effect of a
typical contrast agent is shown on two hypothetical tissues, one with a T1 of 1200 ms, and one with a
T1 of 400 ms. At low concentrations of contrast agent (left side of the graph), it appears that the
contrast agent has a larger effect on the tissue with the longer T1. At higher concentrations of contrast
agent (right side of Fig. 13-3A), both tissues approach approximately the same T1. The simple way to
quantify this effect is to consider the rate of relaxation, 1/T1. In the simplest cases, which characterize
most cases found in medical imaging, the effect of the contrast agent is to increase the rate of
relaxation proportional to the amount of contrast agent:
where T1 is the observed T1 with contrast agent in the tissue, T1o is the T1 prior to addition of the
contrast agent, C is the concentration of contrast agent, and r1 is the longitudinal relaxivity (often just
called relaxivity). The units for r1 are mM-1s-1. Thus the slope of 1/T1 as a function of contrast agent
concentration (Fig. 13-3B) reveals the relaxivity, in this case 4 mM-1s-1. Figure 13-3B shows that effect
on relaxation rate is independent of the initial T1 of the tissue. That is, in terms of relaxation rate, the
contrast agent has the same effect regardless of initial T1.
Transverse, or T2, relaxivity is defined in an analogous way:
For all medically used contrast agents, r2 is larger than r1.
1 of 15 14-04-2010 20:25
The large fluctuating local magnetic field in the vicinity of a paramagnetic center provides an additional
relaxation pathway for solvent nuclei. The magnetic field falls off rapidly with distance away from the
metal ion. Therefore diffusion of water molecules or specific chemical interactions that bring water
molecules near the ion (within 5 Å) are important in transmitting the relaxation effect. Different types of
chemical interactions yield different relaxation efficiencies that are governed by the distance between
the ion and the water hydrogen and the time scale of the interaction. The sum of these interactions
gives the total relaxivity of the paramagnetic species.
page 361
page 362
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
2 of 15 14-04-2010 20:25
Add to lightbox
Add to lightbox
Figure 13-2 Chemical structures of extracellular fluid (ECF) MRI contrast agents that are approved for
use in Europe and/or the US. Note that all contain an octadentate ligand, a gadolinium ion, and single
coordinated water molecule.
It is useful to classify the relevant contributions to water proton relaxivity with respect to three distinct
types of interactions as shown in Figure 13-4. In case A, a water molecule binds in the primary
coordination sphere of the metal ion and transmits the relaxation effect by exchanging with the bulk
solvent. This is referred to as inner-sphere relaxation. Case B represents hydrogen-bonded water
molecules in the second coordination sphere. Because of the difficulty in characterizing the second
sphere interactions (how many waters? what is the H to ion distance?), investigators often do not
distinguish between this mechanism and that of waters diffusing past the metal ion, case C. Cases B
and C are often referred to collectively as outer-sphere relaxation. The total relaxivity can be generally
given by Equation 13-3:
The contribution from inner-sphere relaxivity occurs because the metal-bonded water molecule is
efficiently relaxed. The relaxed water undergoes exchange with water molecules from the solvent. The
exchange rate (~106 per second) is much faster than a typical imaging TR such that many water
molecules are relaxed for each metal ion. The contrast agent can be viewed as a catalyst that changes
the magnetic properties of water. It follows that inner-sphere relaxivity should be dependent on: 1. the
number of water molecules, denoted q, in the inner sphere; 2. how long the water molecule stays
bonded to the metal ion, denoted τm-this time should be short so that many water molecules can be
cycled through and relaxed; and 3. the efficiency of relaxation of the bound water molecule which is
given by T1m, the relaxation time of the bound water. Inner-sphere relaxivity is given by Equation 13-4:
Longitudinal relaxation of the bound water molecule takes place via a dipolar mechanism, Equation
13-5:
3 of 15 14-04-2010 20:25
page 362
page 363
Add to lightbox
Add to lightbox
Figure 13-3 Change in A, longitudinal relaxation time (T1) and B, longitudinal relaxation rate (1/T1) for
hypothetical tissue having initial relaxation times T1 = 1200 ms (dashed line) and T1 = 400 ms (solid
-1 -1
line) as a function of concentration of contrast agent with a relaxivity, r1 = 4 mM s .
Add to lightbox
4 of 15 14-04-2010 20:25
Add to lightbox
Add to lightbox
Figure 13-4 Interactions between water and metal complexes. A, Inner-sphere relaxation resulting from
binding of water molecules directly to the metal ion, i.e. coordinated in the inner sphere. B and C,
Outer-sphere relaxation resulting from water hydrogen-bonded in the second coordination sphere of
the metal (B) and diffusion of water nearby the metal ion but without any binding to the metal complex
(C).
For water bonded to gadolinium, the relaxation depends inversely on the Gd to H distance (rM to H) to
the sixth power. The magnetic moment of the ion is seen in the S(S+1) term where S is the spin
quantum number (S = 7/2 for Gd). There are other physical constants that are the same for all metal
ions: γH is the proton gyromagnetic ratio, ge is the electronic g-factor, and μB is the Bohr magneton.
The term in parentheses includes the proton Larmor frequency, ωH, and the electron Larmor
frequency, ωS = 658ωH; this means that relaxivity is dependent on the applied magnetic field as well
as physical and chemical properties.
The key feature of paramagnetically induced relaxation is that the local magnetic field generated by the
unpaired electrons must fluctuate at proper frequencies to stimulate nuclear relaxation. The time scale
of this fluctuation is characterized by a correlation time, τc; the rate of the fluctuation, 1/τc, is
dominated by the fastest of three rates:
where T1e (also called τs) is the longitudinal electronic relaxation time; τm is the water residency time
as already mentioned; and τR is the rotational tumbling time of the entire metal-water unit. These
processes, all of which alter the magnetic field at the nucleus, are shown in Figure 13-5. In order for
relaxation to be efficient, the local fluctuating magnetic field should be close to the proton Larmor
frequency.
Similar theories exist for outer-sphere relaxation. It should be pointed out that a complete quantitative
understanding of both inner- and outer-sphere relaxivity is limited to the aqua ions at this time.
Nonetheless, the description given above allows one to predict the relaxation behavior of putative
contrast agents.
Relaxivity in Solution
Gadolinium Agents
Table 13-3 lists some relaxivities at 0.47 tesla (20 MHz) for the ECF agents along with some other
metal complexes; structures of the additional ligands are shown in Figure 13-6. Relaxivity is
proportional to the number of inner-sphere water molecules (q), which decreases when multidentate
ligands are bound to the metal ion to reduce its toxicity. For example, chelation of gadolinium with
5 of 15 14-04-2010 20:25
DTPA reduces q from 8 to 1. All the ECF agents approved for use have similar relaxivities and one
inner-sphere water. When q = 0, relaxation is limited to outer-sphere mechanisms. One can see from
Table 13-3 that outer-sphere relaxivity is quite sizable.
page 363
page 364
Add to lightbox
Figure 13-5 Proton relaxation by paramagnetic metal ions. The nuclear spin on the metal-bound water
hydrogens (small vectors) experiences the fluctuating magnetic field of electron spin S (large vector).
The rate of the magnetic fluctuation is determined by the rates of rotation (1/τR), electronic relaxation
(1/T1e), and water exchange (1/τm). The relaxation effect is transmitted to bulk water by rapid water
exchange.
Table 13-3. Longitudinal Relaxivities* (r1) at 20 MHz in the Temperature Range

35 to 39° C for Selected Metal Complexes along with the Number of Coordinated
Waters (q) and the Magnetic Moment, μBM
Complex q μBM (BM) r1 (mM-1s-1)

Aqua ions
Gd(III) 8 8.0 9.1
Mn(II) 6 5.9 8.0
Fe(III) 6 5.9 8.0
Cu(II) 6 1.7 0.84
Dy(III) 8 10.6 0.56
Gadolinium ECF agents
Gd-DTPA 1 8.0 3.8
Gd-DTPA-BMA 1 8.0 3.9
Gd-DO3A-HP 1 8.0 3.7
Gd-DOTA 1 8.0 3.6
Gd-DTPA-BMEA 1 8.0 4.7
Other gadolinium complexes
Gd-EDTA 3 8.0 6.6
Gd-TTHA 0 8.0 2.0
2
*Relaxivities from references 1 and .
6 of 15 14-04-2010 20:25
The complexity of improving the contrast agents can be appreciated from Figure 13-5. One way to
increase relaxivity is to increase q, however this tends to result in unstable metal complexes that may
result in metal loss in vivo. Most contrast agents compromise by leaving one open site for water
coordination. Figure 13-5 and Table 13-3 show that a high magnetic moment is preferred, hence the
reliance on gadolinium. Relaxivities of the aqua ions are proportional to the magnetic moment. A
notable exception to this is dysprosium(III). The reason for the low relaxivity of Dy(III) complexes is an
extremely fast electronic relaxation process. T1e for Dy(III) is about 1 ps and the correlation time, τc,
becomes T1e. The unpaired electrons at the dysprosium ion transit through energy levels so fast that
they create a magnetic field that is fluctuating too fast (~160,000 MHz) for good probability of energy
transfer to, and relaxation of, a hydrogen atom (65 MHz). Paramagnetic ions considered as relaxation
agents-Gd(III), Mn(II), Fe(III)-all have long T1e values.
Add to lightbox
Add to lightbox
Figure 13-6 Chemical structures of Gd-TTHA with no inner-sphere water (q = 0) and Gd-EDTA with 3
inner-sphere waters (q = 3).
Add to lightbox
page 364
page 365
7 of 15 14-04-2010 20:25
Figure 13-7 Calculated inner-sphere longitudinal relaxivities (r1) as a function of magnetic field
(bottom axis) and Larmor frequency (top axis) or NMRD profiles. Each line corresponds to a rotational
correlation time, τR, of 0.1, 0.3, 1, 3, 10, or 100 ns. The 0.1 ns curve is typical of the gadolinium ECF
agents, while the 3 and 10 ns curves are typical of the blood pool agents being developed. The
dispersion (dip) occurring between 0.01 and 0.1 tesla corresponds to the denominator in Equation
13-5, ω Sτc > 1; the dispersion between 1 and 10 tesla corresponds to the denominator in Equation
13-5, ω Hτc > 1. Note that the value of τR affects both the magnitude of relaxivity and the field at which
relaxivity disperses.
Add to lightbox
Figure 13-8 Chemical structures of the blood pool agents MS-325 and B22956 and the multipurpose
agent Gd-BOPTA. The additional organic groups appended to the Gd-DTPA core enable binding to
serum albumin (MS-325, B22956, Gd-BOPTA) and hepatocyte proteins (Gd-BOPTA).
Add to lightbox
Figure 13-9 Chemical structure of the blood pool agent Gadomer. The large size of the compound
results in an increased value of τR.
8 of 15 14-04-2010 20:25
Add to lightbox
Figure 13-10 Chemical structure of the blood pool agent P792. The large size of the compound results
in an increased value of τR.
For metal ions with long T1e values the single most important source of relaxivity enhancement is an
increase in the rotational correlation time, τR. Figure 13-7 shows some theoretical relaxivity curves
simulated for a gadolinium complex with one bound water where the only parameter changing is an
increase in the rotational correlation time. The field dependence in relaxivity arises from two factors.
Electronic relaxation gets slower for Gd(III) as the magnetic field is increased. At low fields, τc ~ T1e,
and relaxivity increases with field. At higher fields τR becomes the dominant correlation time. Eventually
the condition ωHτc > 1 is met, the denominator in Equation 13-5 becomes large, and relaxivity
decreases. The field dependence is often referred to as nuclear magnetic relaxation dispersion
(NMRD). Figure 13-7 demonstrates that the magnitude and functional form of the NMRD profile is
altered when the rotational correlation time is increased (i.e., when tumbling slows). The prominent
peak that forms is noteworthy because it is predicted to occur over the clinically relevant field range.
page 365
page 366
9 of 15 14-04-2010 20:25
Add to lightbox
Figure 13-11 Experimental NMRD data of 0.2 mM Gd-DTPA in 4.5% HSA solution (filled squares),
MS-325 in 4.5% HSA solution (filled circles), and MS-325 in phosphate buffered saline (open circles).
The relaxivity of MS-325 is markedly enhanced in the presence of serum albumin because of protein
binding and a concomitant increase in τR.
The ability to enhance relaxivity by increasing τR is demonstrated by blood pool agents currently in
development. Two approaches have been taken. One is to use a small molecule that can reversibly
interact with a protein-examples of this approach are MS-32521 (EPIX/Schering) and B2295622,23
(Bracco) which bind reversibly to serum albumin (Fig. 13-8). At excess albumin concentrations
approximately 90% of MS-325 is bound to the protein, resulting in a ~100-fold increase in τR.24 The
other approach is to increase the size of the molecule, making it tumble more slowly (increasing
τR)-examples include Gadomer25,26 from Schering AG (Fig. 13-9), and P79227,28 from Guerbet (Fig.
13-10). This is illustrated by the NMRD profiles of MS-325 and Gd-DTPA in either buffer alone or in
human serum albumin (HSA) solution (Fig. 13-11). The relaxivity of MS-325 in buffer is greater than
that of Gd-DTPA because of its larger size and this results in a slight increase in τR. In the presence of
HSA, the relaxivity is much greater for MS-325 and the field dependence predicted above is observed.
There are also compounds with weak protein binding such as Gd-BOPTA (see Fig. 13-8). This
compound is ~10% bound to serum albumin and this results in about a twofold enhancement in
relaxivity compared to Gd-DTPA, but still a much lower enhancement than compared with MS-325.
Iron Agents for T1- and T2-Weighted Imaging

The contrast agents based on iron oxide have to be treated differently. These are not discrete
molecules but crystals of iron oxide (Fe3O4) surrounded by a coating (often dextran). For gadolinium
and manganese contrast agents with multiple ions, the relaxivity is additive for the number of ions in
the compound; the spins of one gadolinium ion do not interact with the spins of another ion. For certain
materials such as ferrites (iron oxides) the individual spins of each iron cooperatively build up to give
the crystal a very large total spin, and thus relaxivity will be a function of the number of spins. Such
materials are called superparamagnetic, to describe the ferromagnetic-like properties, though
superparamagnetic materials do not show any of the hysteresis of ferromagnetic materials.
The iron oxide particles consist of a core of one or more magnetic crystals of Fe3O4 embedded in a
coating. Because these are materials rather than discrete molecules, there is a distribution of sizes.
Ultrasmall particles of iron oxide (USPIO) have a single crystal core and a submicron diameter (e.g.
10 of 15 14-04-2010 20:25
ferumoxtran [Sinerem or AMI-227] has a crystal diameter of 4.3 to 4.9 nm and a global particle
diameter of ~50 nm).29 Small particles of iron oxide (SPIO) have cores containing more than one
crystal of Fe3O4 and are larger than USPIOs but still submicron (e.g., ferumoxide [Endorem or
Feridex] has a crystal diameter of 4.3 to 4.8 nm and a global particle diameter of ~200 nm).29 USPIO
and SPIO are small enough to form a stable suspension and can be administered intravenously. The
size differences result in differences in pharmacokinetic behavior, which will be described below. There
are also large particles which are used for oral applications (e.g., Abdoscan, 50 nm crystals making up
30
a 3 μm particle).
There are no inner-sphere water molecules in iron particles, and relaxation of water arises from the
water molecules diffusing near the particle. However, the mechanism of outer-sphere relaxation is
different than described above. One feature is that the crystals have a net magnetization and as the
external field is increased this magnetization is increased (this is true as well for gadolinium but the
effect is much smaller). The modulation of this net magnetization can cause proton relaxation (so-called
Curie spin relaxation). The theories describing the field dependence of iron oxide relaxivity have been
reported.30
There are some generalities about relaxivity in these particles. For the USPIOs longitudinal relaxivity
(r1) can be quite high and these can function as effective T1 agents. The r2/r1 ratio for USPIOs is
significantly larger than for gadolinium complexes and r2 increases with increasing magnetic field. When
there is aggregation of crystals, which is the case in SPIOs, longitudinal relaxivity tends to decrease (r1
drops) and transverse relaxivity to increase (r2 increases). Thus both for the particles themselves, as
well as aggregates of particles, the ratio of r2/r1 typically increases as the size of the particles or
aggregates increases, though the T2 relaxivity as a function of particle size can be quite complicated.
32
See, for example, references 31 and . The effect of aggregation of crystals is that the aggregate
itself can be considered a large magnetized sphere whose magnetic moment increases with increasing
field strength. This gives rise to susceptibility effects and the SPIOs can act as T2* relaxation agents.
This has important consequences when considering the effects of contrast agent compartmentalization
on imaging (see below).
Relaxivity in Tissue
page 366
page 367
The efficiency by which a metal complex influences tissue relaxation rates depends on three factors:
1. The chemical environment encountered by the complex in vivo. By far the greatest effect is
exerted by binding of the agent to macromolecular structures which can potentially cause
significant relaxivity enhancement. An example of this is shown in Figure 13-11, comparing the
relaxivity of MS-325 in buffer solution and in serum albumin solution.
2. Compartmentalization of the complex in tissue. Generally, tissue water is compartmentalized
into intravascular, interstitial (fluid space between cells and capillaries), and intracellular space
constituting roughly 5%, 15%, and 80% of total water, respectively. Cellular organelles further
subdivide the intracellular component. If water exchange between any of these compartments is
slow relative to the relaxation rate in the compartment with the longest T1, multiexponential
relaxation may result. This can decrease the effective tissue relaxivity of an agent because not
all of the tissue water is encountering the paramagnetic center.
3. The magnetic susceptibility of the contrast agent. The contrast agent causes microscopic field
inhomogeneity on a biological scale of 10 to 1000 nm rather than on the chemical scale of 0.1 to
1 nm. This results in a reduction in apparent T2.
The result of the first two effects is that the simple relaxivity equation (Eq. 13-1) is often not valid in a
biological setting; likewise, describing the effects on an MR pulse sequence with a single tissue
relaxivity can be misleading. Much care has to be taken when trying to estimate concentration of a
contrast agent from signal intensity changes. The effect of chemical composition of the tissue on
relaxation rates and the physical effects of compartmentalization are discussed in further detail below.
11 of 15 14-04-2010 20:25
The actual relaxivity within a compartment might be affected by the biological milieu. For some
compounds, with strong or weak albumin binding (e.g., MS-325, B22956, Gd-BOPTA), local variation
in the albumin concentration will affect the amount of contrast agent bound to albumin and thus affect
the overall relaxivity. For example, in the plasma space, albumin concentration is typically high (0.6 to
0.7 mM) compared to the extracellular space in the normal heart (0.2 to 0.3 mM), and some spaces
(the CSF, for example) have almost no albumin. Figure 13-11 suggests that at 1.5 T the relaxivity of
MS-325 would change from 23 mM-1s-1 to 6 mM-1s-1 depending on whether it is in the plasma or in
CSF.
The hydrophilic ECF agents do not bind to plasma proteins or membrane structures. Tweedle and
colleagues showed that the relaxivity in blood and soft-tissue of Gd-DTPA and Gd-DOTA is the same
within error as the relaxivity in aqueous solution.33 However, in extreme settings, the actual relaxivity
could vary. For example, Stanisz and Henkelman34 showed that in extremely concentrated protein
media, which might characterize some biological compartments, the relaxivity of Gd-DTPA could be
affected when the macromolecular concentration is high enough. However, there is no study to date
that used quantitative methods to compare both the relaxivity and the concentration of contrast agent in
specific biological compartments that has found a relaxivity in vivo that differs from its temperature-
matched in vitro value.35 Some investigators have postulated that binding in specific disease states
might occur in vivo even for the nonspecific ECF agents, thereby increasing relaxivity. As of this date,
however, no reproducible evidence of in vivo binding that affects relaxivity has been shown for any of
the extracellular agents.
Physical compartmentalization makes it more difficult to predict tissue relaxivity and even may make
the term relaxivity irrelevant. Except for pathologic situations, most contrast agents in use today are
36
excluded from intact cells. That is, with the exception of opsonization of iron oxide particles, and the
37
liver-specific agents, most contrast agents are designed to keep the heavy metal out of cells. The
normal situation in most settings is that the contrast agent will be locally concentrated in extracellular
spaces. As a result, the simple relaxivity equations do not necessarily hold. To get an idea of the
effect, consider a 1 mM solution of a gadolinium-based ECF agent. In a simple test tube, it takes an
38
average of about 3 μs for water to diffuse between gadolinium molecules, and thus in the time of a
typical imaging TR, a given water molecule may interact with thousands or millions of gadolinium
molecules, and all water molecules will interact with approximately the same number of gadolinium
ions. However, if that same 1 mM solution is compartmentalized solely within the cardiac
microvasculature, it takes between 2 and 20 s for most of the water in the tissue (85% of it is
extravascular) to physically diffuse into the microvasculature. Thus, most of the water in the tissue
does not have the opportunity to be relaxed by the gadolinium in the typical TR of an imaging
acquisition. As a net result, the actual signal enhancement is typically less than that predicted by using
Equation 13-1 with the assumption that the contrast agent is uniformly distributed throughout the tissue.
To simplify the problem, the concept of "water exchange" and exchange time, τ, between
39
compartments is frequently used. When the contrast agent concentration is different between two or
more compartments because of delivery kinetics, the water exchange rate and the size of the
compartments will determine the effect of the contrast agent on MRI signal. Although the algebra is
straightforward, even in the simplest case of two compartments with a single exchange constant, the
general formula describing the effect of water relaxation is too algebraically complex to be included
here. However, the limiting behavior in two cases can help describe the situations where this exchange
strongly affects contrast and when it can, for all intents and purposes, be ignored.
page 367
page 368
In the first case, water moves so fast between the biological compartments that the net effect is as if
the contrast agent were uniformly spread throughout the whole tissue. This regime, called "fast
exchange," occurs whenever the exchange rate, 1/τ, between the compartments is much faster than
the difference in relaxation rates between the compartments.40 For example, in blood, the red cell has
41
a relatively short exchange time, on the order of 5 to 10 ms. Even though the intact red cell prevents
12 of 15 14-04-2010 20:25
most MR contrast agents from entering the cell, as long as the plasma T1 is longer than 20 ms the two
compartments of the blood (plasma + red cells) will remain in fast exchange, and thus the blood will
behave for MR purposes as if the contrast agent were spread uniformly through the blood. In this
case, in general the effective relaxation rate will be the weighted average of the relaxation in the two
compartments. That is, if for compartment i the volume fraction is fi, the initial T1 is T1i and the
concentration of agent is Ci (which could be zero), the whole tissue together will behave like:
In the second case, water moves much more slowly between the compartments. This case, called
"slow exchange," occurs whenever the water exchange rate is much slower than the difference in
relaxation rates between the compartments. In this case, a single relaxation time, and thus a single
relaxivity, is meaningless, because the two microscopic compartments will relax with their own
relaxation times.
Very few biological compartments show true slow exchange, except at an extremely high concentration
of contrast agent. However, the intermediate case, when exchange is neither slow nor fast
("intermediate exchange") occurs very commonly. In the intermediate case, the relaxation behavior will
also appear biexponential, although both the apparent compartment size and the effective T1 of the
two compartments will vary from their true biological size and T1. It is possible to model the signal
intensity behavior as a function of contrast agent concentration to estimate water exchange times in
vivo (e.g., reference 42). Clearly, characterizing human tissue as having only one or two compartments
is an oversimplification. Nevertheless, models that incorporate two compartments have proved useful
for explaining the effects of biological water mobility on contrast-enhanced scans. 43
Biological compartmentalization also results in susceptibility contrast. The contrast agent causes
microscopic field inhomogeneities. Proton diffusion through these inhomogeneities (often called
"mesoscopic" inhomogeneities44 since the scale of the inhomogeneities is of the same order as the
scale of the Brownian motion of the water) causes the protons to dephase from one another due to the
different magnetic fields that they experience during their random walks. Even in the absence of water
diffusion, the field inhomogeneity causes intravoxel dephasing and thus signal loss on gradient-echo
images due to the different microscopic magnetic fields within the voxel. While the strength of the
perturbing magnetic fields is directly proportional to the agent's concentration, usually expressed
through the agent molar susceptibility (χ) constant, the actual magnitude and even direction of the
magnetic field shifts depend strongly on the size and the shape of the biological compartment in which
45
the contrast agent resides, and the size of the susceptibility contrast effect depends on how the
water diffuses through the tissue.
Because the susceptibility T2 effect does not require water to pass into the hydration sphere of the
contrast agent, large effects can be observed even when the contrast agent is compartmentalized in a
very small tissue compartment. For example, first-pass brain perfusion imaging (so-called PWI46)
relies on the susceptibility effect due to the compartmentalization of currently approved extracellular
gadolinium-based agents. The small blood volume in the brain (4% in gray matter, 2% in white matter)
and the relatively limited water exchange between the extravascular and intravascular spaces in the
brain ultimately limit the size of signal changes due to any T1-based contrast agent at acceptable
doses. The susceptibility-based T2 and T2* effects, however, can be much larger (as much as a 50%
signal drop in normal gray matter at the same dose) due to the "action at a distance" effect possible
with the through-space effect. Thus, especially in cases of slow exchange and small compartments
available for the contrast agent, susceptibility contrast may be the medically relevant contrast
mechanism of choice. At present, these larger effects, however, are accompanied by potentially larger
uncertainty in the absolute quantitation, especially when there is underlying pathology. Nevertheless, as
with the potential limits on quantitation in myocardial perfusion and viability imaging, the appeal of
routine qualitative imaging using these effects has been demonstrated consistently over the past
decade. Iron oxide particles with their much higher magnetic susceptibility are more potent
susceptibility agents.
13 of 15 14-04-2010 20:25
Toxicity
It is important to understand the acute and chronic toxic effects of paramagnetic metal complexes in
view of the likely possibility of routine intravenous administration of such compounds for MRI
examinations in the future. The required doses (roughly 0.5 to 5.0 g per patient) greatly exceed those
of metal ions or complexes used in radioscintigraphy. However, iodine-containing contrast media for
computed tomography and other radiologic procedures are used at much higher doses (~50 to 200 g
per patient). Despite the high dose, gadolinium-based chelates are among the safest injectable
compounds in current medical use and have a reputation for being safer, especially in terms of reduced
nephrotoxicity, than their X-ray contrast agent counterparts.
The safety record of the four longest approved gadolinium-based agents (gadopentetate dimeglumine
, gadodiamide , gadoteridol , and gadoterate meglumine) was recently reviewed by Runge.47 These
four agents had approximately the same overall adverse event rates, with nausea (1% to 2%) and
hives (1%) leading the list. Nearly all adverse events with these agents are transient, mild, and
self-limiting. Nevertheless, there are reports of serious adverse reactions, including life-threatening
anaphylactoid reactions and death, for these agents. The best estimate puts the rate of these events
at between 1 in 200,000 and 1 in 400,000 patients.48 The safety record for iron and manganese 49
based agents is growing, but due in part to their much smaller market share, is currently less well
documented. The gadolinium-based ECF contrast agents have been studied and most are approved
for pediatric use in children above 2 years, though there are differences in their approval wording. The
package insert for these agents should always be consulted for the latest safety information.
page 368
page 369
Table 13-4. Acute LD50 Values for Metal Salts, Metal Complexes, and Free
Ligands Administered Intravenously to Mice
Compound LD50 (mmol/kg) Reference
GdCl3 0.4 33
(MEG)[Gd(EDTA)(H2O)3] 0.3* 51
(MEG)2[Gd(DTPA)(H2O)] >10 33
Na2[Gd(DTPA)(H2O)] >10 52
(MEG)2[Gd(DOTA)(H2O)] >10 33
Na2[Gd(DOTA)(H2O)] >10 33
[GdDTPA-BMA(H2O)] 34 53
[GdHP-DO3A(H2O)] 12 53
(MEG)3H2DTPA 0.15 33
(MEG)2H2DOTA 0.18 33
Na3[Ca(DTPA)] 3.5 52
MnCl2 0.22* 53
Na2[MnEDTA(H2O)] 7.0* 53
*animal, rat.
MEG, N-methylglucamine.
For compounds in development, possible toxic effects merit further discussion. Toxicity and stability
are considered together since the dissociation of the complex generally leads to a higher degree of
toxicity stemming from the free metal ion or free chelating ligand. For example, the DTPA ligand and
gadolinium chloride both have LD50 of 0.5 mmol/kg in rats, while the Gd-DTPA complex has almost a
14 of 15 14-04-2010 20:25
20-fold higher safety margin, its LD50 being 8 mmol/kg.50 LD50 is the dose that has an approximately
50% chance of causing death. In addition to metal and free ligand based toxicity, one must also
consider the toxicity of the intact complex or any metabolites.
A survey of available toxicologic data points to the importance of metal complex dissociation as a key
factor. Table 13-4 lists acute LD50 values determined for some metal ions, complexes, and ligands in
mice. The fact that metal ions and free ligands tend to be more toxic than the corresponding metal
chelate can be understood by considering that the complexation step "neutralizes" the coordinating
properties of both the metal ion and ligand to some degree, decreasing their avidity for binding to
proteins, enzymes, or membranes. Table 13-4 also shows that the toxic effect of the free ligand can
be muted by formulation as the calcium salt.
In the simplest view, the degree of toxicity of a metal chelate is related to its degree of dissociation in
vivo before excretion. A good example of this is the comparison between Gd-EDTA and Gd-DTPA.
The latter complex is stable (metal-ligand formation constant,54 log KML = 22.5), excreted intact by the
kidneys, and exhibits a low degree of toxicity (LD50 = 8 to 20 mmol/kg). Gd-EDTA on the other hand
54
has a lower stability constant (log KML = 17.4) and a higher toxicity that is comparable to GdCl3
(LD50 ~0.5 mmol/kg). The straightforward explanation is that Gd-EDTA dissociates in vivo, yielding the
toxicity of the free metal ion. Stability constants are one measure of predicting whether a complex will
be stable in vivo, but these are thermodynamic measurements. Equally important is kinetic inertness.
51
For instance, Gd-DTPA-BMA has a similar stability constant (log KML = 16.9) to Gd-EDTA but it is
much less toxic (LD50 = 34 mmol/kg) and is used clinically. This is because the rate at which the
gadolinium ion dissociates is much slower for Gd-DTPA-BMA.
55
The toxicity of metal ions has been extensively reviewed. The coordination of ions to oxygen,
nitrogen, and sulfur heteroatoms in macromolecules and membranes alters the dynamic equilibria
necessary to sustain life. The gadolinium ion can bind to calcium-binding sites, often with higher affinity
owing to its greater charge/radius ratio.
The toxicity of free ligands is less well understood. It stems from sequestration of essential metal ions
such as calcium and zinc, in addition to "organic" toxicity.
The toxicity of intact metal complexes can stem from a wide variety of specific and nonspecific effects.
At the high doses required in LD50 determinations of relatively nontoxic hydrophilic chelates like
Gd-DTPA, the nonspecific hypertonic effect is thought to be important. A difference in osmolality
between intracellular and extracellular compartments is established after injection of large quantities of
ionic complexes and appropriate counter ions. Water is drawn out of the cells as a result of the
osmotic gradient, causing cellular and circulatory damage. The nonionic gadolinium agents-
Gd-DTPA-BMA (gadodiamide , Omniscan), Gd-DO3A-HP (gadoteridol , ProHance), and
Gd-DTPA-BMEA (gadoversetamide , OptiMARK)-were developed to reduce the osmolality of the
injected formulations.56 Judging from the excellent safety record of ionic Gd-DTPA,47 it is not clear
whether the nonionic concept is medically relevant or just a convenient marketing tool. Low osmolality
may reduce the apparent acute toxicity in animals at large doses (100 times the clinical dose, Table
13-4), but the relevance of these findings to the clinical setting is uncertain.
© 2010 Elsevier
15 of 15 14-04-2010 20:25
BIODISTRIBUTION
page 369
page 370
Add to lightbox
Figure 13-12 Principal distribution sites and excretion pathways for intravenously administered soluble
metal complexes.
Targeting a paramagnetic agent to a particular site within the body is one of the most challenging
aspects of MRI contrast agent design. The diagnostic utility of a contrast-enhanced MRI examination
depends on the absolute concentration of the agent in the desired tissue and the selectivity of the
distribution relative to other tissues. True targeting is rarely achieved. After administration, the agent
equilibrates in several body compartments before excretion; preferential distribution of the agent to the
desired site is all that can be expected in most circumstances. MRI agents are similar to
radiopharmaceuticals or iodinated CT agents in that the MR image enhancement depends on the
concentration of the paramagnetic metal complex. The principles of distribution governing these other
agents are directly applicable to MRI agents. One key difference of MRI agents is the dependence of
relaxivity on the chemical environment. By targeting a complex to desired sites where binding to a
macromolecule occurs, the target/nontarget ratio in terms of relaxation rate changes may be greater
than the ratio in terms of concentration. This so-called receptor-induced magnetization enhancement
(RIME)24,57 effect has been put into practice for some liver agents and intravascular agents (discussed
below).
Figure 13-12 illustrates potential distribution sites and excretion pathways relevant for soluble metal
complexes. An intravenously administered chelate rapidly equilibrates in the intravascular and interstitial
(space between cells) fluid compartments; these are referred to collectively as the extracellular
compartment. Depending on its structure, the complex may also be distributed into various intracellular
environments (including that of liver and kidney) by passive diffusion or specific uptake processes.
Clinically available contrast agents and those currently undergoing clinical trials are targeted primarily
by their distribution: extracellular fluid agents (ECF agent, also called ECS agents-extracellular space),
liver agents, and intravascular or blood pool agents (for further detailed information, consult Chapter 14
on Tissue-Specific Contrast Agents). These will be discussed in detail in the following sections.
Compounds that are specifically targeted to biopolymers or receptors are usually referred to as
molecular imaging agents. These are all currently at the preclinical stage of development and will be
discussed briefly here. They are discussed in more detail in Chapter 15.
Extracellular Fluid (ECF) Agents

The approved ECF agents are shown in Figure 13-2. These compounds exhibit three similar features:
they all contain gadolinium, they all contain an 8-coordinate ligand binding to Gd(III), and they all
contain a single water molecule coordination site to Gd(III). The multidentate ligand is required for
safety.2 It provides high thermodynamic stability and kinetic inertness with respect to metal loss, and
1 of 7 14-04-2010 20:27
enables the contrast agent to be excreted intact. The presence of charged or hydrogen-bonding
groups such as carboxylates and the lack of large hydrophobic groups ensure minimal interaction with
plasma proteins, other macromolecules, and membranes. As discussed above (Relaxivity section), the
gadolinium ion and coordinated water molecule are essential to providing contrast. The high magnetic
moment of Gd(III) and its slow electronic relaxation rate make it an excellent relaxor of water protons.
The proximity of the coordinated water molecule leads to efficient relaxation. The coordinated water
molecule is in rapid chemical exchange (106 exchanges per second) with solvating water molecules.58
This results in a catalytic effect whereby the gadolinium complex effectively shortens the relaxation
times of the bulk solution.
Table 13-5. European or US approved (November 2004) MRI Contrast Agents-

Relaxivity, Osmolality, and Viscosity
r1, 0.5 r2, 0.5
Product Chemical T 37° T 37° Osmolality* Viscosity*
Generic Name Name Abbreviation C C (osmol/kg) (cP)
Gadopentetate Magnevist Gd-DTPA 3.860 1.9661 2.961
Gadoterate Dotarem Gd-DOTA 3.662 4.862 1.3563 2.063
(4.02)63 (11.3)63
Gadodiamide Omniscan Gd-DTPA-BMA 63 61 61
3.9 0.79 1.4
63 63
(1.90) (3.9)
Gadoteridol Prohance Gd-HPDO3A 3.763 0.6363 1.363
(1.91)63 (3.9)63
Gadobutrol Gadovist Gd-DO3A-butrol 3.660 0.5760 1.460
(1.39)61 (3.7)61
Gadoversetamide Optimark Gd-DTPA-BMEA 64 65 65
4.7 1.11 2.0
Gadobenate Multihance Gd-BOPTA 66 66 67 67
4.4 5.6 1.97 5.3
Gadoxetic acid Primovist Gd-EOB-DTPA 5.3107 6.2107 0.88 N/A
disodium
Mangafodipir Teslascan Mn-DPDP 68 68 69 69
1.9 2.2 0.29 0.7
Ferumoxide Feridex IV AMI-25 29 29 70 70
24 107 0.2 0.34
Endorem
Ferucarbotran Resovist SHU555a 71 71 72 72
20 190 0.33 1.0
page 370
page 371
*All concentrations 0.5M except those in parentheses, 1M, and Mn-DPDP (0.01M),AMI-25 (0.2M),
Gd-EOB-DTPA (0.25M).
The extracellular agents have very similar properties. They are all very hydrophilic complexes with
similar relaxivities and excellent safety profiles, and can be formulated at high concentrations. Because
of the close similarity in their pharmacologic behavior, MRI medical usage often just refers to these
compounds as "gadolinium." An injection of any of these agents (with rare exceptions59) yields the
same diagnostic information. Table 13-5 lists the relaxivities, osmolalities, and viscosities of the
contrast agents currently approved in the US and/or Europe.
There are some differences among the physical properties. The diamide complexes have considerably
2 of 7 14-04-2010 20:27
lower thermodynamic stability (log K~17 vs log K >21 for other gadolinium complexes),73,74 although
this does not seem to affect their safety profile relative to the other agents. The nonionic (neutral)
compounds (gadodiamide , gadoteridol , gadoversetamide , gadobutrol) were designed to minimize
the osmolality of the formulation as discussed above (Toxicity section). One benefit of the nonionic
75
compounds is the ability to formulate them at high concentration (1M) without drastically increasing
the osmolality or viscosity (Table 13-5). These high-concentration formulations may be useful in fast
dynamic studies such as brain perfusion and dynamic MR angiography (MRA).
A major use of these nonspecific agents is in the detection of cerebral capillary breakdown or the
enhancement of tissues with an increased extracellular volume.76 Both applications stem from the
dependence of the bulk tissue 1/T1 on the volume of distribution of the paramagnetic agent. If an agent
equilibrates to roughly the same concentration in the extracellular space, then 1/T1 in the extracellular
space is relatively constant in different tissues. Tissues that have a larger fraction of extracellular
volume will give an increased signal change post contrast because there is more contrast agent within
the voxel. This finding has been observed for tumors and abscesses, which often exhibit increased
interstitial volume.77 The most dramatic enhancement of lesions is seen in the brain. Here the normal
tissue exhibits little enhancement because of the impermeable nature of brain capillaries (the
blood-brain barrier) and the small intravascular volume of distribution (5%) of the agent. The capillaries
of tumors allow passage of the agent into the tumor's interstitial space resulting in selective
enhancement.
Many other applications have been developed for ECF agents. The renal excretion of these agents
yields the obvious application of imaging the kidneys, both for structural and for functional
78 46
information. The status of blood flow to a tissue (perfusion) is another application of these agents.
This requires the use of fast imaging techniques to follow the rapid passage through the tissue. ECF
agents have been used for dynamic MRA studies.79 Another application is delayed enhancement in
80
cardiac imaging to identify areas of infarct.
Liver Agents
Add to lightbox
Add to lightbox
Figure 13-13 Chemical structures of the liver imaging agents Mn-DPDP and Gd-EOB-DTPA.
3 of 7 14-04-2010 20:27
The ECF agents, like X-ray contrast agents, are cleared almost exclusively through the kidneys by
glomerular filtration. It was recognized early on that altering the excretion pathway could allow liver
imaging. The diagnostic utility of this class of MRI agents includes: selective enhancement of normal,
functioning liver tissue to aid in the detection of small lesions such as metastatic tumors (focal liver
disease); indication of the status of liver function to detect diffuse liver disease such as cirrhosis;
high-resolution visualization of bile ducts and the gallbladder.
Liver agents include the gadolinium-based compounds Gd-BOPTA (gadobenate, MultiHance)81-84 and
Gd-EOB-DTPA (Primovist),85 the manganese complex Mn-DPDP (mangafodipir, Teslascan, Fig.
13-13),86 and the iron particle formulations AMI-25 (Feridex I.V., Endorem), 87 AMI-227 (Combidex,
Sinerem),88 and SHU555a (Resovist).85 All of these compounds are available clinically, either in Europe
and/or the US, with the exception of Sinerem which is in an advanced stage of development. The
different metal types have different mechanisms of action. The gadolinium compounds are taken up by
hepatocytes89 and cleared intact via the hepatobiliary system. The gadolinium complexes provide
positive contrast (T1 weighted) of the hepatobiliary system. Mn-DPDP undergoes partial dissociation in
vivo.90 It is believed that endogenous zinc replaces the manganese ion. The free manganese is
absorbed by the pancreas and hepatocytes in the liver and enables T1-weighted imaging of these
organs. The relaxivity of the manganese bound to liver proteins is much higher than that of Mn-DPDP.
The iron oxide particles are taken up by the reticuloendothelial system and accumulate in the Kupffer
88
cells in the liver. As a result, even though these iron oxide agents generally create "dark" signal, they
create positive contrast for liver tumors by leaving undiminished the signal from tumors that have a
paucity of Kupffer cells. These iron agents are so-called superparamagnetic iron oxides (SPIO). These
particles have a much greater effect on T2 and T2* than on T1 (Tables 13-5 and 13-6). The SPIO
agents are used with T2- or T2*-weighted sequences.
Blood Pool Agents

page 371
page 372
Table 13-6. Iron Particle Contrast Agents-Relaxivities at 20 MHz (37° C)

-1 -1 -1 -1
Generic Name Product Name Chemical Abbreviation Particle r1 (mM s ) r2 (mM s )
Ferumoxide Feridex IV AMI-25 SPIO 2429 10729
Endorem SHU555a SPIO 2071 19071
Ferucarbotran Resovist AMI-227 USPIO 2329 5129
Ferumoxtran Sinerem NC100150 USPIO 2591 41
91
Clariscan VSOP-C63 USPIO 3092 39

92
Since most contrast agents are administered intravenously, they are all potentially capable of imaging
the blood vessels, and the ECF agents described above are used routinely, if off-label, for
angiography. The major drawback of the ECF agents for MRA is their pharmacokinetics. ECF agents
rapidly leak out of the vascular space into all the interstitial spaces of the body. Angiography with ECF
agents is thus typically limited to dynamic arterial studies. There has been considerable effort toward
designing specific blood pool agents that would be tailored for vascular imaging. The ideal blood pool
agent should remain in the vascular compartment and not leak out into the extracellular space. It should
be capable of being given as a bolus such that a dynamic arterial image can be obtained. At the same
time it should have sufficient relaxivity and blood half-life that it is possible to obtain high-resolution
steady-state images. There are currently (November 2004) no approved blood pool agents. However,
there are several in various stages of clinical development. Three approaches have been taken to
design blood pool agents: protein binding, increased size, and ultrasmall iron oxide particles. These are
4 of 7 14-04-2010 20:27
discussed below.
MS-32521 and B2295622,23 are gadolinium-based compounds (see Fig. 13-8) that bind reversibly to
serum albumin. Albumin is the most abundant protein in plasma, and its concentration is high enough
(600 to 700 μM) to bind enough contrast agent to have significant effects on blood T1. Reversible
albumin binding serves four purposes: 1. the albumin slows the leakage of the contrast agent out of the
intravascular space; 2. the reversible binding still allows a path for excretion-the unbound fraction can
be filtered through the kidneys or taken up by hepatocytes; 3. the bound fraction is "hidden" from the
liver and kidneys leading to an extended plasma half-life; and 4. the relaxivity of the contrast agent is
increased 4- to 10-fold upon binding to albumin (see below). (Gd-BOPTA and Gd-EOB-DTPA have
weak affinity for albumin [~10% bound], which leads to a modest relaxivity increase relative to the
ECF agents.)
The binding and relaxivity features of the gadolinium-based blood pool agents are listed in Table 13-7.
Since binding affinity is moderate to weak for these compounds, the fraction bound to albumin will
depend on the concentrations of albumin and the contrast agent. Immediately following injection, when
the concentration of the contrast agent is high relative to albumin, there will be a greater free fraction.
As the concentration of the contrast agent begins to stabilize (at ~0.5 mM) the fraction bound will
become constant. The observed relaxivity will depend on the fraction bound; unlike ECF agents, T1
change in plasma is not linearly related to contrast agent concentration. Among the albumin-binding
agents, MS-325 has a somewhat lower albumin affinity than B22956, although the majority of both is
bound under steady-state conditions. The relaxivity of albumin-bound MS-325 is higher than that of
B22956. MS-325 is mainly cleared by the kidneys while B22956 has significant biliary clearance as
well as renal excretion.
Table 13-7. Albumin Binding and Observed Relaxivities (20 MHz) of MS-325,
B22956, Gd-BOPTA, Gadomer, and P792 at 37° C
MS-325* B22956† Gd-BOPTA Gadomer‡ P792§
Agent Type Strong protein Strong protein Weak protein Increased Increased
binding binding binding size size
r1 buffer 6.621 6.523 4.496 3928
(mM-1s-1)
r1 plasma 5021 2723 9.793 18.794 44.528
-1 -1
(mM s )
% bound 21 23
91 95
plasma
page 372
page 373
*Data at 0.1mM.
†
Data at 0.5mM.
‡
Relaxivity per Gd.
§
4% HSA.
Early work on blood pool compounds involved gadolinium covalently linked to macromolecules such as
polylysine, dextran, or modified bovine serum albumin.2 The large size of these compounds severely
restricted diffusion out of the vascular space and led to very good vascular imaging properties. One
major drawback was the very slow clearance of these agents in preclinical studies, as well as potential
immunologic response. This approach was modified by the synthesis of compounds that were large
enough to be kept in the vascular compartment, but small enough to still be eliminated by glomerular
filtration in the kidneys. Gadomer (sometimes referred to as Gadomer-17, see Fig. 13-9) is an
example of this type of blood pool agent.25,26 Gadomer is a large (acting with the hydrodynamic
5 of 7 14-04-2010 20:27
properties of a ~17 kDa dendrimer) dendrimer that contains 24 gadolinium complexes covalently
linked. The dendritic (branching) approach to synthesis results in a compound that is approximately
globular. The per-gadolinium relaxivity of Gadomer is much higher than its monomeric units because of
the slow tumbling of the molecule (see below). Using multiple gadolinium chelates to increase the size
-1 -1
also increases the molecular relaxivity (24 Gd × 18.7 = molecular relaxivity of 450 mM s ), which in
turn means that lower doses can be given. Gadomer is a neutral hydrophilic compound that has little
affinity for plasma proteins. It appears to be excreted renally.
Another blood pool agent in clinical trials is P792 from Guerbet, see Figure 13-10.27,28 P792 can be
viewed as a modified Gd-DOTA, where each acetate arm contains a large hydrophilic group.
Increasing the molecular weight also increases the relaxivity, which like the albumin-bound MS-325 or
Gadomer means that lower doses of gadolinium-compared to ECF agents-can be given to obtain
comparable contrast. P792 has little affinity for plasma proteins and has predominantly renal
clearance.
It should be noted that all of the gadolinium-based vascular agents described above are not "true"
blood pool compounds. Although far superior to the ECF agents in terms of extravasation and
relaxivity, there is still some fraction of the compound that leaks out into the extracellular space. Iron
oxide particles, on the other hand, are true blood pool agents.
The SPIO particles used for liver imaging are large enough to be recognized by the reticuloendothelial
system and rapidly removed from the blood stream. It was found that the smallest size fraction of
these particles, the so-called ultrasmall iron oxide particles (USPIO), evaded the reticuloendothelial
30
system and could be used to image the blood pool. Although smaller, these are still particles that are
too large to passively leak out of the vascular space, and they make very good blood pool agents.
Making ultrasmall particles not only changes the biodistribution of the compound, but also changes the
relaxation phenomena. SPIO have a much greater effect on T2 than T1 (large r2/r1) and are used as
T2 or T2* agents. USPIO have very good T1 relaxation properties (smaller r2/r1) and can be used for
bright blood imaging (T1 weighted). The iron oxide particles are not excreted; the iron is eventually
resorbed into the body. The iron particles in clinical trials are summarized in Table 13-6 along with their
longitudinal (r1) and transverse (r2) relaxivities.
Molecular Imaging
Molecular imaging has been defined as "the in vivo characterization and measurement of biologic
processes at the cellular and molecular level."95 Much work has been done in the radiotracer field of
SPECT and PET imaging. With the advent of genomics and proteomics there is a growing number of
protein-based targets to molecularly address disease. While these proteins and cell surface receptors
can often be targeted with gamma or positron emitting nuclei, these targets are not often amenable to
MRI detection. MRI is a relatively insensitive technique and rather high concentrations of contrast
media are required (micromolar) whereas most biological targets are present in the nanomolar
concentration range. Moreover the compartmentalization effects discussed earlier can also limit
sensitivity. Nevertheless, the exquisite resolution of MRI coupled with the ability to simultaneously
obtain an anatomic image in addition to a "molecular" image has spurred research in this area.
One approach has been to increase the local concentration of the signal-generating group by using
assemblies of paramagnetic ions. Weissleder and co-workers have developed chemistry for making
modified iron oxides, the so-called cross-linked iron oxides (CLIO). This has enabled a number of
96-100
targeted molecular imaging agents. Wickline and colleagues have used emulsion-type particles to
non-covalently assemble hundreds of gadolinium complexes. The particle is then directed by an
antibody to the target. They have demonstrated this in a canine thrombus model. 101
An alternative approach is to identify molecular targets present at high concentrations. Fibrin is present
at high concentration (>50 μM) in thrombi. Scientists at EPIX Pharmaceuticals identified a peptide that
binds selectively to fibrin and labeled it with four gadolinium chelates. The compound, EP-2104R, binds
selectively to fibrin in vivo. Figure 13-14 shows an image of a mural carotid artery thrombus in a rabbit
6 of 7 14-04-2010 20:27
model. The corresponding blood pool image shows diffuse vessel damage but neither specifically
identifies the thrombus nor shows its extent. The field of molecular imaging will be considered in
greater detail in Chapter 15.
© 2010 Elsevier
7 of 7 14-04-2010 20:27
CHEMICAL EXCHANGE SATURATION TRANSFER

Another interesting area of contrast agent development is that of chemical exchange saturation
transfer (CEST).102,103 These are reagents that affect image contrast by a different mechanism than
T1 or T2 change. CEST agents contain an exchangeable hydrogen atom or atoms. These hydrogen
atoms resonate at a different Larmor frequency than that of water. If an RF pulse is applied at the
frequency of the exchangeable water, this resonance becomes saturated. When the saturated
hydrogen exchanges with water hydrogens, it transfers its magnetization to the water. Hence the name
chemical exchange saturation transfer; these agents are sometimes called magnetization transfer
agents. The net effect is a loss of magnetization of the water resonance.
These agents can be diamagnetic or paramagnetic. The concentration requirement for the current crop
of CEST agents is still above 1 mM. However, improvements continue to be made by increasing the
number of exchangeable hydrogens and by improving the exchange rate. These agents offer the
possibility of being able to "turn on" the contrast agent effect with an RF pulse by employing the proper
pulse sequence.
page 373
page 374
Add to lightbox
1 of 2 14-04-2010 20:29
Add to lightbox
Figure 13-14 T1-weighted MRIs of an injured carotid artery in a rabbit. The injury produces a
nonocclusive thrombus in the carotid. The main images are maximum-intensity projections; insets are
single axial slices through the site of injury. On the left is an image obtained 30 minutes after
administration of the fibrin-binding contrast agent EP-2104R. Arrows indicate the site of injury. On the
right is the same animal, imaged using a blood pool contrast agent. The blood pool image
demonstrates the patency of the vessel, as well as diffuse injury in the model, but does not specifically
demonstrate the location or extent of thrombus. EP-2104R, currently being developed by EPIX
Pharmaceuticals (Cambridge, MA, USA) is a gadolinium-based, fibrin-specific MRI contrast agent,
which binds to intact fibrin selectively without binding to circulating fibrinogen.
© 2010 Elsevier
2 of 2 14-04-2010 20:29
HYPERPOLARIZED CONTRAST AGENTS

The lack of sensitivity in MRI stems from the very small degree of polarization among the nuclear
spins. In a magnetic field there is a net magnetization but this is small: approximately 0.0006% of the
spins are polarized. A technique called spin-exchange using a high-powered laser (also called optical
pumping) can increase the polarization by 4 to 5 orders of magnitude (hyperpolarization).104 Isotopes
possessing long T1 values can be hyperpolarized and used as contrast agents. The long T1 is
necessary to maintain the contrast medium in the hyperpolarized state long enough to image before
the spins relax back to the equilibrium value.
Gases often have long T1 values, and isotopes of the noble gases helium ( 3He) and xenon (129Xe)
have been used for imaging. The biggest application has been imaging the lung.104,105 Recently
contrast agents with hyperpolarized 13C were reported. Svensson and co-workers106 described a
13
C-enriched water-soluble compound, bis-1,1-(hydroxymethyl)-1-(13)C-cyclopropane-D(8), that had
long relaxation times-in vitro: T1 approximately 82 s, T2 approximately 18 s; in vivo: T1 approximately
38 s, T2 approximately 1.3 s. This could be formulated at a 13C concentration of 200 mM and
hyperpolarized to 15%. The authors used this material for a contrast-enhanced magnetic resonance
angiogram (CE-MRA) in rats.
A major benefit of hyperpolarized contrast media is the excellent sensitivity and lack of background
(high S/N). Challenges include the distribution and availability of the hyperpolarization equipment and
imaging hardware compatibility for imaging nonhydrogen nuclei (not available on all clinical scanners).
© 2010 Elsevier
1 of 1 14-04-2010 20:29
CONTRAST AGENT USE AT HIGH FIELD

page 374
page 375
At the time of writing (November 2004), most clinical imagers operate at 1.5 T and most contrast
agents used are ECF agents. However, there is increasing growth in 3 T scanners and there is a push
to move clinical imaging to still higher fields. In terms of relaxivity, longitudinal relaxivity (r1) will usually
decrease as the field strength is increased while transverse relaxivity (r2) will usually increase. For the
ECF agents, the effect is small. However, for slow tumbling agents like the blood pool agents
described above, the field dependence is quite marked (see Fig. 13-7). The r2 of iron oxide particles
can increase dramatically with increasing field strength. It is important to keep in mind that relaxivities
change as contrast media applications are moved to higher field strengths.
REFERENCES
1. Lauffer RB: Paramagnetic metal complexes as water proton relaxation agents for NMR imaging: Theory and design. Chem
Rev 87:901-927, 1987.
2. Caravan P, Ellison JJ, McMurry TJ, et al: Gadolinium(III) chelates as MRI contrast agents: Structure, dynamics, and
applications. Chem Rev 99:2293-2352, 1999. Medline Similar articles
3. Bloch F, Hansen WW, Packard M: The nuclear induction experiment. Phys Rev 70:474, 1948.
4. Kubo R, Tomita K: Paramagnetic relaxation. J Phys Soc Jpn 9:888, 1954.
5. Solomon I: Relaxation processes in a system of two spins. Phys Rev 99:559, 1955.
6. Bloembergen N, Purcell EM, Pound RV: Relaxation effects in nuclear magnetic resonance absorption. Phys Rev 73:679,
1948.
7. Bloembergen N: Proton relaxation times in paramagnetic solutions. J Chem Phys 27:572, 1957.
8. Eisinger J, Shulman RG, Blumberg WE: Relaxation enhancement by paramagnetic iron binding in deoxyribonucleic acid
solutions. Nature 192:963, 1961. Medline Similar articles
9. Dwek RA: Nuclear Magnetic Resonance (NMR) in Biochemistry. Oxford: Oxford University Press, 1973.
10. Burton DR, Forsen S, Karlstrom G, et al: Proton relaxation enhancement (PRE) in biochemistry: a critical survey. Prog NMR
Spectrosc 13:1, 1979.
11. Mildvan AS. Proton relaxation enhancement. Ann Rev Biochem 43:357, 1974. Medline Similar articles
242:190, 1973.
13. Damadian R, Goldsmith M, Minkoff L: NMR in cancer: XVI. FONAR image of the live human body. Physiol Chem Phys
9:97-100, 108, 1977. Medline Similar articles
14. Hinshaw WS, Bottomley PA, Holland GN: Radiographic thin-section image of the human wrist by nuclear magnetic
resonance. Nature 270:722-723, 1977. Medline Similar articles
15. Andrew ER, Bottomley PA, Hinshaw WS, et al: NMR images by the multiple sensitive point method: application to larger
biological systems. Phys Med Biol 22:971-974, 1977. Medline Similar articles
16. Lauterbur PC, Mendonca Dias MH, Rudin AM: Augmentation of tissue water proton spin-lattice relaxation rates by in vivo
addition of paramagnetic ions. In Dutton PO, Leigh J, Scarpa A (eds): Frontiers of Biological Energetics. Academic Press,
New York, NY, 1978, pp. 752-759.
17. Brady TJ, Gebhardt MC, Pykett IL, et al: Proton nuclear magnetic resonance imaging of regionally ischemic canine hearts:
effect of paramagnetic proton signal enhancement. Radiology 144:343-347, 1982. Medline Similar articles
18. Goldman MR, Brady TJ, Pykett IL, et al: Quantification of experimental myocardial infarction using nuclear magnetic
resonance imaging and paramagnetic ion contrast enhancement in excised canine hearts. Circulation 66:1012-1016, 1982.
19. Young IR, Clarke GJ, Bailes DR, et al: Enhancement of relaxation rate with paramagnetic contrast agents in NMR imaging.
J Comput Tomogr 5:543-547, 1981. Medline Similar articles
20. Carr DH, Brown J, Bydder GM, et al: Intravenous chelated gadolinium as a contrast agent in NMR imaging of cerebral
tumours. Lancet 1:484-486, 1984. Medline Similar articles
21. Lauffer RB, Parmelee DJ, Dunham SU, et al: MS-325: albumin-targeted contrast agent for MR angiography. Radiology
22. La Noce A, Stoelben S, Scheffler K, et al: B22956/1, a new intravascular contrast agent for MRI: first administration to
humans-preliminary results. Acad Radiol 9:S404-S406, 2002.
23. Cavagna FM, Lorusso V, Anelli PL, et al: Preclinical profile and clinical potential of gadocoletic acid trisodium salt
(B22956/1), a new intravascular contrast medium for MRI. Acad Radiol 9:S491-S494, 2002.
24. Caravan P, Cloutier NJ, Greenfield MT, et al: The interaction of MS-325 with human serum albumin and its effect on proton
1 of 4 14-04-2010 20:30
relaxation rates. J Am Chem Soc 124:3152-3162, 2002. Medline Similar articles

25. Dong Q, Hurst DR, Weinmann HJ, et al: Magnetic resonance with Gadomer-17. An animal study original investigation.
Invest Radiol 33:699-708, 1998. Medline Similar articles
26. Nicolle GM, Toth E, Schmitt-Willich H, et al: The impact of rigidity and water exchange on the relaxivity of a dendritic MRI
contrast agent. Chemistry 8:1040-1048, 2002. Medline Similar articles
27. Gaillard S, Kubiak C, Stolz C, et al: Safety and pharmacokinetics of p792, a new blood-pool agent: results of clinical testing
in nonpatient volunteers. Invest Radiol 37:161-166, 2002. Medline Similar articles
28. Port M, Corot C, Raynal I, et al: Physicochemical and biological evaluation of P792, a rapid-clearance blood-pool agent for
magnetic resonance imaging. Invest Radiol 36:445-454, 2001. Medline Similar articles
29. Jung CW, Jacobs P: Physical and chemical properties of superparamagnetic iron oxide MR contrast agents: ferumoxides
, ferumoxtran, ferumoxsil. Magn Reson Imaging 13:661-674, 1995. Medline Similar articles
30. Muller RN, Roch A, Colet J-M, et al: Particulate magnetic contrast agents. In Merbach AE, Tóth E (eds): Chemistry of
Contrast Agents in Medical Magnetic Resonance Imaging. John Wiley, Chichester, UK, 2001, pp. 417-435.
31. Weisskoff RM, Zuo CS, Boxerman JL, et al: Microscopic susceptibility variation and transverse relaxation: theory and
experiment. Magn Reson Med 31:601-610, 1994. Medline Similar articles
32. Hardy P, Henkelman RM: On the transverse relaxation rate enhancement induced by diffusion of spins through
inhomogeneous fields. Magn Reson Med 17:348-356, 1991. Medline Similar articles
33. Tweedle MF, Gaughan GT, Hagan J, et al: Considerations involving paramagnetic coordination compounds as useful NMR
contrast agents. Nucl Med Biol 15:31-36, 1988.
34. Stanisz GJ, Henkelman RM: Gd-DTPA relaxivity depends on macromolecular content. Magn Reson Med 44:665-667,
35. Donahue KM, Burstein D, Manning WJ, et al: Studies of Gd-DTPA relaxivity and proton exchange rates in tissue. Magn
36. Moore A, Weissleder R, Bogdanov A Jr: Uptake of dextran-coated monocrystalline iron oxides in tumor cells and
macrophages. J Magn Reson Imaging 7:1140-1145, 1997. Medline Similar articles
37. Clement O, Siauve N, Cuenod CA, et al: Contrast agents in magnetic resonance imaging of the liver: present and future.
Biomed Pharmacother 52:51-58, 1998. Medline Similar articles
38. Donahue KM, Weisskoff RM, Burstein D: Water diffusion and exchange as they influence contrast enhancement. J Magn
39. Hazlewood CF, Chang DC, Nichols BL, et al: Nuclear magnetic resonance transverse relaxation times of water protons in
skeletal muscle. Biophys J 14:583-606, 1974. Medline Similar articles
40. McLaughlin AC, Leigh JS: Relaxation times in systems with chemical exchange. J Magn Reson 9:296-304, 1973.
41. Wright GA, Hu BS, Macovski A: Estimating oxygen saturation of blood in vivo with MR imaging at 1.5 T. J Magn Reson
Imaging 1:275-283, 1991.
42. Landis CS, Li X, Telang FW, et al: Equilibrium transcytolemmal water-exchange kinetics in skeletal muscle in vivo. Magn
43. Donahue KM, Weisskoff RM, Chesler DA, et al: Improving MR quantification of regional blood volume with intravascular T1
contrast agents: accuracy, precision, and water exchange. Magn Reson Med 36:858-867, 1996. Medline
Similar articles
44. Sukstanskii AL, Yablonskiy DA: Theory of FID NMR signal dephasing induced by mesoscopic magnetic field
inhomogeneities in biological systems. J Magn Reson 151:107-117, 2001. Medline Similar articles
45. Chu SC, Xu Y, Balschi JA, et al: Bulk magnetic susceptibility shifts in NMR studies of compartmentalized samples: use of
paramagnetic reagents. Magn Reson Med 13:239-262, 1990. Medline Similar articles
46. Rosen BR, Belliveau JW, Vevea JM, et al: Perfusion imaging with NMR contrast agents. Magn Reson Med 14:249-265,
47. Runge VM: Safety of approved MR contrast media for intravenous injection. J Magn Reson Imaging 12:205-213, 2000.
48. Carr JJ: Magnetic resonance contrast agents for neuroimaging: Safety issues. Neuroimaging Clin N Am 4:43-54, 1994.
49. Federle M, Chezmar J, Rubin DL, et al: Efficacy and safety of mangafodipir trisodium (MnDPDP) injection for hepatic
MRI in adults: results of the U.S. Multicenter phase III clinical trials. Efficacy of early imaging. J Magn Reson Imaging
50. Gries H: Extracellular MRI contrast agents based on gadolinium. Top Curr Chem 221:1-24, 2002.
51. Weinmann HJ, Brasch RC, Press WR, et al: Characteristics of gadolinium-DTPA complex: a potential NMR contrast agent.
Am J Roentgenol 142:619, 1984.
52. Tweedle MF, Brittain HG, Eckelman WC, et al: Principles of Contrast-Enhanced MRI. In Partain CL (ed): Magnetic
Resonance Imaging. 2nd ed, W. B. Saunders, Philadelphia, 1988, pp. 793-809.
53. Wolf GL, Burnett KR, Goldstein EJ, et al: Contrast agents for magnetic resonance imaging. In Kressel H (ed): Magnetic
2 of 4 14-04-2010 20:30
Resonance Annual. Raven, New York, 1985, pp. 231-266.

54. Martell AE, Smith RM, Motekaitis RJ: NIST Critically Selected Stability Constants of Metal Complexes: NIST Standard
Reference Data 46. Washington, DC: National Institute of Standards and Technology, 1997.
55. Lucky TD, Venugopal B: Metal Toxicity in Mammals. New York: Plenum, 1977.
56. Watson AD, Rocklage SM, Carvlin MJ: Contrast Agents. In Stark DD, Bradley WG (eds): Magnetic Resonance Imaging.
Mosby, St Louis, 1992, pp. 372-437.
57. Lauffer RB: Targeted relaxation enhancement agents for MRI. Magn Reson Med 22:339, 1991. Medline
Similar articles
58. Powell DH, et al: High-pressure NMR kinetics. Part 74. Structural and dynamic parameters obtained from 17O NMR, EPR,
and NMRD studies of monomeric and dimeric Gd3+ complexes of interest in magnetic resonance imaging: An integrated and
theoretically self-consistent approach. J Am Chem Soc 118:9333-9346, 1996.
59. Bashir A, Gray ML, Boutin RD, et al: Glycosaminoglycan in articular cartilage: in vivo assessment with delayed Gd(DTPA)
(2-)-enhanced MR imaging. Radiology 205:551-558, 1997. Medline Similar articles
60. Vogler H, Platzek J, Schuhmann-Giampieri G, et al: Pre-clinical evaluation of gadobutrol: a new, neutral, extracellular
contrast agent for magnetic resonance imaging. Eur J Radiol 21:1-10, 1995. Medline Similar articles
61. Tweedle MF, Hagan JJ, Kumar K, et al: Reaction of gadolinium chelates with endogenously available ions. Magn Reson
page 375
page 376
62. Aime S, et al: Synthesis, characterization, and 1/T1 NMRD profiles of gadolinium(III) complexes of monoamide derivatives
of DOTA-like ligands. X-ray structure of the 10-[2-[[2-hydroxy-1-(hydroxymethyl)ethyl]amino]-1-[(phenylmethoxy)methyl]-2-
oxoethyl]-1,4,7,10-tetraazacyclododecane-1,4,7-triacetic acid-gadolinium(III) complex. Inorg Chem 31:2422-2428, 1992.
63. Tweedle MF: Physicochemical properties of gadoteridol and other magnetic resonance contrast agents. Invest Radiol
64. Periasamy M, White D, de Learie L, et al: The synthesis and screening of nonionic gadolinium(III) DTPA-bisamide
complexes as magnetic resonance imaging contrast agents. Invest Radiol 26:S217-S220, 1991.
65. OptiMARK package insert. St. Louis: Mallinckrodt, Inc.
66. Uggeri F, et al: Novel contrast agents for magnetic resonance imaging. Synthesis and characterization of the ligand
BOPTA and its Ln(III) complexes (Ln = Gd, La, Lu). X-ray structure of disodium (TPS-9-145337286-C-S)-[4-Carboxy-5,8,11-
tris(carboxymethyl)-1-phenyl-2-oxa- 5,8,11-triazatridecan-13-oato(5-)]gadolinate(2-) in a mixture with its enantiomer. Inorg
Chem 34:633-642, 1995.
67. Multihance package insert. Konstanz: Bracco-Byk Gulden.
68. Tirkkonen B, Aukrust A, Couture E, et al: Physicochemical characterisation of mangafodipir trisodium . Acta Radiol
69. Teslascan package insert. Princeton: Amersham Health.
70. FeridexIV package insert. Wayne, NJ: Berlex.
71. Bremer C, Allkemper T, Baermig J, et al: RES-specific imaging of the liver and spleen with iron oxide particles designed
for blood pool MR-angiography. J Magn Reson Imaging 10:461-467, 1999. Medline Similar articles
72. Resovist package insert. Berlin: Schering Diagnostics.
73. White DH, deLearie LA, Moore DA, et al: The thermodynamics of complexation of lanthanide(III) DTPA-bisamide
complexes and their implication for stability and solution structure. Invest Radiol 26:S226-S228, 1991.
74. Kumar K, Chang CA, Tweedle MF: Equilibrium and kinetic studies of lanthanide complexes of macrocyclic polyamino
carboxylates. Inorg Chem 32:587-593, 1993.
75. Tombach B, Heindel W: Value of 1.0-M gadolinium chelates: review of preclinical and clinical data on gadobutrol. Eur J
Radiol 12:1550-1556, 2002.
76. Roberts TP, Chuang N, Roberts HC: Neuroimaging: do we really need new contrast agents for MRI? Eur J Radiol
77. Brasch RC, Weinmann HJ, Wesby GE: Contrast enhanced NMR imaging: animal studies using gadolinium-DTPA
complex. Am J Roentgenol 142:625, 1984.
78. Grenier N, et al: Functional MRI of the kidney. Abdom Imaging 28:164-175, 2003. Medline Similar articles
79. Rajagopalan S, Prince MR: Magnetic resonance angiographic techniques for the diagnosis of arterial disease. Cardiol
Clin 20:501-512, 2002. Medline Similar articles
80. Kim RJ, Shah DJ, Judd RM: How we perform delayed enhancement imaging. J Cardiovasc Magn Reson 5:505-514, 2003.
81. Kirchin MA, Pirovano GP, Spinazzi A: Gadobenate dimeglumine (Gd-BOPTA). An overview. Invest Radiol 33:798-809,
82. Davies BE, et al: Pharmacokinetics and safety of gadobenate dimeglumine (multihance) in subjects with impaired liver
function. Invest Radiol 37:299-308, 2002. Medline Similar articles
3 of 4 14-04-2010 20:30
83. Kirchin MA, Pirovano GP, Spinazzi A: Gadobenate dimeglumine (Gd-BOPTA). Invest Radiol 33:798-809, 1998.
84. Davies BE, Kirchin MA, Bensel K, et al: Pharmacokinetics and safety of gadobenate dimeglumine (multihance) in subjects
with impaired liver function. Invest Radiol 37:299-308, 2002. Medline Similar articles
85. Mintorovitch J, Shamsi K: Eovist injection and Resovist injection: two new liver-specific contrast agents for MRI. Oncology
(Huntingt) 14:37-40, 2000.
86. Rofsky NM, Earls JP: Mangafodipir trisodium injection (Mn-DPDP). A contrast agent for abdominal MR imaging. Magn
Reson Imaging Clin N Am 4:73-85, 1996. Medline Similar articles
87. Clement O, Siauve N, Cuenod CA: Liver imaging with ferumoxides (Feridex): fundamentals, controversies, and practical
aspects. Top Magn Reson Imaging 9:167-182, 1998. Medline Similar articles
88. Wang YX, Hussain SM, Krestin GP: Superparamagnetic iron oxide contrast agents: physicochemical characteristics and
applications in MR imaging. Eur J Radiol 11:2319-2331, 2001.
89. Pascolo L, Cupelli F, Anelli PL, et al: Molecular mechanisms for the hepatic uptake of magnetic resonance imaging
contrast agents. Biochem Biophys Res Commun 257:746-752, 1999. Medline Similar articles
90. Schmidt PP, Toft KG, Skotland T, et al: Stability and transmetalation of the magnetic resonance contrast agent MnDPDP
measured by EPR. J Biol Inorg Chem 7:241-248, 2002. Medline Similar articles
91. Kellar KE, Fujii DK, Gunther WH, et al: NC100150 Injection, a preparation of optimized iron oxide nanoparticles for
positive-contrast MR angiography. J Magn Reson Imaging 11:488-494, 2000. Medline Similar articles
92. Taupitz M, Schnorr J, Abramjuk C, et al: New generation of monomer-stabilized very small superparamagnetic iron oxide
particles (VSOP) as contrast medium for MR angiography: preclinical results in rats and rabbits. J Magn Reson Imaging
93. de Haën C, Gozzini L: Soluble-type hepatobiliary contrast agents for MR imaging. J Magn Reson Imaging 3:179-186,
94. Clarke SE, Weinmann HJ, Dai E, et al: Comparison of two blood pool contrast agents for 0.5-T MR angiography:
experimental study in rabbits. Radiology 214:787-794, 2000.
95. Weissleder R, Mahmood U: Molecular imaging. Radiology 219:316-333, 2001. Medline Similar articles
96. Wunderbaldinger P, Josephson L, Weissleder R: Crosslinked iron oxides (CLIO): a new platform for the development of
targeted MR contrast agents. Acad Radiol 9:S304-S306, 2002.
97. Schellenberger EA, Hogemann D, Josephson L: Annexin V-CLIO: a nanoparticle for detecting apoptosis by MRI. Acad
Radiol 9:S310-S311, 2002.
98. Kang HW, Josephson L, Petrovsky A, et al: Magnetic resonance imaging of inducible E-selectin expression in human
endothelial cell culture. Bioconjug Chem 13:122-127, 2002. Medline Similar articles
99. Josephson L, Kircher MF, Mahmood U, et al: Near-infrared fluorescent nanoparticles as combined MR/optical imaging
probes. Bioconjug Chem 13:554-560, 2002. Medline Similar articles
100. Ichikawa T, Hogemann D, Saeki Y, et al: MRI of transgene expression: correlation to therapeutic gene expression.
Neoplasia 4:523-530, 2002. Medline Similar articles
101. Flacke S, Fischer S, Scott MJ, et al: Novel MRI contrast agent for molecular imaging of fibrin: implications for detecting
vulnerable plaques. Circulation 104:1280-1285, 2001. Medline Similar articles
102. Ward KM, Aletras AH, Balaban RS: A new class of contrast agents for MRI based on proton chemical exchange
dependent saturation transfer (CEST). J Magn Reson 143:79-87, 2000. Medline Similar articles
103. Zhang S, Merritt M, Woessner DE, et al: PARACEST agents: modulating MRI contrast via water proton exchange. Acc
Chem Res 36:783-790, 2003. Medline Similar articles
104. Moller HE, Chen XJ, Saam B, et al: MRI of the lungs using hyperpolarized noble gases. Magn Reson Med 47:1029-1051,
105. Salerno M, Altes TA, Mugler JP 3rd, et al: Hyperpolarized noble gas MR imaging of the lung: potential clinical
applications. Eur J Radiol 40:33-44, 2001.
106. Svensson J, Mansson S, Johansson E, et al: Hyperpolarized 13C MR angiography using trueFISP. Magn Reson Med
50:256-262, 2003.
107. Weinmann H-J, Schuhmann-Ginmpieri G, Schmitt-Willich H, et al: A new lipophilic gadolinium chelate as a tissue-specific
contrast medium for MRI. Magn Reson Med 22:233-237, 1991. Medline Similar articles
© 2010 Elsevier
4 of 4 14-04-2010 20:30
ISSUE PECIFIC ONTRAST GENTS

Peter Reimer
Thomas Helmberger
Wolfgang Schima
INTRODUCTION
The use of low molecular weight extracellular non-specific gadolinium chelates in magnetic resonance
1-3
imaging (MRI) is well established and a number of different contrast agents are clinically approved.
This class of contrast agents as well as the underlying mechanisms, which are also valid for tissue-
specific contrast agents, are described in Chapter 13 in detail. These non-specific contrast agents may
enhance both normal and diseased tissues, which led to the concept of developing contrast agents
with tissue specificity for improving the detection and characterization of disease. Contrast agents may
be directed to either normal tissue or diseased tissue. However, directing contrast agents to normal
tissue or normal structures is much easier than directing them specifically to diseased tissue such as
inflammation, degeneration, tumor or gene expression of disease. Currently, attempts to achieve
receptor, optical or gene specificity are propeling the development of a new field called molecular
imaging and optical imaging which is described in Chapter 15. Contrast agents for the GI tract are
covered in Chapter 13 since they are lumen-filling agents without tissue specificity.
This chapter will cover contrast agents with tissue specificity, focusing on those which are clinically
approved or are in advanced clinical trials. We will briefly discuss the different approaches taken and
highlight the three main applications of contrast agents directed to the reticuloendothelial system
(RES), the hepatobiliary system, and the vascular system. Each of these agents has unique properties
that offer advantages over unenhanced and nonspecific gadolinium chelate-enhanced MR imaging. Use
of these agents, however, requires an understanding of their current and potential clinical indications
and inherent limitations. The purpose of this chapter is to provide information on the properties, the
clinical development, and clinical applications of tissue-specific contrast agents which have been
approved or are likely to gain approval in the near future. In addition, experimental concepts and
4
preclinical concepts will be covered.
© 2010 Elsevier
1 of 1 14-04-2010 20:31
TISSUE SPECIFICITY
The principle of tissue specificity involves directing a contrast agent towards a certain tissue within the
body. Contrast agents with this specificity are either superparamagnetic iron oxides (SPIO) or
paramagnetic agents with gadolinium or manganese as the magnetic component.1,5 As when imaging
RES-containing organs such as the liver, spleen, lymph nodes or bone marrow, iron oxides of different
sizes are used. The larger the iron oxides, the higher and faster the uptake in liver and spleen and the
lower the uptake in lymph nodes and bone marrow. A higher uptake into the lymph nodes and the bone
marrow can be achieved by downsizing particles. In addition, smaller "ultrasmall" iron oxides (USPIO)
may also be utilized for vascular imaging based upon their prolonged circulating time and T1 effects as
compared to larger iron oxides. Furthermore, by binding specific markers to the magnetic components,
further tissue specificity may be achieved. This also applies to paramagnetic-based contrast agents,
which are directed to different tissues by means of a modification of the chemical structure resulting in
specific attachment to a particular tissue component such as hepatocytes or albumin.
page 377
page 378
Table 14-1. Contrast Agents for Specific MR Imaging

Acronym Generic Brand name Company Availability
Nonspecific* (ECS)
Gd-DTPA Gadopentetate Magnevist Schering AG, worldwide
dimeglumine Berlin,
Germany
Gd-DOTA Gadoterate Dotarem Guerbet, Europe,
meglumine Aulney- Australia
sous-Bois,
France
Gd-DTPA-BMA Gadodiamide Omniscan Amersham worldwide
Health,
London, UK
Gd-HP-DO3A Gadoteridol ProHance Bracco worldwide
Imaging SpA,
Milan, Italy
Gd-BOPTA Gadobenate MultiHance Bracco Europe
dimeglumine Imaging SpA,
Milan, Italy
Gd-DO3A-butrol Gadobutrol Gadovist Schering AG, Europe,
Berlin, Australia
Germany
Gd-DTPA- Gadoversetamide OptiMARK Mallinckrodt US
bis-methoxyethylamide Medical, St
Louis, USA
Hepatocyte specific
MnDPDP Mangafodipir Teslascan Nycomed worldwide
trisodium Amersham,
Oslo, Norway
Gd-BOPTA Gadobenate MultiHance Bracco Europe
dimeglumine Imaging SpA,
Milan, Italy
1 of 4 14-04-2010 20:33
Gd-EOB-DTPA Gadoxetate Primovist Schering AG, Europe

Berlin,
Germany
RES specific
AMI-25 Ferumoxides Endorem/Feridex Guerbet, worldwide
Aulney-
sous-Bois,
France
SH U 555 A Ferucarbotran Resovist Schering AG, Europe,
Berlin, Japan
Germany experimental
SBPA Bracco
Research
Geneva,
Switzerland
Blood pool specific
AMI-227/Code-7228 Combidex Advanced pending
Sinerem Magnetics,
Cambridge,
MA; Cytogen,
Princeton NJ,
USA Guerbet,
Aulney-
sous-Bois,
France
SH U 555 C Ferucarbotran Supravist Schering AG, phase 3
Berlin,
Germany
finished
MS-325 to be determined Schering AG pending
Berlin
Germany/Epix,
Cambridge,
MA, USA
NC-100150 Feruglose Clariscan Nycomed terminated
Amersham
Health,
Lomdon, UK
SH L 643 A Gadomer-17 Schering AG, phase 2/3
Berlin,
Germany
B- 22956/1 Gadocoletic acid Bracco phase 2/3
trisodium Imaging SpA,
Milan, Italy
P-792 Vistarem Guerbet, phase 2/3
Aulney-
sous-Bois,
France
under way
*for abdominal imaging
2 of 4 14-04-2010 20:33
The most thoroughly investigated and widely approved tissue-specific contrast agents relate to hepatic
MRI. A variety of liver contrast agents have been developed for contrast-enhanced MRI of the liver,
which are designed to overcome the limitations of nonspecific tissue uptake by extracellular low
molecular weight gadolinium chelates. 1,3 The two main classes of liver-specific contrast agents are
6-30
superparamagnetic iron oxides (SPIO), with uptake via the RES mainly into the liver and spleen,
and hepatobiliary contrast agents with uptake into hepatocytes followed by variable biliary
excretion.31-56 Two hepatobiliary contrast agents, mangafodipir trisodium (Teslascan, Amersham
Health, Oslo, Norway) and gadobenate dimeglumine (MultiHance, Bracco Imaging SpA, Milan, Italy),
are already clinically approved in many countries (Table 14-1). A third hepatobiliary contrast agent,
Gd-EOB-DTPA (Primovist, Schering AG, Berlin, Germany), is likely to get approval shortly.
These agents exhibit different features for the detection and characterization of liver tumors.
Enhancement during the distribution phase of contrast agents mainly depends on tumor vascularity
(hypovascular versus hypervascular) and its blood supply while enhancement on delayed images is
characterized by the cellular specificity of MR contrast agents (extracellular versus intracellular-
hepatocyte phase or accumulation phase). Therefore, enhancement characteristics of hepatobiliary
contrast agents are applicable to the diagnosis of primary hepatocellular liver tumors. It has been
predicted that these intracellular agents may enable grading of hepatocellular carcinomas
(differentiated versus undifferentiated) because of their hepatocyte-specific uptake with active uptake
into differentiated carcinoma cells and delayed elimination.44,57
Another group of cell-specific contrast agents is the ultrasmall superparamagnetic iron oxides (USPIO).
Historically, USPIO have been prepared first by ultrafiltration of AMI-25. USPIO particles have a
stronger affinity for bone marrow and lymph nodes than SPIO. Plasma relaxation time measurements
13,58
demonstrated a persistent, dose-dependent decrease in both T1 and T2. Since USPIO exhibit a
longer plasma circulation time and stronger T1 shortening characteristics than SPIO, these compounds
can also be used as blood pool agents with angiographic effects.59 Newer USPIO formulations such as
AMI-227 (Advanced Magnetics, Cambridge, MA) have been prepared by direct synthesis and are
currently in phase 3 clinical trials for contrast-enhanced MRI of the bone marrow and lymph nodes.
60-63
Physical properties (Table 14-2) and results of clinical trials results have been described in detail.
USPIO particles might be a valuable diagnostic tool for both vascular and tissue-specific imaging. In
contrast to SPIO particles, USPIO particles present an increased blood pool circulation with a blood
half-life of over 200 minutes. The blood clearance finally occurs via macrophages of the liver, spleen,
lymph nodes, and bone marrow. Independent of the administered dose, chest pain, dyspnea, and rash
are reported as adverse events in less than 5%.64,65
page 378
page 379
Table 14-2. Physicochemical Properties of Different Contrast Agents

Relaxivity
(s/mM)
Dissociation
Thermodynamic factor Osmolality Viscosity
-1
Compound stability (LOG Keq) (k[obs]s ) (osmol/kg) (mPas) R1 R2
Gd-DTPA 22.1 -3 1.96 2.9 4.1 4.6
1.2 × 10
Gd-DOTA 25.8 -5 1.35 2.0 3.4 4.27
2.1 × 10
Gd-DTPA-BMA 16.9 -2 0.78 1.9 3.9 4.8
> 2 × 10
Gd-HP-DO3A 23.8 6.3 × 10-5 0.63 1.3 5.3 6.6
Gd-DO3A-butrol 6 × 10-3 1.603 4.96 5.2 6.1
Gd-DTPA-BMEA 1100 2.0 4.7 5.2
3 of 4 14-04-2010 20:33
MnDPDP 15.1 290 2.0 1.9 2.2

Gd-BOPTA 22.6 1.97 5.3 9.7 12.5
Gd-EOB-DTPA 6.9 8.7
Ferumoxides 338 1.2 4.5 33
Ferucarbotran 319 1.03 7.4 95
NSR 0430 35 239
SBPA 12.8 468
AMI-227 24 53
MS-325 >40
NC-100150 19.5 36
Gadomer-17 16 19
P-792 25 48.5
We are grateful to B Bonemain, Ch DeHaen, W Ebert, M Rohrer, and T Skotland for providing us with
data unavailable in the current literature (Helmberger 2001; Semelka 2001; Rohrer 2004). In general,
measurements may be performed at different field strengths, temperatures, and within different
solutions.
After initial reports on USPIO as a compound with a potential specificity for the lymphatic system,
hepatic and vascular imaging also entered the focus of USPIO imaging. 64,66-68 After USPIO
administration the signal loss of the hepatic parenchyma on T2- and T1-weighted imaging is
comparable to SPIO-enhanced imaging, while a positive vascular enhancement on T1-weighted
imaging can be appreciated even with Gd chelate known specific contrast effects such as ring
enhancement in malignant lesions.14,22,64,65 Nevertheless, USPIO particles are not yet approved for
hepatic MRI.
The heterogeneous group of "blood-specific" paramagnetic blood pool contrast agents is characterized
by a prolonged circulation time within the blood pool.69 Typical representatives of this type of contrast
agents are low molecular weight protein-binding Gd complexes (e.g., MS-325) and large polymeric Gd
63,70-73
complexes (e.g., Gadomer-17, Schering; P-792, Guerbet). The strong binding to albumin makes
some of these Gd complexes stay within the blood pool for a prolonged time in comparison to "classic"
Gd chelates. The compounds demonstrated an excellent enhancement of the vascular space in clinical
trials which makes them suitable for MR angiography and perfusion studies.64 These compounds may
also enable new applications such as monitoring of tumor therapy or monitoring during or after
74-77
intravascular interventions.
© 2010 Elsevier
4 of 4 14-04-2010 20:33
EXPERIMENTAL CONCEPTS
Target-Specific Concepts
In general, target-specific contrast agents consist of two components: a magnetic label capable of
altering the signal intensity on MR images and a target-specific carrier molecule having a characteristic
affinity for a specific type of cell, a specific binding site or both. Such agents show preferential
accumulation over time in the organ containing the target cell or binding site. In order to cause a
measurable change in signal intensity, the density of local binding sites and the relaxivity of the agent
have to be high enough. Model agents for a variety of applications like the liver78-91 and spleen,92-94 the
95 96-98 99,100 101-103
adrenal glands, the pancreas, myocardium, the nervous system, and many others
have been studied. Here, we elaborate on some concepts which appear promising for pertinent clinical
applications.
Lymph Nodes
Imaging of lymph nodes has mainly been explored with iron oxides58,104,105 and to a lesser degree with
106
T1 agents. Iron oxides used for liver imaging do not substantially enhance normal lymph nodes
following intravenous injection within the dose range utilized for liver imaging. Subsequently, interstitial
applications were studied showing decreased signal intensity of normal lymph nodes, whereas
metastatic lymph nodes showed no significant change in signal intensity.107 Smaller iron oxides
(USPIO) were evaluated as an intravenous contrast agent for lymph nodes. MR imaging of animal
model of nodal metastases confirmed the hypothesis that intravenously administered USPIO decrease
58,108-110
signal intensity of normal but not metastatic nodes. Following this original observation, various
USPIO have been developed and tested in animal models following intravenous injection in order to
enhance all lymph nodes.105,111-114 Clinical trials (Fig. 14-1) have been under way for more than a
67,115-133
decade but clinical approval is still uncertain. Initial results appear quite promising for
distinguishing normal from tumor-infiltrated nodes. For instance, in a recent study high-resolution MRI
with magnetic nanoparticles allowed the detection of small and otherwise undetectable lymph-node
metastases in patients with prostate cancer.428
page 379
page 380
Add to lightbox
Figure 14-1 Normal and metastatic lymph node with USPIO. Pelvic MRI with spoiled GRE is shown
before (left) and following (right) IV injection of AMI-227 in a cancer patient. The post-contrast image
(right) shows a normal lymph node with uptake of USPIO leading to a hypointense lymph node (dark
arrow) in the right pelvis and a metastatic lymph node in the left pelvis without uptake or signal change
compared to precontrast (white arrow). (Images courtesy of S Saini, MGH, Boston, MA.)
More recently, blood pool agents have also been experimentally tested for interstitial
lymphography.134-136 A new approach describes magnetic nanoparticle-based MR contrast agents that
have a near-infrared fluorescence (NIRF) that is activated by certain enzymes. The probes are
prepared by conjugation of arginyl peptides to cross-linked iron oxide amine (amino-CLIO), either by a
disulfide or a thioether linkage, followed by the attachment of the indocyanine dye Cy5.5. The NIRF of
1 of 49 14-04-2010 20:37
disulfide-linked conjugates are activated by DTT (dithiothreitol), while the NIRF of thioether-linked
conjugates is activated by trypsin. Fluorescent quenching of the attached fluorochrome occurs in part
due to the interaction with iron oxide, as evidenced by the activation of fluorescence with DTT when
nanoparticles that have less than one dye attached per particle. With a subcutaneous injection of the
probe, axillary and brachial lymph nodes were darkened on MR images and easily delineated by NIRF
imaging. The probes may provide the basis for a new class of so-called smart nanoparticles, capable
of pinpointing their position through their magnetic properties, while providing information on their
environment by optical imaging techniques. 137,138
Bone Marrow
Uptake of USPIO into macrophages within bone marrow was shown by electron microscopy with
transmigration of the capillary wall by means of vesicular transport and through interendothelial
junctions. This was achieved by a size decrease of the particles subsequently prolonging the circulation
time after intravenous administration: 3.6% of the injected dose per gram of tissue was found in lymph
13
nodes, 2.9% per gram in bone marrow, 6.3% per gram in liver, and 7.1% per gram in spleen. The
experimental evaluation in an animal model of an intramedullary tumor demonstrated the potential of
USPIO to enable differentiation between tumor and normal red marrow. USPIO-enhanced MR imaging
improves the detection of smaller tumors and allows differentiation of tumor deposits from islands of
hyperplastic or normal red marrow.139
Subsequently, patients with cancer of the hematopoietic system were studied to determine the
differentiation of normal, hypercellular, and neoplastic bone marrow based on its MR enhancement
after intravenous administration of superparamagnetic iron oxide. In this particular study, 18 patients
with cancer of the hematopoietic system underwent MRI of the spine before and after infusion of
ferumoxides and ferumoxtran. Changes in bone marrow signal intensity after iron oxide administration
were more pronounced on STIR images as compared with T1- and T2-weighted TSE images. The
STIR images showed a strong signal decline of normal and hypercellular marrow 45-60 minutes after
iron oxide infusion with only a minor signal decline of neoplastic bone marrow lesions.
Superparamagnetic iron oxides are taken up by normal and hypercellular reconverted bone marrow but
not by neoplastic bone marrow lesions. Therefore, superparamagnetic iron oxides may be useful to
differentiate normal and neoplastic bone marrow.140
Cell Labeling
Mechanisms of cell uptake of contrast agents and modes of intracellular trafficking were investigated
75,138,141-143
by different groups and with different strategies. Dextran-coated monocrystalline iron oxide
modified with rhodamine as a fluorescent label and opsonized with albumin (RMA) was exposed to a
phagocytic C6 cell line and murine bone marrow macrophages. Immediately after cell contact, RMA
localized to the lysosomal compartment and at long time points remained in vesicles that by
morphology and distribution appeared to be terminal lysosomes. Iron oxides therefore demonstrated
metabolism via the lysosomal pathway. The mechanism of cellular uptake of a prototypical opsonized
iron oxide label was consistent with receptor-mediated endocytosis. 144
It has been shown that mammalian and stem cells may be labeled by combining commercially available
transfection agents (TAs) with SPIO. When transfected ferumoxides or monocrystalline iron oxide
nanocompound (MION)-46L were used, intracytoplasmic particles stained with Prussian blue were
detected for all cell lines with a labeling efficiency of nearly 100%. Limited or no uptake was observed
for cells incubated with ferumoxides or MION-46L alone. Cell viability was not affected by endosomal
incorporation of SPIO nanoparticles.145
Human hematopoietic progenitor cells were also labeled with ferumoxides , ferumoxtran, magnetic
polysaccharide nanoparticles-transferrin, P7228 liposomes, and gadopentetate dimeglumine
liposomes. For all contrast agents, intracellular cytoplasmic uptake was demonstrated.146
Necrosis
2 of 49 14-04-2010 20:37
page 380
page 381
Metalloporphyrins were initially described as tumor-specific agents particularly within the liver.
Schmiedl et al compared manganese (III) mesoporphyrin (Mn-mesoporphyrin) and manganese
tetrakis-(4 sulfonatophenyl) porphyrin (Mn-TPPS4) for their hepatic MRI properties. Liver abscesses
and tumors were induced in rats. Mn-mesoporphyrin (0.035 mmol/kg) caused significant enhancement
of normal liver parenchyma and increased the lesion-to-liver contrast in both the models of hepatic liver
abscess and metastatic liver disease. Mn-TTPS4 (0.04 mmol/kg) typically enhanced both lesion and
normal liver parenchyma and therefore did not improve the lesion-to-liver contrast.147-149
Ni et al investigated the tumor specificity of gadolinium mesoporphyrin (Gd-MP) and manganese

tetraphenylporphyrin (Mn-TPP). In their experiments, both metalloporphyrins initially behaved as
nonspecific agents, similar to gadopentetate dimeglumine , and enhanced the tumor by perfusion and
diffusion. However, metalloporphyrins, but not gadopentetate dimeglumine , caused a delayed (= 3 h)
enhancement in some compartments of certain lesions such as necrosis, cystic formation or
thrombosis. Metalloporphyrins did not prove to be tumor specific. However, their observed affinity for
150
non-viable tissue has elicited other potential applications for these agents.
Hoffman and colleagues investigated the molecular mechanism by which metalloporphyrins such as
gadophrin-2 bind to necrosis. Within a given tumor, the agent preferentially localized in the periphery of
necrotic areas. Within these regions gadophrin-2 was bound to interstitial albumin and no other
proteins, lipids or DNA. It was speculated that tumoral accumulation of gadophrin-2 occurs through its
151
binding to plasma albumin and subsequent slow extravasation into the tumor interstitium. This
152
hypothesis was not confirmed for all metalloporphyrins. Nevertheless, metalloporphyrins attracted
99,153-164
research interest for cardiac imaging by directly contrasting infarcts. This concept has been
replaced by the late enhancement approach by means of low molecular weight gadolinium
chelates.158,165-177
Tumor
Several studies have investigated the accumulation and cellular uptake of different contrast agents into
tumor tissue in order to develop a model for vector delivery in malignant tumors. Dextran-coated iron
oxide preparations have been shown to accumulate in macrophages and tumor cells. To explore
nonspecific and specific mechanisms, USPIO such as MION particles were labeled with fluorescein
isothiocyanate or radio-iodinated and purified by gel permeation chromatography. MION and plasma-
opsonized MION were used and opsonization resulted in C3, vitronectin, and fibronectin association
with MION. Incubation of cells with fluorescent MION showed active uptake of particles in
macrophages both before and after opsonization. In C6 tumor cells, however, intracellular MION was
125
only detectable in dividing cells. Quantitatively, I-labeled MION was internalized into cells.
Opsonization increased MION uptake into macrophages sixfold, whereas it increased the uptake in C6
tumor cells only twofold. Results from uptake inhibition assays suggested that cellular uptake of
non-opsonized (dextran-coated) MION particles is mediated by fluid-phase endocytosis, whereas
receptor-mediated endocytosis is presumably responsible for the uptake of opsonized (protein-coated)
particles.178 Uptake into tumor cells and tumor-associated macrophages was confirmed for different
179
cells lines. Uptake into tumor cells appears to be proportional to cellular proliferation rates.
A human study was performed with ferumoxides and ferumoxtran in a small number of patients with
intracranial tumors. No significant T1 or T2 signal intensity changes were seen after ferumoxide
administration at either examination time. Fifteen of 17 patients given ferumoxtran had T1 and/or T2
shortening consistent with iron penetration into the tumor. The histologic examination revealed minimal
iron staining of the tumor with strong staining at the periphery of the tumors. Histologic examination
showed cellular uptake primarily by parenchymal cells at the tumor margin.180 Tumor targeting may
also be achieved via surface receptors. Since some tumors of epithelial origin express the high-affinity
folate receptor, a folate-conjugated dendrimer polychelate was investigated showing accumulation in
tumors expressing hFR.181 Furthermore, different studies demonstrated that macromolecular contrast
3 of 49 14-04-2010 20:37
agents may be useful to assess vascular permeability and tumor capillary permeability. 182-186,417,418
Kresse and colleagues coupled human transferrin covalently to USPIO particles to target the
transferrin receptor, which is overexpressed in many tumors. The MR evaluation of tumor signal
intensity over time showed a 40% signal reduction 150 minutes after injection, with the reduction
persisting for at least 8 hours. Control experiments using the parent USPIO compound or USPIO
labeled with a nonspecific human serum albumin (HSA-USPIO) showed a change of only 10% in tumor
signal intensity over time. The results demonstrated that a combination of the USPIO relaxivity
properties with the specificity of transferrin-mediated endocytosis allows the in vivo detection of tumors
by MR imaging.187 The receptor was genetically modified in subsequent experiments, providing proof
of the principle that imaging of gene expression is feasible by MR imaging. 188-192 A more recent
approach mimics FDG-PET.193,194
Vascular
page 381
page 382
Contrast agents that reduce the T1 relaxivity of blood efficiently increase intravascular signal. The use
of intravenous low and high molecular gadolinium chelates for contrast-enhanced magnetic resonance
angiography (MRA) is limited by the rapid equilibration of these agents between the intravascular and
extravascular, extracellular compartments. Whether this actually leads to relevant disadvantages
remains to be proven. In theory, MR contrast agents confined to the intravascular space, so-called
blood pool agents, may change the way vessels are currently imaged by means of MRA. Furthermore,
if these agents also exhibit a prolonged plasma half-life, additional applications within the field of MRA
but also beyond MRA may open new avenues. Blood pool agents are particularly promising in
contrasting smaller vessels, vessels with slow flow, and vessels with complex flow. In addition, they
may be used for perfusion imaging, functional imaging or tumor imaging such as the demonstration of
angiogenesis.195
Several classes of blood pool agents are under development: paramagnetic gadolinium attached to
large molecules (macromolecules), ultrasmall iron oxides, gadolinium-based molecules with reversible
protein binding enhancing T1 relaxivity, and gadolinium-based synthetic molecules/polymers. 59,66,196-206
Whereas the design of iron oxides suited for blood pool imaging is well understood and developed, a
variety of gadolinium-based agents with different pharmacologic profiles is under development (see
Tables 14-1 and 14-2).
The initial development of blood pool agents focused on macromolecules with high relaxivity based on
the slow rotation of these molecules in blood. Different molecules were evaluated, including albumin,
dextran, polylysine, and polymers.207 Concerns linger about the potential immunogenicity and excretion
of these agents, particularly after repeated injections.208 Gadobenate dimeglumine also exhibits some
weak protein binding, which increases intravascular signal on contrast-enhanced MRA.209,210 Some of
the agents with synthetic molecules/polymers exhibiting different blood half-lives are under
experimental or clinical development. The focus of these concepts has already gone beyond pure
MRA, approaching new applications such as cardiac MRA, organ perfusion, tumor assessment or
functional imaging.63,77,135,199,211-215
Plaque
Conventional imaging techniques such as X-ray angiography show the arterial lumen but do a poor job
of characterizing the vessel wall, including the severity and composition of atherosclerotic plaque.
Magnetic resonance imaging offers the noninvasive ability to characterize plaque in vivo.429 The
"vulnerable plaque" is of particular interest because such lesions may be at great risk for sudden
rupture and vessel occlusion. Potential clinical applications of plaque imaging include risk stratification
and helping to design therapies to prevent stroke and acute coronary syndromes. Imaging of plaque
has traditionally been done using multi-spectural pulse sequences without contrast enhancement.
4 of 49 14-04-2010 20:37
Standard gadolinium chelates diffusely enhance the aortic wall but the enhancement patterns may not
be specific for normal wall versus atherosclerotic involvement. However, tissue-specific contrast
agents may have a role to play in the identification and characterization of plaque. For instance,
Gadoflourine (Schering AG, Berlin, Germany) is a lipophilic, macrocyclic (1528 Da), water-soluble,
gadolinium chelate complex (Gd-DO3A derivative) with a perflourinated side chain. It has high
relaxivity, long plasma half-life, water solubility, and lipophilicity compared with Gd-DTPA.430 The
mechanism of plaque enhancement is uncertain, perhaps relating to enhanced endovascular
permeability or an increase in the vasa vasorum feeding plaque neovasculature. In animal studies,
Gadofluorine enhances the imaging of atherosclerotic plaques and enables improved plaque detection
431
of even nonstenotic lesions that are not visible on unenhanced MRL. Ultrasmall iron oxide particles,
which are taken up by macrophages associated with inflammation, cause susceptibility-based signal
loss within ruptured and ruture-prone plaques. 432,433
Inflammation
Imaging of inflammation has been studied by plain iron oxide particles with nonspecific uptake by
216-218 219,220
macrophages within areas of inflammation and targeted contrast agents. Human polyclonal
immunoglobulin G (IgG) was attached to a MION. In an animal model of myositis, MION-IgG caused
reduced signal intensity at the site of inflammation. No change in signal intensity existed after an
injection of unlabeled MION. Site-specific localization of MION-IgG was corroborated with scintigraphic
imaging by indium-111 IgG and MION-111In-IgG and was confirmed histologically with iron staining.
219
These results indicate that inflammation-specific antibody MRI is feasible in vivo.
Alternatively, it has been demonstrated by Gupta et al that a nontargeted, long-circulating, synthetic

polymer accumulates in areas of inflammation, with high capillary permeability and increased regional
blood flow. Methoxy poly(ethylene glycol)-poly-L-lysine (PL)-diethylenetriaminepenta-acetic acid
(MPEG-PL-DTPA) was labeled with technetium-99m for scintigraphy and with gadolinium for MRI.
99m
Tc-labeled MPEG-PL-DTPA demonstrated nearly eightfold higher accumulation in Escherichia
coli-infected muscle when compared with normal muscle. Scintigrams and MR images showed areas
of inflammation with peak accumulation at 24 hours after injection of 99mTc- or gadolinium-labeled
221
MPEG-PL-DTPA.
Tissue-Specific Contrast Agents with Completed Clinical Trials

Blood Pool Agents-Vascular
With the advent of rapid three-dimensional imaging sequences combined with existing extracellular
gadolinium-based contrast agents, MRA has shown promise to become a time-efficient and
cost-effective tool for the complete assessment of many vascular regions or clinical referrals such as
peripheral vascular disease.71,419-427 Alternative paramagnetic71,205,222-224 and
59,66,225-232
superparamagnetic agents for enhancing MR angiographic images, known as blood pool
agents, have been designed and some are under clinical investigation. There are two blood pool
agents with completed clinical trials. The most advanced agent with a filed FDA application is
gadofosveset trisodium (EPIX Medical, Cambridge, MA; Schering, Berlin, Germany), formerly
identified with the code name MS-325.72,73 The second agent, SH U 555 C (Schering, Berlin,
226,229,233-235
Germany), an ultrasmall iron oxide, is derived from Ferucarbotran.
page 382
page 383
Gadofosveset Trisodium
Gadofosveset is a gadolinium-based small molecule (molecular weight, 975.88) contrast agent
designed specifically for MR angiography.4,63,71-73,76,136,201,224,236-271 Gadofosveset is noncovalently
bound to albumin (80%-96%) in human plasma and is primarily excreted renally. 72 This reversible
albumin binding of gadofosveset enhances the paramagnetic effectiveness of gadolinium and allows
lower contrast agent doses than are needed with conventional MR agents.72 In plasma, gadofosveset
5 of 49 14-04-2010 20:37
exhibits a relaxivity at 0.5 T that is approximately 6-10 times that of gadopentetate dimeglumine .72
Gadofosveset trisodium has been studied in different vascular territories (Fig. 14-2) and for various
applications.4,63,71,72,76,224,271 Recently, data from a phase 2 trial looking at peripheral vascular disease
were published.73 Within this study, the dose response and safety of gadofosveset-enhanced MRA
compared with nonenhanced 2D time-of-flight MRA, with X-ray angiography as the standard of
reference, were evaluated. The study was designed as a double-blind, multicenter, placebo-controlled
trial and patients (n = 238) were randomly assigned to receive a single intravenous dose of one of the
following: placebo or 0.005, 0.01, 0.03, 0.05, or 0.07 mmol gadofosveset/kg bodyweight. The
evaluation focused on aortoiliac arterial disease in patients who either received a diagnosis of the
disease or were suspected of having it on the basis of the physical examination results and medical
history. All readers revealed increased sensitivity with gadofosveset-enhanced MRA compared with
nonenhanced MRA and gadofosveset-enhanced MRA versus X-ray angiography for detection of
stenosis of 50% or greater in the aortoiliac region.
Add to lightbox
6 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-2 MS-325 enhanced MRA (0.03 mmol/kg gadofosveset/kg bw) of foot during an early (A)
and later blood pool phase (B) shows the capability to image small vessels with high vessel/tissue
contrast in this patient with PAOD and multiple stenoses of the lateral plantar artery.
This study demonstrated that improvement in diagnostic effectiveness is dose dependent and that 0.03
mmol/kg is the minimally effective dose of gadofosveset for detection of aortoiliac occlusive disease.
The side-effect profile was clinically acceptable. The binding to serum albumin enhances the
paramagnetic effectiveness of gadolinium and allows lower doses than are required with conventional
MR imaging contrast agents. Protein binding also increases the intravascular residence time of the
71
contrast agent. The result is extended imaging time, higher spatial resolution, and greater anatomic
coverage. The route of clearance of gadofosveset is renal excretion, which is a desired route for MR
imaging contrast agents.73
Similar results were obtained for the carotid arteries.256 Within this phase 2 trial 50 carotid arteries in
26 patients were imaged with 3D-spoiled GRE MRA at 5 and 50 minutes after injection of
gadofosveset at doses of 0.01, 0.03 or 0.05 mmol/kg bodyweight. Again, conventional contrast
catheter angiography was used as the standard of reference. Overall accuracy for MS-325-enhanced
carotid MRA performed during steady-state conditions approximately 5 minutes after injection was high
256
(88%-100%) at 0.03 and 0.01 mmol/kg as determined by blinded reading.
Ferucarbotran C
page 383
page 384
SH U 555 C (Supravist, Schering AG, Berlin, Germany) as an optimized bolus-injectable formulation of

Ferucarbotran has been proposed for equilibrium-phase MRA following encouraging results in animal
studies.226,272 Allkemper et al reported an excellent T1 effect for various Ferucarbotran formulations of
different overall particle size and best results for a formulation with a mean particle size of 21 nm. A
prolonged signal enhancement over time with increasing doses up to 40 μmol Fe/kg bodyweight was
demonstrated.
SH U 555 C is a sterile, bolus-injectable, ready-to-use formulation, provided in a concentration of 0.5

mmol Fe/mL. Electron microscopy and X-ray diffraction studies showed a mean core particle size of
about 3-5 nm and dynamic laser light scattering (DLS) a mean hydrodynamic diameter of about 20 nm
in an aqueous environment. Relaxivity measurements yield a R1 of 22 s/mM and a R2 of 45 s/mM at
7 of 49 14-04-2010 20:37
40° and 20 MHz in water.273
Within phase 1 studies healthy volunteers274 and elderly volunteers with risk factors for arterial
vascular disease were studied.234 Placebo-controlled, double-blind studies at doses of 5, 10, 20, 40,
and 80 μmol Fe/kg bodyweight were conducted. The injection rate of SH U 555 C was 0.5 mL/s
followed by 20 mL saline flush (0.9%) at a flow rate of 3.0 mL/s using an automatic bolus injector.
Following first-pass scans with an estimated delay, serial 3D MRAs of different vascular regions were
performed followed by serial 3D-MRA data sets every 6 minutes up to 42 minutes.
Add to lightbox
Add to lightbox
Figure 14-3 SH U 555 C enhanced MRA (40 μmol Fe/kg bw) of the thigh and knee during first-pass
(A) and blood pool phase (B) shows the capability to image both first pass upon bolus injection and
the blood pool phase.
The lowest effective dose of 40 μmol Fe/kg bodyweight for first-pass and equilibrium-phase MRA was
determined based upon qualitative and quantitative analysis (Fig. 14-3). Cardiac perfusion studies in a
limited number of patients at a dose of 40 μmol Fe/kg bodyweight demonstrated a significant
enhancement compared to baseline within the right ventricle, the left ventricle, and the left ventricular
myocardium. Signal changes within the myocardium of the left ventricle showed an initial increase
during first pass followed by a small decrease and subsequent equilibrium. The IV bolus injection of SH
U 555 C was well tolerated by all volunteers. No relevant changes in vital signs (blood pressure, heart
8 of 49 14-04-2010 20:37
rate) occurred during the observation period.274 More recently, phase 3 trials in patients with
peripheral arterial disease and renal vascular disease were completed. Catheter angiography was
obtained for comparison and more than 200 patients were enrolled in each study. SH U 555
C-enhanced MRA provides a dose-dependent first-pass and equilibrium-phase effect without relevant
cardiovascular side-effects. Furthermore, cardiac perfusion studies using SH U 555 C are feasible,
274
demonstrating the significant potential for organ perfusion studies.
Tissue-Specific Contrast Agents with Clinical Approval

Superparamagnetic Iron Oxides
A variety of parenterally administered iron oxides have been developed for contrast-enhanced MR
imaging of the liver and spleen.275,276 The SPIO agents efficiently accumulate in liver, with
approximately 80% of the injected dose, and spleen, with 5-10% of the injected dose, within minutes
19,275
after administration. Following sequestration by phagocytic cells, these agents mainly decrease
liver and spleen signal intensity within several minutes. Malignant tumors are typically devoid of a
substantial number of phagocytic cells so they appear as hyperintense/bright lesions contrasted
against the hypointense/dark liver on T2-weighted sequences.277 Tumors with a substantial number of
phagocytic cells, such as focal nodular hyperplasia, hepatocellular adenoma, well-differentiated
hepatocellular carcinoma, and/or a significant blood pool (hemangioma or hypervascular lesions) may
show sufficient uptake of SPIO with decreasing signal intensity on T2-weighted sequences. The signal
278,279
decrease is related to the Kupffer cell activity or tumor vascularity. Furthermore, SPIO show
signal changes in T1-weighted sequences, both during the perfusion phase (Ferucarbotran) and
accumulation phase (Feridex and Ferucarbotran), providing additional information.
Two different classes of iron oxides are currently clinically approved or in phase 3 trials (see Table
14-1): SPIO with a high R2/R1 relaxivity ratio and short blood half-life (<10 min) and USPIO with a
lower R2/R1 relaxivity ratio and longer blood half-life (<2 hours).13,58,130,280 The higher the R2/R1
relaxivity ratio, the higher the T2 effect and signal decrease on T2-weighted images (see Tables 14-1
and 14-2).
page 384
page 385
Clinical trials with intravenously administered Feridex (see Table 14-2) were initiated as early as
1987.18,281 Early studies were performed with bolus injections of relatively high doses (= 50 μmol
Fe/kg bodyweight) of an initial formulation of AMI-25, causing significant hypotension. Subsequently,
the agent was reformulated to achieve iso-osmolarity and is currently administered by drip infusion
over a period of 30 minutes since side-effects (facial flush, dyspnea, rash, lumbar pain) depend on the
18,275,281
dose rate. AMI-25 is a relatively safe drug if drip infused in glucose or saline at a dose of
10-15 μmol Fe/kg bodyweight over 30 minutes with a flow rate of approximately 3 mL/min. 9,282 Clinical
283-286
trials with Feridex have mainly emphasized lesion conspicuity, but the characterization of focal
liver lesions is equally important in order to provide a comprehensive clinical work-up. 8,284,287 Dynamic
MR imaging with extracellular gadolinium chelates for characterization of focal liver lesions is based on
intravenous bolus injections and fast MRI techniques. Comparable bolus injections with SPIO have not
been feasible with prototype compounds such as AMI-25 or AMI-227 because of cardiovascular
18,275,281
side-effects and acute lumbar pain. More recently, the new SPIO formulation Ferucarbotran
did not show side-effects following rapid intravenous injection.7
The second SPIO preparation available in many countries is Ferucarbotran (SH U 555 A, Schering AG,
Germany) consisting of SPIO microparticles coated with carboxydextran (see Tables 14-1 and 14-2).7
Physical properties, safety data, and results of clinical trials have been described in detail. 19,288 In
phase 3 studies, preloading the contrast agent into a connecting IV line (8-12 μmol Fe/kg bodyweight;
0.9-1.4 mL) and flushing the V line with 10 mL saline within 3 s showed no cardiovascular
side-effects.7 The R1 relaxivity of SH U 555 A varies considerably within the range of diagnostically
applied field strengths. However, R1 relaxivity of 12.3 s/mM at 40 MHz is still four times higher than R1
9 of 49 14-04-2010 20:37
relaxivity of low molecular gadolinium chelates such as gadopentetate dimeglumine . The R2 relaxivity
of SH U 555 A is stable within the range of diagnostically applied field strengths with 188 s/mM at 40
MHz. The very high R2/R1 ratio is characteristic for superparamagnetic colloids of the SPIO type. This
ratio varies from 6 to 15 as the Larmor frequency increases from 10 MHz to 40 MHz, a fact which
might favor the use of lower imaging fields to better utilize the T1 effect of these materials. Phantom
experiments with SH U 555 A in human plasma demonstrated positive enhancement at low
concentrations <700 μmol Fe/L of iron oxide with T1-weighted pulse sequences. Enhancement
increased with the degree of T1-weighting and shifted towards higher concentrations with shorter echo
times. The T1 effect of SH U 555 A is a function of dose and concentration in plasma and can be
monitored during a time window at which the plasma concentration stays below a certain level
depending on pulse sequence parameters.289
Similar to Gd chelate-enhanced MRI, heavily T1-weighted GRE images will display a positive
enhancement in vessels and in hypervascularized lesions after administration of SPIO or USPIO
particles. This positive T1 effect is caused by the ultrasmall subfraction within SPIO preparations
containing iron oxide particles with a hydrodynamic diameter below 50 nm (USPIO).14,22 Dependent on
their concentration, these smaller particles are reducing the T1 relaxivity similar to Gd chelates.
Additionally, they are recognized by the macrophages significantly later than the bigger particles. The
resulting increased circulation time leads to a prolonged T1 and T2* enhancement which allows for
blood pool imaging.289 Incorporating this effect into the concept of imaging and image interpretation,
SPIO particles provide the potential to assess both "static" tissue characterization ("Kupffer cell
storage phase") and vascularization. This is especially true if particles suitable for rapid injection are
used, as demonstrated by Reimer et al in dynamic imaging after bolus injection of SPIO particles
(Ferucarbotran).290 This parallel presence of T1, T2 and T2* contrast effects may address more
pathophysiologic processes than Gd chelates, but image interpretation might become more complex.
Recently presented experimental data could prove that this problem may be overcome by a new
superparamagnetic blood pool agent without a positive T1 signal enhancement (SPBA, Fe3O4 coated
with phospholipids and the nonionic surfactant, ethylene oxide-propylene oxide block co-polymer,
Bracco Research, Switzerland).291
Side-Effects of SPIO Particles

The administered free iron is incorporated into the normal iron metabolism but the administered dosage
of SPIO does not significantly affect the body's iron pool (e.g., 15 μmol Fe/kg ~ 840 ng
Fe/bodyweight). Before the degradation of the clustered iron particles, the superparamagnetic effects
last for several days, providing an imaging window of several hours up to several days after
administration. Minor adverse events have been described, among which back pain is the most
common (5%-30%). Back pain mostly occurs when the SPIO particles (especially ferumoxides ) are
infused too rapidly. However, in most cases the SPIO administration can be reinitiated after reduction
of the injection rate. Allergic reactions may also be caused by stabilizing substances (e.g., dextran or
carboxydextran).107,275
Contrast Effects at Clinical MR Imaging

SPIO: FERIDEX AND FERUCARBOTRAN
The iron crystals clustered within phagolysosomes cause a significant superparamagnetic disturbance
of the local magnetic field, resulting in a pronounced shortening of both T2 and T1 relaxivity. Therefore,
the fundamental principle of hepatic and splenic contrast enhancement after administration of SPIO
particles goes along with the presence or relative absence of Kupffer cells within the normal hepatic
parenchyma and specific lesions. A significant decrease of the signal in tissue containing Kupffer cells,
such as normal liver, can be appreciated after SPIO administration, resulting in a relative signal
increase in tissues with a lack of Kupffer cells. Hypovascular liver metastases and liver cysts are the
typical example for this contrast pattern. Nevertheless, the persistence of Kupffer cells in some
hepatocellular tumors can be beneficial in differential diagnostic considerations. These contrast
patterns apply to the accumulation phase of SPIO imaged more than 10 minutes following injection.
10 of 49 14-04-2010 20:37
Ferucarbotran may be used for imaging the perfusion phase of lesions, providing additional
characterization information especially in hypervascular lesions. However, also due to the much lower
dose, the contrast yield is lower than that observed for low molecular weight gadolinium chelates. 7
page 385
page 386
Add to lightbox
Figure 14-4 Scheme of the contrast behavior in different hepatic lesions pre and post SPIO
administration on typical sequences.
Benign hepatic lesions derived from hepatocytes such as focal nodular hyperplasia, hepatocellular
adenomas, regenerative nodules, and dysplastic nodules may embody various amounts of Kupffer
cells. Therefore, the relative signal loss after SPIO administration in these lesions may parallel the
signal loss in normal hepatic parenchyma, most probably indicating a Kupffer cell containing benign
lesion (Figs. 14-4 to 14-6).
Hemangiomas also show signal changes, which are believed to be based upon the blood pool and
uptake into sinusoidal cells within hemangiomas. Therefore, the relative signal loss after SPIO
administration in these lesions may parallel the signal loss in normal hepatic parenchyma, most likely
indicating a Kupffer cell containing benign lesion. On mild to heavily T2-weighted images, hemangiomas
exhibit decreased signal intensity. T1-weighted sequences may show a signal increase (Fig. 14-7).
The contrast pattern in malignant liver tumors also depends on the hepatocyte content of the lesions.
All metastases are devoid of RES cells and thus hypovascular metastases show stable signal and thus
increased conspicuity (Fig. 14-8). Hypervascular metastases may exhibit perfusion changes and
pooling during the accumulation phase, resulting in a temporary signal decrease or increase dependent
on T1 effects. Hepatocyte-derived malignant lesions reveal more complex patterns. Poorly or
moderately differentiated hepatocellular carcinomas are almost completely devoid of Kupffer cells.
Therefore they exhibit a contrast pattern comparable to metastases concomitant with their vascularity
(hypovascular versus hypervascular). Well-differentiated hepatocellular carcinomas and hyperplastic
292,293
regenerative nodules may also contain Kupffer cells. Subsequently, these tumors may show
uptake into Kupffer cells and thus a signal decrease proportional to the content and activity on
accumulation-phase images. This may cause a decreased conspicuity during the accumulation phase
compared to plain images. Furthermore, the perfusion phase may be imaged with additional
information regarding lesion vascularity (Figs. 14-9 and 14-10).
To further complicate the contrast patterns, an in vitro tumor model with dextran-coated MION has
demonstrated uptake into the interstitium of a tumor and into the tumor cells. 178 However, it is currently
unclear whether this effect has any impact upon in vivo imaging.
11 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
12 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-5 Focal nodular hyperplasia (FNH) with SPIO. On nonenhanced T1- (A, C) and FS
T2-weighted (B) imaging, FNH (seg. 4a/8, biopsy proven) presents almost the same signal intensity
as the surrounding hepatic tissue. A central "dot" can be appreciated representing the nidus of the
lesion. In dynamic 3D T1- (D, E) and standard T1-weighted imaging (F) the lesion is only visible on the
arterial phase; however, on later images the iron uptake is comparable to the rest of the liver, making
the lesion vanish. On the other hand, the lesion can still be seen due to a slightly higher signal intensity
on more susceptibility sensitive T2- and T2*-weighted (G, H) sequences based on the higher amount
of nonenhancing fibrous tumor stroma.
SPIO AND USPIO IMAGING OF THE SPLEEN

page 386
page 387
Add to lightbox
13 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-6 Hepatocellular adenoma (HCA) with SPIO. This HCA (seg. 7, biopsy proven) presented
on nonenhanced T1 and opposed-phase (A, B) and fat-suppressed T2-weighted (C) imaging a slightly
higher signal intensity than the normal liver tissue without signs of an increased fat content (found in ca.
60% of HCA). Fifteen minutes after SPIO administration (Resovist) there is a significant signal loss
within the lesion on T1-, T2- and T2*-weighted imaging (D-F), reflecting the increased content of
Kupffer cells. Note the high signal in the center of the lesion paralleling the signal of surrounding
vessels that can mimic a nidus as often seen in FNH.
14 of 49 14-04-2010 20:37
In general there is no absolute indication for MRI of the spleen but there may be rare indications such
as proliferative lymphatic diseases and rare primary splenic tumors. Since the spleen is composed of
numerous vascular channels and lymphatic tissue, literally representing a huge lymph node, the organ
is perfectly suited for iron oxide-enhanced imaging. The higher content of macrophages in the spleen in
comparison to the amount of Kupffer cells in the liver will result in a higher rate of phagocytosis in the
spleen followed by a theoretically higher signal loss. However, with diagnostically relevant imaging
sequences this signal loss cannot be appreciated in general. Furthermore, from a diagnostic point of
view it would make no sense to adjust the injected dosage to the potential maximum uptake of the liver
or the spleen.
In cases of severe hepatic cirrhosis the splenic signal loss after iron oxide administration far surpasses
the signal loss in the liver, due to the loss of functioning hepatic Kupffer cells. This phenomenon is
known as "colloid shift" from nuclear medicine studies.18,294 After USPIO administration the spleen
presents the highest uptake (7.1%/g after 24 h) in comparison to liver (6.3%/g), lymph nodes
(3.6%/g), and bone marrow (2.9%/g).13,58 However, due to the very limited indications, this fact might
be of low diagnostic significance.
Clinical Applications
Add to lightbox
Add to lightbox
Add to lightbox
15 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-7 Hemangioma with SPIO. The hemangioma (seg. 2), characterized on plain T1-weighted
(A) by low and on T2-weighted MRI (F) by high signal intensity, presents on dynamic SPIO-enhanced
(Resovist) T1-weighted MRI the typical peripheral nodular enhancement progressing to the center of
the lesion (B-E). Note the high signal intensity on T2*-weighted imaging (G) of the vascular channels
paralleling the signal behavior of other vascular structures.
There is general agreement that SPIO particles increase lesion conspicuity followed by an improved
9,12,282,295,296
detection rate and tissue characterization. However, some authors argue the benefit of
283,284
using iron oxides. In various studies SPIO-enhanced MRI confirmed its superiority over contrast-
8,285,295-297
enhanced CT and spiral CT and also over CTAP. Apart from the varying designs applied in
different studies, comparative studies seem to be limited by the different dosages approved for use in
the United States, Europe, and Japan. For ferumoxides , a dosage of 15 μmol/kg is approved in
Europe and Japan in contrast to 10 μmol/kg in the USA. These different dosages will result in
considerably varied and pronounced T1 and T2* effects, especially when gradient-echo sequences are
applied.286 This phenomenon might be even more striking with Ferucarbotran, in which dosages up to
16 of 49 14-04-2010 20:37
20 μmol/kg produce the maximum contrast effects298; however, only a fixed dose of 0.5 mmol Fe/mL
7
(524 mg Ferucarbotran = 28 mg Fe) is approved for clinical use.
Recommendations for Clinical Protocols

page 387
page 388
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
17 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-8 Metastasis in colorectal cancer with SPIO. Based on SPIO-induced T1 effects, there is an
increasing contrast of the multiple low signal intensity metastases on SPIO (Resovist) enhanced
arterial (B) and portal-venous (C) gradient-echo 3D imaging and ring enhancement on delayed
imaging (D) in comparison to the nonenhanced T1-weighted MRI (A). Note the wedge-shaped
enhancement on delayed T1- (D) and T2*-weighted MRI (G), indicating a peripheral disturbance of
arterial and portal perfusion together with a loss of Kupffer cell activity. No further information is
provided by T2-weighted post-contrast imaging (F) in comparison with nonenhanced T2-weighted
imaging (E).
Add to lightbox
18 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-9 Regenerative nodule with SPIO. Regenerating nodules may present with high signal
intensity on nonenhanced T1-weighted MRI (A) due to an increased content of Fe or Mn. In this case,
increased fatty content was excluded by opposed-phase imaging and by nonfat suppressed HASTE
T2-weighted imaging (D). Therefore, the high signal is still present on dynamic SPIO (Resovist)
enhanced imaging (B, C) and is not caused by hypervascularization as confirmed by T2*-weighted
imaging (E).
Prior to contrast injection T1- and T2-weighted sequences should be obtained, at least in the
transverse plane. Specific parameters and strategies (breath-hold versus triggered) are somewhat
dependent on vendor, scanner type, and software release. In order to utilize the information provided
by looking at lesion perfusion, the weak perfusion effect of Ferucarbotran should be studied by either
T1- or T2*-weighted sequences. For T2* weighting, it appears advantageous not to maximize for T2*
in order not to give up too much signal to noise. Initial timing may follow the usual time-points of the
arterial phase and portal-venous phase followed by delayed acquisitions up to 10 minutes.
19 of 49 14-04-2010 20:37
Accumulation-phase imaging may start as early as 10 minutes post injection with T2-weighted
sequences including a GRE sequence with proton-density characteristics for detection. Mild and
heavily T2-weighted TSE sequences also provide useful information on the signal decrease within
RES-containing tumors.
Paramagnetic Hepatobiliary Contrast Agents

Mangafodipir Trisodium
page 388
page 389
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-10 Hepatocellular carcinoma with SPIO. On nonenhanced T1- (A) and T2-weighted imaging
(B) only an inhomogeneous pattern of the liver parenchyma can be discerned in the dome of the liver.
After SPIO administration multiple hepatocellular tumor nodules (biopsy proven) can be delineated on
20 of 49 14-04-2010 20:37
T1- (C) and T2*-weighted MRI (D) due to the lack of Kupffer cells within the tumors.
Mangafodipir trisodium (formerly known as Mn-DPDP; Teslascan, Amersham Health, Oslo), a

manganese (Mn) chelate (manganese (II)-N,N'-dipyridoxylethylenediamine-N,N'-diacetate-
5,5'-bis(phosphate)sodium salt), is a paramagnetic agent that has been developed as an MR contrast
material for the hepatobiliary system.299 Following IV administration, mangafodipir trisodium is
metabolized by dephosphorylation to Mn-DPMP and Mn-PLED and transmetallated by zinc to the
300,301 2+
corresponding compounds. Mn ions are released from mangafodipir, bound by
α2-macroglobulin and transported into the liver. The T1 relaxivity (R1) of mangafodipir in aqueous
solution is similar to gadolinium chelates (Gd-DOTA, Guerbet; R1: mangafodipir 2.8 mmol/L/s,
2+
Gd-DOTA: 3.8 mmol/L/s). Due to the intracellular uptake of Mn ions, the R1 of mangafodipir in liver
tissue is three times greater than that of gadolinium chelates (21.7 mmol/L/s versus 6.7 mmol/L/s).41
Biodistribution studies of mangafodipir in rats have shown that at 30 minutes after IV injection, 13% of
the paramagnetic agent and its metabolites are present in the liver, 9% in the small intestine, 3% in the
blood, 1.3% in the kidneys, and less than 1% in other organs. In rats, fecal excretion amounts to 47%
and renal elimination to 43% after 6 hours, with 6% retained in the body after 7 days.41 According to
the manufacturer's brochure, in humans, manganese (II) ions of mangafodipir are eliminated in the
urine by 15% in 24 hours and by 59% in the feces in 5 days.
Manganese is an essential trace metal in humans with a normal whole-body content of 12-20 mg in
adults. Although the IV dose of manganese during administration of mangafodipir far exceeds the
recommended daily dose and is close to the total whole-body amount of manganese ,302 no acute or
subchronic toxicity has been observed. The LD50 (the dose that is lethal for 50% of animals exposed)
of mangafodipir in mice is 5.4 mmol/kg, compared to 5.5-10 mmol/kg for Gd-DTPA. The range of
safety, as expressed by the ratio of LD50 to the dose used for clinical imaging, is 540 for mangafodipir
and 60-100 for Gd-DTPA.41
The following adverse events have been observed in clinical trials: feeling of warmth and flushing,
nausea, pounding heart, and dizziness.38,303 In a phase 2 study, facial flushing was reported by 55% of
38 304
patients and nausea by 9%. In a European phase 3 study, 17% suffered from adverse events. In
another phase 3 study, only 2.5% of patients reported facial flushing and 1.7% reported mild adverse
events other than flushing.305 In the study by Bernardino et al, facial flushing was more often seen after
38
rapid bolus injection (over 1 minute) than after infusion. However, facial flushing is not significantly
increased if a slow bolus injection over 2-2.5 minutes is used rather than infusion over 10-20
minutes.306 In general, it has been shown that mangafodipir is a safe MR contrast agent at a dosage
307
of 5 μmol/kg bodyweight.
In the US, mangafodipir is administered as an IV bolus over 1 minute. According to the manufacturer's
brochure, imaging can begin "within minutes" after injection. In Europe, mangafodipir is administered as
a short IV infusion over 10-20 minutes and imaging should begin 20 minutes after the start of infusion.
Near-maximal enhancement is achieved at 20 minutes after the start of the infusion and persists for
several hours. Thus, post-contrast imaging may begin as soon as 15-20 minutes after the start of the
308
infusion, but longer scan delays are possible because of a plateau-like enhancement.
LIVER
Mangafodipir trisodium was approved at a dose of 5 μmol/kg in many European (EU) countries and the
US in 1997, in Hong Kong in 1998, Canada and South Korea in 2000, in China and Turkey in 2001 and
in Russia in 2002. Mangafodipir as a hepatocyte-specific MR contrast agent is not specifically taken up
by non-hepatocellular lesions. Thus, lesion-to-liver contrast is significantly improved after administration
309
of mangafodipir. Mangafodipir-enhanced MRI showed more lesions than nonenhanced MRI in up to
21 of 49 14-04-2010 20:37
36% of patients (Fig. 14-11).305 In European phase 3 trials, mangafodipir-enhanced scans showed
305
more lesions than contrast-enhanced CT in 31% and fewer lesions in 13% of cases (Fig. 14-12). In
310
US phase 3 trials, mangafodipir-enhanced MRI was comparable or superior to CT. In more recent
311 312
studies, contrast-enhanced MRI was equivalent or superior to helical CT in lesion detection.
Mangafodipir-enhanced MRI is likely to influence patient management in surgical candidates with liver
tumors by detecting small metastases not shown by CT.312 Preliminary experience with current MRI
techniques indicates that mangafodipir-enhanced MRI is at least equal or even superior to contrast-
enhanced multidetector row CT for lesion detection.
page 389
page 390
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-11 Mangafodipir-enhanced MRI of liver metastases from breast cancer. The nonenhanced
T1-weighted GRE image shows some low signal intensity lesions, suggestive of metastases (A). The
fat-suppresssed T2-weighted TSE image displays only two solid subcapsular lesions (B). The
mangafodipir-enhanced T1-weighted 3D GRE image (C) reveals multiple small liver metastases as
nonenhancing filling defects (arrows). The other nonenhancing structures are demonstrated to be
tubular on adjacent slices (not shown), indicative of portal and hepatic venous branches (C).
22 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Figure 14-12 Colorectal liver metastasis: comparison of mangafodipir-enhanced MRI and helical CT.
Contrast-enhanced MR clearly shows metastatic lesion in the dome of the liver (A). Contrast-enhanced
helical CT reveals inhomogeneity of parenchyma (arrow), but no definite lesion (B).
23 of 49 14-04-2010 20:37
In general, tumors not derived from hepatocytes, such as liver metastases, do not take up the contrast
313
agent, which increases tumor conspicuity post contrast. Occasionally, a peripheral rim of
enhancement of metastases and primary liver cancers may be found at early (20 min post) and
delayed (4-24 h post) imaging.314,315 Rim enhancement of metastases after mangafodipir does not
correlate with rim enhancement seen with gadolinium chelates and iodinated contrast agents, which
largely reflects the vascularity of tumors. Peripheral rim enhancement of the surrounding parenchyma
or a wedge-shaped area of increased perilesional enhancement is probably due to tumor compression
of surrounding parenchyma or functional biliary obstruction with subsequent retention of contrast
314
agent (Fig. 14-13). Other nonhepatocellular liver lesions, such as hemangiomas and cysts, do not
take up the agent either. Peripheral rim enhancement as a sign of peripheral parenchymal compression
due to lesion growth is very rarely seen in benign lesions, such as cysts and hemangiomas. Because of
the lack of enhancement of all hepatocellular lesions with mangafodipir, differentiation between
cysts/hemangiomas and metastases is primarily based on T2-weighted pulse sequences.32,316 Very
rarely, enhancement of neuroendocrine tumor metastases has been seen. The pathophysiology of
enhancement of these metastases is not yet clear, but may be related to the increased metabolism of
317
neuroendocrine tumors.
page 390
page 391
Add to lightbox
24 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-13 Peripheral rim enhancement of liver metastasis: comparison of mangafodipir-enhanced
MRI and gadolinium-enhanced MRI. Mangafodipir-enhanced T1-weighted GRE image (A) shows thin
rim of contrast agent uptake in the surrounding compressed liver parenchyma (arrows), suggestive of
lesion growth. The lesion itself does not enhance. Consecutive application of nonspecific gadolinium
chelate (B) shows enhancement of vital tumor (arrows) in the periphery of the metastasis.
The reduced number of hepatocytes and the reduced function of remaining hepatocytes cause a
significant reduction in liver enhancement as compared to patients with normal liver. 318-320
Regenerative nodules contain functioning hepatocytes and display an uptake of mangafodipir roughly to
the same degree as the surrounding parenchyma. Therefore, regenerative nodules may be isointense
on contrast-enhanced images or slightly hyperintense in cases where the nodule is already
hyperintense on nonenhanced T1-weighted images. The number of functioning hepatocytes in
hepatocellular carcinoma (HCC) is variable. Thus, HCCs tend to show a considerable variation in
mangafodipir uptake: well-differentiated HCCs may show significant uptake of the contrast,321,322
sometimes to an even greater degree than the surrounding parenchyma. It has been suggested that
the hepatocellular function of absorption of the contrast agent is retained, but biliary excretion may be
impaired by the presence of abnormal bile ducts (Fig. 14-14). However, enhancement of
well-differentiated HCCs is often heterogeneous with sparing of the pseudocapsule. It has been
reported that metastatic lymph nodes of HCC may show enhancement after administration of
mangafodipir, which may help to differentiate between reactive lymph nodes, which are commonly
present in cirrhosis,310 and the presence of tumor lymph nodes.323 Likewise, tumor thrombus in the
portal vein may also show uptake of contrast agent.324 Poorly differentiated HCCs lack functioning
hepatocytes. Thus, tumor enhancement after mangafodipir administration is minimal or absent, which
increases tumor conspicuity (Fig. 14-15).57
Conflicting results regarding the value of mangafodipir for HCC detection have been reported. In a
study by Murakami et al, detection of HCC was only minimally improved by mangafodipir, 57 whereas in
a European phase 3 trial, mangafodipir-enhanced MRI was significantly superior to unenhanced MRI
325
(sensitivity 81% versus 48%) (see Figs. 14-14 and 14-15). Kim et al compared ferumoxide- and
mangafodipir-enhanced MRI with regard to HCC detection. In their study, ferumoxide-enhanced MRI
had superior overall sensitivity for detection of HCCs (93% versus 86%). For HCCs smaller than 10
25 of 49 14-04-2010 20:37
mm, ferumoxide was much better than mangafodipir (sensitivity 86% versus 44%).326 However, the
number of minute HCCs (n = 12) in this study was too small to draw general conclusions.
Nevertheless, differentiation between small (and well-differentiated) HCCs and regenerative nodules
may be difficult with mangafodipir-enhanced MRI.
Confluent mass-like fibrosis is seen in up to 14% of patients undergoing liver transplantation for
327
advanced cirrhosis. Cirrhosis commonly involves the anterior segments of the right lobe or the
medial segment of the left lobe and presents as a wedge-shaped lesion, which is hyperintense on
T2-weighted and hypointense on T1-weighted images. The differential diagnosis includes primary
hepatic malignancies, which may have the same appearance. In general, the diagnosis of confluent
fibrosis is a diagnosis of exclusion after biopsies negative for cancer. Mangafodipir-enhanced MRI
helps in the differentiation between fibrosis and tumor. The fibrotic strands, which are devoid of
functioning hepatocytes, typically do not enhance, making the pattern of fibrosis with underlying
cirrhosis more visible.
page 391
page 392
Add to lightbox
Add to lightbox
26 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-14 Well-differentiated HCC with mangafodipir trisodium . T1-weighted GRE (A) in a
patient with sudden right upper abdominal pain shows an encapsulated mass with focal area of
subacute hemorrhage (methemoglobin) in this patient with subtle signs of cirrhosis. On the
fat-suppressed T2-weighted TSE image (B), the lesion is slightly hyperintense with areas of
hemorrhage. There is homogeneous mangafodipir uptake (C) of the well-differentiated HCC with
better delineation of the pseudocapsule. Uptake of mangafodipir is indicative of hepatocellular lesion.
The lesion size and the presence of a pseudocapsule render diagnosis of regenerative nodule
unlikely. Hemorrhage of tumor is most often seen in adenoma and HCC. At surgery, a
well-differentiated HCC was resected.
Add to lightbox
27 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
page 392
page 393
Figure 14-15 Undifferentiated HCC: lesion detection and characterization with mangafodipir trisodium
. Nonenhanced T1-weighted GRE image (A) shows cirrhosis and several ill-defined low signal
intensity lesions (arrows). On the T2-weighted fat-suppressed TSE image (B), the lesions are not well
seen. The mangafodipir-enhanced T1-weighted GRE image (C) displays enhancement of cirrhotic
parenchyma and only minimal enhancement of several masses (arrows), which are better delineated
than before contrast administration. Undifferentiated HCC shows only minimal or no uptake of
mangafodipir because of loss of hepatocellular function.
Add to lightbox
Add to lightbox
28 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-16 Large FNH in a young female patient imaged with mangafodipir trisodium .
Nonenhanced T1-weighted GRE image (A) shows a large mass in liver segment 4. Lesion is
isointense to parenchyma (arrows). T2-weighted TSE image (B) shows lesion to be isointense
(arrows) with central hyperintense scar (arrowhead). Mangafodipir-enhanced T1-weighted GRE image
(C) shows enhancement of the lesion to a slightly higher degree than parenchyma. A central scar is
visualized, which does not take up the hepatobiliary contrast agent (arrowhead).
Benign hepatocyte-derived tumors such as focal nodular hyperplasia (FNH) or adenoma show
interesting contrast patterns based upon enhancement characteristics which are useful for lesion
characterization. FNH is the most common hepatocellular lesion in the noncirrhotic liver. Especially in
patients with breast cancer, the diagnosis of FNH versus hypervascular metastatic lesions is a
challenge with contrast-enhanced CT. FNHs typically enhance homogeneously following mangafodipir
administration and are isointense or minimally hyperintense to surrounding parenchyma in 89% of
cases328 (Fig. 14-16). If present, a central scar is hyperintense on T2-weighted MRI and does not
enhance after mangafodipir. The radial appearance of the central scar is better displayed on enhanced
images (Fig. 14-17). The imaging features of (near) isointensity on unenhanced T1-weighted and
T2-weighted images and the presence of a central T2-weighted hyperintense scar are suggestive of
FNH. The presence of homogeneous enhancement after mangafodipir (to approximately the same
degree as the parenchyma) completes the triad. The presence of this triad of imaging features is
328,329
pathognomonic for FNH and helps to avoid biopsy in these lesions.
Hepatocellular adenomas tend to contain fat or hemorrhagic areas. Adenomas are considered surgical
lesions because of their propensity to bleed and their potential for malignant degeneration. No large
series has studied mangafodipir enhancement features of adenomas. Adenomas display considerable
variability in enhancement. Some lesions show homogeneous mangafodipir enhancement to the same
degree as, or even more than, surrounding parenchyma (Fig. 14-18). However, the lack of a
T2-weighted hyperintense central scar is most important in order to avoid a misdiagnosis of FNH.
Some adenomas show only minimal uptake of mangafodipir, which does not allow reliable
differentiation between adenoma and well-differentiated HCC. However, mangafodipir-enhanced MRI
is useful for characterizing hepatocellular lesions as non-FNH type, in which case histologic verification
is mandatory.329
29 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
30 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-17 Small FNH in a patient with breast cancer imaged with mangafodipir trisodium .
T1-weighted (A) and T2-weighted images (B) show a small exophytic mass near the gallbladder
fossa, which is nearly isointense on both images (arrows). Contrast-enhanced T1-weighted 2D GRE
image (C; 8 mm slice thickness) shows homogeneous uptake of mangafodipir with delineation of
central scar (arrow). Uptake of mangafodipir helps to rule out metastatic lesion. Mangafodipir-
enhanced thin slice (D; 3 mm) T1-weighted 3D GRE image reveals better delineation of central radial
scar (arrow). Imaging features are typical for FNH.
In general, mangafodipir-enhanced MRI outperforms dual-phase helical CT for classification of focal

liver lesions as hepatocellular or nonhepatocellular (accuracy 90% versus 64%) and benign or
malignant (accuracy 90% versus 64%).330 A specific histologic diagnosis of hepatocellular lesions is
not always possible due to the overlap of enhancement features between different hepatocellular
331 321,332,333
lesions. However, mangafodipir improves the differentiation between FNH and HCC.
page 393
page 394
Add to lightbox
Add to lightbox
31 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-18 Adenoma with fatty infiltration imaged with mangafodipir trisodium . T1-weighted GRE
in-phase image (A) shows slightly hypointense lesion anteriorly (arrow). Note that there is a lobulated
lesion posteriorly, which was proven to be a hemangioma. T1-weighted opposed-phase image (B)
reveals signal intensity drop of lesion, indicative of fatty infiltration (arrow). On the T2-weighted image
(C) the lesion is isointense (arrow) and not visible. The hemangioma posteriorly in the right lobe is very
hyperintense. On the mangafodipir-enhanced T1-weighted in-phase image (D) the lesion shows
homogeneous uptake, indicative of hepatocellular origin (arrow). The hemangioma does not take up
mangafodipir. Mangafodipir-enhanced T1-weighted opposed-phase image (E) reveals signal intensity
drop of fatty lesion (arrow). Homogeneous mangafodipir enhancement is suggestive of a diagnosis of
FNH. However, fatty infiltration of lesion renders FNH unlikely. Biopsy yielded diagnosis of adenoma.
BILIARY SYSTEM
32 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Figure 14-19 Mangafodipir-enhanced postoperative follow-up after orthotopic liver transplantation.
Coronal HASTE MR cholangiography (A) shows normal-sized bile ducts (arrowheads) after orthotopic
liver transplantation. There is a fluid retention (arrow) around the bile ducts, which cannot be reliably
characterized (hematoma versus biloma due to leakage). Mangafodipir-enhanced T1-weighted fat-sat
GRE image (B) shows excretion of contrast material into biliary radicles (arrowheads). The common
bile duct (arrow) is filled with contrast agent and mangafodipir-enhanced MR cholangiography helps to
rule out biliary leakage.
Due to the strong biliary excretion of mangafodipir metabolites, enhancement of the biliary system and
drainage into the duodenum are visualized at 20 minutes after administration in patients without biliary
obstruction. This effect has been used to perform mangafodipir-enhanced MR cholangiography (MRC),
similar to CT cholangiography after infusion of biliary iodinated contrast agents, as an adjunct to classic
MRC with T2-weighted pulse sequences (Fig. 14-19). For visualization of the mangafodipir-enhanced
biliary system, a T1-weighted GRE (preferably 3D-GRE) pulse sequence in the coronal or oblique
334,335
plane provides the best overview and spatial resolution. Clinical applications include the
preoperative evaluation of the biliary system of potential living-related liver donors. In these patients,
the detection of biliary duct variants is of utmost importance for the preoperative planning of split-liver
surgery.334,336 Mangafodipir cholangiography also helps in the detection of leakage after biliary
33 of 49 14-04-2010 20:37
surgery, particularly if T2-weighted MRC fails to characterize fluid retentions around the bile
ducts.337,338 Demonstration of biliary excretion into the duodenum helps to rule out biliary extravasation
(see Fig. 14-9). However, biliary excretion of mangafodipir may be delayed in cases of biliary
obstruction. Thus, compression of bile ducts by a biloma can impair extravasation of contrast material
from a leak. In this case, biliary excretion with extravasation of contrast may be seen only after several
hours' delay.
page 394
page 395
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-20 Mangafodipir-enhanced MRI of pancreatic cancer with liver metastases. Unenhanced
T1-weighted GRE image (A) shows a large tumor within the pancreatic body and tail (arrow).
Mangafodipir-enhanced T1-weighted GRE image (B) reveals better delineation of the nonenhancing
tumor (arrows). Mangafodipir-enhanced image (C) at a higher level reveals two small liver metastases
with lack of enhancement (arrows). Coronal enhanced T1-weighted image (D) provides a good
34 of 49 14-04-2010 20:37
overview of tumor encasing the superior mesenteric artery and vein (arrow) and small liver metastases
(arrowheads).
PANCREAS
There is also uptake of mangafodipir into the pancreatic parenchyma. Gehl et al were the first to
demonstrate that IV mangafodipir infusion of 10 μmol/kg (double the dose now recommended for
clinical use) results in a 98% enhancement of the pancreas.339 Enhancement of the pancreas started 3
minutes after administration and reached its peak after 1.5 hours. Parenchyma enhancement lasts for
several hours, with 50% enhancement still present 6 hours post administration. There is only very little,
if any, enhancement within pancreatic adenocarcinomas. This results in a reliable increase of
pancreas/tumor contrast of more than 200%,339 which improves the detection of pancreatic cancer.
Mangafodipir at the registered dose of 5 μmol/kg still increases the tumor/pancreas contrast by up to
70% on T1-weighted fatsat TSE and by 80% on T1-weighted GRE images.340,341 The studies by
Schima et al and Romijn et al have shown that mangafodipir-enhanced MRI is at least equal to
contrast-enhanced helical CT for the detection and staging of pancreatic cancer (Fig. 14-20).340,342
Mangafodipir and nonspecific gadolinium chelates have been compared in pancreatic MR imaging.
Bolus injection of gadolinium chelates provides better enhancement of pancreatic parenchyma than
mangafodipir (73% versus 36% on T1-weighed GRE images).343-345 However, this does not
necessarily mean that the tumor/pancreas contrast is higher for gadolinium than for mangafodipir. In a
study by Diehl et al, a significant increase in tumor contrast was found only on manganese-enhanced
346
T1-weighted GRE images but not on gadolinium-enhanced images.
Since 2001, mangafodipir trisodium has been licensed for use in pancreatic MR imaging in several
European (EU) countries, but it is not licensed for this indication in the US and Asia.
Gadobenate Dimeglumine
MultiHance is a sterile solution for intravenous injection approved for use in MR imaging of focal liver
disease and in MR examinations of the CNS in some countries (see Tables 14-1 and 14-2).
Gadobenate dimeglumine (Gd-BOPTA, Atlana-Bracco, Italy, Milano), the active ingredient of
MultiHance, is an octadentate chelate of the paramagnetic ion gadolinium. Its kinetic properties
resemble those of conventional iodinated contrast media347 and can be described by a bi-exponential
function comprising a distribution phase and an elimination phase.348,349
Studies have shown that this agent differs from other available gadolinium chelates in that it not only
distributes to the extracellular fluid space (ECF) but is selectively taken up by functioning hepatocytes
and excreted into the bile by the so-called canalicular multispecific organic anion transporter shared
with bilirubin.35,36,348-350 However, whereas the biliary excretion rate is 55% in rats and 25% in rabbits,
351
it is only 3% to 5% in humans. Nevertheless, this level of uptake is sufficient to bring about specific,
long-lasting enhancement of MR signal intensity in normal liver parenchyma while at the same time
demonstrating the plasma kinetics of agents distributing solely to the ECF.
The liver parenchyma enhancement obtained with gadobenate dimeglumine is equivalent to that
32
achieved with purely liver-specific contrast media. However, gadobenate dimeglumine has a higher
relaxivity than equimolar formulations of other approved extracellular contrast agents such as
gadopentetate dimeglumine (Magnevist; Schering AG, Berlin, Germany), gadodiamide (Omniscan;
Amersham Health, Oslo, Norway), and gadoterate meglumine (Dotarem; Guerbet, Aulnay-sous-Bois,
France),352 due to its more lipophilic structure and its capacity for weak and transient interaction with
353
serum albumin. In the liver, the estimated relaxivity is about 30 mmol/s, compared with calculated
values of 16.6 mmol/s for Gd-EOB-DTPA (Schering AG, Berlin, Germany) and 21.7 mmol/s for
mangafodipir trisodium .10 This effect is thought to be due more to increased intracellular
36
microviscosity within the hepatocytes than to transient interactions with intracellular proteins.
35 of 49 14-04-2010 20:37
Gadobenate dimeglumine has been approved in many European countries for MRI of the liver since
1998; in 2001 this contrast agent was administered to about 100,000 patients. So far, a total of 2540
adult and pediatric subjects have been included in clinical trials. 354
CLINICAL APPLICATIONS
page 395
page 396
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
36 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-21 Dynamic imaging of the liver with gadobenate dimeglumine. Unenhanced T2-weighted
(A) and T1-weighted (B) images show signs of liver cirrhosis with multiple regenerating nodules and
hypertrophy of the left liver lobe and splenomegaly with small foci of low signal intensity corresponding
to Gamna-Gandy bodies. However, no focal liver lesion can be distinguished. In the arterial phase
(approx. 20 s following the start of CM injection) after bolus injection of 0.05 mmol/kg gadobenate
dimeglumine (C), multiple hypervascular liver lesions can be identified (arrows) that show a persistent
enhancement in the portal-venous phase (D). In the equilibrium phase (E) the lesions are again
isointense compared to the cirrhotic liver parenchyma. This case shows the importance of dynamic
imaging even with a cell-specific contrast agent. The arterial phase is the most sensitive phase after
CM administration to detect hypervascular lesions like HCC in a cirrhotic liver. (Reprinted with
permission from European Radiology/Springer.)
Gadobenate dimeglumine can be used for rapid dynamic imaging of the liver following bolus injection
1,355
(Fig. 14-21) in the same way in which other nonliver specific contrast materials are used. This
approach exploits the differences in blood flow between lesions and normal liver parenchyma. The
results are comparable to other conventional ECF contrast materials, particularly for the improved
visualization of hypervascular lesions.42 Furthermore, as with these other ECF contrast materials,
19,356-359
gadobenate dimeglumine allows improved assessment of lesion hemodynamics ; however, the
improvement in the detection of hypovascular lesions is not that impressive with any of these
agents.360
Add to lightbox
37 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-22 Detection of hypovascular lesions with gadobenate dimeglumine during the
hepatocellular phase. The unenhanced T1-weighted scan (A) shows three hypointense liver lesions in
the dome of the liver in a 48-year-old female patient with breast cancer. Lesions are obscured on the
arterial phase (B) image (0.05 mmol gadobenate dimeglumine/kg bw) and portal-venous phase
images (not shown). However, in the equilibrium phase (C) two of the lesions that were apparent on
the unenhanced scan are again detectable (arrows), presenting with a peripheral hypointense rim
surrounding the lesions. This sign is highly specific for metastases and is observed due to a washout
of CM in the peripheral vital parts of the tumor whereas the contrast medium in the more central parts
that show regressive changes demonstrates a delayed uptake most likely due to diffusion of the
contrast medium. In the hepatocellular phase 60 min post gadobenate dimeglumine injection, when the
contrast medium is taken up by functioning hepatocytes (D), an increase of the signal intensity of
normal liver parenchyma is visible. In addition, a higher contrast between normal liver tissue and the
liver metastases is observed, now allowing the detection of a total of seven liver metastases (arrows).
(Reprinted with permission from European Radiology/Springer.)
Unlike the situation with conventional ECF agents, the fraction of gadobenate dimeglumine taken up
into the hepatocytes manifests itself as a strong enhancement of the liver parenchyma signal intensity
in T1-weighted images between 40 and 120 minutes post injection (delayed scans).36,45 This effect is
dose dependent up to a dose of 0.1 mmol/kg36; the recommended standard dose for gadobenate
38 of 49 14-04-2010 20:37
dimeglumine-enhanced MRI of the liver is 0.05 mmol/kg bodyweight. The fraction taken up by the
hepatocytes leads to a markedly increased contrast of nonhepatocellular lesions, which are unable to
take up the contrast agent.351,361 An increased rate of detection of hypovascular metastases is
therefore possible on delayed or "static" hepatobiliary liver imaging with gadobenate dimeglumine
40-120 minutes post injection (Fig. 14-22). In a multicenter, multireader study (214 patients) published
in 2000, Petersein et al reported a significantly increased number of lesions detected on delayed
post-contrast images. Furthermore, the average size of the detected lesions decreased considerably
45
while the conspicuity of all lesions improved.
page 396
page 397
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
39 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Figure 14-23 Lesion characterization with gadobenate dimeglumine. Unenhanced T2-weighted (A)
and T1-weighted (B) images show a large, homogeneous lesion of the left liver lobe (arrows), slightly
hyperintense on T2- and slightly hypointense on T1-weighted images in a 30-year-old female patient
with suspicion of a large tumor within the left liver lobe on ultrasound. Dynamic arterial-phase imaging
(C) following IV injection of 0.05 mmol gadobenate dimeglumine/kg bw shows a markedly
hypervascular lesion with a persisting hyperintensity during the portal-venous phase (D) and isointense
appearance on equilibrium phase (E). No central scar is visible and the most likely diagnosis would be
an FNH with atypical appearance. This diagnosis is confirmed on a delayed, hepatobiliary scan 60
min post injection of contrast medium (F), when the lesion takes up contrast medium to a higher
degree than normal liver tissue, now appearing hyperintense. This indicates a lesion containing
functioning hepatocytes with an abnormal biliary drainage as is the case in FNH. Adenoma,
fibrolamellar carcinoma (FLC) or hypervascular metastases, which would represent the main
differential diagnoses, would not show an uptake of contrast medium and thus the confidence to make
the diagnosis of FNH is clearly increased by adding the information of the hepatobiliary scan.
(Reprinted with permission from European Radiology/Springer.)
Primary benign liver tumors exhibit variable enhancement.362 All types of cystic lesions show no signal
changes following IV injection of gadobenate dimeglumine. Mesenchymal tumors show signal changes
proportional to their vascularity during the perfusion phase. Hemangiomas often demonstrate a nodular
enhancement pattern followed by isointensity or hypointensity on delayed images. Focal nodular
hyperplasia shows a significant signal increase during the arterial phase at dynamic scanning and
persistent enhancement during the hepatocellular phase, indicating functioning hepatocytes typically
without a hyperintense scar.
Morana et al363 examined 249 patients with a variety of primary and secondary hypervascular tumors
(both dynamic and delayed imaging). They found that delayed imaging gave additional information for
lesion characterization with high accuracy in distinguishing true benign lesions like FNH (Fig. 14-23) and
364
regenerative hyperplasia from other lesion types (sensitivity 79.7%, specificity 96.1%). Grazioli et al
studied a subset of patients with FNH comparing gadobenate dimeglumine to ferumoxides . They
noted that 57/60 lesions displayed typical enhancement characteristics after gadobenate dimeglumine
(and 60/60 were identified correctly), whereas after ferumoxides only 27/43 lesions showed typical
enhancement (with 71.6 correctly identified as FNH). Adenomas also enhance during the perfusion
phase at dynamic scanning with a washout during the portal-venous phase and equilibrium phase.
Delayed images are characterized by hypointensity due to the absence of biliary ducts.
Malignant liver tumors also exhibit variable enhancement, which mainly depends on their origin as either
primary or secondary malignant.362 Dynamic imaging with gadopentetate dimeglumine shows signal
40 of 49 14-04-2010 20:37
changes proportional to the vascularity of malignant tumors as known from extracellular gadolinium
chelates. Hepatocellular phase imaging is particularly interesting for tumors of hepatocellular origin,
while other tumors are hypointense to normal liver. Moderately differentiated and well-differentiated
HCCs may show hyperintense features compared to poorly differentiated HCCs. The degree of
residual hepatocytic activity determines hyperintensity on delayed images. Kuwatsuru et al published a
study in 2001 in which they compared gadobenate dimeglumine with gadopentetate dimeglumine in
257 patients suspected of having malignant liver tumors.365 They concluded that the contrast efficacy
on dynamic post-contrast images was comparable, while on delayed images gadobenate dimeglumine
was significantly superior to gadopentetate dimeglumine in terms of improvement over the
nonenhanced scans (44.5 versus 19.0% on breath-hold gradient-echo sequences). Schneider et al
similarly reported a tendency toward more accurate results after gadobenate dimeglumine than after
34
gadopentetate dimeglumine in a study of 43 patients published in 2003.
Gd-EOB-DTPA
page 397
page 398
Gd-EOB-DTPA (gadolinium-ethoxybenzyl-diethylenetriamine-penta-acetic-acid; SH L 569 B; gadoxetic

acid disodium) is a paramagnetic hepatobiliary contrast agent with hepatocellular uptake via the anionic
transporter protein and a molecular weight of 725.71 Da (C23H28GdN3Na2O11).50,366 Gd-EOB-DTPA
exhibits a T1 relaxivity as measured in water at 0.47 T of 4.9 mM/s comparable to gadopentetate
dimeglumine with 3.7 mM/s but a higher T1 relaxivity in human plasma (R1 8.2 mM/s) than
gadopentetate dimeglumine (R1 5.0 mM/s). This may be explained by the greater degree of protein
binding (10.7 ± 3.4%) compared to gadopentetate dimeglumine (1.6 ± 4.2%). At 37° C, the solution
has an osmolality of 0.89 osmol/kg H2O and a viscosity of 1.22 mPas. Gd-EOB-DTPA is an aqueous
formulation with a concentration of 0.25 mol/L and a thermodynamic stability constant of 1020.
153
Biodistribution studies in dogs using Gd-labeled Gd-EOB-DTPA revealed a dose-dependent renal
(40.9 ± 2.35%) and biliary (57.0 ± 2.49%) excretion without signs for metabolism and an enterohepatic
recirculation of approximately 2.1 ± 0.56%.50,366 Preclinically, Gd-EOB-DTPA did not show acute
hepatotoxicity in an isolated perfused rat liver model, cardiovascular effects in the anesthetized rat or
signs of an IgG immune response. Hemodynamic effects were only observed after femoral bolus
367
injection at relatively high dosages of 0.3-0.5 mmol Gd/kg.
Gd-EOB-DTPA was well tolerated within phase 1 trials at doses of 10, 25, 50, and 100 μmol Gd/kg,
42
with no important side-effects or changes in laboratory parameters. During subsequent clinical phase
2 trials, the diagnostic efficacy and safety of Gd-EOB-DTPA were explored at five doses (3.0, 6.0,
12.5, 25, and 50 μmol Gd-EOB-DTPA/kg bw) as compared to placebo (0.9% saline) in patients with
known focal liver lesions. Clinical phase 2 trials revealed a lowest effective dose of 25 μmol Gd-EOB-
DTPA/kg bw (injection volumes of 7.0 mL in a 70 kg patient).368,369 The number of patients with more
lesions being visualized on post-contrast than on precontrast scans increased with increasing dose up
to 12.5 μmol (at 20 min post MRI: 12.1% (3.0 μmol); 15.6% (6.0 μmol); 30.3% (12.5 μmol); 31.4% of
patients (25.0 μmol)). The enhancement of hepatic vessels and liver parenchyma during dynamic
imaging following bolus injection at 2 mL/s also increased with increasing dose. There was no
significant improvement in quantitative or qualitative efficacy parameters at 45 minutes post injection
compared to 20 minutes post injection. Dynamic imaging can start immediately after bolus injection and
368,369
accumulation-phase imaging can be performed at 20 minutes post injection.
In a total of 171 patients studied in phase 2 multicenter trials, no serious adverse events (AEs) were
reported and adverse events were seen in six patients. There was no dose dependency seen in the
occurrence of AEs. Of these, four AEs (4%) were considered as possibly or probably drug related. All
AEs were mild except one (anxiety). No significant changes in vital signs or laboratory parameters
were observed.369 Within the subsequent phase 3 multicenter trials, no clinically relevant changes in
hemodynamic or laboratory parameters were observed. No death or any AE leading to discontinuation
of the study was reported. The most frequent reported symptoms of definitely, possibly or probably
related AEs were nausea, vasodilatation, headache, taste perversion, and injection site pain.370
41 of 49 14-04-2010 20:37
LIVER
As already described for gadobenate dimeglumine, following bolus injection tumors show non-specific
uptake based upon their vascularity during the perfusion phase. Following washout, in the accumulation
phase the contrast agent stays within hepatocytes and hepatocyte-derived tumors. Tumors not derived
from hepatocytes such as liver metastases do not take up the contrast agent, increasing tumor
conspicuity post contrast.
Liver imaging with Gd-EOB-DTPA consists of the acquisition of precontrast T1-weighted and
T2-weighted images followed by bolus injection and dynamic T1-weighted imaging of the perfusion
phase. Since the injected dose is just one quarter of the regular gadolinium dose, contrast
enhancement is somewhat weaker than experienced with nonspecific gadolinium chelates.
Nevertheless, the perfusion phase allows for the assessment of perfusion patterns of liver lesions.
Accumulation-phase images may be obtained upon washout of nonspecifically perfused liver lesions
and may start as early as 8 minutes following injection as tested in phase 3. 370 Acquiring T2-weighted
images following perfusion-phase imaging shortens the examination time with reproducible image
370
quality. Several experimental studies and selected clinical evaluations have been
reported.10,39,40,42,44,50,53-55,366-397
More recently, phase 3 results have been published.370 In this trial, three blinded readers detected
215, 197, and 191 correctly matched lesions (193, 183, and 177 lesions = 1 cm; 22, 14, and 14
lesions <1 cm) on precontrast scans. On post-contrast MR images, a correct match was made for
240, 207, and 212 lesions (205, 190, and 189 lesions = 1 cm; 35, 17, and 23 lesions <1 cm). These
results show that the detection of liver lesions is improved for a spectrum of lesions and that especially
smaller lesions are detected. These results apply mainly to the accumulation-phase images, while
perfusion-phase images are more useful for lesion characterization. For characterization, the per
patient sensitivity was significantly (P < .05) higher on post-contrast images alone or on combined
images compared to the precontrast MR images.370
METASTASES
Add to lightbox
Add to lightbox
42 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
page 398
page 399
Figure 14-24 Liver metastasis with Gd-EOB-DTPA. Precontrast T1-weighted MRI (A) shows a round
lesion with homogeneous hypointense signal intensity in segment 2. The lesion is moderately
hyperintense on T2-weighted MRI (B). Administration of Gd-EOB-DTPA caused rim enhancement of
the lesion in the arterial phase (C). Delayed images during the hepatocyte phase 20 minutes after
contrast injection (D) showed the lesion hypointense without any visible intratumoral enhancement.
Note the signal increase of the liver parenchyma (B compared to D) due to the uptake into functioning
hepatocytes. (Courtesy of the AKH Vienna, Austria, and provided by J Breuer, Schering AG, Berlin,
Germany.)
During the perfusion phase (Fig. 14-24), liver metastases show highest enhancement 90-120 s
following IV injection of Gd-EOB-DTPA. Tumor enhancement then gradually decreases and stabilizes
after 10 minutes. Liver enhancement is stronger than enhancement of metastases after 3 minutes
following IV injection of Gd-EOB-DTPA. Therefore, lesion/liver contrast of liver metastases gradually
increases from 2 minutes to 10 minutes post contrast. The increase from precontrast and early
post-contrast (2 min) signal intensities to later post-contrast signal intensities (= 10 min) is significant
(P = .05); however, differences among later post-contrast time-points (10, 20, and 45 min) were not
significant.386 In phase 2, the dose of 50 μmol Gd-EOB-DTPA/kg showed less liver/lesion contrast
than 25 μmol Gd-EOB-DTPA/kg, which was explained by a more pronounced extracellular effect of the
contrast medium at higher doses.393 Gd-EOB-DTPA provides improved detection and characterization
of liver metastases compared to nonenhanced and Gd-enhanced MRI, both experimentally and
clinically.378,381,386,393,395
In a recently completed US phase 3 trial, almost no metastasis showed complete enhancement during
dynamic imaging. Moreover, in some cases, there is partial, often rim enhancement in metastases (see
Fig. 14-24) which do not possess a uniform and homogeneously distributed vasculature. In the
accumulation phase (see Fig. 14-24) most metastases did not enhance in both the clinical and blinded
reader studies.398
HEPATOCELLULAR CARCINOMA
43 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
44 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-25 Hepatocellular carcinoma with Gd-EOB-DTPA. Precontrast T1-weighted MRI (A) shows
a hypointense lesion with a layering appearance and a central hyperintensity. The T2-weighted MRI
(B) shows a moderately hyperintense lesion with some areas of strong central hyperintensities.
Following administration of Gd-EOB-DTPA, dynamic imaging shows enhancement within the lesion
(arterial phase C, portal-venous phase D, equilibrium phase E). The hepatocyte phase (F) is
characterized by signal enhancement of both the lesion and liver parenchyma. Intratumoral signal
changes are due to central bleeding. (Courtesy of the Institute of Clinical Radiology, Ludwig-
Maximilians-University Munich, Klinikum Großhadern, Munich, Germany and provided by J Breuer,
Schering AG, Berlin, Germany.)
During dynamic imaging, hepatocellular carcinomas (Fig. 14-25) typically demonstrate enhancement
during the initial distribution phase almost similar to liver parenchyma with subsequently prolonged
enhancement compared to metastases and liver parenchyma. Depending on tumor vascularity, some
tumors may enhance stronger and some may enhance weaker than normal liver parenchyma as also
happens with gadolinium chelates.
Experimental studies with Gd-EOB-DTPA showed specific tumor enhancement in only two of 79
primary hepatomas, both of which were highly differentiated. 384 Another experimental study revealed
44
no intracellular uptake within hepatocellular carcinomas irrespective of their differentiation.
Furthermore, the enhancement pattern of chemically induced hepatocellular carcinomas was compared
with the uptake pattern of technetium-99m labeled iminodiacetic acid (IDA), a hepatobiliary radioactive
tracer. The organic anion transporter drives the hepatocyte uptake of both the contrast agent and the
scintigraphic agent. Tumors enhanced less than liver after Gd-EOB-DTPA administration, whereas the
99m
Tc-IDA uptake of differentiated HCCs exceeded that of the liver at 30 minutes and 3 hours after
administration. The enhancement pattern of a differentiated HCC with Gd-EOB-DTPA is not
99m 53
comparable to Tc-IDA.
In another experimental study, Gd-EOB-DTPA was compared to SPIO. Seventeen mice with 66
chemically induced HCCs underwent MRI with both Gd-EOB-DTPA (30 μmol/kg bw) and SPIO (10
μmol/kg bw). Lesion/liver contrast of 47 detected HCCs increased negatively from 3.7 ± 10.7 (mean ±
SD) to -55.1 ± 25.8 with Gd-EOB-DTPA (P < .001) and increased positively from 10.4 ± 10.4 to 26.1
± 16.3 with SPIO (P < .001). The improvement after administration of SPIO was less in smaller lesions
(<4 mm), whereas that after administration of Gd-EOB-DTPA was independent of lesion size.
However, Gd-EOB-DTPA positively enhanced four HCCs (8.5%), both highly differentiated (grade 1)
and moderately differentiated (grade 2).
Gd-EOB-DTPA allows the conspicuous detection of small HCCs; however, moderately differentiated
HCCs occasionally may be positively enhanced.374 Clinically, hepatocellular carcinomas studied are
mostly isointense or hyperintense on precontrast images and show a significant decrease in lesion/liver
386
contrast post contrast. In a recently completed US phase 3 trial, HCCs were usually described as
enhancing heterogeneously in the dynamic phase. Enhancement was greater during the dynamic
phases than during the hepatocyte phase.398
HEMANGIOMA
page 399
page 400
45 of 49 14-04-2010 20:37
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 14-26 Hemangioma with Gd-EOB-DTPA. Plain T1-weighted MRI (A) shows an
inhomogeneous predominantly hypointense lesion with a markedly hypointense area in the center. On
T2-weighted MRI (B) the lesion appears hyperintense, especially in the central area. Dynamic images
with 25 μmol Gd-EOB-DTPA/kg bw (C arterial phase, D portal-venous phase, E equilibrium phase)
show a centripetal filling pattern with peripheral-nodular increase in signal intensity suggestive of
hemangioma. No contrast enhancement of the lesion is observed in the hepatocyte phase (F),
whereas the liver tissue exhibits a hyperintense signal due to contrast uptake by functioning
46 of 49 14-04-2010 20:37
hepatocytes. (Courtesy of Hospital Clinico y Provincial de Barcelona, Barcelona, Spain and

provided by J Breuer, Schering AG, Berlin, Germany.)
Enhancement of liver hemangioma is significantly stronger than for metastases or hepatocellular

carcinomas during the perfusion phase up to 10 minutes following IV injection of Gd-EOB-DTPA. In a
recently completed US phase 3 trial, an increasing proportion of hemangiomas displayed partial
enhancement as time elapsed in the dynamic phase (Fig. 14-26). In the accumulation phase, the
398
majority of hemangiomas did not enhance or only partially enhanced. Delayed images at 45 minutes
have already shown no significant difference of hemangiomas compared to metastases or
hepatocellular carcinomas. Hemangiomas initially demonstrate a slight decrease in lesion/liver contrast
due to relative tumor enhancement following IV injection of Gd-EOB-DTPA. The increase from
precontrast and early post-contrast values up to 10 minutes was not significant and lesion/liver contrast
still increased between 20 and 45 minutes (P = .05).386
FNH AND ADENOMA
Add to lightbox
Add to lightbox
Add to lightbox
47 of 49 14-04-2010 20:37
Add to lightbox
Figure 14-27 Focal nodular hyperplasia with Gd-EOB-DTPA. Plain T1-weighted MRI (A) shows a
well-defined mass with a hypointense central scar in the left liver lobe. Dynamic T1-weighted MR
images with 25 μmol Gd-EOB-DTPA/kg bw (B arterial phase and C portal-venous phase) show a
hypervascular lesion with strong enhancement sparing the scar. Accumulation phase T1-weighted MRI
(D) demonstrates persisting hyperintensity with an isointense appearance compared to normal liver.
The scar does not show contrast enhancement.
Hepatocyte-derived benign liver tumors exhibit prolonged tumor enhancement (Fig. 14-27) due to
395,399
specific intracellular uptake of Gd-EOB-DTPA. In a recently completed US phase 3 trial, the vast
majority of FNH lesions enhanced completely during dynamic imaging. During the accumulation phase,
most FNHs showed either complete or partial enhancement. 376,398 No large series have studied
Gd-EOB-DTPA enhancement features of adenomas. Adenomas display considerable variability in
enhancement, during both dynamic imaging and accumulation-phase imaging. Some lesions show
homogeneous enhancement to the same degree as, or even more than, surrounding parenchyma. The
lack of a T2-weighted hyperintense central scar is important in order to avoid a misdiagnosis of FNH.
Some adenomas show only minimal uptake, which does not allow reliable differentiation between
adenoma and well-differentiated HCC so far. However, Gd-EOB-DTPA enhanced MRI is useful for
characterizing hepatocellular lesions as non-FNH type, in which a histologic verification is
376,398
mandatory.
BILIARY SYSTEM
Imaging of biliary excretion and hepatic function with Gd-EOB-DTPA enhanced MRI has been
371,372,377,383,389,390,392
investigated experimentally in normal liver and diseased liver. Due to the strong
biliary excretion of Gd-EOB-DTPA, enhancement of the biliary system and drainage into the duodenum
are visualized within several minutes after administration in patients without biliary obstruction. Analysis
of the course of enhancement in bile ducts and gallbladder revealed that the common bile duct showed
dose-dependent intense signal enhancement beginning 5-16 minutes after injection (mean 10 min),
persisting for more than 2 hours. Intrahepatic bile ducts were hyperintense as compared with liver
parenchyma in all subjects receiving lower doses while at higher dose (25 μmol Gd/kg bw) intrahepatic
bile ducts were not displayed, probably because of strong enhancement of liver parenchyma. The
gallbladder showed signal enhancement beginning 7-33 minutes after injection (mean 19 min) and
remained visible for up to 6 hours in 94% of subjects.371
page 400
page 401
In a subsequent study of patients with liver masses, the 20 minute post-contrast delay after
Gd-EOB-DTPA administration (25 μmol Gd/kg) appeared to be sufficient for adequate biliary
enhancement. Residual hepatic enhancement does not appear to interfere with biliary visualization.372
A comparison with nonenhanced MRC showed that combining nonenhanced MRC and EOB-MRC
improved biliary visualization over either test alone. However, improvement was small and the
perceived added value of Gd-EOB-DTPA was considered modest.400
EFFECT OF HEPATIC AND RENAL IMPAIRMENT
The impact of hepatic and/or renal impairment was investigated in a special population study. Groups
of volunteer patients with various levels of impaired hepatic function, impaired renal function, coexistent
hepatic and renal impairment, and a control were enrolled. Pharmacokinetic parameters of
Gd-EOB-DTPA derived from the serum level data were markedly altered only in endstage renal failure.
Mild to severe hepatic impairment and moderate renal impairment did not markedly change the
pharmacokinetic parameter values. Considering the safety profile, the observed differences in the
patients with endstage renal failure may not be clinically relevant. Hepatic parenchymal enhancement
with a standard dose of Gd-EOB-DTPA was accomplished in all patient groups. For clinical use, the
enhancement was sufficient above 50%, and of sufficient duration with 60 minutes or more. No special
considerations appear to be required in the presence of end-organ disease. Gd-EOB-DTPA, as a
single rapid intravenous dose of 25 μmol/kg bw, was safe and well tolerated by all patients in this
48 of 49 14-04-2010 20:37
study. This included patients who were hepatic impaired, renal impaired (including endstage renal
disease), patients with both hepatic and renal impairment and volunteer patients, male and female,
with normal liver and kidney function under 65 and over 65 years of age. 401
MRI WITH HEPATOCYTE-SPECIFIC HEPATOBILIARY CONTRAST AGENTS
Bolus-injectable paramagnetic hepatobiliary contrast agents combine features of extracellular agents,

e.g., initial tumor perfusion and enhancement on delayed images of tumors with a large blood pool or
with hepatocytes maintaining cell membrane function. The detection and characterization of focal liver
disease are significantly improved compared to nonenhanced MRI, MRI with nonspecific contrast
agents, and contrast-enhanced CT.370,376,378,395,398 Innovative rapid acquisition techniques with near
isotropic 3D pulse sequences with fat saturation parallel the technical progress made by multislice
computed tomography (MSCT) combined with an unparalleled improvement in tumor/liver
334,402-404
contrast.
Hepatobiliary contrast agents now come in three different "flavors." The individual decision on which
one to use is partly based on personal preferences. The capability for bolus injection appears to be
advantageous, allowing characterization of lesions by means of dynamic imaging during the perfusion
phase with gadobenate dimeglumine and Gd-EOB-DTPA. Following a variable time window, all
contrast agents offer improved lesion-to-liver contrast during the hepatocyte phase. Typically, scans
following mangafodipir trisodium are obtained as early as 20 minutes post-contrast after start of the
injection. The delay time for gadobenate dimeglumine is of the order of >60 minutes and for
45,370
Gd-EOB-DTPA as early as 8-10 minutes with validated data also for a 20 minute time-point. The
persistent contrast accumulation in the liver provides a long imaging window of several hours. Delayed
imaging after 24 hours has even been shown to improve lesion delineation of metastases for
mangafodipir. However, it has not been proven that significantly more lesions are detected by
additional delayed imaging, which is therefore recommended in selected cases only. 315
© 2010 Elsevier
49 of 49 14-04-2010 20:37
COST-EFFECTIVENESS AND COMPARISON OF LIVER MAGNETIC

RESONANCE IMAGING WITH TISSUE-SPECIFIC CONTRAST MEDIA
The prognosis of malignant liver tumors is, in general, poor. However, 25% of patients with liver
metastases from colorectal cancer may be surgical candidates, with a 5-year survival of up to 40%. In
a study by Mann et al, the clinical and cost-effectiveness of mangafodipir-enhanced MRI and contrast-
enhanced helical CT were compared in potential surgical candidates. In this study, mangafodipir-
enhanced MRI showed significantly more metastases than CT, altering the management of patients in
74%.312 Unnecessary surgery was avoided in one third of the patients because of MRI evidence of
312
widespread disease. Thus, total cost savings were at least $140,940 in this analysis. Comparable
data were presented for SPIO compared to nonenhanced MRI and contrast-enhanced CT.405
Various studies have demonstrated the improved lesion detection and characterization by contrast-
enhanced MRI compared to non-enhanced MRI, contrast-enhanced CT or contrast-enhanced MRI with
nonspecific contrast agents.30,296,297,326,406-413
As for hepatobiliary contrast agents alone, Schima et al published a comparison of gadobenate

dimeglumine and mangafodipir trisodium for liver enhancement and lesion/liver contrast on
T1-weighted SE and GRE images in 51 patients (three groups of 17 patients each: gadobenate
dimeglumine (0.1 or 0.05 mmol/kg) or mangafodipir trisodium (5 μmol/kg)). Overall, GRE images
were superior to SE images for signal-to-noise ratio and contrast-to-noise ratio. Mangafodipir
trisodium and gadobenate dimeglumine (0.1 mmol/kg) provided equal liver enhancement and lesion
32
conspicuity post-contrast.
page 401
page 402
Kuwatsuru et al365 reported on the comparison of gadobenate dimeglumine with gadopentetate

dimeglumine for MRI of the liver in 257 patients suspected of having malignant liver tumors. Both
contrast agents were administered at a dose of 0.1 mmol Gd/kg bw. Dynamic-phase GRE images, SE
images obtained within 10 minutes of injection, and delayed images obtained 40-120 minutes after
injection were acquired. Contrast efficacy for dynamic-phase imaging was moderately or markedly
improved in 90.9% (110/121) and 87.9% (109/124) of patients for gadobenate dimeglumine and
gadopentetate dimeglumine respectively. At 40-120 minutes after injection, there were significant
improvements with gadobenate dimeglumine, with 21.7% (26/120) versus 11.6% (14/121) for SE and
44.5% (53/119) versus 19.0% (23/121) for breath-hold GRE, respectively. Adverse events were seen
in 4.7% (6/128) and 1.6% (2/127) of patients receiving gadobenate dimeglumine and gadopentetate
dimeglumine , respectively. The difference was not statistically significant. During the dynamic phase,
both contrast agents were equally effective. However, gadobenate dimeglumine was superior during
the delayed (40-120 min) phase of contrast enhancement.365
Recently, comparative studies with different liver-specific contrast agents have been performed.
Schneider et al compared gadobenate dimeglumine and SPIO (Ferucarbotran) for the detection and
characterization of hypervascular liver tumors (32 patients with 67 lesions). A greater confidence for
414
the visualization of lesions was achieved with gadobenate dimeglumine. Kim et al compared SPIO
(ferumoxides ) and mangafodipir trisodium in 64 patients. Ferumoxides performed better in the
detection of focal hepatic lesions, mainly hepatocellular carcinomas. Mangafodipir trisodium was
advantageous for the differentiation of hepatocellular and non-hepatocellular origin of the focal liver
lesions.415 Marugami et al compared Gd-EOB-DTPA and SPIO (ferumoxides ) enhanced MRI for the
detection of malignant liver tumors in 14 patients with 21 lesions; Gd-EOB-DTPA was superior to
416
SPIO-enhanced MRI.
© 2010 Elsevier
1 of 1 14-04-2010 20:39
CONCLUSION
Numerous specific MR contrast agents have been developed and experimentally explored. Few have
successfully passed clinical trials and received approval. Four approved specific contrast agents are
already available for hepatic MRI.7,8,308 Some of these are also suited for splenic or pancreatic MRI. It
can be expected that contrast-enhanced MRI of lymph nodes and bone marrow may become available
in the near future. Blood pool compounds will be the next group of approved specific contrast agents.
REFERENCES
1. Brasch RC: New directions in the development of MR imaging contrast media. Radiology 183(1):1-11, 1992.
2. Runge VM: Safety of approved MR contrast media for intravenous injection. J Magn Reson Imaging 12(2):205-213, 2000.
3. Bellin MF, Vasile M, Morel-Precetti S: Currently used non-specific extracellular MR contrast media. Eur Radiol
13(12):2688-2698, 2003.
4. Earls JP, Bluemke DA: New MR imaging contrast agents. Magn Reson Imaging Clin North Am 7(2):255-273, 1999.
5. Seltzer SE: The role of liposomes in diagnostic imaging. Radiology 171(1):19-21, 1989.
6. Wang YX, Hussain SM, Krestin GP: Superparamagnetic iron oxide contrast agents: physicochemical characteristics and
applications in MR imaging. Eur Radiol 11(11):2319-2331, 2001.
7. Reimer P, Balzer T: Ferucarbotran (Resovist): a new clinically approved RES-specific contrast agent for contrast-enhanced
MRI of the liver: properties, clinical development, and applications. Eur Radiol 13(6):1266-1276, 2003.
8. Reimer P, Tombach B: Hepatic MRI with SPIO: detection and characterization of focal liver lesions. Eur Radiol
8(7):1198-1204, 1998.
9. Winter TC 3rd, Freeny PC, Nghiem HV, et al: MR imaging with i.v. superparamagnetic iron oxide: efficacy in the detection of
focal hepatic lesions. Am J Roentgenol 161(6):1191-1198, 1993.
10. Schuhmann-Giampieri G: Liver contrast media for magnetic resonance imaging. Interrelations between pharmacokinetics
and imaging. Invest Radiol 28(8):753-761, 1993.
11. Ishida T, Murakami T, Kato N, et al: Superparamagnetic iron oxide enhanced magnetic resonance imaging of rat liver with
hepatocellular carcinoma and benign hyperplastic nodule. Invest Radiol 32(5):282-287, 1997.
12. Bellin MF, Zaim S, Auberton E, et al: Liver metastases: safety and efficacy of detection with superparamagnetic iron oxide
in MR imaging. Radiology 193(3):657-663, 1994.
13. Weissleder R, Elizondo G, Wittenberg J, et al: Ultrasmall superparamagnetic iron oxide: characterization of a new class of
contrast agents for MR imaging. Radiology 175(2):489-493, 1990.
14. Mergo PJ, Helmberger T, Nicolas AI, Ros PR: Ring enhancement in ultrasmall superparamagnetic iron oxide MR imaging:
a potential new sign for characterization of liver lesions. Am J Roentgenol 166(2):379-384, 1996.
15. Schima W: [Organ specific MRI contrast media in general practice]. Wien Med Wochenschr 113 (Suppl):8-11, 2002.
16. Ferrucci JT: Iron oxide enhanced MR imaging of the liver and spleen: review of the first five years. Keio J Med
40(4):206-214, 1991.
17. Clement O, Firja G, Chambon C, et al: Liver tumors in cirrhosis: experimental study with SPIO-enhanced MR imaging.
Radiology 180(1):31-36, 1991.
18. Ferrucci JT, Stark D: Iron oxide-enhanced MR imaging of the liver and spleen: review of the first 5 years. Am J Roentgenol
155(5):943-950, 1990.
19. Hamm B, Staks T, Taupitz M, et al: Contrast-enhanced MR imaging of liver and spleen: first experience in humans with a
new superparamagnetic iron oxide. J Magn Reson Imaging 4(5):659-668, 1994.
20. Ward J, Naik KS, Guthrie JA, et al: Hepatic lesion detection: comparison of MR imaging after the administration of
superparamagnetic iron oxide with dual-phase CT by using alternative-free response receiver operating characteristic
analysis. Radiology 210(2):459-466, 1999.
21. Ferrucci JT: Advances in abdominal MR imaging. Radiographics 18(6):1569-1586, 1998.
22. Mergo PJ, Engelken JD, Helmberger T, et al: MRI in focal liver disease: a comparison of small and ultra-small
superparamagnetic iron oxide as hepatic contrast agents. J Magn Reson Imaging 8(5):1073-1078, 1998.
23. Terrier F, Tourniaire J, Baclenger J, et al: Magnetic resonance imaging with superparamagnetic iron oxide particles to
evaluate hepatic macrophage-monocytic phagocytosis after arterial devascularization in minipigs. Acad Radiol 2(7):565-575,
1995.
24. Blakeborough A, Ward J, Wilson D, et al: Hepatic lesion detection at MR imaging: a comparative study with four
sequences. Radiology 203(3):759-765, 1997.
25. Taylor PM, Hawnaur JM, Hutchinson CE: Superparamagnetic iron oxide imaging of focal liver disease. Clin Radiol
50(4):215-219, 1995.
26. Chen F, Ward J, Robinson PJ: MR imaging of the liver and spleen: a comparison of the effects on signal intensity of two
superparamagnetic iron oxide agents. Magn Reson Imaging 17(4):549-556, 1999.
1 of 16 14-04-2010 20:40
27. Lwakatare F, Yamashita Y, Nakayama M, Takahashi M: SPIO-enhanced MR imaging of focal fatty liver lesions. Abdom
Imaging 26(2):157-160, 2001.
28. Furuhata T, Okita K, Tsuruma T, et al: Efficacy of SPIO-MR imaging in the diagnosis of liver metastases from colorectal
carcinomas. Dig Surg 20(4):321-325, 2003.
29. Takahama K, Amano Y, Hayashi H, et al: Detection and characterization of focal liver lesions using superparamagnetic iron
oxide-enhanced magnetic resonance imaging: comparison between ferumoxides-enhanced T1-weighted imaging and
delayed-phase gadolinium-enhanced T1-weighted imaging. Abdom Imaging 28(4):525-530, 2003.
30. Ward J, Guthrie JA, Wilson D, et al: Colorectal hepatic metastases: detection with SPIO-enhanced breath-hold MR
imaging-comparison of optimized sequences. Radiology 228(3):709-718, 2003.
31. Rofsky NM, Weinreb JC: Manganese (II) N,N'-dipyridoxylethylenediamine-N,N'-diacetate 5,5'-bis(phosphate): clinical
experience with a new contrast agent. Magn Reson Q 8(3):156-168, 1992.
32. Schima W, Petersein J, Hahn PF, et al: Contrast-enhanced MR imaging of the liver: comparison between Gd-BOPTA and
Mangafodipir. J Magn Reson Imaging 7(1):130-135, 1997.
33. Schima W, Saini S, Petersein J, et al: MR imaging of the liver with Gd-BOPTA: quantitative analysis of T1-weighted
images at two different doses. J Magn Reson Imaging 10(1):80-83, 1999.
34. Schneider G, Maas R, Schuitze Kool L, et al: Low-dose gadobenate dimeglumine versus standard dose gadopentetate
dimeglumine for contrast-enhanced magnetic resonance imaging of the liver: an intra-individual crossover comparison. Invest
Radiol 38(2):85-94, 2003.
35. Spinazzi A, Lorusso V, Pirovano G, Kirchin M: Safety, tolerance, biodistribution, and MR imaging enhancement of the liver
with gadobenate dimeglumine: results of clinical pharmacologic and pilot imaging studies in nonpatient and patient volunteers.
Acad Radiol 6(5):282-291, 1999.
36. Spinazzi A, Lorusso V, Pirovano G, et al: Multihance clinical pharmacology: biodistribution and MR enhancement of the
liver. Acad Radiol 5 (Suppl 1):S86-89; discussion S93-94, 1998.
37. Aicher KP, Laniado M, Kopp AF, et al: Mn-DPDP-enhanced MR imaging of malignant liver lesions: efficacy and safety in
20 patients. J Magn Reson Imaging 3(5):731-737, 1993.
38. Bernardino ME, Young SW, Lee JK, Weinreb JC: Hepatic MR imaging with Mn-DPDP: safety, image quality, and
sensitivity. Radiology 183(1):53-58, 1992.
39. Clement O, Muhler A, Vexler V, et al: Gadolinium-ethoxybenzyl-DTPA, a new liver-specific magnetic resonance contrast
agent. Kinetic and enhancement patterns in normal and cholestatic rats. Invest Radiol 27(8):612-619, 1992.
page 402
page 403
40. Clement O, Muhler A, Vexler VS, et al: Comparison of Gd-EOB-DTPA and Gd-DTPA for contrast-enhanced MR imaging of
liver tumors. J Magn Reson Imaging 3(1):71-77, 1993.
41. Elizondo G, Fretz CJ, Stark DD, et al: Preclinical evaluation of MnDPDP: new paramagnetic hepatobiliary contrast agent
for MR imaging. Radiology 178(1):73-78, 1991.
42. Hamm B, Staks T, Muhler A, et al: Phase I clinical evaluation of Gd-EOB-DTPA as a hepatobiliary MR contrast agent:
safety, pharmacokinetics, and MR imaging. Radiology 195(3):785-792, 1995.
43. Kopp AF, Laniado M, Aicher KP, et al: [Manganese DPDP as a contrast medium for MR tomography of focal liver lesions.
Tolerance and image quality in 20 patients]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 157(6):539-547, 1992.
44. Marchal G, Zhang X, Ni Y, et al: Comparison between Gd-DTPA, Gd-EOB-DTPA, and Mn-DPDP in induced HCC in rats: a
correlation study of MR imaging, microangiography, and histology. Magn Reson Imaging 11(5):665-674, 1993.
45. Petersein J, Spinnazzi A, Giovagnionit A, et al: Focal liver lesions: evaluation of the efficacy of gadobenate dimeglumine in
MR imaging-a multicenter phase III clinical study. Radiology 215(3):727-736, 2000.
46. Runge VM, Kirsch JE, Wells JW, et al: Enhanced liver MR: contrast agents and imaging strategy. Crit Rev Diagn Imaging
34(1-2):1-30, 1993.
47. Runge VM, Lee C, Williams NM: Detectability of small liver metastases with gadolinium BOPTA. Invest Radiol
32(9):557-565, 1997.
48. Runge VM: Liver MRI with contrast enhancement. Crit Rev Diagn Imaging 38(3):207-230, 1997.
49. Rummeny EJ, Torres CG, Kurdziel JC, et al: MnDPDP for MR imaging of the liver. Results of an independent image
evaluation of the European phase III studies. Acta Radiol 38(4 Pt 2):638-642, 1997.
50. Schuhmann-Giampieri G, Schmitt-Willich H, Press WR, et al: Preclinical evaluation of Gd-EOB-DTPA as a contrast agent
in MR imaging of the hepatobiliary system. Radiology 183(1):59-64, 1992.
51. Thorstensen O, Isberg B, Jorulf H, et al: MnDPDP enhancement in rabbit liver after intravenous bolus injection and slow
infusion. Acta Radiol 38(4 Pt 2):717-723, 1997.
52. Torres CG, Lundby B, Sterud AT, et al: MnDPDP for MR imaging of the liver. Results from the European phase III studies.
Acta Radiol 38(4 Pt 2):631-637, 1997.
53. Van Beers BE, Grandin C, Pauwels S, et al: Gd-EOB-DTPA enhancement pattern of hepatocellular carcinomas in rats:
comparison with Tc-99m-IDA uptake. J Magn Reson Imaging 4(3):351-354, 1994.
54. Vander Elst L, Maton F, Laurent S, et al: A multinuclear MR study of Gd-EOB-DTPA: comprehensive preclinical
characterization of an organ specific MRI contrast agent. Magn Reson Med 38(4):604-614, 1997.
2 of 16 14-04-2010 20:40
55. Vander Elst L, Maton F, Laurent S, et al: Multinuclear MR characterization of a new hepatobiliary contrast agent.
Preliminary results. Acta Radiol 412(Suppl):135-138, 1997.
56. Wang C, Gordon PB, Hustvedt SO, et al: MR imaging properties and pharmacokinetics of MnDPDP in healthy volunteers.
Acta Radiol 38(4 Pt 2):665-676, 1997.
57. Murakami T, Baron RL, Peterson MS, et al: Hepatocellular carcinoma: MR imaging with mangafodipir trisodium
(Mn-DPDP). Radiology 200(1):69-77, 1996.
58. Weissleder R, Elizondo G, Wittenberg J, et al: Ultrasmall superparamagnetic iron oxide: an intravenous contrast agent for
assessing lymph nodes with MR imaging. Radiology 175(2):494-498, 1990.
59. Frank H, Weissleder R, Brady T: Enhancement of MR angiography with iron oxide: Preliminary studies in whole-blood
phantom and in animals. Am J Roentgenol 162:209-213, 1994.
60. Anzai Y, Prince MR, Chenevert TL, et al: MR angiography with an ultrasmall superparamagnetic iron oxide blood pool
agent. J Magn Reson Imaging 7(1):209-214, 1997.
61. Loubeyre P, Zhao S, Canet E, et al: Ultrasmall superparamagnetic iron oxide particles (AMI 227) as a blood pool contrast
agent for MR angiography: experimental study in rabbits. J Magn Reson Imaging 7(6):958-962, 1997.
62. Rohl L, Ostergaard L, Simonsen CZ, et al: NC100150-enhanced 3D-SPGR MR angiography of the common carotid artery
in a pig vascular stenosis model. Quantification of stenosis and dose optimization. Acta Radiol 40(3):282-290, 1999.
63. Corot C, Violas X, Robert P, et al: Comparison of different types of blood pool agents (P792, MS325, USPIO) in a rabbit
MR angiography-like protocol. Invest Radiol 38(6):311-319, 2003.
64. Saini S, Sharma R, Baron RL, et al: Multicentre dose-ranging study on the efficacy of USPIO ferumoxtran-10 for liver MR
imaging. Clin Radiol 55(9):690-695, 2000.
65. Sharma R et al: Safety profile of ultrasmall superparamagnetic iron oxide ferumoxtran-10: phase II clinical trial data. J Magn
Reson Imaging 9(2):291-294, 1999.
66. Saini S, Edelman RR, Sharma P, et al: Blood pool MR contrast material for detection and characterization of focal hepatic
lesions: initial clinical experience with ultrasmall superparamagnetic iron oxide (AMI-227). Am J Roentgenol
164(5):1147-1152, 1995.
67. Guimaraes R, Clement O, Bittoun J, et al: MR lymphography with superparamagnetic iron nanoparticles in rats: pathologic
basis for contrast enhancement. Am J Roentgenol 162(1):201-207, 1994.
68. Small WC, Nelson RC, Bernardino ME: Dual contrast enhancement of both T1- and T2-weighted sequences using
ultrasmall superparamagnetic iron oxide. Magn Reson Imaging 11(5):645-654, 1993.
69. Saeed M, Wendland MF, Higgins CB: Blood pool MR contrast agents for cardiovascular imaging. J Magn Reson Imaging
12(6):890-898, 2000.
70. Port M, Meyer D, Bonnemain B, et al: P760 and P775: MRI contrast agents characterized by new pharmacokinetic
properties. Magma 8(3):172-176, 1999.
71. Grist TM, Korosec FR, Peters PG, et al: Steady-state and dynamic MR angiography with MS-325: initial experience in
humans. Radiology 207(2):539-544, 1998.
72. Lauffer RB, Parmelle DJ, Dunham SU, et al: MS-325: albumin-targeted contrast agent for MR angiography. Radiology
207(2):529-538, 1998.
73. Perreault P, Edelman MA, Baum RA, et al: MR angiography with gadofosveset trisodium for peripheral vascular disease:
phase II trial. Radiology 229(3):811-820, 2003.
74. Weinmann HJ, Ebert W, Missewitz B, Schmitt-Willich H: Tissue-specific MR contrast agents. Eur J Radiol 46(1):33-44,
2003.
75. Taupitz M, Schmitz S, Hamm B: [Superparamagnetic iron oxide particles: current state and future development]. Rofo
Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 175(6):752-765, 2003.
76. Adzamli K, Yablonskiy DA, Chicoine MR, et al: Albumin-binding MR blood pool agents as MRI contrast agents in an
intracranial mouse glioma model. Magn Reson Med 49(3):586-590, 2003.
77. Kim S, Kim JH, Moon WK, Min BG: Lesion characterization using vessel permeability map to new blood pool contrast
agent calculated from dynamic magnetic resonance images. J Digit Imaging 13(2 Suppl 1):193-195, 2000.
78. Unger EC, Winokur T, MacDougall P, et al: Hepatic metastases: liposomal Gd-DTPA-enhanced MR imaging. Radiology
171(1):81-85, 1989.
79. Unger EC, Shen DK, Fritz TA: Status of liposomes as MR contrast agents. J Magn Reson Imaging 3(1):195-198, 1993.
80. Unger EC, MacDougall P, Cullis P, et al: Liposomal Gd-DTPA: effect of encapsulation on enhancement of hepatoma
model by MRI. Magn Reson Imaging 7(4):417-423, 1989.
81. Weissleder R, Reimer P, Lee AS, et al: MR receptor imaging: ultrasmall iron oxide particles targeted to asialoglycoprotein
receptors. Am J Roentgenol 155(6):1161-1167, 1990.
82. Josephson L, Groman EV, Menz E, et al: A functionalized superparamagnetic iron oxide colloid as a receptor directed MR
contrast agent. Magn Reson Imaging 8(5):637-646, 1990.
83. Reimer P, Weissleder R, Lee R, et al: Receptor imaging: application to MR imaging of liver cancer. Radiology
177(3):729-734, 1990.
84. Reimer P, Weissleder R, Lee R, et al: Asialoglycoprotein receptor function in benign liver disease: evaluation with MR
3 of 16 14-04-2010 20:40
imaging. Radiology 178(3):769-774, 1991.

85. Reimer P, Weissleder R, Brady TJ, et al: Experimental hepatocellular carcinoma: MR receptor imaging. Radiology
180(3):641-645, 1991.
86. Groman EV, Enriquez PM, Jung C, Josephson L: Arabinogalactan for hepatic drug delivery. Bioconjug Chem
5(6):547-556, 1994.
87. Schaffer BK, Linker C, Papisov M, et al: MION-ASF: biokinetics of an MR receptor agent. Magn Reson Imaging
11(3):411-417, 1993.
88. Kobayashi H, Kawamoto S, Saga T, et al: Novel liver macromolecular MR contrast agent with a polypropylenimine
diaminobutyl dendrimer core: comparison to the vascular MR contrast agent with the polyamidoamine dendrimer core. Magn
Reson Med 46(4):795-802, 2001.
89. Weissleder R, Stark DD, Compton CC, et al: Ferrite-enhanced MR imaging of hepatic lymphoma: an experimental study in
rats. Am J Roentgenol 149(6):1161-1165, 1987.
90. Weissleder R, Saini S, Stark DD, et al: Dual-contrast MR imaging of liver cancer in rats. Am J Roentgenol 150(3):561-566,
1988.
91. Weissleder R, Saini S, Stark DD, et al: Pyogenic liver abscess: contrast-enhanced MR imaging in rats. Am J Roentgenol
150(1):115-120, 1988.
92. Weissleder R, Hahn PF, Stark DD, et al: MR imaging of splenic metastases: ferrite-enhanced detection in rats. Am J
Roentgenol 149(4):723-726, 1987.
93. Weissleder R, Stark DD, Ruromeny EJ, et al: Splenic lymphoma: ferrite-enhanced MR imaging in rats. Radiology
166(2):423-430, 1988.
94. Harisinghani M, Saini S, Weissleder R, et al: Splenic imaging with ultrasmall superparamagnetic iron oxide ferumoxtran-10
(AMI-7227): preliminary observations. J Comput Assist Tomogr 25(5):770-776, 2001.
95. Reimer P, Weissleder R: [Development and experimental use of receptor-specific MR contrast media]. Radiologe
36(2):153-163, 1996.
96. Blume J, Palmie S, Ave B, et al: Pancreatic contrast enhancement with Mn-bis pyridoxal ethylene diamine diacetic acid
and the influence of a hormonal stimulation of the pancreas in pigs. Acad Radiol 1(3):253-260, 1994.
97. Reimer P, Weissleder R, Shen TT, et al: Pancreatic receptors: initial feasibility studies with a targeted contrast agent for
98. Shen TT, Bogdanov A, Bodanov A, et al: Magnetically labeled secretin retains receptor affinity to pancreas acinar cells.
Bioconjug Chem 7(3):311-316, 1996.
99. Lee SS, Goo HW, Park SB, et al: MR imaging of reperfused myocardial infarction: comparison of necrosis-specific and
intravascular contrast agents in a cat model. Radiology 226(3):739-747, 2003.
100. Saeed M, Wagner S, Wendland MF, et al: Occlusive and reperfused myocardial infarcts: differentiation with Mn-DPDP-
enhanced MR imaging. Radiology 172(1):59-64, 1989.
page 403
page 404
101. Lonser RR, Weil RJ, Morrison PF, et al: Direct convective delivery of macromolecules to peripheral nerves. J Neurosurg
89(4):610-615, 1998.
102. Enochs WS, Schaffer B, Bhide PG, et al: MR imaging of slow axonal transport in vivo. Exp Neurol 123(2):235-242, 1993.
103. Van Everdinger KJ, Enochs WS, Bhide PG, et al: Determinants of in vivo MR imaging of slow axonal transport. Radiology
193(2):485-491, 1994.
104. Vassallo P, Matei C, Heston WD, et al: AMI-227-enhanced MR lymphography: usefulness for differentiating reactive from
tumor-bearing lymph nodes. Radiology 193(2):501-506, 1994.
105. Hamm B, Taupitz M, Hussmann P, et al: MR lymphography with iron oxide particles: dose-response studies and pulse
sequence optimization in rabbits. Am J Roentgenol 158(1):183-190, 1992.
106. Harika L, Weissleder R, Poss K, et al: MR lymphography with a lymphotropic T1-type MR contrast agent: Gd-DTPA-PGM.
Magn Reson Med 33(1):88-92, 1995.
107. Weissleder R, Elizondo G, Josephson L, et al: Experimental lymph node metastases: enhanced detection with MR
lymphography. Radiology 171(3):835-839, 1989.
108. Lee AS, Weissheder R, Brady TJ, et al: Lymph nodes: microstructural anatomy at MR imaging. Radiology
178(2):519-522, 1991.
109. Bengele HH, Palmacci S, Rogers J, et al: Biodistribution of an ultrasmall superparamagnetic iron oxide colloid, BMS
180549, by different routes of administration. Magn Reson Imaging 12(3):433-442, 1994.
110. Clement O, Rety F, Cuenod CA, et al: MR lymphography: evidence of extravasation of superparamagnetic nanoparticles
into the lymph. Acad Radiol 5 (Suppl 1):S170-172; discussion S183-184, 1998.
111. Elste V, Wagner S, Taupitz M, et al: Magnetic resonance lymphography in rats: effects of muscular activity and
hyperthermia on the lymph node uptake of intravenously injected superparamagnetic iron oxide particles. Acad Radiol
3(8):660-666, 1996.
112. Taupitz M, Wagner S, Hamm B, et al: Interstitial MR lymphography with iron oxide particles: results in tumor-free and VX2
4 of 16 14-04-2010 20:40
tumor-bearing rabbits. Am J Roentgenol 161(1):193-200, 1993.

113. Taupitz M, Wagner S, Hamm B, et al: MR lymphography using iron oxide particles. Detection of lymph node metastases in
the VX2 rabbit tumor model. Acta Radiol 34(1):10-15, 1993.
114. Taupitz M, Wagner S, Hamm B: [Contrast media for magnetic resonance tomographic lymph node diagnosis (MR
lymphography)]. Radiologe 36(2):134-140, 1996.
115. Anzai Y, McLachlan S, Morris M, et al: Dextran-coated superparamagnetic iron oxide, an MR contrast agent for assessing
lymph nodes in the head and neck. Am J Neuroradiol 15(1):87-94, 1994.
116. Anzai Y, Blackwell KE, Hirschowitz SL, et al: Initial clinical experience with dextran-coated superparamagnetic iron oxide
for detection of lymph node metastases in patients with head and neck cancer. Radiology 192(3):709-715, 1994.
117. Anzai Y, Piccoli CW, Outwater EK, et al: Evaluation of neck and body metastases to nodes with ferumoxtran 10-enhanced
MR imaging: phase III safety and efficacy study. Radiology 228(3):777-788, 2003.
118. Bellin MF, Beigelman C, Precetti-Morel S: Iron oxide-enhanced MR lymphography: initial experience. Eur J Radiol
34(3):257-264, 2000.
119. Bellin MF, Roy C, Kinkel K, et al: Lymph node metastases: safety and effectiveness of MR imaging with ultrasmall
superparamagnetic iron oxide particles-initial clinical experience. Radiology 207(3):799-808, 1998.
120. Bellin MF, Lebleu L, Meric JB: Evaluation of retroperitoneal and pelvic lymph node metastases with MRI and MR
lymphangiography. Abdom Imaging 28(2):155-163, 2003.
121. Gellissen J, Axmann CH, Prescher A, et al: Extra- and intracellular accumulation of ultrasmall superparamagnetic iron
oxides (USPIO) in experimentally induced abscesses of the peripheral soft tissues and their effects on magnetic resonance
imaging. Magn Reson Imaging 17(4):557-567, 1999.
122. Harisingani MG, Saini S, Hahn PF, et al: MR imaging of lymph nodes in patients with primary abdominal and pelvic
malignancies using ultrasmall superparamagnetic iron oxide (Combidex). Acad Radiol 5 (Suppl 1):S167-169, discussion
S183-184, 1998.
123. Harisinghani MG, Saini S, Weissleder R, et al: MR lymphangiography using ultrasmall superparamagnetic iron oxide in
patients with primary abdominal and pelvic malignancies: radiographic-pathologic correlation. Am J Roentgenol
172(5):1347-1351, 1999.
124. Hudgins PA, Anzai Y, Morris MR, et al: Ferumoxtran-10, a superparamagnetic iron oxide as a magnetic resonance
enhancement agent for imaging lymph nodes: a phase 2 dose study. Am J Neuroradiol 23(4):649-656, 2002.
125. Ishiyama K, Sashi R, Katayose Y, et al: [Sentinel lymph node after intramammary injection of superparamagnetic iron
oxide]. Nippon Igaku Hoshasen Gakkai Zasshi 62(13):744-746, 2002.
126. Laniado M, Kopp AF: [Current status of the clinical development of MR contrast media]. Rofo Fortschr Geb Rontgenstr
Neuen Bildgeb Verfahr 167(6):541-550, 1997.
127. Lind K, Kresse M, Debus NP, et al: A novel formulation for superparamagnetic iron oxide (SPIO) particles enhancing MR
lymphography: comparison of physicochemical properties and the in vivo behaviour. J Drug Target 10(3):221-230, 2002.
128. Mack MG, Balzer JO, Straub R, et al: Superparamagnetic iron oxide-enhanced MR imaging of head and neck lymph
nodes. Radiology 222(1):239-244, 2002.
129. McCauley TR, Rifkin MD, Ledet CA: Pelvic lymph node visualization with MR imaging using local administration of
ultra-small superparamagnetic iron oxide contrast. J Magn Reson Imaging 15(4):492-497, 2002.
130. McLachlan SJ, Morris MR, Lucas MA, et al: Phase I clinical evaluation of a new iron oxide MR contrast agent. J Magn
Reson Imaging 4(3):301-307, 1994.
131. Michel SC, Keller TM, Frohlick JM, et al: Preoperative breast cancer staging: MR imaging of the axilla with ultrasmall
superparamagnetic iron oxide enhancement. Radiology 225(2):527-536, 2002.
132. Nguyen BC, Stanford W, Thompson BH, et al: Multicenter clinical trial of ultrasmall superparamagnetic iron oxide in the
evaluation of mediastinal lymph nodes in patients with primary lung carcinoma. J Magn Reson Imaging 10(3):468-473, 1999.
133. Sigal R, Vogl T, Casselman J, et al: Lymph node metastases from head and neck squamous cell carcinoma: MR imaging
with ultrasmall superparamagnetic iron oxide particles (Sinerem MR)-results of a phase-III multicenter clinical trial. Eur Radiol
12(5):1104-1113, 2002.
134. Misselwitz B, Scmitt-Willich H, Michaelis M, Oellinger JJ: Interstitial magnetic resonance lymphography using a polymeric
t1 contrast agent: initial experience with Gadomer-17. Invest Radiol 37(3):146-151, 2002.
135. Herborn CU, Vogt FM, Lauevstein TC, et al: Assessment of normal, inflammatory, and tumor-bearing lymph nodes with
contrast-enhanced interstitial magnetic resonance lymphography: preliminary results in rabbits. J Magn Reson Imaging
18(3):328-335, 2003.
136. Herborn CU, Lavenstein TC, Vogt FM, et al: Interstitial MR lymphography with MS-325: characterization of normal and
tumor-invaded lymph nodes in a rabbit model. Am J Roentgenol 179(6):1567-1572, 2002.
137. Josephson L, Kircher MF, Mahmood U, et al: Near-infrared fluorescent nanoparticles as combined MR/optical imaging
probes. Bioconjug Chem 13(3):554-560, 2002.
138. Wunderbaldinger P, Josephson L, Weissleder R: Tat peptide directs enhanced clearance and hepatic permeability of
magnetic nanoparticles. Bioconjug Chem 13(2):264-268, 2002.
139. Seneterre E, Weissleder R, Jaramillo D, et al: Bone marrow: ultrasmall superparamagnetic iron oxide for MR imaging.
Radiology 179(2):529-533, 1991.
5 of 16 14-04-2010 20:40
140. Daldrup-Link HE, Rummeny EJ, Ihssen B, et al: Iron-oxide-enhanced MR imaging of bone marrow in patients with
non-Hodgkin's lymphoma: differentiation between tumor infiltration and hypercellular bone marrow. Eur Radiol
12(6):1557-1566, 2002.
141. Zhao M, Kircher MF, Josephson L, et al: Differential conjugation of tat peptide to superparamagnetic nanoparticles and its
effect on cellular uptake. Bioconjug Chem 13(4):840-844, 2002.
142. Bhorade R, Weissleder R, Nakakoshi T, et al: Macrocyclic chelators with paramagnetic cations are internalized into
mammalian cells via a HIV-tat derived membrane translocation peptide. Bioconjug Chem 11(3):301-305, 2000.
143. Ahrens ET, Feili-Hariti M, Xu H, et al: Receptor-mediated endocytosis of iron-oxide particles provides efficient labeling of
dendritic cells for in vivo MR imaging. Magn Reson Med 49(6):1006-1013, 2003.
144. Schulze E, Ferrucci JT, Poss K, et al: Cellular uptake and trafficking of a prototypical magnetic iron oxide label in vitro.
Invest Radiol 30(10):604-610, 1995.
145. Frank JA, Miller BR, Arbab AS, et al: Clinically applicable labeling of mammalian and stem cells by combining
superparamagnetic iron oxides and transfection agents. Radiology 228(2):480-487, 2003.
146. Daldrup-Link HE, Rudelius M, Oostendorp RA, et al: Targeting of hematopoietic progenitor cells with MR contrast agents.
Radiology 228(3):760-767, 2003.
147. Nelson JA, Schmiedl U: Porphyrins as contrast media. Magn Reson Med 22(2):366-371; discussion 378, 1991.
148. Schmiedl UP, Nelson JR, Starr FL, Schmidt R: Hepatic contrast-enhancing properties of manganese-mesoporphyrin and
manganese-TPPS4. A comparative magnetic resonance imaging study in rats. Invest Radiol 27(7):536-542, 1992.
149. Schiedl UP, Nelson JA, Teng L, et al: Magnetic resonance imaging of the hepatobiliary system: intestinal absorption
studies of manganese mesoporphyrin. Acad Radiol 2(11):994-1001, 1995.
150. Ni Y, Marchal G, Yu J, et al: Localization of metalloporphyrin-induced "specific" enhancement in experimental liver tumors:
comparison of magnetic resonance imaging, microangiographic, and histologic findings. Acad Radiol 2(8):687-699, 1995.
151. Hofmann B, Bogdanov A, Marecos E, et al: Mechanism of gadophrin-2 accumulation in tumor necrosis. J Magn Reson
Imaging 9(2):336-341, 1999.
152. Ni Y, Adzamli K, Miao Y, et al: MRI contrast enhancement of necrosis by MP-2269 and gadophrin-2 in a rat model of liver
infarction. Invest Radiol 36(2):97-103, 2001.
153. Saeed M: New concepts in characterization of ischemically injured myocardium by MRI. Exp Biol Med (Maywood)
226(5):367-376, 2001.
154. Ju Lee H, Kim IQ, Hyung Kim S, et al: Dynamic enhancement features of gadophrin-2 on magnetic resonance imaging: an
experimental model of VX2 carcinoma and bacterial abscess in rabbit thigh. Invest Radiol 37(12):663-671, 2002.
155. Saeed M, Watzinger N, Kirombach GF, et al: Left ventricular remodeling after infarction: sequential MR imaging with oral
nicorandil therapy in rat model. Radiology 224(3):830-837, 2002.
page 404
page 405
156. Saeed M, Lund G, Wendland MF, et al: Magnetic resonance characterization of the peri-infarction zone of reperfused
myocardial infarction with necrosis-specific and extracellular non-specific contrast media. Circulation 103(6):871-876, 2001.
157. Kim TK, Choi BI, Park SW, et al: Gadolinium mesoporphyrin as an MR imaging contrast agent in the evaluation of tumors:
an experimental model of VX2 carcinoma in rabbits. Am J Roentgenol 175(1):227-234, 2000.
158. Choi S, Choi SH, Kim ST, et al: Irreversibly damaged myocardium at MR imaging with a necrotic tissue-specific contrast
agent in a cat model. Radiology 215(3):863-868, 2000.
159. Saeed M, Bremerich J, Wendland MF, et al: Reperfused myocardial infarction as seen with use of necrosis-specific
versus standard extracellular MR contrast media in rats. Radiology 213(1):247-257, 1999.
160. Marchal G, Ni Y, Herijgers P, et al: Paramagnetic metalloporphyrins: infarct avid contrast agents for diagnosis of acute
myocardial infarction by MRI. Eur Radiol 6(1):2-8, 1996.
161. Pislaru SV, Ni Y, Pislaru C, et al: Noninvasive measurements of infarct size after thrombolysis with a necrosis-avid MRI
contrast agent. Circulation 99(5):690-696, 1999.
162. Ni Y, Pislaru C, Bosmans H, et al: Intracoronary delivery of Gd-DTPA and Gadophrin-2 for determination of myocardial
viability with MR imaging. Eur Radiol 11(5):876-883, 2001.
163. Saeed M, Wendland MF, Bremerich GL, et al: Assessment of myocardial viability using standard extracellular and
necrosis specific MR contrast media. Acad Radiol 9 (Suppl 1):S84-8, 2002.
164. Ni Y, Dymarkuwski S, Chen F, et al: Occlusive myocardial infarction enhanced or not enhanced with necrosis-avid
contrast agents at MR imaging. Radiology 225(2):603-605; author reply 605-606, 2002.
165. Choudhury L, Manrholdt H, Wagner A, et al: Myocardial scarring in asymptomatic or mildly symptomatic patients with
hypertrophic cardiomyopathy. J Am Coll Cardiol 40(12):2156-2164, 2002.
166. Rehwald WG, Fieno DS, Chen EL, et al: Myocardial magnetic resonance imaging contrast agent concentrations after
reversible and irreversible ischemic injury. Circulation 105(2):224-229, 2002.
167. Jeong AK, Choi SI, Kim DH, et al: Evaluation by contrast-enhanced MR imaging of the lateral border zone in reperfused
myocardial infarction in a cat model. Korean J Radiol 2(1):21-27, 2001.
168. Wu E, Judd RM, Vargas JD, et al: Visualisation of presence, location, and transmural extent of healed Q-wave and
6 of 16 14-04-2010 20:40
non-Q-wave myocardial infarction. Lancet 357(9249):21-28, 2001.

169. Kim RJ, Wu E, Rafael A, et al: The use of contrast-enhanced magnetic resonance imaging to identify reversible
myocardial dysfunction. N Engl J Med 343(20):1445-1453, 2000.
170. Choi SI, Jiang CZ, Lim KH, et al: Application of breath-hold T2-weighted, first-pass perfusion and gadolinium-enhanced
T1-weighted MR imaging for assessment of myocardial viability in a pig model. J Magn Reson Imaging 11(5):476-480, 2000.
171. Choi S, Lee SS, Choi SI, et al: Occlusive myocardial infarction: investigation of bis-gadolinium mesoporphyrins-enhanced
T1-weighted MR imaging in a cat model. Radiology 220(2):436-440, 2001.
172. Kim RJ, Chen EL, Lima JA, et al: Myocardial Gd-DTPA kinetics determine MRI contrast enhancement and reflect the
extent and severity of myocardial injury after acute reperfused infarction. Circulation 94(12):3318-3326, 1996.
173. Wagner A, Mahrholdt H, Sechtem U, et al: MR imaging of myocardial perfusion and viability. Magn Reson Imaging Clin
North Am 11(1):49-66, 2003.
174. Goo HW, Kim DH, Lee SS, et al: The significance of perfusion defect at myocardial perfusion MR imaging in a cat model
of acute reperfused myocardial infarction. Korean J Radiol 3(4):235-239, 2002.
175. Bello D, Shah DJ, Fatah GM, et al: Gadolinium cardiovascular magnetic resonance predicts reversible myocardial
dysfunction and remodeling in patients with heart failure undergoing beta-blocker therapy. Circulation 108(16):1945-1953,
2003.
176. Sandstede J: Assessment of myocardial viability by MR imaging. Eur Radiol 13(1):52-61, 2003.
177. Flacke S, Allen JS, Chia JM, et al: Characterization of viable and nonviable myocardium at MR imaging: comparison of
gadolinium-based extracellular and blood pool contrast materials versus manganese-based contrast materials in a rat
myocardial infarction model. Radiology 226(3):731-738, 2003.
178. Moore A, Weissleder R, Bogdanov A Jr: Uptake of dextran-coated monocrystalline iron oxides in tumor cells and
macrophages. J Magn Reson Imaging 7(6):1140-1145, 1997.
179. Moore A, Marecos E, Gogdanov A, et al: Tumoral distribution of long-circulating dextran-coated iron oxide nanoparticles
in a rodent model. Radiology 214(2):568-574, 2000.
180. Varallyay P, Nesbit G, Muldoon LL, et al: Comparison of two superparamagnetic viral-sized iron oxide particles
ferumoxides and ferumoxtran-10 with a gadolinium chelate in imaging intracranial tumors. Am J Neuroradiol 23(4):510-519,
2002.
181. Konda SD, Aref M, Brechbiel M, et al: Development of a tumor-targeting MR contrast agent using the high-affinity folate
receptor: work in progress. Invest Radiol 35(1):50-57, 2000.
182. Su MY, Wang Z, Carpenter PM, et al: Characterization of N-ethyl-N-nitrosourea-induced malignant and benign breast
tumors in rats by using three MR contrast agents. J Magn Reson Imaging 9(2):177-186, 1999.
183. Su MY, Muhler A, Lao X, et al: Tumor characterization with dynamic contrast-enhanced MRI using MR contrast agents of
various molecular weights. Magn Reson Med 39(2):259-269, 1998.
184. Roberts T, Roberts H, Brasch R: Optimizing imaging techniques to reduce errors in microvascular quantitation with
macromolecular MR contrast agents. Acad Radiol 5 (Suppl 1):S133-136; discussion S145-146, 1998.
185. Roberts TP, Turetschek K, Preda A, et al: Tumor microvascular changes to anti-angiogenic treatment assessed by MR
contrast media of different molecular weights. Acad Radiol 9 (Suppl 2):S511-513, 2002.
186. Bednarski MD, Lee JW, Callstrom MR, et al: In vivo target-specific delivery of macromolecular agents with MR-guided
focused ultrasound. Radiology 204(1):263-268, 1997.
187. Kresse M, Wagner S, Pfefferer D, et al: Targeting of ultrasmall superparamagnetic iron oxide (USPIO) particles to tumor
cells in vivo by using transferrin receptor pathways. Magn Reson Med 40(2):236-242, 1998.
Neoplasia 4(6):523-530, 2002.
189. Hogemann D, Josephson L, Weissleder R, et al: Improvement of MRI probes to allow efficient detection of gene
expression. Bioconjug Chem 11(6):941-946, 2000.
190. Moore A, Basilion JP, Chiocc EA, et al: Measuring transferrin receptor gene expression by NMR imaging. Biochim
Biophys Acta 1402(3):239-249, 1998.
191. Moore A, Josephson L, Bhorade L, et al: Human transferrin receptor gene as a marker gene for MR imaging. Radiology
221(1):244-250, 2001.
192. Weissleder R, Moore A, Mahmood U, et al: In vivo magnetic resonance imaging of transgene expression. Nat Med
6(3):351-355, 2000.
193. Luciani A, Olivier JC, Clement O, et al: Glucose-receptor MR imaging of tumors: study in mice with PEGylated
paramagnetic niosomes. Radiology 231(1):135-142, 2004.
194. Reimer P: Tumor-targeted MR contrast agents: hype or future hope? Radiology 231(1):1-2, 2004.
195. Choyke P, Dwyer A, Knopp M: Functional tumor imaging with dynamic contrast-enhanced magnetic resonance imaging. J
Magn Reson Imaging 17(5):509-520, 2003.
196. Brasch RC: Rationale and applications for macromolecular Gd-based contrast agents. Magn Reson Med 22(2):282-287;
discussion 300-303, 1991.
197. Wiener EC, Brechbiel MW, Brothers H, et al: Dendrimer-based metal chelates: a new class of magnetic resonance
7 of 16 14-04-2010 20:40
imaging contrast agents. Magn Reson Med 31(1):1-8, 1994.

198. Lu ZR, Parker DL, Goodrich KC, et al: Extracellular biodegradable macromolecular gadolinium(III) complexes for MRI.
Magn Reson Med 51(1):27-34, 2004.
199. Daldrup-Link H, Brasch R: Macromolecular contrast agents for MR mammography: current status. Eur Radiol
13(2):354-365, 2003.
200. Sato N, Kobayashi H, Kiraga A, et al: Pharmacokinetics and enhancement patterns of macromolecular MR contrast
agents with various sizes of polyamidoamine dendrimer cores. Magn Reson Med 46(6):1169-1173, 2001.
201. Kroft L, de Roos A: Blood pool contrast agents for cardiovascular MR imaging. J Magn Reson Imaging 10(3):395-403,
1999.
202. Kroft L, Doornbos J, Van der Geest RJ, et al: Ultrasmall superparamagnetic particles of iron oxide (USPIO) MR imaging
of infarcted myocardium in pigs. Magn Reson Imaging 16(7):755-763, 1998.
203. Shibata DK, Schmied UP, Yvan C, et al: Two prototype blood pool agents for contrast-enhanced magnetic resonance
angiography of the portal vein in pigs. Acad Radiol 2(8):705-708, 1995.
204. Abe Y, Yamashita Y, Namimoto T, et al: The value of fast and ultrafast T2-weighted MR imaging sequences in hepatic
enhancement with ferumoxides: comparison with conventional spin-echo sequence. Radiat Med 18(2):97-105, 2000.
205. Frank H, Weissleder R, Bogdanor AA, et al: Detection of pulmonary emboli by using MR angiography with MPEG-PL-
GdDTPA: an experimental study in rabbits. Am J Roentgenol 162(5):1041-1046, 1994.
206. Schmiedl U, Brasch RC, Ogan MD, et al: Albumin labeled with Gd-DTPA. An intravascular contrast-enhancing agent for
magnetic resonance blood pool and perfusion imaging. Acta Radiol 374 (Suppl):99-102, 1990.
207. Kobayashi H, Kawamoto S, Star RA, et al: Activated clearance of a biotinylated macromolecular MRI contrast agent from
the blood pool using an avidin chase. Bioconjug Chem 14(5):1044-1047, 2003.
208. Baxter A, Melnikoff S, Shites DP, et al: AUR Memorial Award 1991. Immunogenicity of gadolinium-based contrast agents
for magnetic resonance imaging. Induction and characterization of antibodies in animals. Invest Radiol 26(12):1035-1040,
1991.
209. Knopp M, Schoenberg SO, Rehm C, et al: Assessment of gadobenate dimeglumine for magnetic resonance
angiography: phase I studies. Invest Radiol 37(12):706-715, 2002.
210. Lohr HA, Froehlich JM, Pfyffer M, et al: Comparison of Gd-BOPTA and Gd-DOTA for peripheral CE-MRA: a double-blind
clinical study. Acad Radiol 9 (Suppl 2):S421-424, 2002.
211. Corot C, Violas X, Robert P, Port M: Pharmacokinetics of three gadolinium chelates with different molecular sizes shortly
after intravenous injection in rabbits: relevance to MR angiography. Invest Radiol 35(4):213-218, 2000.
212. Kobayashi H, Kawamoto S, Jo SK, et al: Renal tubular damage detected by dynamic micro-MRI with a dendrimer-based
magnetic resonance contrast agent. Kidney Int 61(6):1980-1985, 2002.
page 405
page 406
213. Aref M, Brechbiel M, Wiener E: Identifying tumor vascular permeability heterogeneity with magnetic resonance imaging
contrast agents. Invest Radiol 37(4):178-192, 2002.
214. Kobayashi H, Kawamoto S, Saga T, et al: Micro-MR angiography of normal and intratumoral vessels in mice using
dedicated intravascular MR contrast agents with high generation of polyamidoamine dendrimer core: reference to
pharmacokinetic properties of dendrimer-based MR contrast agents. J Magn Reson Imaging 14(6):705-713, 2001.
215. Blustajn J, Cuenod CA, Clement O, et al: Measurement of liver blood volume using a macromolecular MRI contrast agent
at equilibrium. Magn Reson Imaging 15(4):415-421, 1997.
216. Dardzinski BJ, Schmithorst VJ, Holland SK, et al: MR imaging of murine arthritis using ultrasmall superparamagnetic iron
oxide particles. Magn Reson Imaging 19(9):1209-1216, 2001.
217. Kaim A, Jundt G, Wischer T, et al: Functional-morphologic MR imaging with ultrasmall superparamagnetic particles of iron
oxide in acute and chronic soft-tissue infection: study in rats. Radiology 227(1):169-174, 2003.
218. Jo S, Hu X, Kobayshi H, et al: Detection of inflammation following renal ischemia by magnetic resonance imaging. Kidney
Int 64(1):43-51, 2003.
219. Weissleder R, Lee AS, Fischam AJ, et al: Polyclonal human immunoglobulin G labeled with polymeric iron oxide: antibody
220. Pirko I, Ciric B, Johnson AJ, et al: Magnetic resonance imaging of immune cells in inflammation of central nervous
system. Croat Med J 44(4):463-468, 2003.
221. Gupta H, Willinson RA, Bogdanov AA, et al: Inflammation: imaging with methoxy poly(ethylene glycol)-poly-L-lysine-DTPA,
a long-circulating graft copolymer. Radiology 197(3):665-669, 1995.
222. Burtea C, Laurent S, Coler JM, et al: Development of new glucosylated derivatives of gadolinium
diethylenetriaminepentaacetic for magnetic resonance angiography. Invest Radiol 38(6):320-333, 2003.
223. Desser T, Rubin DL, Muller HH, et al: Dynamics of tumor imaging with Gd-DTPA-polyethylene glycol polymers:
dependence on molecular weight. J Magn Reson Imaging 4(3):467-472, 1994.
224. Sharafuddin MJ, Stolpen AH, Dang YM, et al: Comparison of MS-325- and gadodiamide-enhanced MR venography of
iliocaval veins. J Vasc Interv Radiol 13(10):1021-1027, 2002.
225. Kuehne T, Saeed M, Moore P, et al: Influence of blood pool contrast media on MR imaging and flow measurements in the
8 of 16 14-04-2010 20:40
presence of pulmonary arterial stents in swine. Radiology 223(2):439-445, 2002.

226. Allkemper T, Bremer C, Matuszewski L, et al: Contrast-enhanced blood pool MR angiography with optimized iron oxides:
effect of size and dose on vascular contrast enhancement in rabbits. Radiology 223(2):432-438, 2002.
227. Hoffmann U, Schima W, Herold C: Pulmonary magnetic resonance angiography. Eur Radiol 9(9):1745-1754, 1999.
228. Schmitz S, Albrecht T, Wolf K: MR angiography with superparamagnetic iron oxide: feasibility study. Radiology
213(2):603-607, 1999.
229. Wacker FK, Wendt M, Ebert W, et al: Use of a blood pool contrast agent for MR-guided vascular procedures: feasibility of
ultrasmall superparamagnetic iron oxide particles. Acad Radiol 9(11):1251-1254, 2002.
230. Schmitz SA, Albrecht T, Jenson K, Wolf KJ: [Superparamagnetic iron oxide enhanced MR angiography of the portal vein
system]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 172(1):51-54, 2000.
231. Weishaupt D, Hetzer FH, Ruehm SG, et al: Three-dimensional contrast-enhanced MRI using an intravascular contrast
agent for detection of traumatic intra-abdominal hemorrhage and abdominal parenchymal injuries: an experimental study. Eur
Radiol 10(12):1958-1964, 2000.
232. Weissleder R, Bogdanov A, Tung CH, et al: Size optimization of synthetic graft copolymers for in vivo angiogenesis
imaging. Bioconjug Chem 12(2):213-219, 2001.
233. Clarke SE, Weinman HJ, Dai E, et al: Comparison of two blood pool contrast agents for 0.5-T MR angiography:
experimental study in rabbits. Radiology 214(3):787-794, 2000.
234. Tombach B, Reimer P, Mahler M, et al: First-pass and equilibrium phase MRA following intravenous bolus injection of SH
U 555 C: Phase I clinical trial in elderly volunteers with risk factors for arterial vascular disease. Acad Radiol 9 (Suppl
2):S425-427, 2002.
235. Wacker FK, Reither K, Ebert W, et al: MR image-guided endovascular procedures with the ultrasmall superparamagnetic
iron oxide SH U 555 C as an intravascular contrast agent: study in pigs. Radiology 226(2):459-464, 2003.
236. Bayes M, Rabasseda X, Prous J: Gateways to clinical trials. Methods Find Exp Clin Pharmacol 25(9):747-771, 2003.
237. Kotek J, Lebduskovai P, Herman P, et al: Lanthanide(III) complexes of novel mixed carboxylic-phosphorus acid
derivatives of diethylenetriamine: a step towards more efficient MRI contrast agents. Chemistry 9(23):5899-5915, 2003.
238. Jerosch-Herold M, Hu X, Murthy NS, et al: Magnetic resonance imaging of myocardial contrast enhancement with MS-325
and its relation to myocardial blood flow and the perfusion reserve. J Magn Reson Imaging 18(5):544-554, 2003.
239. Krause MH, Kwong KK, Xiong J, et al: MRI of blood volume with MS 325 in experimental choroidal melanoma. Magn
Reson Imaging 21(7):725-732, 2003.
240. Maravilla KR, Cao Y, Heiman JR, et al: Serial MR imaging with MS-325 for evaluating female sexual arousal response:
determination of intrasubject reproducibility. J Magn Reson Imaging 18(2):216-224, 2003.
241. Lei T, Udupa JK, Odhner D, et al: 3DVIEWNIX-AVS: a software package for the separate visualization of arteries and
veins in CE-MRA images. Comput Med Imaging Graph 27(5):351-362, 2003.
242. Caravan P, Astashkin A, Raitsimring A: The gadolinium(III)-water hydrogen distance in MRI contrast agents. Inorg Chem
42(13):3972-3974, 2003.
243. Suh DD, Yang CC, Cao Y, et al: Magnetic resonance imaging anatomy of the female genitalia in premenopausal and
postmenopausal women. J Urol 170(1):138-144, 2003.
244. Deliganis A, Maravilla KR, Heiman JR, et al: Female genitalia: dynamic MR imaging with use of MS-325 initial
experiences evaluating female sexual response. Radiology 225(3):791-799, 2002.
245. Jiang Y, Zhao JJ, Tang H, et al: MRI of antigen-induced arthritis in rabbits: improved contrast and disease characterization
with the blood pool agent MS-325. Acad Radiol 9 (Suppl 2):S480, 2002.
246. Herborn CU, Waldschuetz R, Lavenstein TC, et al: Contrast-enhanced magnetic resonance imaging (MS-325) in a murine
model of systemic lupus erythematosus. Invest Radiol 37(8):464-469, 2002.
247. Krombach GA, Higgens CB, Baireuther R, et al: [MR contrast media for cardiovascular imaging]. Rofo Fortschr Geb
Rontgenstr Neuen Bildgeb Verfahr 174(7):819-829, 2002.
248. Bluemke D, Kraichman DL, Heldman A, et al: Optimal characterization of myocardial perfusion with AngioMARK. Acad
Radiol 9 (Suppl 1):S78, 2002.
249. Weinmann H, Bauer H, Ebert W, et al: Comparative studies on the efficacy of MRI contrast agents in MRA. Acad Radiol 9
(Suppl 1):S135-136, 2002.
250. Lei T, Udupa JK, Saha PK, et al: 3D MRA visualization and artery-vein separation using blood pool contrast agent
MS-325. Acad Radiol 9 (Suppl 1):S127-133, 2002.
251. Caravan P, Cloutier NJ, Greenfield MT, et al: The interaction of MS-325 with human serum albumin and its effect on
proton relaxation rates. J Am Chem Soc 124(12):3152-3162, 2002.
252. Kraitchman D, Chin BB, Heldman AW, et al: MRI detection of myocardial perfusion defects due to coronary artery
stenosis with MS-325. J Magn Reson Imaging 15(2):149-158, 2002.
253. Turetschek K, Floyd E, Helbich T, et al: MRI assessment of microvascular characteristics in experimental breast tumors
using a new blood pool contrast agent (MS-325) with correlations to histopathology. J Magn Reson Imaging 14(3):237-242,
2001.
254. Lahti KM, Lauffer RB, Chan T, Weisskoff RM: Magnetic resonance angiography at 0.3 T using MS-325. Magma
9 of 16 14-04-2010 20:40
12(2-3):88-91, 2001.
255. Caravan P, Comuzzi C, Crooks W, et al: Thermodynamic stability and kinetic inertness of MS-325, a new blood pool
agent for magnetic resonance imaging. Inorg Chem 40(9):2170-2176, 2001.
256. Bluemke DA, Stillman AE, Bis KG, et al: Carotid MR angiography: phase II study of safety and efficacy for MS-325.
Radiology 219(1):114-122, 2001.
257. Arildsen H, Christiansen EH, Pedersen AK, et al: Reproducibility of QT parameters derived from 24-hour ambulatory
ECG recordings in healthy subjects. Ann Noninvasive Electrocardiol 6(1):24-31, 2001.
258. Mahfouz AE: Ms-325 Epix. Curr Opin Investig Drugs 1(4):476-480, 2000.
259. Straub V, Donahue KM, Allamand V, et al: Contrast agent-enhanced magnetic resonance imaging of skeletal muscle
damage in animal models of muscular dystrophy. Magn Reson Med 44(4):655-659, 2000.
260. Deliganis AV, Maravilla KR, Heiman JR, et al: Lecture 3: dynamic MR imaging of the female genitalia using AngioMARK:
initial experience evaluating the female sexual response. Int J Impot Res 12 (Suppl 3):S45, 2000.
261. Kraitchman D, Bluemke DA, Chin BB, et al: A minimally invasive method for creating coronary stenosis in a swine model
for MRI and SPECT imaging. Invest Radiol 35(7):445-451, 2000.
262. Aime S, Chiaussa M, Digilio G, et al: Contrast agents for magnetic resonance angiographic applications:1H and 17O
NMR relaxometric investigations on two gadolinium(III) DTPA-like chelates endowed with high binding affinity to human serum
albumin. J Biol Inorg Chem 4(6):766-774, 1999.
263. Stuber M, Botnar RM, Danias PG, et al: Contrast agent-enhanced, free-breathing, three-dimensional coronary magnetic
resonance angiography. J Magn Reson Imaging 10(5):790-799, 1999.
264. Prasad PV, Cannillo J, Chavez DR, et al: First-pass renal perfusion imaging using MS-325, an albumin-targeted MRI
contrast agent. Invest Radiol 34(9):566-571, 1999.
265. Li D, Dolan RP, Walovitch RC, Lauffer RB: Three-dimensional MRI of coronary arteries using an intravascular contrast
agent. Magn Reson Med 39(6):1014-1018, 1998.
266. Prasad PV, Cannillo J, Chavez DR, et al: Contrast-enhanced MR angiography and first-pass renal perfusion imaging
using MS-325, an intravascular contrast agent. Acad Radiol 5 (Suppl 1):S219-222; discussion S226-227, 1998.
267. Parmelee DJ, Walovitch RC, Ouellet HS, Lauffer RB: Preclinical evaluation of the pharmacokinetics, biodistribution, and
elimination of MS-325, a blood pool agent for magnetic resonance imaging. Invest Radiol 32(12):741-747, 1997.
268. Lin W, Abendschein DR, Celik A, et al: Intravascular contrast agent improves magnetic resonance angiography of carotid
arteries in minipigs. J Magn Reson Imaging 7(6):963-971, 1997.
269. Lin W, Abendschein D, Haacke E: Contrast-enhanced magnetic resonance angiography of carotid arterial wall in pigs. J
270. Lauffer RB, Parmelee DJ, Ouellet HS, et al: MS-325: a small-molecule vascular imaging agent for magnetic resonance
imaging. Acad Radiol 3 (Suppl 2):S356-358, 1996.
page 406
page 407
271. Daldrup HE, Roberts TP, Muhler A, et al: [Macromolecular contrast media for MR mammography. A new approach to
characterizing breast tumors]. Radiologe 37(9):733-740, 1997.
272. Knollman FD, Bock JC, Rautenberg K, et al: Differences in predominant enhancement mechanisms of
superparamagnetic iron oxide and ultrasmall superparamagnetic iron oxide for contrast-enhanced portal magnetic resonance
angiography. Preliminary results of an animal study original investigation. Invest Radiol 33(9):637-643, 1998.
273. Lawaczeck R, Bauer H, Frenzel T, et al: Magnetic iron oxide particles coated with carboxydextran for parenteral
administration and liver contrasting. Pre-clinical profile of SH U555A. Acta Radiol 38(4 Pt 1):584-597, 1997.
274. Reimer P et al: Myocardial perfusion and contrast-enhanced 3D MRA of the chest with a novel intravascular iron-oxide
based contrast agent (SH U 555 C): results of a placebo-controlled clinical phase I study. Radiology 231:474-481, 2004.
275. Weissleder R: Liver MR imaging with iron oxides: toward consensus and clinical practice. Radiology 193(3):593-595,
1994.
276. Mandeville JB, Moore J, Chesler DA, et al: Dynamic liver imaging with iron oxide agents: effects of size and
biodistribution on contrast. Magn Reson Med 37(6):885-890, 1997.
277. Josephson L, Lewis J, Jacobs P, et al: The effects of iron oxides on proton relaxivity. Magn Reson Imaging 6(6):647-653,
1988.
278. Hahn P, Stark D, Ferrucci J: Accumulation of iron oxide particles around liver metastases during MR imaging.
Gastrointest Radiol 17(2):173-174, 1992.
279. Hahn P, Saini S: Liver-specific MR imaging contrast agents. Radiol Clin North Am 36(2):287-297, 1998.
280. Mayo-Smith WW, Saini S, Slater G, et al: MR contrast material for vascular enhancement: value of superparamagnetic
iron oxide. Am J Roentgenol 166(1):73-77, 1996.
281. Stark DD, Weissleder R, Elizondo G, et al: Superparamagnetic iron oxide: clinical application as a contrast agent for MR
imaging of the liver. Radiology 168(2):297-301, 1988.
282. Ros PR, Freeny PC, Harms SE, et al: Hepatic MR imaging with ferumoxides: a multicenter clinical trial of the safety and
efficacy in the detection of focal hepatic lesions. Radiology 196(2):481-488, 1995.
10 of 16 14-04-2010 20:40
283. Duda SH, Laniado M, Kopp AF, et al: Superparamagnetic iron oxide: detection of focal liver lesions at high-field-strength
MR imaging. J Magn Reson Imaging 4(3):309-314, 1994.
284. Deys A, Arrive L, Marx C, et al: Hepatic tumors: detection and characterization at 1-T MR imaging enhanced with AMI-25.
Radiology 193(3):665-669, 1994.
285. Strotzer M, Gmeinwieser J, Schmidt J, et al: Diagnosis of liver metastases from colorectal adenocarcinoma. Comparison
of spiral-CTAP combined with intravenous contrast-enhanced spiral-CT and SPIO-enhanced MR combined with plain MR
imaging. Acta Radiol 38(6):986-992, 1997.
286. Oudkerk M, Van den Heuvel AG, Wielopolsk PA, et al: Hepatic lesions: detection with ferumoxide-enhanced T1-weighted
287. Hahn PF, Stark DD, Weissleder R, et al: Clinical application of superparamagnetic iron oxide to MR imaging of tissue
perfusion in vascular liver tumors. Radiology 174(2):361-366, 1990.
288. Shamsi K, Balzer T, Saini C, et al: Superparamagnetic iron oxide particles (SH U 555 A): evaluation of efficacy in three
doses for hepatic MR imaging. Radiology 206(2):365-371, 1998.
289. Reimer P, Muller M, Marx C, et al: T1 effects of a bolus-injectable superparamagnetic iron oxide, SH U 555 A:
dependence on field strength and plasma concentration-preliminary clinical experience with dynamic T1-weighted MR imaging.
Radiology 209(3):831-836, 1998.
290. Reimer P, Allkemper T, Matuszewski L, Balzer T: Contrast-enhanced 3D-MRA of the upper abdomen with a bolus-
injectable SPIO (SH U 555 A). J Magn Reson Imaging 10(1):65-71, 1999.
291. Sayegh Y, Pochons S, Vallee JP, et al: Detection of experimental hepatic tumors using long circulating
superparamagnetic particles. Invest Radiol 36(1):15-21, 2001.
292. Tanaka M, Nakashima O, Wada Y, et al: Pathomorphological study of Kupffer cells in hepatocellular carcinoma and
hyperplastic nodular lesions in the liver. Hepatology 24(4):807-812, 1996.
293. Grandin C, Van Beers BE, Robert A, et al: Benign hepatocellular tumors: MRI after superparamagnetic iron oxide
administration. J Comput Assist Tomogr 19(3):412-418, 1995.
294. Bremer C, Allkemper T, Baernig J, Reimer P: RES-specific imaging of the liver and spleen with iron oxide particles
designed for blood pool MR-angiography. J Magn Reson Imaging 10(3):461-467, 1999.
295. Ba-Ssalamah A, Heinz-Peer G, Schima W, et al: Detection of focal hepatic lesions: comparison of unenhanced and SHU
555 A-enhanced MR imaging versus biphasic helical CTAP. J Magn Reson Imaging 11(6):665-672, 2000.
296. Helmberger T, Gregor M, Holzknecht N, et al: [Comparison of dual-phase helical CT with native and ferum oxide-
enhanced magnetic resonance imaging in detection and characterization of focal liver lesions]. Radiologe 39(8):678-684,
1999.
297. Seneterre E, Taourel P, Bouvier Y, et al: Detection of hepatic metastases: ferumoxides-enhanced MR imaging versus
unenhanced MR imaging and CT during arterial portography. Radiology 200(3):785-792, 1996.
298. Kato N, Takahashi M, Tsuji T, et al: Dose-dependency and rate of decay of efficacy of Resovist on MR images in a rat
cirrhotic liver model. Invest Radiol 34(9):551-557, 1999.
299. Young SW, Simpson BB, Ratner AV, et al: MRI measurement of hepatocyte toxicity using the new MRI contrast agent
manganese dipyridoxal diphosphate, a manganese/pyridoxal 5-phosphate chelate. Magn Reson Med 10(1):1-13, 1989.
300. Grant D, Toft KG, Martinson I, et al: Tissue distribution and general safety of MnDPDP in male beagle dogs, with or
without total common bile duct obstruction. Acta Radiol 38(4 Pt 2):732-739, 1997.
301. Toft KG, Hustvedt SO, Grant D, et al: Metabolism and pharmacokinetics of MnDPDP in man. Acta Radiol 38(4 Pt
2):677-689, 1997.
302. Misselwitz B, Mühler A, Weinmann H: A toxicologic risk for using manganese complexes? A literature survey of existing
data through several medical specialties. Invest Radiol 30:611-620, 1995. Medline Similar articles
303. Lim KO, Stark DD, Leese PT, et al: Hepatobiliary MR imaging: first human experience with MnDPDP. Radiology
178(1):79-82, 1991.
304. Wang C, Ahlstrom H, Ekholm S, et al: Diagnostic efficacy of MnDPDP in MR imaging of the liver. A phase III multicentre
study. Acta Radiol 38(4 Pt 2):643-649, 1997.
305. Padovani B, Leesne R, Raffaelli C, et al: Tolerability and utility of mangafodipir trisodium injection (MnDPDP) at the
dose of 5 mumol/kg body weight in detecting focal liver tumors: results of a phase III trial using an infusion technique. Eur J
Radiol 23(3):205-211, 1996.
306. Bernhard C et al: Safety of mangafodipir administration at abdominal MR imaging: bolus injection vs. slow infusion. Eur
Radiol 12 (Suppl 1):291, 2002.
307. Rummeny E, Ehrenheim C, Gehl HB, et al: Manganese-DPDP as a hepatobiliary contrast agent in the magnetic
resonance imaging of liver tumors. Results of clinical phase 2 trials in Germany including 141 patients. Invest Radiol
26:S142-145, 1991.
308. Reimer P, Schneider G, Schima W: Hepatobiliary contrast agents for contrast-enhanced MRI of the liver: properties,
clinical development and applications. Eur Radiol 14(4):559-578, 2004.
309. King LJ, Burkill GJ, Scurr ED, et al: MnDPDP enhanced magnetic resonance imaging of focal liver lesions. Clin Radiol
11 of 16 14-04-2010 20:40
310. Federle ME, Chezmar J, Rubin DL, et al: Efficacy and safety of mangafodipir trisodium (MnDPDP) injection for hepatic
MRI in adults: results of the U.S. multicenter phase III clinical trials. Efficacy of early imaging. J Magn Reson Imaging
311. Braga HJ, Choti MA, Lee VS, et al: Liver lesions: manganese-enhanced MR and dual-phase helical CT for preoperative
detection and characterization comparison with receiver operating characteristic analysis. Radiology 223(2):525-531, 2002.
312. Mann GN, Marx HF, Lai LL, et al: Clinical and cost effectiveness of a new hepatocellular MRI contrast agent, mangafodipir
trisodium , in the preoperative assessment of liver resectability. Ann Surg Oncol 8(7):573-579, 2001.
313. Ham B, Vogl TJ, Branding G, et al: Focal liver lesions: MR imaging with Mn-DPDP-initial clinical results in 40 patients.
Radiology 182(1):167-174, 1992.
314. Ni Y, Marchal G, Yu J, et al: Experimental liver cancers: Mn-DPDP-enhanced rims in MR-microangiographic-histologic
correlation study. Radiology 188(1):45-51, 1993.
315. Kane PA, Ayton V, Walters HL, et al: MnDPDP-enhanced MR imaging of the liver. Correlation with surgical findings. Acta
Radiol 38(4 Pt 2):650-654, 1997.
316. Ito K, Mitchell DG, Outwater EK, et al: Hepatic lesions: discrimination of nonsolid, benign lesions from solid, malignant
lesions with heavily T2-weighted fast spin-echo MR imaging. Radiology 204:729-737, 1997. Medline Similar articles
317. Mathieu D, Coffin C, Kobeiter H, et al: Unexpected MR-T1 enhancement of endocrine liver metastases with mangafodipir.
J Magn Reson Imaging 10(2):193-195, 1999.
318. Murakami T, Baron RL, Federle MP, et al: Cirrhosis of the liver: MR imaging with mangafodipir trisodium (Mn-DPDP).
Radiology 198(2):567-572, 1996.
319. Coley BD, Mattrey RF, Baker KG, et al: MR imaging assessment of experimental hepatic dysfunction with Mn-DPDP. J
320. Marti-Bonmati L, Longedo E, Poyatos C, Casillas C: MnDPDP enhancement characteristics and differentiation between
cirrhotic and noncirrhotic livers. Invest Radiol 33:717-722, 1998. Medline Similar articles
321. Vogl TJ, Ham B, Schnell B, et al: Mn-DPDP enhancement patterns of hepatocellular lesions on MR images. J Magn
Reson Imaging 3(1):51-58, 1993.
322. Liou J, Lee JK, Borello JA, Brown JJ, et al: Differentiation of hepatomas from nonhepatomatous masses: use of
MnDPDP-enhanced MR images. Magn Reson Imaging 12(1):71-79, 1994.
323. Burkill G, Mannion E, Healy J: Technical report: lymph node enhancement at MRI with MnDPDP in primary hepatic
carcinoma. Clin Radiol 56(1):67-71, 2001.
324. Marti-Bonmati L, Lonjedo E, Mathieu D, et al: Tumoural portal vein thrombosis. Enhancement with MnDPDP. Acta Radiol
38(4 Pt 2):655-659, 1997.
325. Bartolozzi C, Donati F, Cioni D, et al: MnDPDP-enhanced MRI vs dual-phase spiral CT in the detection of hepatocellular
carcinoma in cirrhosis. Eur Radiol 10(11):1697-1702, 2000.
326. Kim SK, Kim SH, Lee WJ, et al: Preoperative detection of hepatocellular carcinoma: ferumoxides-enhanced versus
mangafodipir trisodium-enhanced MR imaging. Am J Roentgenol 179(3):741-750, 2002.
page 407
page 408
327. Ohtomo K, Baron RL, Dodd GD, et al: Confluent hepatic fibrosis in advanced cirrhosis: evaluation with MR imaging.
328. Ba-Ssalamah A, Schima W, Schmook MT, et al: Atypical focal nodular hyperplasia of the liver: imaging features of
non-specific and liver-specific MR contrast agents. Am J Roentgenol 179(6):1447-1456, 2002.
329. Sahani DV, O'Malley ME, Bhat S, et al: Contrast-enhanced MRI of the liver with mangafodipir trisodium: imaging
technique and results. J Comput Assist Tomogr 26(2):216-222, 2002.
330. Oudkerk M, Torres CG, Song B, et al: Characterization of liver lesions with mangafodipir trisodium-enhanced MR
imaging: multicenter study comparing MR and dual-phase spiral CT. Radiology 223(2):517-524, 2002.
331. Coffin CM, Diche T, Mahfouz A, et al: Benign and malignant hepatocellular tumors: evaluation of tumoral enhancement
after mangafodipir trisodium injection on MR imaging. Eur Radiol 9(3):444-449, 1999.
332. Helmberger TK, Laubenberger J, Rummeny E, et al: MRI characteristics in focal hepatic disease before and after
administration of MnDPDP: discriminant analysis as a diagnostic tool. Eur Radiol 12:62-70, 2002. Medline
Similar articles
333. Rofsky NM, Weinreb JC, Bernardino ME, et al: Hepatocellular tumors: characterization with Mn-DPDP-enhanced MR
imaging. Radiology 188(1):53-59, 1993.
334. Lee VS, Rofsky NM, Morgan GR, et al: Volumetric mangafodipir trisodium-enhanced cholangiography to define
intrahepatic biliary anatomy. Am J Roentgenol 176:906-908, 2001.
335. Lee VS, Morgan GR, Teperman LW, et al: MR imaging as the sole preoperative imaging modality for right hepatectomy:
a prospective study of living adult-to-adult liver donor candidates. Am J Roentgenol 176(6):1475-1482, 2001.
336. Goldman J, Florman S, Varotti G, et al: Noninvasive preoperative evaluation of biliary anatomy in right-lobe living donors
with mangafodipir trisodium-enhanced MR cholangiography. Transplant Proc 35(4):1421-1422, 2003.
337. Vitellas KM, El-Deib A, Vaswani K, et al: Detection of bile duct leaks using MR cholangiography with mangfodipir
12 of 16 14-04-2010 20:40
trisodium (Teslascan). J Comput Assist Tomogr 25(1):102-105, 2001.

338. Vitellas KM, El-Deib A, Vaswani K, et al: Using contrast-enhanced MR cholangiography with IV mangafodipir trisodium
(Teslascan) to evaluate bile duct leaks after cholecystectomy: a prospective study of 11 patients. Am J Roentgenol
179(2):409-416, 2002.
339. Gehl HB, Vorwerk D, Klose KC, et al: Pancreatic enhancement after low-dose infusion of Mn-DPDP. Radiology
180(2):337-339, 1991.
340. Romijn MG, Stoker J, Van Eijck CH, et al: MRI with mangafodipir trisodium in the detection and staging of pancreatic
cancer. J Magn Reson Imaging 12(2):261-268, 2000.
341. Hamoud AK, Khan MF, Aboalaali N, et al: Mangan-enhanced MR imaging for the detection and localisation of small
pancreatic insulinoma. Eur Radiol 14:923-925, 2004
342. Schima W, Fugger R, Schober E, et al: Diagnosis and staging of pancreatic cancer: comparison of mangafodipir
trisodium-enhanced MR imaging and contrast-enhanced helical hydro-CT. Am J Roentgenol 179(3):717-724, 2002.
343. Kettritz U, Warshouer DM, Brown ED, et al: Enhancement of the normal pancreas: comparison of manganese-DPDP and
gadolinium chelate. Eur Radiol 6(1):14-18, 1996.
344. Kettritz U, Semelka R: Contrast-enhanced MR imaging of the pancreas. Magn Reson Imaging Clin North Am 4(1):87-100,
1996.
345. Kettritz U, Schlund JF, Wilbur K, et al: Comparison of gadolinium chelates with manganese-DPDP for liver lesion
detection and characterization: preliminary results. Magn Reson Imaging 14(10):1185-1190, 1996.
346. Diehl SJ, Lehmann KJ, Gaa J, et al: MR imaging of pancreatic lesions. Comparison of manganese-DPDP and
gadolinium chelate. Invest Radiol 34(9):589-595, 1999.
347. Watson A: Theory and mechanisms of contrast-enhancing agents. In Higgins CB, Hricak H, Helms CA (eds): Magnetic
Resonance Imaging of the Body. New York: Raven Press, 1992.
348. de Haen C, Lorusso V, Tirone P: Hepatic transport of gadobenate dimeglumine in TR-rats. Acad Radiol 3 (Suppl
2):S452-454, 1996.
349. de Haen C, Lorusso V, Luzzani F, et al: Hepatic transport of the magnetic resonance imaging contrast agent gadobenate
dimeglumine in the rat. Acad Radiol 2(3):232-238, 1995.
350. Kirchin M, Pirovano G, Spinazzi A: Gadobenate dimeglumine (Gd-BOPTA). An overview. Invest Radiol 33(11):798-809,
1998.
351. de Haen C, Gozzini L: Soluble-type hepatobiliary contrast agents for MR imaging. J Magn Reson Imaging 3(1):179-186,
1993.
352. de Haen C, La Ferla R, Maggioni F: Gadobenate dimeglumine 0.5 M solution for injection (MultiHance) as contrast agent
for magnetic resonance imaging of the liver: mechanistic studies in animals. J Comput Assist Tomogr 23 (Suppl 1):S169-179,
1999.
353. Cavagna FM, Maggioni F, Castelli PM, et al: Gadolinium chelates with weak binding to serum proteins. A new class of
high-efficiency, general purpose contrast agents for magnetic resonance imaging. Invest Radiol 32(12):780-796, 1997.
354. Kirchin MA, Pirovano G, Venetianer C, et al: Safety assessment of gadobenate dimeglumine (MultiHance): extended
clinical experience from phase I studies to post-marketing surveillance. J Magn Reson Imaging 14(3):281-294, 2001.
355. Saini S: Contrast-enhanced MR imaging of the liver. Radiology 182(1):12-14, 1992.
356. Mahfouz A, Hamm B, Wolf K: Peripheral washout: a sign of malignancy on dynamic gadolinium-enhanced MR images of
focal liver lesions. Radiology 190(1):49-52, 1994.
357. Nakamura H, Murakami T, Ishida T, et al: 3DFT-FISP MRI with gadopentetate dimeglumine in differential diagnosis of
small liver tumors. J Comput Assist Tomogr 18(1):49-54, 1994.
358. Powers C, Ros PR, Stoupis C, et al: Primary liver neoplasms: MR imaging with pathologic correlation. Radiographics
14(3):459-482, 1994.
359. Semelka RC, Brown ED, Ascher SM, et al: Hepatic hemangiomas: a multi-institutional study of appearance on
T2-weighted and serial gadolinium-enhanced gradient-echo MR images. Radiology 192(2):401-406, 1994.
360. Hamm B, Mahfouz AE, Taupitz M, et al: Liver metastases: improved detection with dynamic gadolinium-enhanced MR
imaging? Radiology 202(3):677-682, 1997.
361. Cavagna FM et al: Hepatobiliary contrast agents for MRI. In Ferrucci J, Stark D (eds): Liver Imaging: Current Trends and
New Techniques. Boston: Andover Medical Publishers, 1990.
362. Schneider G, Grazioli L, Saini S: MRI of the liver. Milan: Springer, 2003.
363. Morana G, Grazioli L, Schneider G, et al: Hypervascular hepatic lesions: dynamic and late enhancement pattern with
Gd-BOPTA. Acad Radiol 9 (Suppl 2):476-479, 2002.
364. Grazioli L, Morana G, Kirchin MA, et al: MRI of focal nodular hyperplasia (FNH) with gadobenate dimeglumine
(Gd-BOPTA) and SPIO (ferumoxides ): an intra-individual comparison. J Magn Reson Imaging 17(5):593-602, 2003.
365. Kuwatsuru R, Katoya M, Ohtoma K, et al: Comparison of gadobenate dimeglumine with gadopentetate dimeglumine for
magnetic resonance imaging of liver tumors. Invest Radiol 36(11):632-641, 2001.
366. Weinman HJ, Schulmann-Giampieri G, Schmitt-Willich H, et al: A new lipophilic gadolinium chelate as a tissue-specific
contrast medium for MRI. Magn Reson Med 22(2):233-237; discussion 242, 1991.
13 of 16 14-04-2010 20:40
367. Muhler A, Clement O, Saeed M, et al: Gadolinium-ethoxybenzyl-DTPA, a new liver-directed magnetic resonance contrast
agent. Absence of acute hepatotoxic, cardiovascular, or immunogenic effects. Invest Radiol 28(1):26-32, 1993.
368. Reimer P, Rummeny EJ, Shamsi K, et al: Phase II clinical evaluation of Gd-EOB-DTPA: dose, safety aspects, and pulse
sequence. Radiology 199(1):177-183, 1996.
369. Shamsi K et al. Gd-EOB-DTPA (Eovist), a liver specific contrast agent for MRI: results of a placebo controlled, double
blind dose ranging study in patients with focal liver lesions. Proceedings of the Tenth Scientific Meeting and Exhibition of the
International Society for Magnetic Resonance in Medicine. Honolulu, 2002.
370. Huppertz A, Balzer T, Blakeborough A, et al: Improved detection of focal liver lesions at MR imaging: multicenter
comparison of gadoxetic acid-enhanced MR images with intraoperative findings. Radiology 230(1):266-275, 2004.
371. Bollow M, Taupitz M, Hamm B, et al: Gadolinium-ethoxybenzyl-DTPA as a hepatobiliary contrast agent for use in MR
cholangiography: results of an in vivo phase-I clinical evaluation. Eur Radiol 7(1):126-132, 1997.
372. Carlos RC, Branam JD, Dona Q, et al: Biliary imaging with Gd-EOB-DTPA: is a 20-minute delay sufficient? Acad Radiol
9(11):1322-1325, 2002.
373. Clement O, Muhler A, Vexler VS, et al: Evaluation of radiation-induced liver injury with MR imaging: comparison of
hepatocellular and reticuloendothelial contrast agents. Radiology 185(1):163-168, 1992.
374. Fujita M, Yamamoto R, Fritz-Zieroth B, et al: Contrast enhancement with Gd-EOB-DTPA in MR imaging of hepatocellular
carcinoma in mice: a comparison with superparamagnetic iron oxide. J Magn Reson Imaging 6(3):472-477, 1996.
375. Giovagnoni A, Paci E: Liver. III: Gadolinium-based hepatobiliary contrast agents (Gd-EOB-DTPA and
Gd-BOPTA/Dimeg). Magn Reson Imaging Clin North Am 4(1):61-72, 1996.
376. Hammerstingl R, Zangos S, Schwatz R, et al: Contrast-enhanced MRI of focal liver tumors using a hepatobiliary MR
contrast agent: detection and differential diagnosis using Gd-EOB-DTPA-enhanced versus Gd-DTPA-enhanced MRI in the
same patient. Acad Radiol 9 (Suppl 1):S119-120.
377. Hirohashi S, Hitohashi R, Uchida H, et al: Optimal dose of hepatobiliary contrast agent for MR cholangiography:
experimental study in rats. J Magn Reson Imaging 8(4):847-852, 1998.
378. Imam K, Bluemke D: MR imaging in the evaluation of hepatic metastases. Magn Reson Imaging Clin North Am
8(4):741-756, 2000.
379. Kato N, Yokawa T, Tamura A, et al: Gadolinium-ethoxybenzyl-diethylenetriamine-pentaacetic acid interaction with clinical
drugs in rats. Invest Radiol 37(12):680-684, 2002.
380. Kim T, Murakami T, Hasuike Y, et al: Experimental hepatic dysfunction: evaluation by MRI with Gd-EOB-DTPA. J Magn
Reson Imaging 7(4):683-688, 1997.
381. Muhler A, Clement O, Vexler VS, et al: Hepatobiliary enhancement with Gd-EOB-DTPA: comparison of spin-echo and
STIR imaging for detection of experimental liver metastases. Radiology 184(1):207-213.
382. Muhler A, Freise CE, Kuwatsuru R, et al: Acute liver rejection: evaluation with cell-directed MR contrast agents in a rat
transplantation model. Radiology 186(1):139-146, 1993.
383. Murakami T, Kim T, Gotoh M, et al: Experimental hepatic dysfunction: evaluation by MR imaging with Gd-EOB-DTPA.
Acad Radiol 5 (Suppl 1):S80-82, 1998.
page 408
page 409
384. Ni Y, Marchal G, Yu J, et al: Prolonged positive contrast enhancement with Gd-EOB-DTPA in experimental liver tumors:
potential value in tissue characterization. J Magn Reson Imaging 4(3):355-363, 1994.
385. Ni Y, Marchal G, Lukito G, et al: MR imaging evaluation of liver enhancement by Gd-EOB-DTPA in selective and total bile
duct obstruction in rats: correlation with serologic, microcholangiographic, and histologic findings. Radiology 190(3):753-758,
1994.
386. Reimer P, Rummeny EJ, Daldrup HE, et al: Enhancement characteristics of liver metastases, hepatocellular carcinomas,
and hemangiomas with Gd-EOB-DTPA: preliminary results with dynamic MR imaging. Eur Radiol 7(2):275-280, 1997.
387. Runge VM: A comparison of two MR hepatobiliary gadolinium chelates: Gd-BOPTA and Gd-EOB-DTPA. J Comput
Assist Tomogr 22(4):643-650, 1998.
388. Schmitz SA, Muhler A, Wagner S, Wolf KJ: Functional hepatobiliary imaging with gadolinium-EOB-DTPA. A comparison
of magnetic resonance imaging and 153gadolinium-EOB-DTPA scintigraphy in rats. Invest Radiol 31(3):154-160, 1996.
389. Shimada M, Hayashi S, Senoo A, et al: [Dynamic MR hepatocholangiography with the SIP Fast GRE (saturation inversion
projection fast gradient echo) method]. Nippon Igaku Hoshasen Gakkai Zasshi 57(11):681-683, 1997.
390. Shimizu J, Dono K, Gotah M, et al: Evaluation of regional liver function by gadolinium-EOB-DTPA-enhanced MR imaging.
Dig Dis Sci 44(7):1330-1337, 1999.
391. Shuter B, Tofts PS, Wang SC, et al: The relaxivity of Gd-EOB-DTPA and Gd-DTPA in liver and kidney of the Wistar rat.
392. Shutter B, Wang SC, Roche J, et al: Relaxivity of Gd-EOB-DTPA in the normal and biliary obstructed guinea pig. J Magn
Reson Imaging 8(4):853-861, 1998.
393. Stern W, Schick F, Kopp AF, et al: Dynamic MR imaging of liver metastases with Gd-EOB-DTPA. Acta Radiol
41(3):255-262, 2000.
394. Tanimoto A, Saron Y, Yuasa Y, et al: Performance of Gd-EOB-DTPA and superparamagnetic iron oxide particles in the
14 of 16 14-04-2010 20:40
detection of primary liver cancer: a comparative study by alternative free-response receiver operating characteristic analysis. J
395. Vogl TJ, Kummel S, Hammerstingl R, et al: Liver tumors: comparison of MR imaging with Gd-EOB-DTPA and Gd-DTPA.
Radiology 200(1):59-67, 1996.
396. Yoshikawa K, Inoue Y, Akahane M, et al: Phantom and animal studies of a new hepatobiliary agent for MR imaging:
comparison of Gd-DTPA-DeA with Gd-EOB-DTPA. J Magn Reson Imaging 18(2):204-209, 2003.
397. Zangos S, Hammerstingl R, Mack MG, et al: [Renal enhancement and excretion of the hepatobiliary contrast agent
Gd-EOB-DTPA]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 173(11):1034-1040, 2001.
398. Breuer J: A Multicenter (US) Open-Label Study of Gd-EOB-DTPA with a Single Intravenous Injection (25 mmol/kg body
weight) in Patients with Known or Suspected Focal Liver Lesions and the Corresponding Blinded Reading to Evaluate the
Diagnostic Efficacy of Gd-EOB-DTPA in the Characterization of Focal Liver Lesions. Germany: Schering AG, 2003.
399. Vogl T et al. Enhancement of malignant and benign focal liver lesions using Gd-EOB-DTPA and Gd-DTPA enhanced MRI
in the same patient. Proceedings of the International Conference on MRI of the Abdomen and Pelvis. Bremen, Germany:
Hauschild, 1995.
400. Carlos RC, Hussain HK, Song JH, Francis IR: Gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid as an
intrabiliary contrast agent: preliminary assessment. Am J Roentgenol 179(1):87-92, 2002.
401. Breuer J: Study of the Safety of Gd-EOB-DTPA and the Effect of Hepatic and Renal Impairment on the Pharmacokinetics
(PK) and the Pharmacodynamics (PD) of Gd-EOB-DTPA after a Single Intravenous Dose in Volunteer Patients. Germany:
Schering AG, 2003.
402. Schima W, Fugger R: Evaluation of focal pancreatic masses: comparison of mangafodipir-enhanced MR imaging and
contrast-enhanced helical CT. Eur Radiol 12(12):2998-3008, 2002.
403. Wang C, Johansson L, Western A, et al: Sequence optimization in mangafodipir trisodium-enhanced liver and pancreas
MRI. J Magn Reson Imaging 9(2):280-284, 1999.
404. Lee VS, Lavelle MT, Krinsky GA, Rofsky NM: Volumetric MR imaging of the liver and applications. Magn Reson Imaging
Clin North Am 9(4):697-716, v-vi, 2001.
405. Helmberger T, Gregor M, Holzknecht N, et al: [Effects of biphasic spiral CT, conventional and iron oxide enhanced MRI on
therapy and therapy costs in patients with focal liver lesions]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr
172(3):251-259, 2000.
406. Muller RD, Vogel K, Neumann K, et al: SPIO-MR imaging versus double-phase spiral CT in detecting malignant lesions of
the liver. Acta Radiol 40(6):628-635, 1999.
407. Scharf J, Hoffmann V, Lehnert T, et al: Pseudolesions at T1-weighted gradient-echo imaging after administration of
superparamagnetic iron oxide: comparison with portal perfusion abnormalities at CT during arterial portography. Radiology
207(1):67-72, 1998.
408. Yamamoto H, Yamashita Y, Yoshminatsu S, et al: Hepatocellular carcinoma in cirrhotic livers: detection with unenhanced
and iron oxide-enhanced MR imaging. Radiology 195(1):106-112, 1995.
409. Ward J, Guthrie JA, Scott DJ, et al: Hepatocellular carcinoma in the cirrhotic liver: double-contrast MR imaging for
diagnosis. Radiology 216(1):154-162, 2000.
410. Ward J, Robinson PJ: How to detect hepatocellular carcinoma in cirrhosis. Eur Radiol 12(9):2258-2272, 2002.
411. Vogl TJ, Schwarz W, Blume S, et al: Preoperative evaluation of malignant liver tumors: comparison of unenhanced and
SPIO (Resovist)-enhanced MR imaging with biphasic CTAP and intraoperative US. Eur Radiol 13(2):262-272, 2003.
412. Vogl TJ, Hammerstingl R, Schwarz W, et al: Superparamagnetic iron oxide-enhanced versus gadolinium-enhanced MR
imaging for differential diagnosis of focal liver lesions. Radiology 198(3):881-887, 1996.
413. Kim MJ, Kim JH, Chung JJ, et al: Focal hepatic lesions: detection and characterization with combination gadolinium- and
superparamagnetic iron oxide-enhanced MR imaging. Radiology 228(3):719-726, 2003.
414. Schneider G et al: Gd-BOPTA versus SH U 555 A: two different MR contrast agents for the detection and DDx of
hypervascular liver tumors. Proceedings of the Radiological Society of North America, Chicago, p. 305, 2003.
415. Kim J et al: Detection and characterization of focal hepatic lesions: manganese vs SPIO-enhanced MR imaging.
Proceedings of the Radiological Society of North America, Chicago, p. 307, 2003.
416. Marugami N et al: Detection of malignant hepatic tumors: comparison of Gd-EOB-DTPA-enhanced and SPIO-enhanced
MR imaging. Proceedings of the Radiological Society of North America, Chicago, p. 305, 2003.
417. Desser TS, Rubin DL, Muller HH, et al: Dynamics of tumor imaging with Gd-DTPA-polyethylene glycol polymers:
dependence on molecular weight. J Magn Reson Imaging 4(3):467-472, 1994.
418. Schwickert HC, Stiskal M, Roberts TP, et al: Contrast-enhanced MR imaging assessment of tumor capillary permeability:
effect of irradiation on delivery of chemotherapy. Radiology 198(3):893-898, 1996.
419. Poon E, Yucel E, Pagan-Marin H, Kayne H: Iliac artery stenosis measurements: comparison of two-dimensional time-of-
flight and three-dimensional dynamic gadolinium-enhanced MR angiography. Am J Roentgenol 169(4):1139-1144, 1997.
420. Snidow JJ: Interpretations and treatment decisions based on MR angiography versus conventional arteriography in
symptomatic lower extremity ischemia. J Vasc Interv Radiol 6(4):595-603, 1995.
421. Omary RA, Henseler KP, Unal O, et al: Comparison of intraarterial and IV gadolinium-enhanced MR angiography with
digital subtraction angiography for the detection of renal artery stenosis in pigs. Am J Roentgenol 178(1):119-123, 2002.
15 of 16 14-04-2010 20:40
422. Hany TF, Carroll JJ, Omary RA, et al: Aorta and runoff vessels: single-injection MR angiography with automated table
movement compared with multiinjection time-resolved MR angiography-initial results. Radiology 221(1):266-272, 2001.
423. Frayne R, Grist TM, Swan JS, et al: 3D MR DSA: effects of injection protocol and image masking. J Magn Reson Imaging
12(3):476-487, 2000.
424. Grist TM: MRA of the abdominal aorta and lower extremities. J Magn Reson Imaging 11(1):32-43, 2000.
425. Quinn SF, Sheley RC, Semowsen KG, et al: Aortic and lower-extremity arterial disease: evaluation with MR angiography
versus conventional angiography. Radiology 206(3):693-701, 1998.
426. Baumn RA, Rutter CM, Sunshine JH, et al: Multicenter trial to evaluate vascular magnetic resonance angiography of the
lower extremity. American College of Radiology Rapid Technology Assessment Group. JAMA 274(11):875-880, 1995.
427. Hirsch AT, Criqui MH, Treat-Jacobson D, et al: Peripheral arterial disease detection, awareness, and treatment in primary
care. JAMA 286(11):1317-1324, 2001.
428. Harishinghani MG, Barentsz J, Hahn PF, et al: Noninvasive detection of clinically occult lymph-node metastases in
prostate cancer. N Engl J Med 348(25):2491-2499, 2003.
429. Choi CJ, Kramer CM: MR imaging of atherosclerotic plaque. Radiol Clin North Am 40(4):887-898, 2002.
430. Barkhausen J, Ebert W, Heyer C, Debatin JF, et al: Detection of atherosclerotic plaque with Gadoflourine-enhanced
magnetic resonance imaging. Circulation 108(5):605-609, 2003.
431. Sirol M, Itskovich VV, Mani V, et al: Lipid-rich atherosclerotic plaques detected by gadoflourine-enchanced in vivo
magnetic resonance imaging. Circulation 108(23):2890-2896, 2004.
432. Ruehm SG, Corot C, Vogt P, Kolb S, Debatin JF: Magnetic resonance imaging of atherosclerotic plaque with ultrasmall
superparamagnetic particles of iron oxide in hyperlipidemic rabbits. Circulation 103(3):415-422, 2001.
433. Kooi ME, Cappendijk VC, Cleutjens KB, et al: Accumulation of ultrasmall superparamagnetic particles of iron oxide in
human atherosclerotic plaques can be detected by in vivo magnetic resonance imaging. Circulation 107(19):2453-2458, 2003.
© 2010 Elsevier
16 of 16 14-04-2010 20:40
OLECULAR AND ELLULAR MAGING

Jonathan B. Kruskal
INTRODUCTION
The ability to image at the cellular or molecular level represents an entirely new paradigm for the
diagnosis and characterization of many disease processes. MR imaging has advanced beyond
functioning as a sensitive diagnostic imaging tool and is able to accurately characterize molecular
events at the cellular level. These advances have been brought about in part by the design and
synthesis of an entirely new generation of innovative or so-called "smart contrast agents" capable of
targeting unique cellular sites or of being activated by the local presence of specific enzymes or ions.
For diagnostic purposes, these agents permit pathologic processes to be accurately localized and their
metabolic activity and function can be evaluated noninvasively prior to initiating treatment. Furthermore,
MR imaging is becoming a useful adjunct in the selection, optimization, and follow-up of therapies, also
permitting the in vivo study of basic intracellular events and signaling.
As technologies compete for imaging at the cellular and molecular levels, much attention is being
focused on magnetic resonance as the technique most likely to provide the resolution and sensitivity
required for providing the types of answers to meet expanding clinical demands. While much research
has centered on the technical development of MR techniques, this attention has also spurred numerous
technical advances in MR resolution and in in vivo spectroscopy and microscopy. This chapter will
focus on the MR techniques currently being used to image at the cellular level, as well as the host of
novel concepts being applied to contrast agent development for these specific purposes.
© 2010 Elsevier
1 of 1 14-04-2010 20:41
EMERGING OPPORTUNITIES FOR MOLECULAR IMAGING USING MAGNETIC

RESONANCE
To highlight the current status and future opportunities for MR imaging at the cellular and molecular
level, it is important to summarize the types of molecular events that can be imaged and where further
research opportunities exist. The recent unraveling of the human genome has been an enormous
stimulus to the entire field of molecular medicine and has directly and indirectly raised the stature and
importance of molecular imaging. A greater understanding of the molecular causes and results of
disease processes, coupled with emerging new molecular therapies, has led to a demand for
noninvasive indices for studying these fields.
As an example, gene-based therapy continues to advance primarily at the preclinical level, with much
imaging research focusing on the development of techniques capable of evaluating expression of
desired gene products in vivo. MR is one of several logical techniques being actively explored for this
purpose. As the field of molecular medicine continues to expand, hosts of MR imaging opportunities
are being identified. These include but are certainly not limited to identification of cell receptors for
selection of drug therapies, imaging of apoptosis or programmed cell death to evaluate efficacy of
treatment, and evaluation of angiogenic activity for tumor staging and evaluation of treatment efficacy.
page 410
page 411
As our understanding of the molecular mechanisms of disease and their associated treatments
improves, it is widely anticipated that many new opportunities for MR imaging at the molecular and
cellular level will become available. Each of these opportunities encompasses technical advances not
only in image resolution but also in the engineering of novel contrast materials. As examples of the
above-mentioned scenarios, MR is already able to image delivery of the β-galactosidase reporter
1 2
gene and can image the breast cancer Her-2/neu receptor or the phosphatidylserine only present on
3
the surface of apoptotic cells and the neoendothelial αvβ3 integrin only present on the surface of newly
4
formed endothelial cells.
However, before these emerging opportunities can be fully embraced, several essential components
must be addressed, including contrast agent specificity and sensitivity, signal amplification systems,
and molecular reporter and detection systems. Below we illustrate the current status of each of these
components.
© 2010 Elsevier
1 of 1 14-04-2010 20:42
CONTRAST AGENTS FOR MOLECULAR IMAGING

The majority of MR contrast agents currently in clinical use serve as intravascular or interstitial agents,
enhancing the signal intensity of each of these pools, respectively. Other agents are able to target
function of specific cell populations (such as phagocytic or reticuloendothelial activity). Contrast agents
in current clinical use are used to enhance differences in perfusion or flow between the vascular and
interstitial compartments. As such, these are only able to provide an indirect estimate of abnormal
physiology. For imaging at the cellular or molecular level using MRI, many important conceptual and
structural advances have been achieved in the development of contrast agents. Broadly, these
advances have encompassed the design of new classes of contrast agents, engineered both to target
unique sites or activities and to enhance delivery to these sites.
Add to lightbox
Add to lightbox
Figure 15-1 Schematic of the transition of EgadMe from a weak to a strong relaxivity state. A,
Schematic diagram representing the site-specific placement of the galactopyranosyl ring on the tetra-
azamacrocycle (side view). Upon cleavage of the sugar residue by β-galactosidase (at red bond), an
3+
inner sphere co-ordination site of the Gd ion becomes more accessible to water. B, Space-filling
molecular model (top view, from above the sugar residue) of the complex before (left) and after
3+
cleavage by the β-gal (right), illustrating the increased accessibility of the Gd ion (magenta) upon
cleavage: white, H; red, O; blue, N; gray, C. (Reprinted with permission of the publisher from
reference 1.)
In order to improve the specificity of targeting, some contrast agents have simply been conjugated or
packaged along with monoclonal antibodies, whereas others have been engineered to be activated by
specific molecules produced at specific sites or in response to specific biological events. An excellent
example of this latter group of so-called "caged" contrast agents is an agent referred to as EgadMe,
developed by Louie et al1 for imaging of reporter gene expression in vivo (see Figs. 15-1 to 15-3). The
authors selected β-galactosidase as the marker enzyme encoded and expressed by a delivered cDNA
gene complex. They constructed their contrast agent such that water access to a chelated
paramagnetic ion is blocked by a substrate (galactopyranose) that is only removed by enzymatic
(β-galactosidase) activity. In the presence of this enzyme, as would occur with successful gene
delivery, the substrate is unblocked, permitting entrance of water and resulting in a change in relaxivity
and thus signal intensity (Fig. 15-1). This is a particularly important study and concept since it highlights
the ability to create designer probes capable of imaging very specific enzyme activity. If the gene for
this enzyme is co-administered along with a desired therapeutic gene, then in vivo detection of an
1 of 3 14-04-2010 20:44
expressed reporter enzyme is typically sufficient to infer co-expression of the therapeutic gene
products. In in vivo studies, this group have also imaged β-galactosidase expression in living Xenopus
laevis embryos (Fig. 15-2) and have also used EgadMe to image expression of the lacZ gene (Fig.
15-3).
Amongst many contributions to the field of molecular and cellular imaging, Weissleder and colleagues
5,6
were instrumental in bringing the ultrasmall superparamagnetic iron oxide preparations to the clinic.
These MR-visible particles can be covalently bound to molecules targeting a spectrum of specific cell
sites or even functions. An advantage of iron oxide-based contrast agents is that their T2 relaxivity
produces strong negative signals at nanomolar concentrations in vitro. 7 The superparamagnetic iron
oxide nanoparticles are available in different size formulations, making it possible to image or evaluate
a spectrum of physiologic events in vivo. Weissleder and colleagues have used these particles for
numerous innovative clinical applications. By labeling these with monoclonal antibodies specific for
8
myosin, Weissleder et al were able to detect infarcted myocardium in vivo. Using the same contrast
agent, they were able to image phagocytic activity in experimental gliomas9 and to image the
10
ubiquitous hepatocyte asialoglycoprotein receptor, present on normal hepatocytes but not on the cell
membrane of intrahepatic tumors.
page 411
page 412
Add to lightbox
Figure 15-2 MRI detection of β-galactosidase mRNA expression in living X. laevis embryos. MR
images of two embryos injected with EgadMe at the two-cell stage. A, Unenhanced MR image. The
embryo on the right was also injected with b-gal mRNA, resulting in the higher intensity regions. The
signal strength is 45-65% greater in the embryo on the right containing b-gal (contrast-to-noise ratio
ranges from 3.5 to 6). The cement gland has intrinsically short T 1, thus is visible as a bright structure
on both embryos. B, Pseudocolor rendering of same image as in (A) with water made transparent.
The image correction makes it possible to recognize the eye and brachial arches in the injected
embryo. d, dorsal; v, ventral; r, rostral; e, eye; c, cement gland; s, somite; b, brachial arches. Scale bar
= 1 mm. (Reprinted with permission of the publisher from reference 1.)
Cross-linked iron oxide (CLIO) nanoparticles are functionalized superparamagnetic preparations that
have been developed for target-specific magnetic resonance imaging.11 These can be attached to
oligonucleotides selected because of their ability to bind to specific cancer cells, messenger RNA or
peptides resulting from specific biological events such as apoptosis.12 The cross-linked structure of the
CLIO molecules permits ongoing repetitive attachment of additional probes to the initially bound
particle, until sufficient oligomerization and signal amplification has occurred and MR imaging is
possible. These CLIO molecules thus represent a self-propagating signal amplification system.
By conjugating antibodies to liposomes containing aqueous paramagnetic contrast agents, so-called
2 of 3 14-04-2010 20:44
antibody-conjugated paramagnetic liposomes (ACPLs) can be directed to specific sites, rendering

these visible to MR imaging. Several clinical applications have been identified to take advantage of this
technology, including imaging of the endothelial leukocyte receptor ICAM-1, which is upregulated on
activated endothelial cells in response to infections or even tumors. 4 Of particular interest has been the
use of ACPLs targeted to the αvβ3 integrin, an endothelial cell marker specific for neoangiogenic
vessels.4 The potential clinical application of this probe is the ability to image tumor angiogenesis at an
early stage which may be useful for staging tumors and for guiding choice of therapy. Also referred to
13
as multivalent polymerized vesicles, these molecules can be engineered to carry either contrast or
therapeutic agents, or both, and when targeted to vascular or tumor endothelium (Fig. 15-4), can
potentially be used to select patients for and to guide vascular-targeted therapies.
Add to lightbox
Add to lightbox
Figure 15-3 EgadMe permits MRI detection of lacZ gene expression. A, MR image of living embryo
injected with plasmids carrying the lacZ gene. Plasmid carrying the lacZ gene was injected to one cell
at the two-cell stage and subsequent enzyme expression was on the left side of the embryo as shown.
Regions of high signal intensity are found in the bright stripe of endoderm (e), regions of the head (h),
and ventrally, including two distinct spots (red arrows) found just ventral to the cement gland (c). B,
Bright-field image of same embryo fixed and stained with X-gal. Whole-mount cytochemistry shows
that regions demonstrating enzyme expression correlate with the regions of high intensity in the MR
image. (Reprinted with permission of the publisher from reference 1.)
14
In similar studies of the vulnerability of atherosclerotic plaques, Flacke et al have synthesized fibrin-
targetable nanoparticles that are sensitive for the detection and localization of cross-linked fibrin within
vulnerable plaques in vivo.
© 2010 Elsevier
3 of 3 14-04-2010 20:44
TRANSLATION OF LABORATORY TECHNIQUES

In an effort to identify new strategies for the clinical imaging of specific molecules, several groups have focused their
attention on translating well-established laboratory imaging techniques into the in vivo imaging setting. As an
example, Artemov et al2 have targeted the breast cancer HER-2/neu receptor by taking advantage of the natural
affinity between streptavidin and biotin (Fig. 15-5). This technique is commonly used in enzyme-linked immunosorbent
assays and in immunohistochemistry. The authors demonstrated specific binding of gadolinium-avidin complexes to
tumor cells and an in vivo breast cancer model, both prelabeled with a biotinylated anti-HER-2/neu antibody (see Fig.
15-5).
page 412
page 413
Add to lightbox
Add to lightbox
Figure 15-4 A, MR images of V2 carcinoma in the thigh muscle of a rabbit and subcutaneously prior to (A) and at 24
hours (B) post anti-αvβ3-labeled AbPV injection. B, MR images of isotype matched controls. (Adapted with
permission of the publisher from reference 13.)
1 of 3 14-04-2010 20:45
Add to lightbox
Add to lightbox
page 413
page 414
Figure 15-5 A, MR T1-weighted images of control EMT-6 and NT-5 tumors obtained before administration of the CA
(avidin-Gd-DTPA conjugate) and at 1, 8, 24, and 48 h after contrast. Arrows show enhanced signal from the tumor at
the 8 and 24 h time points for the HER-2/neu-expressing NT-5 tumor. B, Signal enhancement in the tumor relative to
the muscle tissue at 24 h after contrast. The presented results are means for five NT-5 tumors treated with the
primary biotinylated antibody, three control NT-5 tumors treated with BSA instead of antibody, and three EMT-6
tumors treated with anti-HER-2/neu biotinylated antibody. Error bars represent the SE. The signal enhancement was
significant (P < 0.05) for prelabeled NT-5 tumors. (Reprinted with permission of the publisher from reference 2.)
The sheer number of targets being identified, along with probe engineering, has resulted in the development of
biologically relevant high throughput screening methods15 for identification of specific interactions. Hogemann et al15
have described a technique for rapidly and simultaneously screening libraries consisting of literally thousands of
potential MR reporter agents. Since this is an in vitro assay, targets being evaluated may include whole cells, cell
membranes or intracellular organelles and peptides. As an extension of this technology, bifunctional MR contrast
agents have been developed which can be imaged both with MRI and with optical imaging techniques simultaneously,
providing similar information but at very different resolutions. These optical techniques are essentially serving as in
vivo screening tests for their in vitro counterparts. As an example, Huber et al16 have synthesized a contrast agent
consisting of a chelated gadolinium covalently attached to a fluorescent dye. Without altering function or
bioavailability of these agents, these remain visible both to MR imaging and to optical techniques designed to detect
fluorescence.
The role of these bifunctional agents has been further explored for translational purposes. Modo et al17 labeled stem
cells with a gadolinium-rhodamine-dextran complex identifiable with both MRI and fluorescence microscopy. They
were able to map the temporal and spatial distribution of these labeled cells following engraftment in ischemia-
damaged rat brain.
Reporter Systems
Several innovative systems have been explored for enhancing low-level signal derived from MR contrast agents.
Work from one group has taken advantage of the nonregulatable engineered transferrin receptor 18 whose expression
can be probed with one of several superparamagnetic probes. By co-administration of the gene for this visible
2 of 3 14-04-2010 20:45
receptor using an amplicon (RNA-replicating vector) system, Ichikawa et al19 have confirmed that the engineered
transferrin receptor (RTR) is a sensitive MR marker gene and that RTR gene expression correlates with expression
of the therapeutic genes when these are co-administered.
Amplification Systems
Signal arising from small quantities of contrast agents bound to cellular targets is typically too low to result in any
appreciable change in signal intensity in an MR image. Because of this, attention has focused on developing signal
amplification systems.
Three broad categories have been explored; one has focused on the targeting of receptors where bound MR-visible
ligands do not inhibit subsequent binding through a negative feedback mechanism. Another area has focused on
11
synthesis of contrast agents that oligomerize at a binding site, effectively autoamplifying their presence and thus
11
the signal change resulting from their accumulation. Bogdanov et al have developed a unique strategy based on
enzyme-mediated polymerization of paramagnetic substances into oligomers possessing higher magnetic relaxivity.
More specifically, they bound chelated gadolinium covalently to phenols that serve as electron donors during
enzymatic hydrogen peroxide reduction in the presence of peroxidase. Converted monomers undergo rapid
condensation into paramagnetic oligomers with increased relaxivity. This enzyme-sensing technique is essentially
translating the enzyme-linked immunoabsorbent assays used in vitro to a potentially clinical level.
Of the several novel applications that this group has identified and explored, this system has been shown to be very
sensitive by detecting nanomolar quantities of peroxidase activity and has also been used to image the expression of
E-selectin on the surface of cultured and activated endothelial cells.20 Cross-linked iron oxide nanoparticles are
conjugated with monoclonal antibody F(ab') fragments and, as Bogdanov has stated, 11 this amplification technique
can potentially be used to identify a spectrum of molecular targets simply by attaching visible enzymes to selected
antibodies for a variety of clinical applications.
A further amplification system relies on the intracellular uptake of contrast agents following delivery of a specific
promoter. By manipulating cells to express a modified transferrin receptor with a disrupted negative feedback
21
system, Weissleder et al were able to target these receptors with iron oxide nanoparticles with no subsequent
inhibition of binding. In the in vivo setting, problems with toxicity and delivery of these genes must be overcome
before this concept becomes more universally acceptable, but the basic principle is certainly an elegant and novel
one.
© 2010 Elsevier
3 of 3 14-04-2010 20:45
DELIVERY OF CONTRAST AGENTS

The biodistribution of MR contrast agents depends in large part on their chemical properties, the
nature of any underlying pathologic process, and emerging techniques for enhancing their delivery. The
gadolinium chelates remain primarily in the extracellular pool but can be targeted to the intravascular
pool, to cell membrane receptors or to the function of specific cell types, such as phagocytosis. None
of this permits imaging of any unique molecular structure or event.
Many macromolecular carriers have been bound to gadolinium not only to improve their relaxation
properties, by being able to transport more gadolinium ions,7 but also to enhance their targeted
delivery. For MR imaging to be of any practical use at the molecular level, a challenge that must be
overcome is to deliver MR-visible probes into cells. Other cross-sectional modalities, including PET
scanning, can also image a spectrum of probes targeting cell membrane receptors but the field of
molecular imaging demands detection of intracellular events.
page 414
page 415
One challenging hurdle is to design probes capable of being interiorized or of imaging intracellular
events that manifest as unique and targetable epitopes on the cell membrane. Several strategies have
been studied. The first is the development of probes capable of reaching the intracellular pool. The
second targets cell membrane epitopes arising in response to a specific intracellular event. The third
focuses on mechanical or pharmaceutical enhancement of probe delivery. Of the recognized carriers,
liposomes can be specifically targeted to cell membrane receptors by attaching antibodies to their
surface. Sipkins et al4,22 have taken advantage of the relatively large size of the ensuing intravascular
complex by targeting these MR-visible probes to specific endothelial receptors only expressed on
newly formed endothelial cells.
The use of polylysine has been explored as a molecule capable of enhancing peptide and contrast
delivery into cells. Ligand molecules bound to polylysine can be electrostatically conjugated to DNA and
can bind receptors or antigens on cell membranes with subsequent delivery of the DNA or even MR
contrast agents into these targeted cells.23 In ongoing innovative work performed in the laboratory of
Thomas Meade,23 a new class of macromolecular agent has been developed that is capable of
transfecting genes into cells as well as enhancing the contrast of the targeted cells for MR imaging
purposes. In one example, the delivered DNA both encodes a marker gene and serves as a molecular
scaffold to bind polylysine conjugated to transferrin or a gadolinium chelate (see Figs 15-1 to 15-3).
Arbab et al5 labeled mammalian tumor and stem cells with a superparamagnetic iron oxide
nanoparticle-polylysine complex and were able to demonstrate no short- or long-term toxic effects.
Polyamines are also capable of enhancing targeted delivery of MR contrast agents. In studies aimed
at imaging Alzheimer's amyloid plaques, Poduslo et al24 have synthesized a putrescine-gadolinium-
amyloid-β peptide probe, which not only traverses the blood-brain barrier (a necessary property
conferred by the polyamine moiety) but also binds to the amyloid plaques, selectively enhancing these
for imaging purposes.
Intracellular delivery of contrast agents has been facilitated by the use of the cell-permeating HIV-1 tat
(transactivator) peptide and its derivatives.25 Lewin et al26,27 used short tat peptides bound to
nanoparticles which were then internalized into hematopoietic and neural progenitor stem cells without
altering cell viability, differentiation or proliferation. Not only could the biodistribution of these cells be
followed in vivo, but the cells could also be subsequently retrieved using magnetic extraction techniques
28
for analysis of interactions with target organs. Dodd et al labeled T cells with nanoparticles
derivatized with the tat peptide and, having shown that no interference occurred with activation or
activation-induced cell death, were able to study in vivo trafficking of these cells in a live animal model
using MRI. Other authors have used MR microscopy to study tracking of neural precursors migrating
into periventricular white matter of rat brains.29
During tumor growth, several peptide products of growth-stimulating oncogenes are expressed and
1 of 2 14-04-2010 20:47
released into the cytoplasm. Activation of one of these oncogenes (c-myc) results in the increased
accumulation of c-myc mRNA in the tumor cell cytoplasm; this is a unique peptide sequence expressed
only as a result of this process of oncogene activation. Heckl et al30 have developed a contrast agent
composed of a gadolinium complex, a transmembrane carrier peptide, and an oligonucleotide
sequence (also referred to as a peptide nucleic acid) which binds specifically to the c-myc mRNA.
They were able to show an increased accumulation of their visible complex both in vitro and in vivo,
thus providing evidence that the combination of transmembrane carriers and oligonucleotides may be
used for specific targeting of intracellular cytoplasmic peptides.
Several techniques have been employed for further enhancing local delivery of contrast agents. To
enhance delivery of a manganese-bound paramagnetic contrast agent into the brain, Aoki et al31
increased permeability and thus delivery by infusion of hyperosmolar agents. Thermal energy also
causes transient increases in endothelial permeability, and direct delivery of the endothelial
permeability factor VEGF also increases endothelial pore size to permit extravasation of
macromolecular agents.31
While not immediately applicable to clinical MR imaging, it is anticipated that some of these options
may well be used more widely in the future. Local perturbations in magnetic field strength have not
been fully explored as an option for increasing contrast or macromolecular agent delivery through
induction of alterations in endothelial cell permeability.
© 2010 Elsevier
2 of 2 14-04-2010 20:47
DETECTION SYSTEMS FOR IMAGING AT THE CELLULAR AND MOLECULAR

LEVEL
Field Strength
While the in vivo resolution afforded by MR imaging remains vastly superior to that of competing
molecular imaging technologies such as PET scanning, the sensitivity is far less. While this has resulted
in much research effort aimed at synthesizing visible probes that target specific molecular or cellular
sites or activities, similar progress has been made in the engineering of higher field strength magnets.
The higher field strength magnets used for small animal research are currently of the order of 7-12 T
and are able to resolve anatomic detail at a micrometer level. At the clinical level, magnet strengths
also continue to increase, with 3 T magnets now being widely used in the clinical setting. These
magnets not only provide better resolution and improved anatomic delineation, but are also allowing
techniques such as spectroscopy to finally enter into the clinical arena for many clinical applications.
Magnetic Resonance Spectroscopy

page 415
page 416
For MR spectroscopy, where anatomic detail is clearly not relevant, these research magnets are
capable of detecting signal arising from picomolar quantities of sample. Spectroscopy offers the
potential for an entirely different approach to imaging at the cellular or molecular level. From the
subcellular to the human in vivo level, the technique is well established for determining local
concentrations of molecules and metabolites.
The specific molecules imaged depend entirely on the nucleus being "imaged". This chapter will not
discuss the innumerable applications for metabolic imaging. However, MR spectroscopy can certainly
be used to image products of gene expression and several strategies have been investigated for this
purpose. As an example, several marker enzymes can be activated by specific enzymes to produce
products visible on spectroscopy. A commonly "delivered" marker gene used for these purposes
results in local expression of the enzyme cytosine deaminase which catalyzes the conversion of the
19
chemotherapeutic pro-drug 5-fluorocytosine into 5-fluorouracil, which is visible using F MR
spectroscopy.32 The scientific advantage of this conversion is that not only can an indirect product of
gene expression be imaged in vivo but also the activated pro-drug is toxic and will produce local
33
cytostatic effects. Several similar approaches have been studied and reported.
Magnetic Resonance Microscopy
Add to lightbox
Figure 15-6 Left, A coronal slice through the right kidney from the same data set (50 microns at 7.1 T)
displayed in Fig. 15-2B can be compared to the same level at 100 microns at 2.0 T (in Fig.15-1B) or,
right, the excised kidney scanned at 8× higher resolution (25 microns) at 9.4 T. Note the following
gross structures on the left: (a) liver, (b) stomach, and (c) spleen. On the higher resolution image on the
right, one can see the following: (d) medulla, (e) arcuate veins, (f) papilla, and (g) medullary rays.
(Reprinted with permission of the publisher from reference 34.)
1 of 2 14-04-2010 20:47
MR microscopy is a term which describes the production of images of microscopic details using the
NMR phenomenon. Three-dimensional MR microscopic images have been recorded with isotropic
resolution down to 25 μm3 in isolated organs34 (Figs. 15-6 and 15-7). Acquisition times tend to be
longer than conventional clinical MR, in part to allow for sufficient signal-to-noise ratio (SNR), and much
of the focus in MR microscopic research has aimed to improve the SNR through operating at higher
magnetic fields (now in the order of 12 T), use of more sensitive surface or high-temperature
superconducting coils, and use of more efficient pulse sequences. To date, the major applications have
34
been in morphologic phenotyping of small transgenic mouse models (see Figs. 15-6 and 15-7). An
exciting advance in this field will be MR histology, which will combine high-resolution MR imaging with
MR-visible histologic probes.
© 2010 Elsevier
2 of 2 14-04-2010 20:47
CLINICAL IMAGING OPPORTUNITIES

In order to image unique structures at the molecular level, a series of essential steps must first be met.
First, a contrast agent must be synthesized that will specifically interact with a unique site or peptide or
be activated by a specific molecule. Second, this agent must be delivered in sufficient quantity to the
site being imaged. Third, this agent must reach and interact with its specific corresponding target site,
even if this involves traversing a cell membrane and entering the cell cytoplasm. Fourth, an
amplification system may be required to permit imaging of the agent in vivo; and lastly, an MR
detection system should permit specific and anatomically correct images to be obtained with little
interference from background signal.
Each of these components is being actively investigated and we have highlighted these advances.
However, the ultimate goal is to introduce these emerging MR technologies into the clinical setting and
promising provisional research studies have already been performed and reported.
Add to lightbox
page 416
page 417
Figure 15-7 The unique ability to survey the entire volume is shown in a 512 × 512 × 512 array
acquired through the thorax and abdomen at 7.1 T. A, A single 50-micron axial image shows liver,
stomach, and intestine. B, A magnified view of the upper right quadrant allows one to appreciate the
individual liver lobules. C, The same level volume rendered with a 500-micron slice highlights the
vascular detail. D, The magnified view shows vessels estimated to be 100 microns. (Reprinted with
Advances in our understanding of disease processes, coupled with molecular characterization of

interrelated steps contributing to these processes, have resulted in a host of individual targets which
can potentially be imaged. Tumor growth and treatment options represent an excellent paradigm for
such targets. To illustrate this concept, angiogenesis is an essential requisite for invasive tumor growth
to occur. Angiogenesis is preceded by the so-called "angiogenic switch", a complex series of events
initiated in part by hypoxia and activation of tumor-promoting oncogenes. Imaging at the molecular level
now permits the earliest stages of angiogenesis to be imaged in vivo, as well as the presence of
hypoxia and resulting peptide growth factors expressed by tumor-activating genes.
Moreover, molecular and cellular imaging is now being used not only to optimize tumor treatment
options but also to monitor drug delivery and efficacy. As examples, drug choice is facilitated by
identification of specific targets or receptors present in tumors or on tumor cell surface membranes.
The p-glycoprotein multidrug resistance (MDR) efflux pumps, which contribute to decreased drug
efficacy, can be imaged by their ability to bind specific ligands. Equally importantly, delivery of
molecular medicines, including gene therapy, has stimulated an enormous search for imaging markers
of successful gene delivery, uptake, and function and MRI has already played a central role in this
1 of 2 14-04-2010 20:55
expanding field.
In addition, the recent identification of a host of antiangiogenic therapies has altered the way in which
response to treatment is typically monitored in tumor. Many of these new agents are cytostatic rather
than cytotoxic and traditional measures of change in tumor size must thus be replaced by newer
indices such as apoptosis (cell death) and absent cell division. Indirect estimates of these responses
are provided by contrast perfusion studies but molecular imaging options are becoming available to
directly estimate the effects of such treatments.
Imaging researchers have thus responded to the spectrum of contemporary clinical requirements by
focusing on emerging technology for imaging basic molecular and cellular events in vivo. These
advances have concentrated primarily on the synthesis of novel contrast agents designed specifically
to depict unique molecular or cellular events or structures.
© 2010 Elsevier
2 of 2 14-04-2010 20:55
TARGET SITES AND CLINICAL APPLICATIONS

The availability and identification of suitable targets for imaging purposes are seemingly unlimited. As
old or even newly recognized biological processes become elucidated, imaging targets are defined.
The ability to image these targets is limited only by constraints associated with contrast synthesis. The
major advances in development of targeted contrast agents have focused on imaging of receptors, cell
proteins, genes, specific cell function such as migration or phagocytosis, and identification of specific
ions such as calcium.
Gene Therapy
A major stimulus to the growth of molecular imaging has been the field of gene delivery. Gene therapy
offers an enormous potential for MR not only for assessing response to treatment when therapeutic
genes are employed for tumor treatments, but also for confirming satisfactory delivery, uptake, and
expression of selected gene products.
MR offers several strategies for imaging specific expressed products of delivered genes. There is
33
currently no MR marker for endogenous gene expression and researchers have thus focused on
indirect imaging of gene expression. Since the specific products expressed by delivered genes will
typically possess desired physiologic properties, it is not practical or even advisable to try to image
these directly. For this reason, a host of surrogate marker genes have been explored which can be
co-administered along with the therapeutic gene without affecting its function. In the in vitro setting,
imaging marker genes commonly employed include those encoding β-galactosidase and luciferase.
Several MR contrast agents have been synthesized which are activated or made visible by the local
presence of these enzymes. By co-administering the gene encoding these "visible" marker gene
products along with a desired therapeutic gene, satisfactory delivery, uptake integration, and
expression of the desired gene products are inferred by the presence of the visible gene products.
Thus the biological actions of expressed products can be imaged or visible products can be
co-administered along with the desired therapeutic gene. In addition, an entire generation of what are
routinely referred to as "smart contrast agents" is being synthesized which are capable of being
activated or made visible by physiologic events within a cell. Two representative examples are
compounds that are made visible by the new expression of intracellular enzymes such as
galactosidase (referred to as "EgadMe" by Thomas Meade) (see Figs. 15-1 to 15-3), and agents that
are sensitive to the concentrations of specific intracellular ions such as calcium. As an example, the
caged gadolinium chelate DOPTA-GD is sensitive to intracellular calcium levels and changes its
35
relaxivity according to the calcium ion concentrations. By incorporating genes that express peptides
capable of altering intracellular calcium signaling, it is possible that this technology can be used to
indirectly monitor gene delivery and function.33
Apoptosis
Apoptosis describes a series of molecular steps ultimately resulting in cell death. Specific biochemical
events have been shown to occur at different stages of apoptosis. The ability to image apoptosis
noninvasively is especially important for documenting response of tumors to chemotherapeutic drugs.
Until now, these events could only be documented in ex vivo biopsy samples or in cell culture. As an
example, annexin V is a molecule that recognizes a phosphatidylserine present on the surface of
apoptotic cells. Using nanoparticles conjugated to annexin V, Schellenberger et al11 were able to
successfully probe apoptotic cells in vitro, thus providing proof of the concept that specific steps which
occur during apoptosis can be imaged using MRI. We look forward to animal MR studies confirming
these results.
page 417
page 418
1 of 7 14-04-2010 20:56
Add to lightbox
Add to lightbox
Figure 15-8 pH image of a control female SCID mouse. A, Fat-suppressed, proton density-weighted
reference anatomic image showing a coronal view of the kidneys and nearby tissues. B,
Corresponding calculated pH image. A pH of 6.0 was measured in a urine sample collected
immediately following the imaging experiment. (Reprinted with permission of the publisher from
reference 37.)
Many of the emerging chemotherapeutic and biological agents being used to treat tumors are
cytostatic rather than cytotoxic and new techniques are required to evaluate their efficacy since
changes in tumor size may not occur. Partly as a result of this, attention is being focused on developing
noninvasive imaging techniques for the detection of apoptosis, for which many in vitro laboratories
exist. Cells undergoing apoptosis express anionic phospholipids such as phosphatidylserine on their
surface, providing a specific target for imaging. Zhao et al3 have shown that the C2 domain of
synaptotagmin I, when bound to iron oxide nanoparticles, binds to these lipids and can be imaged with
MR in vivo. Delivery of contrast agents to these sites is further enhanced by additional
pathophysiologic events including increased endothelial permeability that occurs in tumors or in sites of
inflammation.
Calcium Ions
2 of 7 14-04-2010 20:56
Add to lightbox
Add to lightbox
Figure 15-9 pH image of an acetazolamide-treated mouse. A, T1-weighted, post-contrast reference
anatomic image showing the kidneys and nearby tissues. The cortex, medulla, calyx, and proximal
segments of both ureters are visible. B, Corresponding calculated pH image. A pH of 8.2 was
measured in a urine sample collected immediately following the imaging experiment. (Reprinted with
Intracellular calcium plays a central role in the regulation of signal transduction and cell function. Given
the intense research interest being focused on the spectrum of factors controlling cell growth, a variety
of laboratory techniques are currently employed for measuring calcium levels in vitro. Until recently,
techniques for the in vivo measurement of calcium ions have been lacking. In order to study the
regulatory role played by calcium ions in vivo, Li et al35 have studied the MR contrast agent
DOPTA-GD, in which the relaxivity of the complex is controlled by the presence or absence of the
divalent Ca2+ ion. The authors showed that after Ca2+ is bound to DOPTA-GD, the molecule
undergoes a conformational change that opens up the hydrophilic face of the tetra-azacyclododecane
macrocyle. Manganese-based contrast agents accumulate in cells through active transport via
voltage-gated calcium channels,31 thus producing local signal increases in T1-weighted images. This
activity-induced manganese-dependent imaging has been used to document sites of cell activation
following functional stimulation and provides an alternative approach to in vivo imaging of a very
specific calcium-controlled cell transport function.31 It is likely that similar applications will be developed
for the in vivo determination of other metabolically significant ions.
pH Sensitivity
Work from Dean Sherry's group has focused on changes in contrast agent relaxivity at different pHs. A
36
complex was synthesized with a highly pH-dependent proton relaxivity capable of depicting in vivo pH
levels in a color-coded manner (Figs. 15-8 and 15-9).37 More recently, gadolinium complexed to a
hydroxypyridyl ligand showed differential relaxivity at different pH levels, attributed to an increase in the
prototropic exchange of the co-ordinated water molecule at lower pHs and deprotonation of the ligand
3 of 7 14-04-2010 20:56
amide protons at a higher pH.38 The elegant concepts explored by these researchers will have an
impact on in vitro pH determinations, potentially obviating the need for probe insertion into delicate
39
physiologic systems. As an extension of this work, Ward et al have demonstrated that low molecular
weight compounds with slowly exchanging protons (such as -NH or -OH) alter tissue contrast via
chemical exchange saturation transfer (CEST) of presaturated spins to bulk water. This contrast
mechanism permits modulation of the contrast effect via gating of the RF presaturation pulse.
page 418
page 419
Add to lightbox
Figure 15-10 CEST images of phantoms containing 10 mM Eu(1) and either 0, 5, 10 or 20 mM
glucose . PIPES (100 mM) was used to buffer the pH at 7.0. The image parameters were as follows:
TR/TE = 3000/18 ms, FOV = 40 × 40 mm, thickness 2 mm, data matrix 256 × 256, saturation duration
time of 2 s at a power of 1020 Hz at a frequency offset of 50 and 30 ppm. The CEST image was
obtained by subtracting pixel by pixel the image at 50 ppm from that at 30 ppm. (Reprinted with
permission from the publisher from reference 40. Copyright 2003 American Chemical Society.)
Agents have been developed capable of sensing not only pH but also lactate concentrations. More
recently, further work from Sherry's laboratory40 has described a paramagnetic CEST agent that is
able to detect tissue glucose levels (Fig. 15-10). This agent exploits the ability of boronic acids to
bind selectively and reversibly with structures containing cis-diols, such as saccharides and glycated
41
proteins. More recently, Luciano et al have used paramagnetic nonionic vesicles (so-called
"niosomes") coated with polyethylene glycol and glucose conjugates to target and thus image
glucose receptors on tumor cells in vivo (Fig. 15-11).
Angiogenesis
MR imaging with appropriate use of targeted contrast agents offers the real potential to image and
thus evaluate different components of angiogenesis. The antibody-directed nanoparticle complexes
targeting receptors expressed on newly produced endothelial cells have been explored in several in
vivo settings. As alluded to above, MR contrast agents already exist for the identification of newly
expressed endothelial cell receptors in vivo. Those already in use can target either integrins or
selectins and provide far more sensitive and specific information than older compartmental modeling
techniques used to indirectly estimate tumor perfusion and permeability, vascular densities, and global
vascularity.
4 of 7 14-04-2010 20:56
Add to lightbox
Add to lightbox
Figure 15-11 Transverse spin-echo T1-weighted MR images (500/10; section thickness, 3 mm) in two
mice with xenograft PC3 tumors (arrows) 24 hours after injection of contrast agents. A, Saline solution.
B, Glycosylated PEG 4400 niosomes. Image shows increased tumor RI and tumor-to-muscle CNR,
with liver (Li) and gallbladder (Gb) enhancement. Lu, lung; m, muscle; *, heart. (Reprinted with
The imaging of angiogenesis and its regression is not limited to oncology alone. It is important to
consider that angiogenesis is also an integral component of plaque growth and the rapidly emerging
field of proangiogenic therapies seeks noninvasive imaging correlates to document the extent of new
vessel formation in ischemic limbs. In order to document events initiating plaque growth and
subsequent instability, Winter et al42 studied angiogenesis in developing atherosclerotic plaques in
aortic walls of rabbits and were able to specifically identify new endothelial cells in the plaques.
page 419
5 of 7 14-04-2010 20:56
page 420
Cell Tracking
Current understanding of the biodistribution and engraftment of stem cells is limited by the lack of
suitable imaging technologies for visualizing these specific cell populations in vivo. One limitation has
been the ability to label these cells with visible contrast agents without altering cell function, distribution
or differentiation. By labeling purified populations of these cells with MR-visible contrast agents and
ensuring that this labeling does not alter cell trafficking, function or biodistribution, MRI can not only
43
image these cells with standard imaging systems but has provided unique insights into the localization
and delivery patterns of pluripotent stem cells.
44
Bulte et al have developed magnetodendrimers, which they consider a versatile class of magnetic
tags for labeling mammalian stem cells. Through nonspecific membrane adsorption into neural stem
cells, these dendrimers localize to cytoplasmic endosomes without altering cell viability, their capacity
to proliferate or to transform into neurons. As an additional example, using commercially available
contrast agents, several groups continue to study the biodistribution of stem cells in a variety of animal
models in vivo. Jendelova et al45 studied the in vivo fate of bone marrow stromal cells labeled with iron
oxide nanoparticles. They were able to image signal arising from these cells for up to 50 days. Hill et
46
al used MR to study the delivery and tracking of labeled endomyocardial stem cells. These cells
were readily visible in the beating heart. Using micron-scale iron oxide particles, Hinds et al47 have
developed a technique capable of detecting single cells at resolutions that can be achieved in vivo.
Cell-Based Therapies
Biochemotherapeutic protocols now incorporate additional cell types into clinical treatment strategies,
including dendritic cells and lymphocytes, both of which can be engineered to target unique tumors
typically through their membrane receptors. Once administered, the in vivo fate of these cells remains
a mystery and the efficacy must typically be inferred from changes in the size of the targeted
neoplasm.
MR imaging offers the potential to study the biodistribution of these cells in vivo. Kircher et al48 have
developed a biocompatible inert cross-linked nanoparticle that permits intracellular labeling of a variety
of cells at near single-cell resolution. Using labeled cytotoxic T lymphocytes and MRI, the authors were
48
able to image approximately 3 cells/voxel in live mice and to document not only the spatial and
temporal heterogeneity of lymphocyte tracking in vivo but also lymphocyte recruitment by growing
melanoma cells in an animal model.
Dual-Function Contrast Agents

As the specificity of contrast targeting is confirmed in more and more clinical studies, additional
therapeutic roles are being explored for these molecules. Since some of these agents are able to
successfully bind to tumor cell membrane receptors, a logical consideration is for them to be used as
49
carriers of chemotherapeutic or antiproliferative agents. Lanza et al have targeted an antiproliferative
drug to vascular smooth muscle attached to an MR-visible nanoparticle and demonstrated not only that
they were able to target this agent to a very specific site but could quantitate the extent of drug
delivery based on MR signal changes. If the peptides used to target these agents do not interfere with
therapeutic drug function, and the drugs must clearly work through alternative binding sites on the cell
membrane, then this type of dual-function contrast molecule may represent an important new role for
MR imaging.
Perhaps an optimal use for MR will ultimately be the ability to obtain a combination of anatomic,
biochemical, and functional information. As an example of how this may relate to the molecular
evaluation of tumors, Bhujwalla et al50 have illustrated how MR can be used to depict vascular volume
of a tumor and have co-registered these data with tissue permeability and the extracellular pH (Fig.
15-12).
6 of 7 14-04-2010 20:56
© 2010 Elsevier
7 of 7 14-04-2010 20:56
CONCLUSIONS AND FUTURE OPPORTUNITIES

Technical advances in magnet and receiver coil construction, coupled with innovative engineering of
new MR-visible contrast agents, are resulting in MR becoming the premier technique for clinical
imaging at the cellular and molecular level. It is expected that these advances will not be limited to
anatomic delineation alone but will expand into noninvasive determinations of cell membrane enzyme
activities, measurement of receptor densities and evaluation of cell cycling for optimizing selection and
timing of drug therapies, and a host of research applications which will ultimately impact on patient
care. We fully expect that the new paradigm afforded by MR will support an ongoing dramatic pace of
scientific advances.
page 420
page 421
Add to lightbox
Figure 15-12 Co-registration of vascular volume, permeability, and extracellular pH, measured with
3
IEPA 3D reconstructed maps obtained from a single MDA-MB-231 tumor (460 mm ). A, An MRI map
of vascular volume (range 0-200 l/g). B, An MRI map of vascular permeability (range 0-7 l/g-minute). C,
A fused map of vascular volume and permeability. D, A fused map of vascular volume, permeability,
and pH (blue, range 5.3-7.2). E, Hematoxylin and eosin-stained histologic sections. Spatial resolution
of the vascular volume and permeability maps is 0.125 mm in-plane with 1 mm slice thickness. The pH
map was obtained with a spatial resolution of 1 × 1 × 4 mm. (Reprinted with permission of the
publisher from reference 50.)
REFERENCES
1. Louie AY, Huber MM, Ahrens ET, et al: In vivo visualization of gene expression using magnetic resonance imaging. Nat
Biotechnol 18:321-325, 2000. Medline Similar articles
2. Artemov D, Mori N, Ravi R, Bhujwalla ZM: Magnetic resonance molecular imaging of the HER-2/neu receptor. Cancer Res
3. Zhao M, Bearegard DA, Loizou L, et al: Non-invasive detection of apoptosis using magnetic resonance imaging and a
targeted contrast agent. Nat Med 7:1241-1244, 2001. Medline Similar articles
4. Sipkins DA, Cheresh DA, Kazemi MR, et al: Detection of tumor angiogenesis in vivo by alphaVbeta3-targeted magnetic
resonance imaging. Nat Med 4:623-626, 1998. Medline Similar articles
5. Arbab AS, Bashaw LA, Miller BR, et al: Characterization of biophysical and metabolic properties of cells labeled with
1 of 3 14-04-2010 20:57
superparamagnetic iron oxide nanoparticles and transfection agent for cellular MR imaging. Radiology 229:838-846, 2003.
6. Weissleder R, Elizondo G, Wittenberg J, et al: Ultrasmall superparamagnetic iron oxide: characterization of a new class of
contrast agents for MR imaging. Radiology 175:489-493, 1990. Medline Similar articles
7. Artemov D: Molecular magnetic resonance imaging with targeted contrast agents. J Cell Biochem 90:518-524, 2003.
8. Weissleder R, Lee AS, Khaw BA, et al: Antimyosin-labeled monocrystalline iron oxide allows detection of myocardial infarct:
MR antibody imaging. Radiology 182:381-385, 1992. Medline Similar articles
9. Zimmer C, Weissleder R, Poss K, et al: MR imaging of phagocytosis in experimental gliomas. Radiology197:533-538,
1995.
10. Reimer P, Weissleder R, Lee AS, et al: Asialoglycoprotein receptor function in benign liver disease: evaluation with MR
11. Schellenberger EA, Bogdanov A Jr, Hogemann D, et al: Annexin V-CLIO: a nanoparticle for detecting apoptosis by MRI.
Mol Imaging 1:102-107, 2002. Medline Similar articles
12. Bogdanov A Jr, Matuszewski L, Bremer C, et al: Oligomerization of paramagnetic substrates result in signal amplification
and can be used for MR imaging of molecular targets. Mol Imaging 1:16-23, 2002. Medline Similar articles
13. Li KC, Bednarski MD: Vascular-targeted molecular imaging using functionalized polymerized vesicles. J Magn Reson
14. Flacke S, Fischer S, Scott MJ, et al: Novel MRI contrast agent for molecular imaging of fibrin: implications for detecting
vulnerable plaques. Circulation 104:1280-1285, 2001. Medline Similar articles
15. Hogemann D, Nitziachristos V, Josephson L, Weissleder R: High throughput magnetic resonance imaging for evaluating
targeted manoparticle probes. Bioconjug Chem 13:116-121, 2000.
16. Huber MM, Staubli AB, Kustedjo K, et al: Fluorescently detectable magnetic resonance imaging agents. Bioconjug Chem
17. Modo M, Cash D, Mellodew K, et al: Tracking transplanted stem cell migration using bifunctional, contrast-enhanced,
magnetic resonance imaging. Neuroimage 17:803-811, 2002. Medline Similar articles
18. Moore A, Josephson L, Bhorade RM, et al: Human transferrin receptor gene as a marker gene for MR imaging. Radiology
20. Kang HW, Josephson L, Petrovsky A, et al: Magnetic resource imaging of inducible E-selectin expression in human
endothelial cell culture. Bioconjug Chem 13:122-127, 2002. Medline Similar articles
21. Weissleder R, Moore A, Mahmood U, et al: In vivo magnetic resonance imaging of transgene expression. Nat Med
22. Sipkins DA, Gijbels K, Tropper FD, et al: ICAM-1 expression in autoimmune encephalitis visualized using magnetic
resonance imaging. J Neuroimmunol 104:1-9, 2000. Medline Similar articles
23. Kayyem JF, Kumar RM, Fraser SE, Meade TJ: Receptor-targeted co-transport of DNA and magnetic resonance contrast
agents. Chem Biol 2:615-620, 1995. Medline Similar articles
24. Poduslo JF, Wengenack TM, Curran GL, et al: Molecular targeting of Alzheimer's amyloid plaques for contrast-enhanced
magnetic resonance imaging. Neurobiol Dis 11:315-329, 2002. Medline Similar articles
25. Zhao M, Weissleder R: Intracellular cargo delivery using tat peptide and derivatives. Med Res Rev 24:1-12, 2004.
26. Lewin M, Carlesso N, Tung CH, et al: Tat peptide-derivatized magnetic nanoparticles allow in vivo tracking and recovery of
progenitor cells. Nat Biotechnol 18:410-414, 2000. Medline Similar articles
page 421
page 422
27. Josephson L, Tung CH, Moore A, Weissleder R: High-efficiency intracellular magnetic labeling with novel
superparamagnetic-Tat peptide conjugates. Bioconjung Chem 10:186-191, 1999.
28. Dodd CH, Hsu HC, Chu WJ, et al: Normal T-cell response and in vivo magnetic resonance imaging of T cells loaded with
HIV transactivator-peptide-derived superparamagnetic nanoparticles. J Immunol Methods 256:89-105, 2001. Medline
Similar articles
29. Bulte JW, Ben-Hur T, Miller BR, et al: MR microscopy of magnetically labeled neurospheres transplanted into the Lewis
EAE rat brain. Magn Reson Med 50:201-205, 2003. Medline Similar articles
30. Heckl S, Pipkorn R, Waldeck W, et al: Intracellular visualization of prostate cancer using magnetic resonance imaging.
Cancer Res 63:4766-4772, 2003. Medline Similar articles
31. Aoki I, Tanaka C, Takegami T, et al: Dynamic activity-induced manganese-dependent contrast magnetic resonance
imaging (DAIM MRI). Magn Reson Med 48:927-933, 2002. Medline Similar articles
32. Stegman LD: Noninvasive quantitation of cytosine deaminase transgene expression in human tumour xenografts with in
vivo magnetic resonance spectroscopy. Proc Natl Acad Sci USA 96:9821-9826, 1999. Medline Similar articles
2 of 3 14-04-2010 20:57
33. Bell JD, Taylor-Robinson SD: Assessing gene expression in vivo: magnetic resonance imaging and spectroscopy. Gene
Therapy 7:1259-1264, 2000. Medline Similar articles
34. Johnson GA, Cofer GP, Gewalt SL, Hedlund LW: Morphologic phenotyping with MR microscopy: the visible mouse.
35. Li WH, Parigi G, Fragai M, et al: Mechanistic studies of a calcium-dependent MRI contrast agent. Inorg Chem
36. Zhang S, Wu K, Sherry AD: A novel pH-sensitive MRI contrast agent. Angew Chem Int Ed Engl 38:3192-3194, 1999.
37. Raghunand N, Howison C, Sherry AD, et al: Renal and systemic pH imaging by contrast-enhanced MRI. Magn Reson Med
38. Woods M, Zhang S, Ebron VH, Sherry AD: pH-sensitive modulation of the second hydration sphere in lanthanide (III)
tetraamide-DOTA complexes: a novel approach to smart contrast media. Chemistry 9:4634-4640, 2003. Medline
Similar articles
39. Ward KM, Aletras AH, Balaban RS: A new class of contrast agents for MRI based on proton chemical exchange dependent
saturation transfer (CEST). J Magn Reson 143:79-87, 2000. Medline Similar articles
40. Zhang S, Trokowski R, Sherry AD: A paramagnetic CEST agent for imaging glucose by MRI. J Am Chem Soci
125:15288-15289, 2003.
41. Luciani A, Olivieri J-C, Clement O, et al: Glucose-receptor MR imaging of tumors: study in mice with PEGylated
paramagnetic niosomes. Radiology 231:135-142, 2004. Medline Similar articles
42. Winter PM, Morawski AM, Caruthers SD, et al: Molecular imaging of angiogenesis in early-stage atherosclerosis with
alpha (v) beta3-integrin-targeted nanoparticles. Circulation 108:2770-2274, 2003.
43. Daldrup-Link HE, Rudelius M, Oostendorp RA, et al: Targeting of hematopoietic progenitor cells with MR contrast agents.
44. Bulte JW, Douglas T, Witwer B, et al: Magnetodendrimers allow endosomal magnetic labeling and in vivo tracking of stem
cells. Nat Biotechnol 19:1141-1147, 2001. Medline Similar articles
45. Jendelova P, Herynek V, DeCroos J, et al: Imaging the fate of implanted bone marrow stromal cells labeled with
superparamagnetic nanoparticles. Magn Reson Med 50:767-776, 2003. Medline Similar articles
46. Hill JM, Dick AJ, Raman VK, et al: Serial cardiac magnetic resonance imaging of injected mesenchymal stem cells.
Circulation 108:1009-1014, 2003. Medline Similar articles
47. Hinds KA, Hill JM, Shapiro EM, et al: Highly efficient endosomal labeling of progenitor and stem cells with large magnetic
particles allows magnetic resonance imaging of single cells. Blood 102:867-872, 2003. Medline Similar articles
48. Kircher MF, Allport JR, Graves EE, et al: In vivo high resolution three-dimensional imaging of antigen-specific cytotoxic
T-lymphocyte trafficking to tumors. Cancer Res 63:6838-6846, 2003. Medline Similar articles
49. Lanza GM, Yu X, Winter PM, et al: Targeted antiproliferative drug delivery to vascular smooth muscle cells with a magnetic
resonance imaging nanoparticle contrast agent: implications for rational therapy of restenosis. Circulation 106:2842-2847,
50. Bhujwalla ZM, Artemov D, Ballesteros P, et al: Combined vascular and extracellular pH imaging of solid tumors. NMR
© 2010 Elsevier
3 of 3 14-04-2010 20:57
UNCTIONAL ODY
Pottumarthi V. Prasad
MRI provides unique opportunities to probe tissue "function" or "physiologic status." The term "function"
may mean different things to different types of tissues or organs. In neurofunctional MRI (fMRI) the
term may be used in the sense of cognitive brain function. While the actual brain function takes place in
terms of neuronal activity, MRI is very adept at characterizing the associated hemodynamic responses.
Over the past decade the field of neurofunctional MRI has grown in leaps and bounds, and is fast
becoming a regular feature in the news media. Neurofunctional MRI has potential applications from
1,2 3,4
understanding human cognition, to how acupuncture may be beneficial, to serving as an expensive
"lie" detector.5-7 Other organs and tissues within the body also are associated with specific functions
and, while their evaluation may not be of interest yet to television audiences, they may have significant
implications in terms of understanding the physiology and pathophysiology of human disease. This
chapter will elucidate such applications of MRI. MRI has the advantage that the methodology is equally
applicable to small animal models and hence allows for translation of results from preclinical to human
applications.
Functional MRI of tissue is motivated by the following: 1. to better understand physiology and
pathophysiology; 2. to provide more comprehensive characterization of pathologic lesions (e.g., clinical
or functional significance of ischemic/stenotic lesions); and 3. to provide either a more sensitive index
or an early index of disease progression (e.g., loss of proteoglycans prior to irreversible cartilage
degradation). Several MRI-derived indices can provide useful correlates to tissue function that may be
useful for satisfying one or more of these motivating factors. The most commonly used functional
parameter is "perfusion." Although the intended meaning is to measure blood flow to a region of
interest in absolute quantifiable terms (e.g., ml/min/100 g of tissue), MRI is quite capable of providing
semiquantitative or qualitative indices that can be used to monitor changes in "perfusion." There are
many specific means to achieving the end results. This chapter will review the indices that have been
reported in various organs of the body and provide illustrative specific examples. Since many of the
methods discussed have not yet entered routine clinical use, some examples will be based on
preclinical testing to demonstrate feasibility.
© 2010 Elsevier
1 of 1 15-04-2010 20:07
EVALUATION OF PHYSIOLOGIC FUNCTION BY MRI

Perfusion MRI
Blood perfusion is a crucial and fundamental physiologic process that controls delivery of nutrients to
tissue8 and thus is a very important parameter in the assessment of tissue viability and function.
Perfusion is usually expressed as milliliters of blood per minute per 100 g of tissue. Conventional in vivo
9
perfusion measurements are based on the administration of exogenous tracers which can be detected
by several different imaging techniques, such as X-ray computed tomography (CT)10 and positron
11
emission tomography (PET). All these methods are based on the indicator-dilution method introduced
a century ago by Stewart12 and further developed by Zierler.13 In practice, this method involves
introduction of a known quantity of indicator into the system under study and measurement of the
concentration of the indicator as a function of time at one or more points downstream from the location
of the injection.
page 423
page 424
There are a number of applications of nuclear magnetic resonance (NMR) techniques proposed to
image and measure tissue perfusion. Some are based on the indicator-dilution principle and others
depend on inherent MR mechanisms. Even though PET is still considered to be the "gold standard" for
in vivo blood flow measurements, MR perfusion imaging is attractive because MR scanners are more
readily available. Also, no ionizing radiation is involved, and spatial resolution with MR is much superior
to that obtained with PET. Furthermore, MRI offers the unique advantage of obtaining both anatomic
and functional information with a single modality.
Like most applications in MRI, the preliminary application of perfusion imaging was primarily targeted
at the brain. Perfusion MRI of the brain is relatively well established and is part of routine clinical
protocols (see Chapter 12). Here the application of perfusion MRI to other organs of the body is
discussed. While the fundamental principles are similar to the brain application, there are differences in
implementation with different organs such as the heart, kidney, lungs, and liver. Although perfusion MRI
of the heart is increasingly becoming part of clinical protocols, non-neuro-applications remain in the
academic domain.
Before discussing the different techniques that have been proposed to perform perfusion MRI, it is
worthwhile making it clear that many of these techniques do not measure perfusion in classical terms
(i.e., ml/min/100 g of tissue). MR is sensitive to several physiologic parameters that may be related to
perfusion and could be used as indicators of local blood flow. A brief description is given of some of
the parameters that could directly or indirectly influence the MRI signal and thus potentially be deduced
from the MR measurements.
Perfusion (f). If a volume of tissue (V) is supplied with arterial blood at a rate of F (mL/min),
then perfusion f = F/V, which is milliliters of arterial blood delivered per minute per milliliter of
tissue. Even though this is slightly different from the classical description of perfusion, it is more
applicable for imaging studies. Perfusion f is the fundamental rate constant determining delivery
of metabolic substrates to the local tissue and clearance of products of metabolism.
Blood Volume (Vb). Blood volume (Vb) is the fraction of total tissue volume within a voxel
occupied by blood (including arteries, capillaries, and veins). Even though Vb and f are
principally distinct quantities, experimentally they are found to be strongly correlated, at least in
normal brain.14
Blood Velocity (u). Because MR is very sensitive to motion, u is used to describe the perfusion
state of tissue, because changes in u could be correlated with changes in f. However, in general
f, Vb, and u are distinct and independent parameters. However, even though they may be well
correlated in normals the correlation may break down in pathologic states.
Oxygen Extraction Fraction (E). As noted before, f determines the rate of delivery of metabolic
substrates such as oxygen to the tissue, but only a fraction of the supply is actually used for
metabolism; thus, the extraction fraction E is a parameter of interest when considering perfusion
1 of 37 15-04-2010 20:13
imaging, especially as it is known that E could vary with f. By definition E = (Ca - Cv)/Ca, where
C refers to arterial (a) and venous (v) oxygen concentrations.
Tissue-Blood Partition Coefficient (p). When considering the kinetics of an agent as it passes
through a tissue via blood, it is important to know the distribution of the agent between the blood
and tissue compartments. The partition coefficient p usually refers to the ratio of tissue-to-blood
concentration of the agent at equilibrium. For an agent that remains within the blood
(intravascular agent), p equals the blood volume Vb, and for a freely diffusible agent such as
labeled water, p is approximately 1.0.
Mean Transit Time Through the Tissue (τ). Each molecule of an administered agent may trace
a different path through a tissue element, so that there will be a range of transit times. Again,
for agents that remain within the blood, the mean transit time τ is usually only a few seconds,
while for agents that diffuse out of the blood, τ is much longer.
The central volume principle12 relates the terms perfusion (f), blood tissue partition coefficient (p), and
the mean transit time (τ): f = p/τ. This principle applies to any agent whether it is extracted from the
blood or not.
MR Perfusion Imaging Techniques

This section describes a few of the established MR perfusion techniques, including the basic
mechanisms involved and how perfusion information can be derived. For this discussion, it is useful to
subdivide the techniques into two major categories: techniques based on the administration of
exogenous contrast agent; and techniques that obviate the need for exogenous contrast administration.
Each of these approaches has its own advantages and disadvantages.
Techniques Based on Exogenous Contrast Agent Administration

In nuclear medicine, the tissue concentration of a radioactive tracer is measured over time and, using
suitable tracer kinetic models, several physiologic parameters related to blood flow are estimated. 13,15
In principle, a similar approach is possible with MRI if a suitable tracer is available and measurements
could be made with sufficient temporal resolution. With the evolution of ultrafast MRI techniques,16,17
the possibility of monitoring exogenous tracer kinetics in vivo using MRI has become a reality, not only
in the brain but in almost any organ of the body.
page 424
page 425
The currently approved and most widely available MR contrast agents are paramagnetic chelates,
notably gadolinium diethylenetriamine pentetate (Gd-DTPA; Magnevist, Berlex, Wayne, NJ). Gd-DTPA
is a relaxivity agent in that it decreases both T1 and T2 relaxation times. However, because T1 rates
(1/T1) are inherently smaller than T2 rates in most tissues, relaxation time changes on a percentage
basis are much greater for T1 compared to T2. For this reason, these agents are often characterized
as T1 agents. In organs such as the heart, the Gd-DTPA leaves the blood and accumulates in the
interstitial space before being washed away by the venous supply. Thus, it is possible to follow the
accumulation of the agent over time using appropriate MR techniques and to derive perfusion indices.
In the brain, because of the blood-brain barrier, diffusion of the agent is prevented and thus it behaves
as an intravascular agent. This results in T1 effects being small because only a small fraction of the
tissue water can sample the agent which is restricted within a small vascular volume. For this reason,
neuroperfusion MRI is being performed based on T2* contrast. When Gd-DTPA is delivered as a
bolus, it creates a transient drop in signal intensity as it passes through the brain due to magnetic
18
susceptibility effects. The high magnetic moment of gadolinium alters the susceptibility of the blood
compared to the surrounding water in the tissue, and the resulting magnetic field gradients produce
signal loss. Variations in tissue magnetic susceptibility can affect MR images more profoundly than
relaxivity changes when the agent remains in the vascular space. Thus, even though the agent is
compartmentalized within 5% of the volume, the signal drop could be as high as 50%.19
Assuming that signal changes can be quantitatively related to changes in local concentration of
2 of 37 15-04-2010 20:13
Gd-DTPA, tracer kinetic principles could be applied to relate the measured concentration-time curves
to the different physiologic parameters described earlier. From the measurements of the tissue and
arterial concentration curves [Ctissue(t) and Carterial(t)] the volume of distribution of the agent can be
calculated directly as:
Because Gd-DTPA behaves as an intravascular agent in the intact brain, the volume of distribution of
the contrast agent reflects the blood volume Vb within the tissue. The arterial concentration-time curve
can be measured using additional images, including the feeding artery, acquired simultaneously.20
However, in the absence of arterial sampling, relative blood volumes can still be inferred assuming all
the voxels within the image are fed by a single arterial source. It has been shown with intravascular
agents that robust blood volume measurements can be made, but measurement of perfusion is more
difficult.21 The estimate of blood flow based on the central volume principle involves estimates of blood
volume and the mean transit time (τ).
Mean transit time is usually estimated by calculating the first moment of the measured
concentration-time curve. However, as pointed out by Weisskoff et al,22 this formulation is not quite
correct because the indicator-dilution principles really call for concentration at the outlet, whereas MR
(or any other imaging) measurements estimate the tissue concentration. In the same work, however,
the authors point out that flow measurements made using the above formulation can yield a
22
semiquantitative index that could be clinically useful. The measurement of τ using purely intravascular
agents necessitates either ideal delta function bolus injection of the contrast agent or an accurate
measure of the arterial input function which can then be deconvolved from the measured
concentration-time curve. In the case of freely diffusible tracers this is somewhat simplified by the
slower transit times through the tissue; thus, the measurement of τ is less dependent on the arterial
input curve.
APPLICATION IN ORGANS OTHER THAN THE BRAIN.

In all tissues other than the brain, Gd-DTPA leaves the vascular space and distributes within the
interstitial space. Hence, V cannot be equated to the blood volume and for the same reason the signal
intensity-time curve does not fully go back to the equilibrium level over the measurement time. Some
23
useful perfusion indices that can be correlated with blood flow can still be estimated. Wilke et al
estimated Ctissue(t) by fitting an empirically chosen portion of the signal-intensitytissue(t) curve to a
24
gamma variate function. Such a fitting method is also used to avoid contributions due to recirculation.
It is then possible to derive parameters such as apparent transit time (τapp) for the tracer through the
myocardium ("apparent" because it is different from the definition of mean transit time). It was shown
that 1/τapp and slope of the signal intensity-time curve during contrast agent washing exhibit good
correlation with myocardial blood flow, as determined by the radioactive microsphere technique. Thus,
even if absolute flow measurements may not be available, the flow indices that can be estimated using
this technique can provide relative flow distributions which are usually clinically more relevant.
Alternative pharmacokinetic models have been proposed to take into account the diffusion of contrast
agent into the interstitium.9 These are being widely applied for functional evaluation of tumors since
here it is not just the blood flow but also the changes in "permeability" of vessels that are of interest.
With neoangiogenesis there is an increase in vessel density. However, it turns out that these new
vasculatures are relatively more "leaky." Dynamic contrast-enhanced MRI is widely being accepted as
a mainstream clinical tool in clinical oncology.25 Similar methods are also being used in the evaluation
of renal function because most of the currently approved gadolinium chelates for human use are freely
filtered and excreted through the kidneys (see later discussion).26
3 of 37 15-04-2010 20:13
Recent advances in MRI contrast agents have introduced so-called intravascular agents. By virtue of
their physical size or by attaching gadolinium to macromolecules these agents remain within the
vasculature for significantly longer periods of time. While many have been proposed and shown to be
efficacious in preclinical evaluation, not many have progressed towards approval for clinical use. One
agent that has completed phase III clinical trials is MS-325 and it is believed the Food and Drug
Administration (FDA) will approve it in the next year or so.27 Ultrasmall superparamagnetic iron oxides
28
(USPIOs) are currently undergoing phase II clinical trials as potential intravascular contrast agents.
Techniques Based on Endogenous Contrast Mechanisms

page 425
page 426
Techniques relying on exogenous contrast administration have certain limitations. Repeat studies are
limited by the total amount of contrast that can be administered in one sitting. Also, in most patients
follow-up studies are necessary to monitor therapy. Thus, it is desirable to use perfusion MRI
techniques that do not require the use of exogenous contrast agents. Several such techniques are in
existence. These are either based on the inherent motion sensitivity of NMR,29,30 or by using blood
itself as an endogenous contrast agent (taking advantage of the magnetic susceptibility effect of
31-33
deoxyhemoglobin), or by magnetically labeling (using NMR) endogenous water in the large vessels
and subsequently utilizing it as a diffusible tracer.34,35
An early attempt to develop perfusion-sensitive MR methods was the intravoxel incoherent motion
(IVIM) method,29 an outgrowth of the development of the earlier diffusion-weighted imaging. The basic
motivation for this approach is that motion of blood in a randomly oriented capillary network should
have a similar effect on MR signal as random diffusional motion. However, the interpretation of IVIM
36
data has proved to be very difficult, and even in principle IVIM does not measure perfusion directly.
Nevertheless, the IVIM method may still provide information on the manner by which flow increases,
i.e., to address the question of capillary recruitment versus increased velocity.37
It has been established that oxygenated hemoglobin is diamagnetic and it becomes paramagnetic upon
deoxygenation.38 This change in susceptibility of blood leads to the creation of local magnetic field
perturbations which affect the MR signal, leading to an endogenous contrast mechanism. The
phenomenon is known as blood-oxygenation-level dependent (BOLD) contrast. It is widely used in
brain activation studies39,40 and applied in the evaluation of myocardial perfusion.41-43 The basis of
these studies is that increase in regional blood flow is not associated with the proportional increase in
regional oxygen extraction. Thus, with increased flow, the capillary and venous blood is better
oxygenated, producing increased MR signal intensity. This technique is well suited to monitor changes
in perfusion but cannot measure resting or absolute perfusion.
The third approach that makes use of an endogenous contrast mechanism uses magnetically labeled
water as a tracer. A number of approaches based on this basic idea have been proposed. All are
based on manipulating the magnetization of inflowing arterial blood and involve acquiring a
flow-insensitive and a flow-sensitive image and subtracting the former from the latter to remove the
background tissue signals. For each image there is a delay period during which the arterial blood
enters the voxel in proportion to the perfusion (f). These techniques are called arterial spin labeling
(ASL) methods. There are various implementations of this method in practice, but all can be analyzed
in similar ways. For this discussion, the echo planar imaging with signal targeting using alternating
44
radiofrequency (EPISTAR) technique put forward by Edelman et al is considered. In this
implementation, the difference between images with (flow sensitive) and without (flow insensitive)
preinversion of arterial blood using a 180-degree radiofrequency pulse is utilized to obtain perfusion
images.
Detailed quantitative models for the interpretation of flow effects have been developed by combining
single-compartment kinetics with the Bloch equations for relaxation.45 However, the general
quantitative features of inflow effects in these methods can be understood simply by considering the
4 of 37 15-04-2010 20:13
amount of blood entering the voxel and the magnetization carried by that blood. The equilibrium
magnetization in the voxel is given by M0 = m0vCt [where m0 is the equilibrium magnetization of 1 mol
of water, Ct is the concentration of water in tissue (mol/mL), and v is the volume of voxel (mL)], and
the partition coefficient of water p = Ct/Ca (where Ca is the concentration of arterial water). The
magnetization within the voxel can be viewed as being influenced by two conflicting processes:
magnetization is delivered to the voxel by arterial flow and then lost by relaxation and by venous flow.
Because water is nearly 100% extracted from the blood as it passes through the capillary bed, the
rate of loss via venous flow is much less compared to that due to relaxation and, thus, can be
neglected. The analysis can be simplified by assuming that the T1 of blood and tissue are the same.
With these definitions the approximate expressions for ∆M, the measured difference in magnetization,
can be derived.
The arterial magnetization per mole at time t after the inversion pulse is m0 for one image (without
preinversion pulse) and m0[1 - 2e-t/T1] for the other image, so ∆m = 2 m0e-t/T1. Further, the number of
spins which entered the voxel up to the time t after the inversion pulse is N = fvCat. Then, the total
magnetization difference is ∆M = N ∆m, or using the relations defined above:
It can be seen that ∆M is proportional to fT, where T = texp(-t/T1). The technique will be limited by T1,
and the magnitude of change is small. Nevertheless, with state-of-the-art MR imaging systems,
sufficiently high signal-to-noise ratios can be achieved so that these small changes can be readily
measured. In addition, the observed signal changes can be modeled directly in terms of f, thus
providing the potential for obtaining quantitative measurement of perfusion.
IMPLEMENTATION OF THE ASL TECHNIQUE.

There exist in practice several versions of the ASL principle. FAIR (flow-sensitive alternating inversion
recovery) involves a tag image acquired with a nonselective inversion pulse and a control image
acquired with a slice-selective inversion pulse centered on the image slice. 46 If these two types of
inversion pulses are carefully designed to have the same effect on the static spins in the image slice,
then the entering arterial blood is inverted with the nonselective radiofrequency pulse but fully relaxed
with the selective radiofrequency pulse. The static spins in the image slice are identically inverted
(ideally) in both experiments, so their signal subtracts out. Other implementations include variants of
47
the EPISTAR method such as PICORE (proximal inversion with a control for off-resonance effects)
and QUIPPS (quantitative imaging of perfusion with a single subtraction).48 These are more specifically
designed for functional MRI studies in the brain.
page 426
page 427
Specific Examples
Cardiac Perfusion
Next to brain perfusion, myocardial perfusion is the most used in clinical practice. Motivation again
arises from the fact that no other single modality provides both anatomy and function. Since MRI is
now considered the standard for cardiac function evaluation, with the availability of myocardial
perfusion and coronary angiography a comprehensive evaluation for ischemic heart disease can be
performed. Tremendous progress has been made in the past decade on both these fronts. The most
49-52
commonly used method is the first-pass contrast-enhanced perfusion MRI. Since this will be
extensively covered in Chapter 34, discussion here is restricted to some novel methods that have been
reported recently.
One such technique uses manganese-based contrast agent. With five unpaired electrons, manganese
is among the most potent of all paramagnetic ions in increasing water relaxation rates. Its efficacy is
further improved in vivo as it binds to macromolecules, due to decreased tumbling rates.53 Its high
relaxivity prompted the trial of manganese chloride (MnCl2) as the first MR contrast agent.53 Further
interest in manganese as an MR contrast material was spurred by evidence that it is taken up by
5 of 37 15-04-2010 20:13
viable myocytes and retained within intracellular organelles (in particular the mitochondria). 54 Washout
55
is slow, with clearance still incomplete at 1 hour. It was shown to have a short plasma half-life,
approximately 0.8 minutes, and a myocardial distribution that correlates with blood flow as determined
by microspheres.54,56,57 Early studies58 showed that when administered intravenously, it could be used
to distinguish between nonreperfused infarct and normal myocardium on MR images. Its short plasma
lifetime, flow-dependent uptake, and long intracellular retention times also suggest the useful
application of manganese in myocardial perfusion imaging, and offer the possibility of delineating
regions of ischemia with higher resolution and volumetric accuracy than can be achieved with current
59
techniques. Figure 16-1 illustrates an animal model of chronic myocardial ischemia.
Techniques based on endogenous contrast mechanisms have also been proposed to image myocardial
perfusion. These include ASL,42,60 magnetization transfer contrast,61,62 and BOLD MRI.41-43 Figure
16-2 presents BOLD MRI data obtained in a patient with single-vessel disease, both during rest and
dipyridamole-induced stress.
Add to lightbox
Add to lightbox
page 427
page 428
Figure 16-1 A, Images acquired pre- and post-administration of a manganese-based agent (EVP
1001-1; Eagle Vision Pharmaceuticals, Chester Springs, PA) in a chronic ischemia model, using an
inversion-recovery sequence with the inversion time (TI, 700 ms) chosen to null the myocardium prior
to contrast administration. Unlike the gadolinium chelates, manganese is an intracellular agent and
hence the contrast uptake is maintained over a considerably longer period. This allows high-resolution
images to be obtained. Imaging was performed 6 weeks following placement of an ameroid
constrictor on the circumflex branch of the left coronary artery. The left image was acquired before
contrast injection, and the center and right images were obtained 22 and 18 minutes, respectively,
following intravenous administration of EVP 1001-1. These latter images were acquired on two
different days, the center image under resting conditions, and the right image under dobutamine-
induced stress. Signal enhancement is observed throughout the myocardium, but to a lesser extent in
the ischemic territory than in the normal tissue. Note the improved delineation of the lesion (arrows)
under stress. The sequence used was a cardiac-triggered breath-hold segmented inversion-recovery
fast gradient-echo sequence, with TR/TE = 8/4 ms; flip angle = 20 degrees; 23 echos per segment;
161 × 256 matrix; field of view = 285 × 380 mm; and slice thickness = 8 mm. B, Because the
6 of 37 15-04-2010 20:13
enhancement is maintained for several minutes following administration of the contrast agents, the T1
of the myocardium can be measured. R1 (1/T1) maps acquired under resting conditions (left) and
dobutamine-induced stress (right) in the same slice planes as the images in A. Differences in R1
between the ischemic and normal territories are visible even at rest, though they are enhanced under
stress. (Reproduced with permission from Storey P, Danies PG, Post M, et al: Preliminary
evaluation of EVP 1001-1: a new cardiac-specific magnetic resonance contrast agent with kinetics
suitable for steady-state imaging of the ischemic heart. Invest Radiol 38:642-652, 2003.)
Add to lightbox
Add to lightbox
7 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
Add to lightbox
Figure 16-2 Patient with high-grade stenosis of the proximal left anterior descending coronary artery
(LAD). Coronary angiograms in typical A, right anterior oblique (RAO) and B, left anterior oblique
(LAO) projections. LCX, circumflex branch of the left coronary artery. T2* maps (short-axis view of left
ventricle) C, before and D, after administration of dipyridamole . In the latter, the MR examination had
to be interrupted during dipyridamole infusion owing to severe angina in the patient. Note the
reduced T2* in the anteroseptal area (the region which is associated with the diseased vessel). The
dark region at the inferolateral zone (D) is due to susceptibility artifacts arising from the
phrenicomediastinal recess. E, Ten weeks after dilation of the LAD, the measurement (also with
dipyridamole ) was finished without complications. When observing the obtained T2* maps,
differences between regions with reduced T2* values and regions of normal myocardium were less
pronounced than they were before percutaneous transluminal coronary angioplasty. Colors from black
to white reflect T2* values from 15 to 50 ms. (Reprinted from J Am Coll Cardiol, Vol. 41, Wacker CM,
Hartlep AW, Pfleger S, et al, Susceptibility-sensitive magnetic resonance imaging detects human
myocardium supplied by a stenotic coronary artery without a contrast agent, pages 834-840,
Copyright 2003, with permission from the American College of Cardiology.)
Lung Perfusion MRI

Lungs have posed an enormous challenge for MRI due to their high air-tissue interface. However, with
improved gradient hardware that allows for ultrashort echo times, lung MRI is feasible using state-
63,64
of-the-art commercial scanners. Owing to their large blood volume, lungs are good candidates for
perfusion MRI, both by endogenous ASL and exogenous contrast-based techniques. 64-66 Figure 16-3
illustrates lung perfusion imaging in a patient with pulmonary embolism. Perfusion MRI of the lung is
particularly interesting because it can be combined with oxygen-enhanced ventilation MRI to provide
comprehensive ventilation-perfusion evaluation in diseases such as pulmonary embolism. Oxygen-
enhanced ventilation is discussed again later in this chapter, and in greater detail in Chapter 76.
Renal Perfusion MRI

Renal blood flow (RBF) is approximately one-quarter of the cardiac output, the majority being devoted
to the cortex for glomerular filtration. The cortical perfusion is about 500 mL/min/100 g and medullary
8 of 37 15-04-2010 20:13
flow is only 20 mL/min/100 g.67 In clinical practice, measurement of RBF or perfusion could have a
significant effect in the evaluation of renal artery stenoses or nephropathies with microvascular
involvement because flow compromise is a source of hypertension and chronic renal failure. The
technique may also allow for monitoring interventions.
page 428
page 429
Add to lightbox
Figure 16-3 Anatomic images from two coronal slices of a patient with confirmed pulmonary embolism
along with the corresponding FAIR (TI, 1200 ms) and Gd-DPTA contrast-enhanced perfusion images.
Perfusion defect can be observed in the apical region of the right lung (arrows). An inversion recovery
half-Fourier single-shot fast spin-echo sequence was used for image acquisition with an effective TE =
20 ms; echo spacing = 3.6 ms; bandwidth = 250 kHz; slice thickness = 15 mm; field of view = 450 mm;
and a 256 × 128 matrix. For the gadolinium-enhanced perfusion study, the sequence parameters
were: a fast gradient-recalled echo sequence was used with TR/TE = 2.4/0.6 ms; bandwidth = 62.5
kHz; field of view = 420 mm; slice thickness = 12 mm; and a 128 × 96 matrix. (Courtesy of Dr Vu Mai,
Evanston Northwestern Healthcare, Evanston, IL.)
Phase-contrast-based measurement of RBF rate has been used to study the effect of renal artery
68
stenosis. The approach is similar in principle to the Doppler ultrasound method. While the method
does show changes in the flow profiles in the presence of stenotic lesions in the renal artery, its use in
other ischemic situations is limited.69 Also, the total RBF may not truly reflect the regional changes that
may occur within the kidney. Figure 16-4 illustrates intrarenal redistribution of blood flow, a feature that
total RBF measurements cannot detect.
Redistribution of renal flow away from the cortical element towards the medulla has been used to
explain the phenomenon of impaired clearance without a decrease in total RBF, a common observation
during several types of physiologic shock such as sepsis.70 Using invasive Doppler laser probes,
intrarenal redistribution of blood flow has been demonstrated during sepsis-induced bacteremia in
animal models.71 MR perfusion imaging is capable of providing similar information in a noninvasive
9 of 37 15-04-2010 20:13
fashion and hence can be readily performed in human subjects.
Preliminary data were obtained in a small number of subjects undergoing extracorporeal shock wave
lithotripsy (ESWL), which results in a physical shock to the renal parenchyma in the vicinity of the focal
point.72 Figure 16-4 illustrates the signal intensity versus time changes observed in the renal
parenchyma following the bolus injection of Gd-DTPA. Note the absence of corticomedullary
differentiation on the flow-weighted images in the vicinity of the ESWL focal spot. Since Gd-DTPA is
filtered through the kidneys without reabsorption, the signal intensity-time curves are in principle not
amenable to analysis based on central volume principle. However, semiquantitative indices, such as
upslope (initial portion of the signal intensity-time curves), can still be used to compare data from two
kidneys or from different subjects.
page 429
page 430
Add to lightbox
Add to lightbox
Figure 16-4 Gd-DTPA-enhanced first-pass perfusion imaging in a human subject following
extracorporeal shock wave lithotripsy (ESWL). A, Representative perfusion image with B, regions of
interest used for signal intensity-time plots. The blurring of the corticomedullary differentiation in the
lower pole of the right kidney is where the ESWL was focused. Note that the cortical curve in the
affected region is lower than the normal while the medullary curve in the affected region is higher than
normal. Using the slope of these curves as a relative perfusion index, it was shown that there is a
reduction in cortical flow (~30%) with a concomitant increase in medullary flow (34 ± 14%) in the
region where ESWL is focused. Ab_Cor, abnormal cortex; Ab_Med, abnormal medulla; Nor_Cor,
normal cortex; Nor_Med, normal medulla. (Reproduced with permission from Mostafavi MR, Chavez
DR, Cannillo J, et al: Redistribution of renal blood flow after SWL evaluated by Gd-DTPA-enhanced
magnetic resonance imaging. J Endourol 12:9-12, 1998.)
Even with extracellular agents it was shown by combining the upslope measurements with signal
10 of 37 15-04-2010 20:13
intensity versus time data from the arterial blood supply it is possible to derive quantitative absolute
blood flow measurements.73 This model assumes that contrast agent, prior to leaving the kidney,
behaves similarly to microspheres. Accordingly, renal perfusion per unit volume can be estimated using
the following formula:

Such measurement has been validated in an animal model.74
Alternately, with the use of novel intravascular contrast agents, quantitative perfusion indices can be
obtained using the central volume principle. This was demonstrated using MS-325, an agent that has
recently completed phase III clinical trials. MS-325 is the first small-molecule blood-pool contrast
agents for MRI.27 This gadolinium chelate binds reversibly with serum in plasma and leads to high
relaxivity (5-10 times that of Gd-DTPA over a range of magnetic field strengths and concentrations).
Excretion is still efficient because the protein binding is reversible. Due to these properties, MS-325
provides strong, persistent enhancement of the vascular space on T1-weighted MR images. Figure
16-5 shows first-pass perfusion MRI in an animal model of chronic renal artery stenosis.75
Figure 16-6 shows first-pass perfusion MRI using a novel iron-oxide-based contrast agent, ferumoxytol
(Advanced Magnetics, Cambridge, MA). Ferumoxytol consists of nanoparticles of iron oxide with a
dextran coating for biocompatibility, and is currently undergoing phase II clinical trials.28 Gradient-echo
(T2*-weighted) images are shown of the kidney following bolus administration of ferumoxytol (1 mg/kg)
in an anesthetized rabbit. Using a similar agent, quantitative regional blood flow estimations were
76,77
reported using the central volume principles.
RENAL PERFUSION MRI USING ASL.

The basic premise of signal targeting by alternating radiofrequency pulses (STAR) angiography is to
magnetically label blood in a feeding vessel and then to commence imaging after it has flown into a
target vessel. For renal artery imaging, blood in the suprarenal abdominal and thoracic aorta is labeled
using an inversion pulse, concomitant with presaturation of the imaging volume encompassing the renal
arteries (Fig. 16-7). A segmented turbo-FLASH sequence with incremental flip angles (21
lines/segment; TR/TE = 11/8 ms; flip angle = 2-42 degrees in steps of 2 degrees) to acquire the data
following typical inflow times of 200 to 600 ms. The labeling pulse (10.24 ms hyperbolic secant) was
applied on alternate acquisitions, and data were subtracted prior to reconstruction. Subtraction
eliminates all static background tissue signal while maintaining a large intravascular signal (Fig. 16-7).
The basic principle behind STAR angiography can be extended to perform perfusion imaging by simply
increasing the delay between the labeling pulse and the image readout. A single-shot acquisition
technique, such as echo planar imaging (EPI), can be used to allow for averaging. Data such as those
shown in Figure 16-7C can be used to derive quantitative perfusion estimates. 78
Liver Perfusion
page 430
page 431
11 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
Add to lightbox
Figure 16-5 Signal intensity-time profiles obtained from first-pass perfusion MRI in kidneys
instrumented with ameroid constrictors around the renal arteries. This results in a progressive stenosis
of the renal artery over a 5 to 6 week period. Shown are data obtained at three time points during the
12 of 37 15-04-2010 20:13
evolution of the stenotic lesion: from left to right, 1, 3, and 5 weeks following placement of the
constrictor. Note the signal intensity-time profiles are consistent with the intravascular distribution of
the contrast agent. Also obtained are X-ray angiography to document diameter reduction of the
affected renal artery using contralateral artery as a control. As shown, during week 1, stenosis is mild
and the signal intensity-time profiles in either kidney appear symmetric. However, as the stenosis
progresses, asymmetry in the signal intensity-time profiles evolves. (Reproduced with permission
from Prasad PV, Cannillo J, Chavez DR, et al: First-pass renal perfusion imaging using MS-325, an
albumin-targeted MRI contrast agent. Invest Radiol 34:566-571.)
Liver hemodynamics is a main determinant of hepatic function. 79,80 In patients with liver cirrhosis,
modifications of hepatic perfusion contribute to the deterioration of hepatic function by decreasing
81
blood-hepatocyte exchanges. Several methods have been proposed for the noninvasive quantification
of hepatic perfusion, including physiologic tests and imaging methods. Clearance of xenobiotics with
high liver extraction, such as indocyanine green or sorbitol , is flow limited. These compounds can
thus be used to measure hepatic perfusion in healthy subjects.82 In patients with advanced liver
disease, however, the clearance of xenobiotics is limited not only by perfusion but also by intracellular
enzymatic and transport processes. Therefore, the use of xenobiotics as markers of perfusion in liver
83
disease remains controversial. Hepatic perfusion has been measured with nuclear medicine
procedures using SPECT and PET.84,85 Despite promising reports, current nuclear medicine
techniques are limited by their low spatial and temporal resolution. Hepatic blood flow can be
86
measured with Doppler ultrasonography. This noninvasive and inexpensive method is widely available
and can be safely repeated. However, ultrasonography is operator dependent and the reproducibility
needs to be established.87 Dynamic CT has been proposed to quantify hepatic perfusion from
88
first-pass analysis of extravascular contrast agents. CT has been used in preliminary studies to
quantify perfusion abnormalities in liver metastasis, chronic liver diseases, and liver graft rejection. 89-91
MRI is increasingly used for functional imaging of the liver. Liver blood volume has been quantified with
MRI and macromolecular agents.92,93 More recently, hepatic perfusion has been assessed by
pharmacokinetic analysis of dynamic MR images obtained after bolus injection of a commercially
94
available extravascular agent. Figure 16-8 illustrates Gd-DTPA-enhanced first-pass perfusion
imaging in the liver.
Muscle Perfusion
Evaluation of the spatial and temporal aspects of skeletal muscle perfusion is relevant to the
understanding of several common disease states, such as obesity, type 2 diabetes mellitus, and
hypertension. Each of these diseases exhibits a resistance to the action of insulin which stimulates
glucose uptake in skeletal muscle, a primary site of glucose disposal. Existing techniques to
evaluate muscle perfusion include strain-gauge plethysmography,95 radioactive tracer,96 and
thermodilution technique.97 All of these provide bulk blood flow information and do not provide spatial
and temporal distributions. Electromagnetic flow recordings,98 pulsed ultrasound-Doppler, and laser-
Doppler flowmetry99 methods provide temporal perfusion information, but only for localized areas.
Alternately, MR functional imaging techniques not only provide detailed spatial information but also
allow for sufficient temporal resolution to follow regional perfusion changes.
ASL PERFUSION MRI TO ASSESS VASODILATORY RESPONSE IN CALF MUSCLE.

page 431
page 432
13 of 37 15-04-2010 20:13
Add to lightbox
Figure 16-6 Representative series of kidney images in a rabbit obtained using gradient-echo
sequences (TR/TE = 15/3 ms; flip angle = 10 degrees) with a temporal resolution of 2 seconds (right
to left and top to bottom). Ferumoxytol at a dose of 1 mg/kg was administered as a bolus through an
ear vein and simultaneously the acquisition of MRI was initiated. Note the abdominal aorta (white
arrowhead) becomes dark first with the arrival of contrast agent, followed by darkening of the renal
parenchyma, and then the inferior vena cava (thin white arrow). Also, note the relatively fast uptake and
washout in the cortex (long yellow arrow) compared to the medulla (yellow arrowhead). Also note the
return to close-to-precontrast signal levels in the kidney following first-pass consistent with
intravascular distribution of the agent. This is due to the fact that the T2* effects produce a dramatic
reduction in signal intensity only during the first pass when the concentration is high.
Figure 16-9 illustrates the flow-sensitive alternating inversion recovery (FAIR) technique, in which the
selective inversion-recovery images were referred to as control and the nonselective inversion-
recovery images as tag images,46,65 to evaluate perfusion in the calf muscle. The selective inversion
slab was centered on the imaging slice and was twice the thickness of the imaging slice to minimize
slice profile imperfection of the inversion radiofrequency pulse. Inflowing blood from outside the
imaging slice was labeled by a nonselective inversion 180-degree radiofrequency pulse. Inversion was
initiated at a set time delay (tECG), approximately 100 to 500 ms after the detected R wave, and image
acquisition was performed after a time delay of 1000 ms to allow for the inflow of labeled blood.
Depending on the heart rate of the subject, the timing of repetition was adjusted to allow data
acquisition during diastole, and time (TR) was set to five to seven R-R periods. A half-Fourier
single-shot fast spin-echo sequence with an echo spacing of 3.6 ms and an effective echo time of 21
ms was used for image acquisition. A bandwidth of 250 kHz and a slice thickness of 15 mm were
used. A series of 60 pairs of selective and nonselective inversion-recovery images were acquired,
14 of 37 15-04-2010 20:13
totaling 120 images. Two paradigms were used to vary the blood flow to the muscle: 1, after acquiring
10 pairs of selective and nonselective inversion-recovery images, the subject took a nitroglycerine pill
(0.4 mg sublingually) (Fig. 16-9A and B); and 2. on a different day, the same subject underwent a cuff
inflation-deflation paradigm (Fig. 16-9C).
BOLD MRI TO ASSESS REGIONAL CHANGES IN BLOOD FLOW IN THE CALF MUSCLE
DURING CUFF INFLATION AND DEFLATION.
Changes in muscle perfusion using BOLD MRI has been investigated. 100,101 These investigations were
done using echo planar imaging (EPI), which necessitates enhanced-gradient systems and has poor
image quality in terms of spatial resolution, distortions, and sensitivity to susceptibility artifacts.
However, it allows for much higher temporal resolution. Use of multiple gradient-echo sequence allows
for relatively high resolution images with lower temporal resolution. If the interest were following
hyperemic responses (e.g., associated with cuff deflation) the EPI technique would be preferred. On
the other hand, for monitoring changes following administration of pharmacologic agents (e.g.,
vasoactive substances), the higher resolution multiple gradient-echo technique would be preferable.
Also, with the EPI technique, change in rate of spin dephasing (∆R2*) is estimated based on change in
signal intensity (∆SI). Hence, any change in parameters other than R2* (e.g., T1 due to flow
dependence) would result in an artifactual contribution to ∆R2*. With multiple gradient-echo (mGRE),
changes in T1 will not influence R2* measurements because it will affect all the individual images
equally. Figure 16-10 shows data using BOLD MRI acquisitions in muscle.
page 432
page 433
Add to lightbox
Add to lightbox
15 of 37 15-04-2010 20:13
Add to lightbox
page 433
page 434
Figure 16-7 STAR and EPISTAR imaging in an animal model. A, Labeling pulse (vertical slab along
the descending aorta) was used for these acquisitions. For illustrative purposes an axial slice for
image acquisition is shown (horizontal slab). The hatched thick slab is an optional saturation band
applied to remove venous contributions and eliminate any potential signal changes due to involuntary
motion, e.g., peristalsis. B, Projection angiogram obtained with STAR. Note the excellent background
subtraction and the depiction of the stenosis. C, EPISTAR images obtained in a coronal plane with
different delay times in the range of 0.2 to 2.4 s following the labeling pulse. Note the symmetric signal
intensities from either kidney, reflecting similar perfusion because these data were acquired prior to
surgical placement of the constrictor. (From Prasad PV, Kim D, Kaiser AM, et al: Noninvasive
comprehensive characterization of renal artery stenosis by combination of STAR angiography and
EPISTAR perfusion imaging. Magn Reson Med 38:776-787, 1997. Reproduced with permission of
16 of 37 15-04-2010 20:13
Add to lightbox
Figure 16-8 Contrast-enhanced perfusion images in the liver obtained on a 1.5-T scanner using the
following 3D MR technique: TR/TE = 2.3/0.8 ms; flip angle = 9 degrees; 128 × 256 matrix; field of view
= 40 cm; slab thickness = 18 cm, resulting in an interpolated slice thickness of 5 mm. A four-element
phased-array coil was used without parallel imaging. Thirty-six coronal images were acquired every
3.3 seconds for 180 seconds, at first during apnea, then followed by shallow respiration. Coinciding
with the start of the sequence, 7 ml of Gd-DTPA and 20 ml of saline flush were injected at a rate of 5
mL/second using a MR-compatible power injector (Spectris, Medrad). Two representative images are
shown of a splenic artery aneurysm (left) and a hepatocellular carcinoma in a cirrhotic patient (right).
(Courtesy of Drs Krinsky and Lee, NYU.)
Blood Volume Changes Evaluated by Contrast-Enhanced MRI

The first measurement of brain activation with MRI was based on cerebral blood volume changes.102 In
the resting state the subject lay quietly in the dark. In the active state, the subject viewed a flashing
grid of red lights through a goggle system. In each state Gd-DTPA was injected and the kinetic curves
measured in a slice through the calcarine fissure of the visual cortex. The dynamic curve for each voxel
was analyzed to produce a map of cerebral blood volume (CBV) and image voxels showing a
significant change in CBV were highlighted. The result was that CBV increased on average by 24% in
the visual cortex.
Even though for first-pass dynamic contrast-enhanced measurements Gd-DTPA is known to have an
intravascular distribution, the concentration diminishes over time as it extravasates into other tissues
and is excreted via the kidneys. An intravascular contrast agent, such as a USPIO, on the other hand,
will have a sustained presence in the blood pool over several hours. This allows changes in blood
volume to be monitored during the steady-state distribution phase without the need for separate
injections for each state. Since BOLD MRI is now the preferred method to perform activation studies,
there has been no interest in using these agents in the brain. However, in other organs, use of
blood-pool agents may be very attractive. MS-325,27 though not a blood-pool agent in the strict sense,
does provide an extended window where the concentration in the blood is relatively high. This has
provided an opportunity to monitor changes in blood volume by steady-state imaging.
A study aimed at measuring vascular response to sexual stimulation in female genitalia was recently
reported.103 The conventional method to assess vascular changes during sexual arousal in women is
vaginal plethysmography, which involves the use of an intravaginal probe to measure vascular mucosal
changes. This technique is affected by movement artifact and does not provide any anatomic
information.104,105 For sexual arousal, subjects were asked to view a video tape that was divided into
three segments: the first 21 minutes were neutral material, 15 minutes were erotic material, and the
last 9 minutes were neutral material. Serial MR imaging was performed during the entire 45-minute
video presentation. Prior to the start of the video presentation and the administration of contrast agent,
a data set was obtained using a custom-built phased-array coil.106 A custom-built 10-cm phased-array
coil was placed anteriorly to the pubic symphysis, and a larger two-coil array receiver was positioned
posterior to the pelvis. MS-325 was administered and transverse T1-weighted images with spoiled
gradient-echo sequence were obtained using a three-dimensional (3D) acquisition sequence. Figure
16-11 shows representative data during the neutral and erotic video segments. Changes in regional
blood volume (RBV) of 40 ± 10% in the glans of clitoris was recorded during the erotic video
segments.
More recently similar blood volume measurements in the myocardium during different phases of the
107
cardiac cycle were recorded using an iron-oxide-based contrast agent in a small animal model.
Dynamic Contrast-Enhancement Techniques

Renal Function
page 434
page 435
17 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
Add to lightbox
Figure 16-9 A, Signal intensity variation with FAIR perfusion imaging. A series of 60 pairs of selective
and nonselective inversion-recovery (IR) images were acquired, totaling 120 images. After acquiring
10 pairs of selective and nonselective IR images, the subject took a nitroglycerine pill (0.4 mg
sublingually). Note the vasodilatory effect observed around image #50. The difference between each
pair of selective and nonselective images is proportional to the blood flow. The changes post
nitroglycerine can be appreciated. These data acquisitions allow for quantitative estimation of blood
flow. This example is a preliminary demonstration of the sensitivity of the technique to changes in flow
induced in the peripheral muscle. B, Relative flow images (selective-nonselective) from the data set in
A. The left image is the baseline, the middle is during hyperemia, and the right is post nitroglycerine.
Note the clear evidence for dilatation of the major vessels during hyperemia. Also, there are more
shades of yellow (higher flow) in the tissue in the middle image. These data are presented to
demonstrate the feasibility of applying this technique to the peripheral muscle (known to have much
less blood flow compared to organs such as the lung). C, On a different day, the same subject
underwent a study in which a blood pressure cuff was used to stop blood flow to the right leg at the
level of the knee. The "cuff" was initiated after the acquisition of the fifth image, and deflated after the
20th image. Note the clear decrease in flow during inflation followed by a hyperemic response
immediately post deflation. Also included are data from the left leg which served as a control.
18 of 37 15-04-2010 20:13
Kidneys maintain homeostasis by filtering and excreting metabolic waste products, regulating
acid-base balance, and moderating blood pressure and fluid volume. Because decreasing renal
function accompanies renal disease, monitoring renal function permits assessment of disease
progression and is used to guide patient management and therapy. Many noninvasive tests of renal
function are in use but all have their drawbacks. Serum creatinine levels and creatinine clearance are
insensitive measures of global function and cannot provide information about individual renal function.
CT and intravenous urography (IVU) can provide functional and anatomic information but both use
nephrotoxic contrast agents and expose the patient to radiation. MRI is capable of providing functional
information which when combined with the exquisite anatomic detail allows for comprehensive
examination of the kidneys with minimal risk or discomfort to the patient.
page 435
page 436
Add to lightbox
19 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
page 436
page 437
Figure 16-10 A, BOLD MRI data based on echo planar imaging (EPI) in a human subject. The cuff was
placed on the right leg. The increase in rate of spin dephasing (∆R2*) can be appreciated during the
cuff inflation, and returns to baseline abruptly when the cuff is released. The paradigm was repeated
twice. During the same period the left leg shows no appreciable response. B, Data from the same
subject using the BOLD MRI technique based on a mGRE sequence. Note the change in R2* during
the time the cuff was inflated (images #5-21). Also note the abrupt change back towards the baseline
upon release and the undershoot in response, which implies hyperemia. The magnitude of change in
R2* (~6 Hz) during ischemia is consistent with the data obtained using the EPI technique. An increase
in R2* implies an increase in deoxyhemoglobin content and hence reduced blood PO2. Assuming that
blood PO2 is in a dynamic equilibrium with tissue PO2, the observed changes can be interpreted as
changes in muscle PO2. The magnitude of change in R2* is comparable to published values.100 C,
R2* maps of the right calf muscle showing variations during ischemia. The color maps are of the right
calf muscle that underwent a cuff inflation-deflation paradigm (data not shown). For want of space, only
representative time points are shown. The black-and-white image is the anatomic image of the right
calf in which the different muscles can be differentiated [soleus, lateral (LG), and medial
gastrocnemius (MG)].
20 of 37 15-04-2010 20:13
Add to lightbox
Figure 16-11 Contiguous transverse 2-mm MR image sections (TR/TE = 8.6/1.7 ms) obtained in the
same subject at three different time points. The top images were obtained while the subject viewed
neutral video material; the middle images while the subject was aroused by erotic video material; and
the bottom images while the subject viewed the second neutral video segment after having been
aroused by the erotic video presentation. Note the change in size of the crura of the clitoris (arrows) in
response to each video segment. (Reproduced with permission from Deliganis AV, Maravilla KR,
Heiman JR, et al: Female genitalia: dynamic MR imaging with use of MS-325 initial experiences
evaluating female sexual responses. Radiology 225;791-799, 2002.)
Like inulin and iodine contrast agents, gadolinium chelates, such as Gd-DTPA, have a predominant
renal elimination (~98%) by glomerular filtration without tubular secretion or reabsorption. Therefore,
calculation of glomerular filtration rate (GFR) is feasible using Gd-DTPA with the formula:
21 of 37 15-04-2010 20:13
where V is the urine flow rate (mL/min), U is the substance

concentration in urine, and P is the substance concentration in plasma.
page 437
page 438
With MR, calculation of GFR requires measurement of the concentration of the agent via measurement
of relaxivity changes in urine, blood, or kidney, or via measurement of signal intensity within the
kidneys. Ideally relaxation rate measurements allow for estimation of Gd-DTPA concentrations. This
108
approach necessitates fast T1 mapping, such as the look-locker method or fast acquisition
109
relaxation mapping (FARM). These methods may not provide sufficient organ coverage. Under
certain conditions (e.g., low concentration and heavily T1-weighted sequences could eliminate T2
dependence) signal intensity can be related to Gd-DTPA concentration either empirically by measuring
signal intensity of phantoms with different concentrations, or by converting MR signal intensity to
110
Gd-DTPA concentration that uses the following two relationships :
where T1 and T1' are the observed and precontrast relaxation times of the tissue being studied,
respectively, [Gd] is the concentration of gadolinium, and R is the relaxivity of gadolinium. The second
relationship is an empirical one:
where SI is the observed signal intensity, k is a constant multiplicative factor, and f is a monotonic
relationship between SI and the tissue relaxation time, T1.
Since gadolinium chelates are concentrated in the renal medulla and collecting system, it is important
to use low doses of gadolinium. It was shown that 0.025 mmol/kg yields signal intensity-time curves
110
that are similar to the scintigraphic time-activity curves.
With imaging capability the global GFR measurement could be extended to estimating single kidney
GFR:
where EF refers to the extraction fraction. Using the T1 to [Gd] relationship, this can be rewritten as:
Once EF is determined, GFR can be calculated according to the following:
where RBF is the renal blood flow and Hct is the hematocrit.
111-113
Several groups have published results based on this approach to measuring single kidney GFR.
However, it is not a technique viable for routine clinical use because of the need for quantitative
estimation of concentrations in different compartments and determining RBF in small blood vessels.
Alternately, single-kidney GFR can be estimated from monitoring intrarenal kinetics. Baumann and
114
Rudin proposed a first-order kinetic model of the kidney that consists of two compartments, the
cortex and medulla, and a rate constant between the two representing the rate of clearance of tracer
from the cortex:
where [Gd]m,c is the time varying concentrations of gadolinium in the medulla and cortex, respectively,
and k is the flow rate between the two compartments. Preliminary validation in a small animal model
has been reported.115 An alternate two-compartment model116 and a more expansive
117
multicompartmental model have also been proposed.
QUALITATIVE AND SEMIQUANTITATIVE APPROACHES FOR RENAL FUNCTION USEFUL FOR

ROUTINE CLINICAL USE.
22 of 37 15-04-2010 20:13
The signal intensity-time curves obtained with predominantly T1-weighted sequences can be used in
concert with angiotensin-converting enzyme (ACE) inhibition to differentiate kidneys supplied by
stenotic renal arteries.118,119 This approach parallels ACE-I renal scintigraphy. The advantage with the
MRI approach is that it can be readily combined with anatomic information such as MR angiography.
119 118
While preliminary feasibility has been reported in animal models and in human subjects, larger
trials should allow this technique to enter routine clinical use.
Hypertension is a common occurrence with almost 60 million individuals diagnosed with this condition in
the United States. Of these an estimated 1% to 5% have renovascular disease (RVD) as the
underlying cause.120 As one of the few potentially curable causes of hypertension, RVD remains an
important yet challenging diagnosis. Not all patients with renal artery stenosis (RAS) have RVD; in fact,
those with essential hypertension tend to develop accelerated atherosclerosis, which can lead to RAS.
These diagnostic limitations have generated controversies surrounding treatment. Most anatomic tests,
such as conventional angiography and MR and CT angiography, are limited in their ability to diagnose
RVD because they rely on RAS as the sole criterion. ACE-inhibitor renal scintigraphy is the best
predictor of response to therapy because it is a functional test of renal ischemia. It does not, however,
supply anatomic information needed for therapeutic planning. Decreased renal perfusion pressure in
patients with RAS activates the renin-angiotensin system and increases production of angiotensin II.
Angiotensin II causes vasoconstriction of the efferent glomerular arteriole and restores renal perfusion
pressure and glomerular filtration to normal or near-normal levels. This compensated RAS may not
manifest any perfusion or filtration abnormalities on renal scintigraphy or MR renography. Administering
ACE-I lowers GFR in the setting of RVD because it blocks the production of angiotensin II, which
decreases efferent glomerular arteriolar vasoconstriction and reduces perfusion pressure. Figure
16-12 shows captopril MR renography in a patient with unilateral RAS. While this example illustrates
the efficacy of the method, for practical routine clinical use other factors need to be considered. Since
the objective is to combine such functional information with anatomic depiction of the RAS, the protocol
should include MR angiography (MRA). Since contrast-enhanced MRA is the standard in clinical
practice today, the total dose administered is a key issue. Lee et al121 have implemented a protocol
which utilizes a small dose of contrast for MR renography (2 mL or 0.013 mmol/kg of Gd-DTPA). They
also made use of the higher spatial resolution available with MRI (compared to scintigraphy) to
differentiate the signal intensity-time curves independently for the cortex and medulla. Among patients
with normal serum creatinine levels, ACE-I did unmask decreased GFR by depressing medullary
enhancement in patients with RAS (Fig. 16-13).
There are a number of other applications where dynamic contrast-enhancement kinetics could provide
122-124
valuable information for comprehensive evaluation, such as ureteral obstruction (Fig. 16-14), and
125,126
evaluation of renal transplants.
Tumor Assessment
page 438
page 439
23 of 37 15-04-2010 20:13
Add to lightbox
Figure 16-12 Positive captopril test in a patient with a left renal artery stenosis (top to bottom, left to
right). A, Normal and symmetrical kinetics of the contrast agent before administration of captopril: in
both kidneys the medullary phase appears 1 minute after injection and the excretion appears within the
cavities 3 minutes after injection. B, After captopril , signal dynamics are unchanged in the right
kidney but the medullary phase and excretion are not visible in the left kidney (arrows). While this
approach mimics conventional renal scintigraphy, the superior anatomic detail obtained with MRI can
be appreciated. Further, these techniques could be performed with 3D acquisitions. (Reproduced with
permission from Grenier N, Basseau F, Ries M, et al: Functional MRI of the kidney. Abdom
Imaging 28:164-175, 2003. Copyright 2003 Springer-Verlag.)
The concept of functional MRI outside neuroimaging is most widely used in the assessment of tumors.
Angiogenesis is a complex process critical to the growth and metastasis of malignant tumors. Imaging
applications in the study of angiogenesis is a growing area of research. 127-129 MRI has been shown to
be useful in characterizing microvasculature, and providing information about tumor microvessel
130-134
structure and function. While many different techniques are being pursued, dynamic contrast-
enhanced MRI is the most mature in terms of routine clinical use.
Contrast Agent Kinetics

page 439
page 440
When a bolus of paramagnetic, low-molecular-weight contrast agent (such as Gd-DTPA) passes

through a capillary bed it is transiently confined within the vascular space. This first pass includes
arrival of contrast medium and can last many cardiac cycles. In most tissues, except the brain, testes,
and retina, the contrast agent rapidly passes into the extravascular-extracellular space (EES), also
called the leakage space (ve), at a rate determined by the permeability of the microvessels, their
24 of 37 15-04-2010 20:13
surface area, and blood flow.135-137 In tumors, a variable 12% to 45% of the contrast media can leak
138 trans
into the EES during the first pass. The transfer constant (K ) describes the transendothelial
transport of a low-molecular-weight contrast medium by diffusion. Three major factors determine the
behavior of low-molecular-weight contrast media in tissues: blood perfusion, transport of contrast
agent across vessel walls, and diffusion of contrast medium in the interstitial space.
If the delivery of the contrast medium to a tissue is sufficient (flow-limited situations or where
vascular permeability is greater than inflow), then blood perfusion will be the dominant factor
determining contrast agent kinetics and Ktrans approximates to tissue blood flow per unit
volume.139,140 The latter situation is commonly found in tumors. This is important because
necrotic regions with high intrinsic vessel permeability are associated with low Ktrans caused by
poor blood supply.141,142
If tissue perfusion is sufficient and transport of the gadolinium out of the vasculature does not
deplete intravascular contrast medium concentration (non-flow-limited situations, e.g., in areas
of fibrosis or normal brain tissues), then transport across the vessel wall is the major factor that
trans
determines contrast medium kinetics. In this case K is given by the permeability surface
area product per unit volume of tissue.
Add to lightbox
Add to lightbox
25 of 37 15-04-2010 20:13
Add to lightbox
Figure 16-13 MR renography and angiography in a 66-year-old woman with hypertension and a serum
creatinine level of 0.9 mg/mL (80 mmol/L). A, Coronal volume-rendered contrast-enhanced three-
dimensional MR angiographic image (TR/TE 3.8/1.3 ms; flip angle = 25 degrees) demonstrates
moderate left renal artery stenosis (RAS) (arrow), which was confirmed on source images (not shown).
B and C, Relative signal intensity (SI) curves (standardized relative to baseline measurements) for the
renal cortex and renal medulla in the left kidney before (B) and after intravenous injection of the ACE
inhibitor (C). The degree of medullary enhancement following the injection is diminished, which
suggests physiologically significant stenosis on the left. From the initial vascular peak of 0.75,
medullary enhancement increases to 1.5 before the injection of the ACE inhibitor. After the injection,
medullary enhancement increases only from about 1.75 to 2.2. (Reproduced with permission from
Lee VS, Rusinek H, Johnson G, et al: MR renography with low-dose gadopentetate dimeglumine:
feasibility. Radiology 221:371-379, 2001.)
Over time, the contrast medium will diffuse into tissue compartments further removed from the
vasculature, including areas of necrosis and fibrosis. Over a period typically lasting several minutes to
hours, the contrast agent diffuses back into the vasculature (described by the rate constant kep) from
which it is excreted (usually by the kidneys). The rate of clearance from tissue depends on the capillary
permeability. Tissues when necrosed and fibrosed will show persistent delayed enhancement for this
reason.
Many investigators have modeled the dynamic enhancement data that can be generated by repeated
T1-weighted imaging of tissue after injection of gadolinium-labeled tracers. Tofts et al proposed a
standardized nomenclature and unified definitions so that results could be standardized and compared
140
easily across different centers. Accordingly, the tissue concentration-time curve can be described
by the following equation:
trans
where kep = K /ve.
page 440
page 441
26 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
27 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
28 of 37 15-04-2010 20:13
Add to lightbox
page 441
page 442
Figure 16-14 A-F, Functional MR urography (MRU) images (one slice out of eight acquired) 0, 0.5, 7,
18, 23, and 32 minutes after administration of a bolus injection of 0.06 mmol/kg of Gd-DPTA. There is
no signal from the kidneys precontrast; 30 seconds later a blush is seen in the cortex due to perfusion.
Several minutes later (4-7 minutes) contrast starts to appear in the renal pelvis. The pelvis is full of
contrast in both kidneys at 18 to 23 minutes. At 20 minutes MRU was obtained (G) and then 20 mg of
furosemide was administered intravenously. At 32 minutes (F), washout of contrast can be observed
only in the contralateral kidney, implying significant obstruction of the affected kidney. G, Localized
maximum intensity projection of contrast-enhanced MRU obtained with 3D FISP sequence (TR/TE =
3.5/1.4 ms; flip angle = 250 degrees) 20 minutes following the administration of 0.06 mmol/kg
Gd-DTPA. Arrow points to ureteropelvic junction (UPJ). H, Contrast-enhanced MR angiogram
following administration of additional 30 mL Gd-DTPA over 25 seconds. The sequence was the same
as the one used for MRU. Note the accessory renal artery supplying the lower pole of the right kidney
bisecting the right ureter (arrow). This was later confirmed at surgery. (Reproduced with permission
from Mostafavi MR, Saltzman B, Prasad PV: Magnetic resonance imaging in the evaluation of
ureteropelvic junction obstructed kidney. Urology 50:601-602, discussion 602-603, 1997.)
The transfer constant can be physically interpreted as:
under flow-limited conditions;
under permeability-surface-area- or permeability-limited conditions;
and under mixed conditions, where F is perfusion (mL/g/min), Hct is hematocrit, PS is permeability
surface area product per unit mass of tissue (mL/min/g), ρ is the density of tissue (g/mL), and E is the
initial extraction ratio.
In practice, signal enhancement observed via dynamic acquisition of T1-weighted images can be
assessed in two ways: by the analysis of signal intensity changes (semiquantitative) and/or by
quantifying contrast agent concentration changes using pharmacokinetic modeling techniques. 25 Some
of the parameters used in semiquantitative analysis include onset time (time from injection to the arrival
of contrast medium in the tissue), initial and mean gradient of the upslope of the enhancement curves,
maximum signal intensity, and washout gradient. As the rate of enhancement is important for improving
the specificity of examinations, parameters that include an additional time element are also used [e.g.,
maximum intensity time ratio (MITR)143 and maximum focal enhancement at 1 minute144,145].
Semiquantitative parameters have the advantage of being relatively simple to calculate, but are not
without limitations. They do not accurately reflect tissue contrast medium concentration or the vascular
endpoint of interest (tissue perfusion, blood volume, and capillary permeability). They are also subject
to variabilities in the scanner platforms and settings.
Quantitative techniques fit the measured concentration-time curves to mathematical models such as
29 of 37 15-04-2010 20:13
that described earlier and derive useful quantitative parameters. Such parameters are relatively
complicated to derive compared to the semiquantitative parameters. The assumed model may not be
strictly valid for every tissue type.140,146 Nonetheless, if contrast agent concentration can be measured
accurately, and the type, volume, and method of administration are consistent, then it may be possible
to directly compare pharmacokinetic parameters acquired serially in a given patient and in different
147,148
patient images at the same or different scanning sites.
Analysis of enhancement seen on dynamic T1-weighted images is a valuable diagnostic tool in a

number of clinical situations. The most established role is in lesion characterization where it can
distinguish benign from malignant lesions.144,149 Other demonstrated applications include staging of
cancers,150-152 predicting survival,153 and detecting tumor relapse. Dynamic contrast-enhanced MRI is
also able to predict response or monitor the effects of a variety of treatments. These include
neoadjuvant chemotherapy,154-158 radiotherapy,159-162 vascular embolization of fibroids,163-165 and
antiangiogenic/antivascular treatments.166
Other Functional Imaging Approaches

First-pass tracer kinetics can be used to derive perfusion parameters such as relative blood volume,
relative blood flow, and mean transit time. These are typically performed with T2*-weighted sequences
to improve sensitivity. Relative cerebral blood volume (rCBV) mapping can be used to detect areas of
167,168
increased vascularity in brain gliomas. Areas of high tumor rCBV appear to correlate with tumor
168
grade and vascularity. In low-grade gliomas, homogenous low rCBV is found whereas high-grade
168
tumors display both low and high rCBV components. CBV maps can therefore be used to direct
stereotactic biopsy to areas where high tumor grade may be found.169,170 Relative CBV mapping can
171 168
be performed either using gradient-echo or spin-echo sequences. Each of these methods is
sensitive to a different population of vessels, gradient-echo being more sensitive to vessels of all sizes
while spin-echo is maximally sensitive to capillary-sized vessels.32,172,173 Based on this fact, a
combined gradient- and spin-echo acquisition was proposed which allows for calculation of ∆R2*/∆R2
maps which may serve as a marker of vessel diameter and hence are more strongly associated with
174
tumor grade than spin-echo-derived rCBV maps.
Analysis of tumor angiogenesis using macromolecular contrast agents (MMCM) has been shown to be
feasible in the preclinical setting. Hyperpermeability to macromolecules is considered essential to
angiogenesis because it allows plasma proteins to seep into the tumor interstitium forming matrix for
subsequent ingrowth of new capillaries. Data analyses in animal models have shown good correlation
trans
between MRI-derived assays of tumor K and fractional plasma volume (fPV) with histologic
175,176
microvessel density (MVD). Preclinical studies have also demonstrated that MMCM-enhanced
MRI could identify and measure the effect of an anti-vascular endothelial growth factor (VEGF)
antibody on tumor vascular permeability.177 Currently there are no macromolecular contrast agents
approved for human use. Figure 16-15 shows functional MRI applied to the evaluation of a breast
tumor.
page 442
page 443
30 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
Add to lightbox
31 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
Add to lightbox
Figure 16-15 Multifunctional MRI assessment of breast carcinoma. The series of images were
obtained from a 49-year-old woman with multifocal, invasive ductal breast cancer. All images were
obtained in a time period acceptable to most patients (50 minutes). A, Anatomic, sagittal
post-contrast-enhanced, subtraction MR image of the breast shows multiple foci of irregular
enhancement compatible with multifocal breast cancer. See F for explanation of the arrow. B and C,
Images obtained from a T1-weighted acquisition using 0.1 mmol/kg of contrast medium (Gd-DTPA).
B, This transfer constant pixel map shows the permeability surface area product of the vasculature
displayed as a pixel map on the anatomic image (maximum transfer constant displayed is 1.0
mL/mL/min). The high transfer constant values compared with normal breast and heterogeneous
distribution are typical of cancer. C, Leakage space pixel map at the same slice position (maximum
32 of 37 15-04-2010 20:13
leakage space displayed 100%). D and E, Images obtained from a T2*-weighted acquisition using
0.2 mmol/kg of contrast medium. D, Relative blood volume and E, relative blood flow pixel maps
obtained at a slightly lower spatial resolution, but at the same anatomic location. The tumor foci are
markedly hypervascular. Note the good correspondence between the perfusion parameters (D, blood
volume and E, blood flow) and the permeability parameter maps (B, transfer constant). F, Synthetic
R2* maps calculated from a series of blood-oxygen-level-dependent (BOLD) images obtained at the
same level. The BOLD images themselves are not shown. Lighter areas indicate poorer tissue
oxygenation because of greater levels of deoxyhemoglobin. The largest lesion is better oxygenated
than normal tissues. A disparity between tissue oxygenation (greater deoxyhemoglobin) and blood
flow (high flow) is seen in the lowest lesion on its posterior aspect (arrows, A and F). (Reproduced with
permission from Padhani AR. Functional MRI for anticancer therapy assessment. Eur J Cancer
38:2116-2117, 2002.)
In addition to these approaches, molecular imaging is emerging as a new tool for the sensitive and
specific detection of unique "biochemical signatures" that differentiate and characterize tissues beyond
and before their gross anatomic features become obvious. The MR approach to molecular imaging has
advantages over nuclear178,179 and optical180-182 methods in terms of the combined benefits of high
183
spatial resolution with the capability to simultaneously extract physiologic and anatomic information.
At least two different markers have been identified to detect and/or characterize tumors. First,
αvβ3-integrin is a well recognized biomarker of angiogenesis184 that is relatively selective for activated
endothelial cells while essentially unexpressed on mature, quiescent cells. αvβ3-Integrin-targeted
paramagnetic nanoparticles have been shown to increase the MR detection of the developing
185
neovasculature in tumor models. Second, overexpression of endogenous transferrin receptor (TfR)
has been qualitatively described for various cancers and is presumed to be due to malignant
transformation of cells. Based on this, TfR has been considered as a suitable target for application of
molecular imaging technologies to increase detection of small tumors. Iron-oxide particles targeted to
TfR have been shown to be internalized and accumulated in lysosomal vesicles within cells. The extent
of accumulation, and therefore probe efficiency, has been found to be dependent on the nature of the
chemical cross-linking between transferrin and the iron-oxide particle. Based on these observations,
preliminary measurements of TfR overexpression were shown to be feasible in human breast cancer
tissue.186 These novel targeted probe technologies are in their relative infancy, though preliminary
evidence is promising. For a comprehensive review of molecular imaging see Chapter 15.
Oxygenation
Since oxygen is one of the key nutrients necessary for the viability of cells in any tissue, evaluation of
tissue oxygenation is very important. There are number of direct and indirect methods of evaluating
regional oxygenation using MRI. Again, these depend either on endogenous contrast mechanisms or
on exogenous contrast administrations.
BOLD Technique
page 443
page 444
Blood-oxygenation-level-dependent (BOLD) MRI depends on the fact that deoxygenated hemoglobin is

slightly paramagnetic and hence behaves as an intravascular endogenous contrast agent (see Chapter
9).33,38 In principle, changes in tissue oxygenation can be monitored using this technique by assuming
that the regional blood-oxygenation levels are in dynamic equilibrium with the surrounding tissue. The
BOLD signal changes are influenced both by changes in regional blood flow (perfusion) and regional
oxygenation consumption (or extraction). Hence, in principle, BOLD MRI measurements are not very
specific, i.e., changes in perfusion cannot be differentiated from changes in oxygen extraction.
However, by using specific paradigms (physiologic or pharmacologic) it is possible to interpret changes
observed on BOLD MRI as predominantly a perfusion- or oxygenation-related change. With perfusion,
BOLD is sensitive to either blood-volume or -flow change. The BOLD MRI signals are affected in
opposite ways, i.e., an increase in blood volume results in increased signal decay while an increase in
blood flow results in an increase in signal and vice versa. Since most tissue changes in blood volume
parallel those in blood flow, the opposing effects could potentially compromise the net observed
changes in signal intensity. In neurofunctional MRI, the interpretation is based on the premise that
33 of 37 15-04-2010 20:13
observed signal changes are due to changes in local perfusion with very little change in oxygen
extraction. Blood volume enters the picture only to explain the post-stimulus undershoot.187 While the
term BOLD MRI is often used synonymously with neurofunctional MRI, applications in other parts of
the body are being actively pursued, and a few representative examples will be given.
Intrarenal Oxygenation
It is well recognized that the renal medulla operates at a low oxygenation level (renal medullary
hypoxia).188,189 This makes it highly susceptible to even mild reductions in blood flow. Renal ischemia
accounts for almost 50% of observed cases of acute renal failure. 190 Over the years, based on in vitro
and in vivo animal studies, it has been shown that this type of acute renal failure usually involves
hypoxic injury to renal medullary tubules.191-197 Renal dysfunction may also play a role in the
development of all forms of hypertension in humans and laboratory animals.198 Many experiments have
shown that medullary blood flow (and presumably medullary oxygenation status) is reduced in
hypertension, and more importantly, that reduced medullary blood flow is sufficient to produce
hypertension.199-201 All these studies were performed using invasive microelectrodes or Doppler flow
probes in rat kidneys. The availability of a noninvasive technique to monitor renal medullary blood flow
in humans under normal conditions and during physiologic and pharmacologic stresses may extend the
observations to humans.
40,202,203
BOLD MRI has been used extensively in organs such as the brain. As mentioned earlier, the
BOLD MRI technique exploits the fact that the magnetic properties of oxygenated and deoxygenated
hemoglobin differ. This affects the T2* relaxation time of the neighboring water molecules and in turn
influences the MRI signal on T2*-weighted images. The rate of spin dephasing R2* (= 1/T2*) is closely
related to the tissue content of deoxyhemoglobin. Since the oxygen tension (PO2) of capillary blood is
thought to be in equilibrium with the surrounding tissue, changes estimated by BOLD MRI can be
204,205
interpreted as changes in tissue PO2. A strong correspondence has been demonstrated between
renal BOLD MRI measurements in humans and earlier animal data obtained using invasive
microelectrodes. Figure 16-16 illustrates BOLD MRI of intrarenal oxygenation.
Tumor Oxygenation
BOLD contrast can be used to map changes in blood volume fraction and vascular functionality
associated with angiogenesis.206,207 R2* maps can be obtained using the mGRE technique, which is
exclusively sensitive to tissue oxygenation (see Fig. 16-15F). Vascular function can be evaluated by
206,208
analysis of BOLD contrast changes in response to hyperoxia and hypercapnia. Clinical
application of this technique has revealed high signal enhancement in human tumors in response to
carbogen (5% CO2:95% O2) inhalation.209 The major advantage of this technique compared to
dynamic contrast-enhanced MRI is the fact that it does not involve administration of exogenous
contrast and, hence, can be repeated. The major reservation is the limited contrast-to-noise ratio.
Mesenteric Ischemia
The BOLD signal response of the superior mesenteric vein after a meal may be of clinical value in the
diagnosis of chronic mesenteric ischemia. In a healthy individual, the BOLD signal remains stable or
increases after a meal because the greater metabolic demand is balanced by an increase in arterial
blood flow, which provides sufficient oxygen to the tissues. However, with mesenteric ischemia, the
blood flow response is blunted. The resultant ischemic bowel increases the quantity of
deoxyhemoglobin in the superior mesenteric vein.210-212 Such a study can be combined with
gadolinium-enhanced MRA of the mesenteric vessels213 to obtain a comprehensive evaluation.
Effect of Oxygen on Tissue Relaxation

Retinal Oxygenation
The pathogenesis of many blinding diseases, such as diabetic retinopathy and retinopathy of
34 of 37 15-04-2010 20:13
prematurity, is thought to be related to hypoxia. It is commonly hypothesized that retinal hypoxia leads
to the release of angiogenic factors and eventually to neovascularization. The resulting neovascular
growth progresses from retinal detachment, to vitreous hemorrhage and ultimately blindness. The role
of hypoxia in the development of neovascularization is not yet fully understood. 214 It has been shown
that changes in retinal PO2 (∆PO2) following a shift in breathing room air to hyperoxic inhalation can be
measured using T1-weighted signal intensity on MRI.215,216 While these measurements have been
primarily performed in small animal models, in principle they can be extended to human application.
Ventilation MRI
page 444
page 445
Add to lightbox
Add to lightbox
35 of 37 15-04-2010 20:13
Add to lightbox
Add to lightbox
Figure 16-16 BOLD MRI data obtained with an mGRE sequence in a healthy young subject. A, Panel
of 16 T2*-weighted images with different echo times (TE = 7-40.1 ms) (top to bottom and left to right).
B, Anatomical template (same as number 1 in A). Note the decay of signal intensity is appreciably
higher in the medulla compared to the cortex. C, This is reflected as higher R2* (or lower T2*) on the
R2* map. D, R2*map obtained in the same slice location following administration of 20 mg furosemide
intravenously. Note the much reduced R2* values in the medulla. In fact, the numerical mean R2*
value for the medulla approaches that of the cortex. This is thought to be due to reduced oxygen
consumption post furosemide because it inhibits the reabsorptive transport in the medullary thick
ascending limbs which is associated with energy requirements. These results are consistent with
previous data in rat kidneys using invasive microelectrodes.193 In fact the same study documented a
reduction in medullary blood flow post furosemide , which implies that the change in R2* values
cannot be explained by the blood flow change. It is predominantly determined by reduced oxygen
consumption. This example serves as an illustration of the type of study that this technique affords. A
special note is that these data were acquired on a commercial 3.0-T scanner with body imaging
capability. Our preliminary experience suggests that the R2* values at 3.0 T and significantly higher
204,205
compared to 1.5 T consistent with field dependence of susceptibility weighting. At the same
time, the severity of bulk susceptibility artifacts were no worse than those at 1.5 T.
Assessment of ventilation is crucial in the evaluation of a host of pulmonary disorders since sufficient
ventilation of the lung tissue is a major determinant of efficient gas exchange in the lung. Several
imaging methods can evaluate lung ventilation, such as radionuclide scintigraphy and the recently
developed hyperpolarized gas MRI. Radionuclide scintigraphy encounters limitations such as poor
217 3 129
spatial resolution and the need for a radioactive tracer. Hyperpolarized He and Xe MRI has
provided detailed images of the lung, but the high cost of the laser polarizer, the expense of the 3He
129
and Xe gases, and the need for non-proton-imaging apparatus limit its widespread
applications.218-220
An alternative approach to hyperpolarized gas MRI is oxygen-enhanced ventilation imaging, which

exploits the paramagnetic, T1-shortening property of inhaled oxygen.221,222 Compared to
hyperpolarized gas imaging, oxygen-enhanced imaging offers several advantages: 1. oxygen is readily
available as part of the emergency equipment in most MR suites; 2. oxygen-enhanced imaging does
not impose the burden of the high cost of the laser polarizer unit, noble gases, and the need for
non-proton-imaging apparatus; and most importantly, 3. the oxygen ventilation imaging technique
potentially provides a means to directly study oxygen diffusivity from the air space to the pulmonary
vasculature because oxygen is involved in the functional gas exchange.
page 445
36 of 37 15-04-2010 20:13
page 446
The weakly paramagnetic property of molecular oxygen caused by the presence of two unpaired
electrons is the underlying principle of oxygen-enhanced ventilation imaging. Each electron has a
magnetic moment that is 1000 times that of a nucleus, and the resulting fluctuating magnetic field can
produce a greater dipole-dipole interaction than that of the neighboring nuclei, thus causing a faster
223
rate of spin-lattice relaxation (R1). Chiarotti et al first reported that an increase in dissolved oxygen
in water shortens its T1 (1/R1). Young et al224 extended Chiarotti et al's study and reported a
shortening of T1 and an increase in signal intensity of blood in the left ventricle after volunteers inhaled
100% oxygen in their investigation of the potential use of oxygen as a paramagnetic contrast agent.
221
Expanding on these results, Edelman et al first proposed the use of oxygen for ventilation imaging of
the lung. Although oxygen is weakly paramagnetic, its overall effect on the lung is considerable given
the large surface area of the lung and the large difference in partial pressures between room air and
100% oxygen. These two factors facilitate an environment that allows more oxygen to diffuse across
the parenchyma and dissolve into blood.
Oxygen-enhanced ventilation images typically involve acquisition of images during breathing of normal
room air and 100% oxygen, then subtracting the former from the latter. To minimize susceptibility
effects lung imaging is predominantly performed with fast (or turbo) spin-echo sequences. Signal
difference in the lung parenchyma between inhalation of room air (21% oxygen) and 100% oxygen is
222,225,226
considerable, in the order of approximately 10% to 20%. The observed changes in signal
result mainly from the effect of the increased concentration of oxygen dissolved in the blood. This
increase in MR signal intensity, parallel to a decrease in paramagnetic deoxyhemoglobin, has been
referred to as the BOLD effect.31,38 The potential signal changes due to BOLD should be minimal in
oxygen-enhanced ventilation imaging because the fast spin-echo sequence with short echo-spacing
time is quite insensitive to the susceptibility effect (T2*).
The described oxygen-enhanced ventilation imaging technique offers a potential tool in the diagnosis of
pulmonary diseases. It may be directly used to reliably evaluate healthy and diseased lung tissues in
airway obstructive pulmonary pathologies such as emphysema and cystic fibrosis. It may also be used
in conjunction with Gd-DTPA or ASL perfusion imaging to constitute a MR ventilation-perfusion (V/Q)
scan to assess pulmonary diseases such as acute pulmonary embolism (Fig. 16-17).227-232
© 2010 Elsevier
37 of 37 15-04-2010 20:13
EVALUATION OF MECHANICAL FUNCTION

MRI of Myocardial Function
Cardiac Cine MRI
Add to lightbox
page 446
page 447
Figure 16-17 Oxygen-enhanced ventilation, FAIR perfusion, V/Q ratio of the signal intensity, and
contrast-enhanced Gd-DTPA perfusion images from the same patient as in Figure 16.3. Ventilation
and perfusion mismatch can be observed in the apical region of the right lung. The perfusion defect in
the apical region of the right lung in Gd-DTPA confirms the FAIR findings. An inversion recovery
half-Fourier single-shot fast spin-echo sequence was used with an effective TE = 20 ms; echo spacing
= 3.6 ms; bandwidth = 250 kHz; slice thickness = 15 mm; field of view = 450 mm; and 256 × 128
matrix. For the gadolinium-enhanced perfusion study, the sequence parameters are: a fast GRE
sequence was used with a TR/TE = 2.4/0.6 ms; bandwidth = 62.5 kHz; field of view = 420 mm; slice
thickness = 12 mm; 128 × 96 matrix. (Courtesy of Vu Mai, Evanston Northwestern Healthcare,
Evanston, IL.)
Because the prime function of the myocardium is to pump blood efficiently, evaluation of wall motion or
ejection fraction have been used as a diagnostic indicator for the heart. While this has been and
continues to be primarily evaluated using ultrasound, cardiac cine MRI is now considered the standard
of reference for measurements of ventricular volume and various parameters of ventricular function,
such as ejection fraction and ventricular mass.233 Ventricular volume and ejection fraction, however,
may not correlate with myocardial contractility because they are very sensitive to loading conditions.234
Interest in imaging of regional cardiac function has been driven by the fact that ischemic heart disease
is the primary cause of death from cardiovascular disease worldwide.235 In ischemic heart disease,
even a slight oxygen supply-demand imbalance that occurs as a result of coronary artery narrowing
leads to contractile dysfunction of the specific territory involved.236 Regional ventricular function can be
measured with the use of cine MRI and MR tagging techniques, though the image data analysis is time
consuming for routine clinical use. There is a definite need for objective, reproducible assessment of
regional myocardial contractile function.237 Novel data acquisition and analysis methods may allow the
realization of this clinical need in the near future. Here only a brief description of cardiac MRI and its
application in the clinic is provided for completeness as these subjects are discussed in more detail in
other chapters. MRI myocardial function assessment is the most mature of all the methods to evaluate
mechanical function. Some of the methods used in other organs, such as the lung, are direct
extensions from the work in the heart.
Since the last edition of this book, a major advance in cine MRI is the steady-state free-precession
(SSFP) pulse sequence [variants of which are known by the commercial names FIESTA (fast imaging
employing steady-state acquisition), TrueFISP (fast imaging with steady-state precession), and
balanced FFE (fast field echo)]. The use of SSFP sequences in cine MRI has resulted in a substantial
improvement in image quality compared to conventional segmented k-space fast gradient-echo
acquisitions.238-240 With the use of SSFP sequences, acquisitions are less dependent on the inflow of
fresh spins than they are with fast GRE sequences, and signal intensity is related to inherent
1 of 8 15-04-2010 20:15
properties of the tissue (e.g., T2/T1 ratio). The combination of this technique with the multishot
echo-planar approach results in substantial shortening of the acquisition times, resulting in full left
ventricular coverage in less than 2 minutes.241 By reducing the number of k-space sampling in the slice-
select direction, a 3D SSFP-based cine MRI technique has been shown to be feasible. 242 This will
allow for 4D cine MRI (involving three spatial dimensions and a temporal dimension). All these methods
can, of course, benefit from the new developments in partially parallel imaging strategies (see Chapter
8).
Cine MRI pulse sequences allow direct qualitative and semiquantitative assessment of myocardial
regional function. Images acquired with SSFP sequences are substantially better compared to
conventional GRE sequences for visual and semiautomated endocardial contour detection. 238,239
Interstudy variability with SSFP is excellent and interobserver variability is better than with GRE-based
238,239,243
cine techniques. However, in healthy volunteers and patients with heart disease, it has been
demonstrated that analysis by SSFP pulse sequences results in higher end-diastolic and -systolic left
and right ventricular volumes when compared to cine GRE techniques. 239,240,244 This is mostly due to
the improved endocardial border depiction with SSFP techniques.
Add to lightbox
page 447
page 448
Figure 16-18 Scheme of coordinate systems for measuring myocardial strain defined by the finite
strain tensor E. Normal strains (black arrows) are defined in relation to the circumferential, or
short-axis, plane: circumferential shortening (ECC) occurs parallel to the tangent of the myocardium
with respect to the epicardial surface (shown here as the endocardial surface for space reasons);
radial thickening (ERR) occurs perpendicular to the circumferential direction, toward the ventricular
centroid; and longitudinal shorting (ELL) occurs perpendicular to the other two components and
parallel to the longitudinal axis of the left ventricle. Principal strains (white arrows) are defined in
relation to the direction of movement in the main myocyte fiber bundles during systolic deformation.
The maximal principal strain is the greatest elongation (E1) orthogonal to the fiber direction. The
minimal principal strain is the greatest shortening (E 2) parallel to the fiber direction. Principal strains
are referred to the major and minor axes of an ellipse resulting from the deformation of a circle during
systole because of wall shear. Strain that occurs perpendicular to these two principal strains is labeled
E3. The angles between ECC and E2 are defined as α and β, respectively. (Reproduced from Castillo
E, Lima JA, Bluemke DA. Regional myocardial function: advances in MR imaging and analysis.
RadioGraphics 23(Spec No): S127-140, 2003.)
Add to lightbox
2 of 8 15-04-2010 20:15
Figure 16-19 Representative end-systolic time frames of FastHARP acquisition with synthetic tags
and a pseudocolor overlay of ECC are shown in one animal. No change in strain is shown as green,
decreased ECC is shown as red (i.e., decreased circumferential shortening or stretching), and
increased ECC is shown as blue (i.e., increased circumferential shortening). A, Before coronary artery
occlusion, uniform strain is present. After coronary artery occlusion (B), the anterior wall shows
decreased ECC (i.e., red areas) at three heart beats after occlusion (C). D and E, The area of
dysfunction increases with time, concurrent with increased shortening in the opposing wall. The pattern
is visible with free breathing (E). LV, left ventricle. (Reproduced from Kraitchman DL, Sampath S,
Castillo E, et al. Quantitative ischemia detection during cardiac magnetic resonance stress testing
by use of FastHARP. Circulation 107:2025-2030, 2003.)
Since the contraction of the left ventricle involves not just radial but also circumferential and longitudinal
directions, wall motion analysis based on cine MRI is not sufficient to depict the true extent of the
changes in contractility. Myocardial strain imaging is believed to provide a more accurate assessment
of myocardial contraction.245 This involves assessment of local tissue deformation as an indicator of
myocardial contractile function. Strain measurements are expressed as the fractional change in
values.246 Strain is a tensor and is characterized not only by the magnitude of the length change but
also by the direction of the change. Two frames of reference are used in calculating the strain tensor:
one is related to a local cardiac coordinate system, and the other to the local fiber orientation.
Changes during the cardiac cycle in any of the three spatial dimensions relative to the circumferential
plane are considered normal strains (Fig. 16-18). During systole, circumferential shortening occurs in
the short-axis plane along the curved lines parallel to the epicardial surface and is directed clockwise
as viewed from the base. Wall thickening is an indicator of strain that occurs radially, proceeding in a
direction from the long axis toward the epicardium. Base-to-apex shortening occurs longitudinally,
parallel to the long axis. Strain changes that occur in a plane between two of these three initially
orthogonal normal directions are called shear strains. The greatest systolic compression and
expansion in the left ventricular wall occurs in the planes parallel and perpendicular to the direction of
the principal local myocardial fibers.247 Changes in these dimensions, referred to as principal strains,
represent the greatest shortening and the greatest elongation, respectively. Another parameter that is
measured during the assessment of myocardial function with MR tagging is left ventricular systolic
torsion, or twist, and the subsequent diastolic "untwist."246 Torsion involves motion between short-axis
planes that occurs simultaneously with differential rotation around the long axis; i.e., it involves motion
between the heart base (clockwise rotation) and the apex (counterclockwise rotation).
Myocardial Strain MRI

Myocardial MR tagging is performed with cine MRI by applying a special radiofrequency prepulse
immediately following detection of the R wave on the ECG tracing. 248 The prepulse is oriented
perpendicular to the imaging plane and induces a local saturation that is depicted on images as a dark
line superimposed on myocardial tissue. The multiple lines produced by successive applications of the
prepulse are observed as parallel stripes or as a grid on cine MR images of the cardiac cycle. MR
tagging is performed by using segmented k-space GRE pulse sequences with spatial modulation of
magnetization (SPAMM) or delays alternating with nutations for transient excitation (DANTE).249,250 An
251
extension of the SPAMM principle is complementary SPAMM (CSPAMM). This involves two
acquisitions which are subtracted from one another, and allows for tagging both in the systolic and
diastolic phases. Many variations in implementations have been proposed.252
The major hurdle in myocardial strain imaging is the analysis. While highly precise techniques have
been developed,253 until recently the analysis was time consuming. A new technique has been
proposed that overcomes this limitation. This is known as harmonic phase (HARP) analysis254 and it
involves less user intervention and a faster processing time. It enables the tracking of phase changes
from one of the off-center spectral peaks in the Fourier domain. The phase change is related to the
inplane motion of the myocardial tags. If there is no motion, the phase of the sinusoidal tag pattern
remains linear. If there is motion, the sinusoidal tag pattern deviates from linearity in its phase. The
HARP method enables measurement of these deviations to track motion and to compute 2D
3 of 8 15-04-2010 20:15
myocardial strain parameters.
HARP principles have been implemented in the MR pulse sequence, allowing for realtime depiction of
myocardial strain.254 Regional strain changes caused by acute coronary ischemia can be detected
significantly earlier compared to nontagged cine MRI (Fig. 16-19).255
Combined Myocardial Fiber Architecture and Strain Mapping

Wedeen256-260 has pioneered the use of diffusion MRI to map myocardial fiber orientations. Recently
Wedeen et al259 combined diffusion measurements with strain mapping methods to allow for evaluation
of fiber shortening. Strain mapping was performed based on velocity-sensitive imaging. Such a
combination allows for strain mapping in fiber coordinates. This in turn allows for evaluation of
transmural patterns of fiber and cross-fiber shortening. In healthy heart it was shown that fiber
shortening was constant transmurally, while cross-fiber shortening showed a distinct steep gradient. It
is believed that deviations from this norm may allow for detecting diseased hearts. Fiber shortening is
an indicator of myocardial contractility while cross-fiber shortening reflects a dynamic rearrangement of
the myocardial sheet structure to facilitate systolic wall thickening.259 Different cardiac states are
believed to have different effects on these two functions.261
Tagged MRI of Lungs

Strain mapping based on the grid tag pattern was recently shown to be feasible in the lung.262 It is
believed that this type of information may be important in the evaluation of lung compliance or
elasticity, which may change significantly with disease processes such as fibrosis.
MR Elastography
page 448
page 449
Add to lightbox
Figure 16-20 Schematic diagram of MR elastography system. Conventional MRI gradients and
radiofrequency pulses that encode spatial positions are shown at the bottom left. The
electromechanical driver applies transverse acoustic waves to the object to be imaged via a surface
plate (right). The cyclic motion-sensitizing gradients and the acoustic drive are synchronized using
trigger pulses provided by the imager. The phase offset (θ) between the two can be varied. As shown
by the shaded regions, the motion-sensitizing gradients can be superimposed along any desired axis
to detect cyclic motion. (Reprinted from Medical Image Analysis, Vol. 5, Manduca A, Oliphant TE,
Dresner MA, et al, Magnetic resonance elastography: non-invasive mapping of tissue elasticity,
pages 237-254, Copyright 2001, with permission from Elsevier.)
Palpation is still widely used in routine clinical practice because the viscoelastic properties are affected
4 of 8 15-04-2010 20:15
by a variety of disease processes. It is quite common in surgical practice to detect tumors that were
not diagnosed by conventional imaging modalities, including CT, MRI, and ultrasound. The elastic
moduli of various human soft-tissues are known to vary over a wide range (more than four orders of
magnitude). In contrast, most of the physical properties depicted by conventional medical imaging
modalities are distributed over a much smaller numerical range.
Classical measurements of the elastic modulus involve application of a known stress and measurement
of the resulting strain. Extensions of this method involve the use of multiple measurements with varying
stress and/or the application of dynamic rather than static stress.
MR elastrography (MRE) allows direct visualization and quantitative measurement of propagating

263
acoustic strain waves generated by harmonic mechanical excitation. Shear waves at frequencies in
the range of 10 to 1000 Hz are used as a probe because their attenuation is reduced compared to
higher frequencies, their wavelength in tissue is in the useful range of millimeters to tens of millimeters,
and shear modulus varies widely in tissue. A phase-contrast MRI technique can be used to spatially
map and measure the shear-wave-displacement patterns. From such data, the local quantitative shear
modulus can be calculated and elastograms that depict tissue elasticity or stiffness can be generated.
Figure 16-20 illustrates the MRE system. Trigger pulses synchronize an oscillator/amplifier unit that
drives an electromechanical actuator coupled to the surface of the object to be imaged, inducing shear
waves in the object at the same frequency as the motion-sensitizing gradient. Any cyclic motion of the
spins in the presence of these motion-sensitizing gradients causes measurable phase shift in the
received MR signal. From the measured phase shift, it is possible to calculate the displacement at
each voxel, and directly image the acoustic waves within the object.
The phase shift caused by a propagating mechanical wave with a wave constant k within a medium at
263,264
a given frequency (1/T) in the presence of a cyclic motion-encoding gradient is given by :
This accumulated phase shift is proportional to the dot product of the displacement amplitude vector ξ0
and the motion-sensitizing magnetic gradient vector G0, and the relative phase θ of the mechanical and
magnetic oscillations. Particles whose component of motion along the gradient vector are exactly in
phase or out of phase with the magnetic oscillation have maximum phase shifts of opposing polarities.
Particles whose component of motion along the gradient vector is 90 degrees out of phase with it have
no net phase shift. Since the response is also proportional to the number of gradient cycles (N) and the
period of the gradient waveform (T), extreme sensitivity to small amplitude synchronous motion can be
achieved by accumulating phase shifts over multiple cycles of mechanical excitation and the motion-
sensitizing gradient waveform. The quantity γ is the gyromagnetic ratio and vector r relates to the spin
position.
The resulting images reflect the displacement of spins due to acoustic strain wave propagation in the
medium and are termed "wave images." An example of such an image obtained in a tissue-mimicking
gel is shown in Figure 16-21A. The local wavelength of the propagating waves depends on the
elasticity of the material at each location in the object, allowing the construction of an elastogram (Fig.
16-21B).
Figure 16-22 shows an example of the in vivo application of MRE. Tumors within the breast are 5 to 20
times stiffer than normal tissue and can be highlighted on an MRE.
dGEMRIC
page 449
page 450
5 of 8 15-04-2010 20:15
Add to lightbox
Add to lightbox
Figure 16-21 Object with diameter comparable to wavelength. A, Shear waves propagating in a
phantom with an embedded 1.5 cm diameter cylinder of stiffer gel. Shear waves at 300 Hz were
applied at the top margin of the gel block, with transverse motion oriented orthogonal to the plane of
the image. B, Elastogram based on local frequency estimation (LFE) processing clearly depicts the
object, even though it is relatively small in comparison to the wavelength. (Reprinted from Medical
Image Analysis, Vol. 5, Manduca A, Oliphant TE, Dresner MA, et al, Magnetic resonance
elastography: non-invasive mapping of tissue elasticity, pages 237-254, Copyright 2001, with
permission from Elsevier.)
Articular cartilage disease is generally monitored by the loss of tissue substance or mechanical
properties as seen indirectly by joint-space narrowing on radiographs or by directly probing the surface
via arthroscopy. With the numerous techniques in use and on the horizon to repair cartilage injury or
reverse disease, a means of evaluating the state of the cartilage matrix itself becomes important. MRI
has started to play a critical role in the evaluation of cartilage morphologic abnormalities and is widely
used in the evaluation of joint disorders. While these are powerful improvements in the capabilities for
visualizing articular cartilage, MRI also has the potential to evaluate specific information about the
molecular status of the cartilage. Availability of such molecular insight raises the possibility of indirectly
studying the mechanical function of the tissue.
6 of 8 15-04-2010 20:15
Add to lightbox
Add to lightbox
Figure 16-22 A 43-year-old woman with invasive ductal breast carcinoma in the lateral breast. A,
T1-weighted MR image of the breast shows irregular tumor mass (arrow), with thickening of overlying
cutaneous tissue. B, MR elastogram shows focal area of high shear stiffness corresponding to the
location of the known tumor mass. (Reproduced from McKnight AL, Kugel JL, Rossman PJ, et al.
MR elastography of breast cancer: preliminary results. Am J Roentgenol 178:1411-1417, 2002.
Reprinted with permission from the American Journal of Roentgenology.)
The main macromolecular constituents of cartilage are glycosaminoglycan (GAG) and collagen. From
the perspective of GAG, the concentration of the charged moieties of the GAG molecule [the fixed
charge density (FCD)] confers much of the biomechanical stiffness of the tissue. In terms of collagen,
the molecular concentration, molecular structure, and molecular architecture are all contributory factors
to the mechanical properties of the tissue. MRI techniques to image both the GAG and collagen
components of cartilage are in active development. The most direct method of imaging the
concentration of the GAG molecule is through imaging the FCD of the tissue. The basic approach is
based on the work of Maroudas et al265,266 on the theoretical description of how charged ions will
distribute in cartilage in relation to the FCD. Early studies utilized MRI of the charged sodium ion, 267
and in vivo imaging of sodium by MRI was shown to be feasible.268 However, due to the challenges
inherent in quantitating the sodium concentration by MRI in degenerated cartilage269 and the limited
availability of sodium MRI, a more recent approach utilizes the ionic MRI contrast agent Gd-DTPA
(Magnevist, Berlex, NJ), which allows for the much more readily available proton MRI. Collagen has
been investigated through a number of MRI techniques, but mainly using T2 measurements. 270-272 A
combination of the GAG and collagen techniques should prove to be the most useful for a full
evaluation of cartilage integrity and status.
GAGs in cartilage have abundant negatively-charged carboxyl and sulfate groups. Therefore, if mobile
ions are allowed time to distribute in cartilage, they will distribute in relation to the negative fixed
charge density of the cartilage, or effectively in relation to the GAG concentration. Gd-DTPA has the
important characteristic of a negative charge. If Gd-DTPA is allowed to penetrate into cartilage, it will
7 of 8 15-04-2010 20:15
be distributed in higher concentration in areas of cartilage in which the GAG content is relatively low
and be excluded from regions rich in GAG. In fact, the GAG concentration at a particular spatial
location can be quantified by determining the Gd-DTPA concentration at that same location by
measuring regional T1.273,274 This technique is referred to as delayed gadolinium-enhanced MRI of
cartilage (dGEMRIC) the "delay" referring to the time required to allow Gd-DTPA to penetrate the
cartilage tissue. Figure 16-23 shows examples of dGEMRIC images in both symptomatic and
nonsymptomatic individuals. Figure 16-24 illustrates the sensitivity of the technique to monitor changes
over time. This will allow for better understanding of the disease progression and efficacy of
therapeutic interventions.
© 2010 Elsevier
8 of 8 15-04-2010 20:15
OTHER RECENT DEVELOPMENTS

page 450
page 451
Add to lightbox
Figure 16-23 Color scale of T1Gd index used in the delayed gadolinium-enhanced MRI of cartilage
(dGEMRIC) technique. The blue/green end represents high T1Gd and hence high glycosaminoglycan
(GAG) levels. The red end represents low T1Gd values and hence low GAG levels. Global and
functional range of the T1Gd index: A, Sagittal T1Gd image of lateral compartment of a 26-year-old
female professional dancer shows high-range values: mean 547 ± 121 ms for tibial plateau and
weight-bearing zones of the femoral condyle compartments (arrows); B, Coronal T1Gd image of a
28-year-old female asymptomatic volunteer shows T1Gd indices in the mid-range: mean 448 ± 89 ms
across all compartments; C, Medial sagittal image of a 78-year-old woman with moderately severe
medial osteoarthritis shows low-range values in the medial femoral condyle: mean 285 ± 74 ms. Note
that the medial femoral condyle has a focal area (arrows) of low T1Gd index: 243 ± 56 ms, with
surrounding cartilage of 312 ± 75 ms. (Reproduced from Williams A, Gillis A, McKenzie C, et al:
Glycosaminoglycan distribution in cartilage as determined by delayed gadolinium-enhanced MRI
of cartilage (dGEMRIC): potential clinical applications. Am J Roentgenol 182:167-172, 2004.
There are a number of other developments may potentially open up new methods to probe tissue
characterization and function by MRI. These include novel contrast agents that allow for evaluation of
275
regional pH, contrast mechanisms such as chemical exchange dependent saturation transfer
276
(CEST), and hyperpolarized 13C277 which may give rise to novel applications both for imaging and
spectroscopic applications.
1 of 16 15-04-2010 20:17
Add to lightbox
Add to lightbox
Add to lightbox
Figure 16-24 Medial sagittal T1Gd images of a woman recovering from arthroscopic knee surgery.
Images A, before beginning supplements and B, after 6 months of glucosamine and chondroitin
sulfate nutritional supplements (CosaminDS, Nutramax Laboratories, Edgewood, MD) show a 19%
increase in T1Gd. C, Summary plot of 10 volunteers followed for 6 months while taking glucosamine
and chondroitin sulfate supplements. Stars denote significant (P < 0.05) changes over a 6-month
period (two increase, one decreases). Four volunteers who are professional dancers taking
supplements showed no change after 6 months. Increases of 13% and 19% were seen in two
volunteers taking supplements while recovering from arthroscopic surgery. Three other volunteers
recovering from arthroscopic surgery did not change significantly. One volunteer with a history of knee
injuries (patellar fracture and menisectomy) decreased 9%, and had reported that he had stopped
running while taking supplements. (Reproduced from Williams A, Gillis A, McKenzie, C, et al:
Glycosaminoglycan distribution in cartilage as determined by delayed gadolinium-enhanced MRI
of cartilage (dGEMRIC): potential clinical applications. Am J Roentgenol 182:167-172, 2004.
The pH of bodily fluids affects the organism in many ways, especially through its effects on cellular and
plasma proteins. Maintenance of acid-base homeostasis is critical, and occurs at several levels. The
most immediate and local response to an acid or alkali load is through chemical processes, including
intracellular, extracellular, and bone buffers. Certain pathologies are associated with perturbed pH
homeostasis. For example, tumor interstitial fluid has a reduced buffering capacity compared to normal
2 of 16 15-04-2010 20:17
tissue and this, in combination with poor perfusion and increased lactic acid secretion by tumors,278 is
279
believed to result in the acidic extracellular pH of tumors. The acidic pH of tumors has many
consequences, including resistance to radio- and chemo-therapies. 280-284 Pathologically-altered renal
physiology can also manifest with perturbations in both systemic and renal pH. Methodologies to image
the spatial distribution of tissue pH would have considerable biomedical and clinical relevance by
enabling noninvasive assessment of disease extent, progression, and response to therapy.
page 451
page 452
Several approaches have been proposed to measure tissue pH by MR. Some exploit the difference in
31 1 278,285,286
P and H resonances. However, these are limited in terms of imaging applications. Certain
gadolinium complexes (e.g., Gd-DOTA-4AMP; Macrocyclics, Dallas, TX) offer the possibility of
imaging pH with a spatial resolution comparable to standard MRI. 278,285,286 While proof-of-concept has
275
been demonstrated recently, the technique is in its infancy and several practical issues need to be
addressed.
276
Another proposed method to probe tissue pH is using CEST agents. CEST is similar to
magnetization transfer contrast (MTC) which involves exchange between macromolecules and
surrounding water protons. In CEST the exchange is between either an endogenous metabolite or
exogenous agent and the surrounding water. The fundamental advantage of this approach over the pH
measurements based on chemical shift is that the CEST measurement is made with the amplitude of
the water signal resulting in several hundred fold increase in signal to noise ratio. This potentially could
then allow for dynamic studies or eventually image distribution of pH in tissue. More recently
paramagnetic lanthanide complexes have been shown to act as pH sensitive CEST contrast agents for
MRI.287,288 Similarly, these paramagnetic CEST (PARACEST) agents have been shown to be sensitive
289
to presence of lactate and are believed to be able to function as metabolite specific MR contrast
agents.290
The potential advantage of a CEST agents compared to conventional MRI contrast agents is the ability
to turn the contrast effect OFF and ON by the use of pre-irradiation. There are certain other key
differences when compared to commonly used T1 contrast agents. CEST results in loss of signal, but
involves no T2* effects. Since CEST agents need not involve metals, they can potentially be less toxic
and hence may be introduced into cells. Interestingly, a recent report indicated that iopamidol , one of
the most commonly used contrast agents for x-ray computed tomography may be used as a potential
MRI contrast agent.291 While the T1 properties are not useful, the T2 and CEST properties may be
useful especially in combination for certain specific clinical applications. All these studies to-date have
been only performed in in vitro setting and several issues will need to be taken into consideration for
translating into in vivo studies. These will include dose, RF deposition, biocompatibility, and toxicity
issues.
Add to lightbox
Figure 16-25 Parametric map of renal blood flow obtained after venous administration of
13
hyperpolarized C-labeled compound in a rabbit. A trueFISP sequence with a 90° preparation pulse
followed by a train of 180° pulses was employed. Based on these data the cortical blood flow was
3 of 16 15-04-2010 20:17
estimated to be 5.8 ml/min/ml of tissue. (Reproduced with permission from Johansson E, Olsson LE,
Mansson S, et al: Perfusion assessment with bolus differentiation: a technique applicable to
hyperpolarized tracers. Magn Reson Med 52:1043-1051, 2004.)
Hyperpolarized noble gases (3He and 129Xe) are being successfully used for MRI of the lung
airways.292-294 However, the low solubility of these gases in blood makes these agents unsuitable for
most other applications. Recently, it has been shown to be feasible to hyperpolarize 13C as a part of a
water-soluble molecule. Two polarization techniques, parahydrogen-induced polarization (PHIP)295 and
dynamic nuclear polarization (DNP),277 have been shown to be useful in producing hyperpolarized 13C.
High polarizations on the order of 15% are possible with a concentration of 200 mM in water. These
properties make it very promising for MRA and such feasibility was recently demonstrated. 296
Preliminary demonstration of cerebral perfusion has also been reported.297 The major advantage of
using hyperpolarized 13C for perfusion imaging is that the signal is directly related to the concentration
of the agent, which is an important issue with currently available dynamic-susceptibility-based contrast
approaches. However, other issues arise with the use of hyperpolarization. Higher concentrations and
longer relaxation times of the agent may be necessary for cerebral blood flow (CBF)
measurements.297 Also, the depolarization rate has to be measured to allow for quantitative cerebral
blood volume and mean transit time (MTT) estimates. In other tissues such as heart, kidney and lung,
with volume of distribution significantly higher compared to the brain, perfusion imaging will be more
robust. Preliminary experience in evaluating renal perfusion was recently reported.298 Figure 16-25
shows an example of perfusion image in rabbit kidney. This specific study also exploited the application
of RF pulses to fully depolarize the tracer and hence render the technique insensitive to arterial
dispersion.
page 452
page 453
The concept can be further developed to produce hyperpolarized 13C-containing endogenous

13
substances. Preliminary evidence in which C-urea polarized by the DNP method was used to obtain
angiograms supports this approach.299 This could potentially lead to functional/molecular imaging using
299 13
endogenous substances, because MRI of C-labeled glucose was shown to be feasible many
years ago.300 Hyperpolarization could make these techniques more efficacious for routine applications.
Such an approach will combine the superior spatial resolution of MRI and the specificity of techniques
like PET. The major limitation of this technology is that the polarization is relatively short lived, on the
13
order of a minute. The T1 of the C-labeled compound depends greatly on the molecular structure of
the particular compound and is a major determinant of which labeled compounds might have eventual
clinical utility. Also, the hyperpolarizer must be kept in close proximity to the MR scanner room and the
13
C-labeled agent must be rapidly infused into the subject once removed from the polarizer, since the
polarization cannot be maintained in the absence of an external magnetic field and because of the
relatively short T1 of the labeled agent. In this regard the agent is very different from hyperpolarized
noble gases, which can be kept in a polarized state for extended periods of time.
Since NMR spectroscopic methods allow distinguishing 13C nuclei based on their molecular structure, it
is possible to follow metabolic changes. Chemical shift imaging has been used to localize metabolites
of 13C-labeled substances (e.g., glucose , alanine, etc.) within the brain.301 Without hyperpolarization,
these measurements take long time (on the order of minutes). However, with hyperpolarization these
times could be reduced to order of seconds.
It is important to note few fundamentally different properties of hyperpolarized agents that have
significant influence on the imaging protocols: 1. Hyperpolarized nuclei generate signal themselves
rather than just influence the signal from surrounding protons. This results in a simple and linear
relationship between signal intensity and concentration; 2. Owing to the very low natural abundance of
13
C in tissue, hyperpolarized 13C MRI completely lacks any background signal; 3. Unlike the
conventional proton MRI, the longitudinal magnetization consumed during the imaging process cannot
4 of 16 15-04-2010 20:17
be recovered in hyperpolarized MRI. This restricts the window for imaging to few minutes following
administration of the agent. In addition to the natural loss of polarization by T1 decay, application of RF
pulses also result in permanent loss of polarization. Minimizing the number of RF pulses (e.g., single
shot techniques), and the flip angles is crucial for this application; and 4. Since the gyromagnetic ratio
of 13C is four times lower than that for 1H, correspondingly stronger gradients are necessary to
achieve equal spatial resolution within equal time.
REFERENCES
1. Lloyd D: Functional MRI and the study of human consciousness. J Cogn Neurosci 14:818-831, 2002. Medline
Similar articles
2. Stiles J, Moses P, Passarotti A, et al: Exploring developmental change in the neural bases of higher cognitive functions: the
promise of functional magnetic resonance imaging. Dev Neuropsychol 24:641-668, 2003. Medline Similar articles
3. Cho ZH, Chung SC, Jones JP, et al: New findings of the correlation between acupoints and corresponding brain cortices
using functional MRI. Proc Natl Acad Sci USA 95:2670-2673, 1998. Medline Similar articles
4. Cho ZH, Oleson TD, Alimi D, Niemtzow RC: Acupuncture: the search for biologic evidence with functional magnetic
resonance imaging and positron emission tomography techniques. J Altern Complement Med 8:399-401, 2002. Medline
Similar articles
5. Holden C: Polygraph screening: Panel seeks truth in lie detector debate. Science 291:967, 2001. Medline
Similar articles
6. Langleben DD, Schroeder L, Maldjian JA, et al: Brain activity during simulated deception: An event-related functional
magnetic resonance study. Neuroimage 15:727-732, 2002. Medline Similar articles
7. Lo YL, Fook-Chong S, Tan EK: Increased cortical excitability in human deception. Neuroreport 14:1021-1024, 2003.
8. Edvinsson L, MacKenzie ET, McCulloch J: Cerebral Blood Flow and Metabolism. New York: Raven Press, 1993.
9. Kety S, Schmidt CF: Nitrous oxide method for the quantitative determination of cerebral blood flow in man: Theory,
procedure, and normal values. J Clin Invest 27:475-483, 1948.
10. Gur D, Good WF, Wolfson SK, Jr, et al: In vivo mapping of local cerebral blood flow by xenon-enhanced computed
tomography. Science 215:1267-1268, 1982. Medline Similar articles
11. Fox PT, Raichle ME, Mintun MA, Dence C: Nonoxidative glucose consumption during focal physiologic neural activity.
Science 241:462-464, 1988. Medline Similar articles
12. Stewart GN: Researches on the circulation time in organs, and on the influences which affect it: Parts I-III. J Physiol
(London) 15:1-89, 1894.
13. Zeirler KL: Theoretical basis of indicator-dilution methods for measuring flow and volume. Circ Res 10:393-407, 1962.
14. Grubb RL, Jr, Raichle ME, Eichling JO, Ter-Pogossian MM: The effects of changes in PaCO2 on cerebral blood volume,
blood flow, and vascular mean transit time. Stroke 5:630-639, 1974. Medline Similar articles
15. Lassen NA, Perl W: Tracer Kinetic Methods in Medical Physiology. New York: Raven Press, 1979.
16. Davis CP, McKinnon GC, Debatin JF, von Schulthess GK: Ultra-high-speed MR imaging. Eur Radiol 6:297-311, 1996.
17. Nitz WR: Fast and ultrafast non-echo-planar MR imaging techniques. Eur Radiol 12:2866-2882, 2002. Medline
Similar articles
18. Villringer A, Rosen BR, Belliveau JW, et al: Dynamic imaging with lanthanide chelates in normal brain: contrast due to
magnetic susceptibility effects. Magn Reson Med 6:164-174, 1988. Medline Similar articles
19. Rosen BR, Belliveau JW, Vevea JM, Brady TJ: Perfusion imaging with NMR contrast agents. Magn Reson Med
20. Perman WH, Gado MH, Larson KB, Perlmutter JS: Simultaneous MR acquisition of arterial and brain signal-time curves.
21. Lassen NA: Cerebral transit of an intravascular tracer may allow measurement of regional blood volume but not regional
blood flow. J Cereb Blood Flow Metab 4:633-634, 1984. Medline Similar articles
22. Weisskoff RM, Chesler D, Boxerman JL, Rosen BR: Pitfalls in MR measurement of tissue blood flow with intravascular
tracers: which mean transit time? Magn Reson Med 29:553-558, 1993. Medline Similar articles
23. Wilke N, Simm C, Zhang J, et al: Contrast-enhanced first pass myocardial perfusion imaging: correlation between
myocardial blood flow in dogs at rest and during hyperemia. Magn Reson Med 29:485-497, 1993. Medline
Similar articles
24. Thompson HK, Stramer CF, Whalen RE, McIntosh HD: Indicator transit time considered as a gamma variate. Circ Res
14:502, 1964
25. Padhani AR: Dynamic contrast-enhanced MRI in clinical oncology: current status and future directions. J Magn Reson
5 of 16 15-04-2010 20:17

26. Huang AJ, Lee VS, Rusinek H: MR imaging of renal function. Radiol Clin North Am 41:1001-1017, 2003. Medline
Similar articles
28. Ersoy H, Jacobs P, Kent CK, Prince MR: Blood pool MR angiography of aortic stent-graft endoleak. Am J Roentgenol
182:1181-1186, 2004.
29. Le Bihan D, Breton E, Lallemand D, et al: Separation of diffusion and perfusion in intravoxel incoherent motion MR
30. Young IR, Hall AS, Bryant DJ, et al: Assessment of brain perfusion with MR imaging. J Comput Assist Tomogr 12:721-727,
31. Ogawa S, Lee TM, Kay AR, Tank DW: Brain magnetic resonance imaging with contrast dependent on blood oxygenation.
Proc Natl Acad Sci USA 87:9868-9872, 1990. Medline Similar articles
32. Ogawa S, Menon RS, Tank DW, et al: Functional brain mapping by blood oxygenation level-dependent contrast magnetic
resonance imaging. A comparison of signal characteristics with a biophysical model. Biophys J 64:803-812, 1993.
33. Ogawa S, Lee TM, Barrere B: The sensitivity of magnetic resonance image signals of a rat brain to changes in the cerebral
venous blood oxygenation. Magn Reson Med 29:205-210, 1993. Medline Similar articles
34. Detre JA, Leigh JS, Williams DS, Koretsky AP: Perfusion imaging. Magn Reson Med 23:37-45, 1992. Medline
Similar articles
35. Detre JA, Zhang W, Roberts DA, et al: Tissue specific perfusion imaging using arterial spin labeling. NMR Biomed
36. Henkelman RM: Does IVIM measure classical perfusion? Magn Reson Med 16:470-475, 1990. Medline
Similar articles
37. Le Bihan D, Turner R: The capillary network: a link between IVIM and classical perfusion. Magn Reson Med 27:171-178,
39. Kwong KK, Belliveau JW, Chesler DA, et al: Dynamic magnetic resonance imaging of human brain activity during primary
sensory stimulation. Proc Natl Acad Sci U S A 89:5675-5679, 1992. Medline Similar articles
40. Matthews PM, Jezzard P: Functional magnetic resonance imaging. J Neurol Neurosurg Psychiatry 75:6-12, 2004.
41. Li D, Dhawale P, Rubin PJ, et al: Myocardial signal response to dipyridamole and dobutamine: demonstration of the
BOLD effect using a double-echo gradient-echo sequence. Magn Reson Med 36:16-20, 1996. Medline Similar articles
42. Reeder SB, Holmes AA, McVeigh ER, Forder JR: Simultaneous noninvasive determination of regional myocardial
perfusion and oxygen content in rabbits: toward direct measurement of myocardial oxygen consumption at MR imaging.
43. Wacker CM, Hartlep AW, Pfleger S, et al: Susceptibility-sensitive magnetic resonance imaging detects human
myocardium supplied by a stenotic coronary artery without a contrast agent. J Am Coll Cardiol 41:834-840, 2003. Medline
Similar articles
44. Edelman RR, Siewert B, Darby DG, et al: Qualitative mapping of cerebral blood flow and functional localization with
echo-planar MR imaging and signal targeting with alternating radio frequency. Radiology 192:513-520, 1994. Medline
Similar articles
45. Zhang W, Williams DS, Detre JA, Koretsky AP: Measurement of brain perfusion by volume-localized NMR spectroscopy
using inversion of arterial water spins: accounting for transit time and cross-relaxation. Magn Reson Med 25:362-371, 1992.
page 453
page 454
46. Kim SG: Quantification of relative cerebral blood flow change by flow-sensitive alternating inversion recovery (FAIR)
technique: application to functional mapping. Magn Reson Med 34:293-301, 1995. Medline Similar articles
47. Wong EC, Buxton RB, Frank LR: Implementation of quantitative perfusion imaging techniques for functional brain mapping
using pulsed arterial spin labeling. NMR Biomed 10:237-249, 1997. Medline Similar articles
48. Wong EC, Buxton RB, Frank LR: Quantitative imaging of perfusion using a single subtraction (QUIPSS and QUIPSS II).
49. Arai AE: Magnetic resonance first-pass myocardial perfusion imaging. Top Magn Reson Imaging 11:383-398, 2000.
50. Wilke N, Jerosch-Herold M: Assessing myocardial perfusion in coronary artery disease with magnetic resonance first-pass
imaging. Cardiol Clin 16:227-246, 1998. Medline Similar articles
51. Vallee JP, Sostman HD, MacFall JR, Coleman RE: Quantification of myocardial perfusion with MRI and exogenous
6 of 16 15-04-2010 20:17
contrast agents. Cardiology 88:90-105, 1997. Medline Similar articles

52. Wilke N, Jerosch-Herold M, Stillman AE, et al: Concepts of myocardial perfusion imaging in magnetic resonance imaging.
Magn Reson Q 10:249-286, 1994. Medline Similar articles
53. Kang YS, Gore JC, Armitage IM: Studies of factors affecting the design of NMR contrast agents: manganese in blood as
a model system. Magn Reson Med 1:396-409, 1994.
54. Atkins HL, Som P, Fairchild RG, et al: Myocardial positron tomography with manganese-52m. Radiology 133:769-774,
55. Ni Y, Petre C, Bosmans H, et al: Comparison of manganese biodistribution and MR contrast enhancement in rats after
intravenous injection of MnDPDP and MnCl2. Acta Radiol 38:700-707, 1997. Medline Similar articles
56. Schaefer S, Lange RA, Gutekunst DP, et al: Contrast-enhanced magnetic resonance imaging of hypoperfused
myocardium. Invest Radiol 26:551-556, 1991. Medline Similar articles
57. Borg DC, Cotzias GC: Manganese metabolism in man: rapid exchange of MN56 with tissue as demonstrated by blood
clearance and liver uptake. J Clin Invest 37:1269-1278, 1958. Medline Similar articles
58. Brady TJ, Goldman MR, Pykett IL, et al: Proton nuclear magnetic resonance imaging of regionally ischemic canine hearts:
effect of paramagnetic proton signal enhancement. Radiology. 144:343-347, 1982. Medline Similar articles
59. Storey P, Danias PG, Post M, et al: Preliminary evaluation of EVP 1001-1: a new cardiac-specific magnetic resonance
contrast agent with kinetics suitable for steady-state imaging of the ischemic heart. Invest Radiol 38:642-652, 2003.
60. Waller C, Hiller KH, Voll S, et al: Myocardial perfusion imaging using a non-contrast agent MR imaging technique. Int J
Cardiovasc Imaging 17:123-132, 2001. Medline Similar articles
61. Prasad PV, Burstein D, Edelman RR: MRI evaluation of myocardial perfusion without a contrast agent using magnetization
transfer. Magn Reson Med 30:267-270, 1993. Medline Similar articles
62. Tsekos N, Zhang Y, Merkle H, et al: Assessment of myocardial perfusion state with magnetization transfer during coronary
artery occlusion and adenosine vasodilation. In: Proceedings of 2nd Annual Meeting of Society of Magnetic Resonance;
1994:1480.
63. Alsop DC, Hatabu H, Bonnet M, et al: Multi-slice, breathhold imaging of the lung with submillisecond echo times. Magn
64. Hatabu H, Gaa J, Kim D, et al: Pulmonary perfusion: qualitative assessment with dynamic contrast-enhanced MRI using
ultra-short TE and inversion recovery turbo FLASH. Magn Reson Med 36:503-508, 1996. Medline Similar articles
65. Mai VM, Berr SS: MR perfusion imaging of pulmonary parenchyma using pulsed arterial spin labeling techniques: FAIRER
and FAIR. J Magn Reson Imaging 9:483-487, 1999. Medline Similar articles
66. Mai VM, Hagspiel KD, Altes T, et al: Detection of regional pulmonary perfusion deficit of the occluded lung using arterial
spin labeling in magnetic resonance imaging. J Magn Reson Imaging 11:97-102, 2000. Medline Similar articles
67. Grenier N, Basseau F, Ries M, et al: Functional MRI of the kidney. Abdom Imaging 28:164-175, 2003. Medline
Similar articles
68. Schoenberg SO, Just A, Bock M, et al: Noninvasive analysis of renal artery blood flow dynamics with MR cine phase-
contrast flow measurements. Am J Physiol 272:H2477-2484, 1997.
69. de Haan MW, van Engelshoven JM, Houben AJ, et al: Phase-contrast magnetic resonance flow quantification in renal
133
arteries: comparison with Xenon washout measurements. Hypertension. 41:114-118, 2003.
70. Lucas C: Renal support. In Fry D (ed): Multiple System Failure. St Louis: Mosby, 1992, p 255.
71. Garrison RN, Wilson MA, Matheson PJ, Spain DA: Nitric oxide mediates redistribution of intrarenal blood flow during
bacteremia. J Trauma 39:90-96; discussion 96-97, 1995. Medline Similar articles
72. Mostafavi MR, Chavez DR, Cannillo J, et al: Redistribution of renal blood flow after SWL evaluated by Gd-DTPA-enhanced
magnetic resonance imaging. J Endourol 12:9-12, 1998. Medline Similar articles
73. Vallee JP, Lazeyras F, Khan HG, Terrier F: Absolute renal blood flow quantification by dynamic MRI and Gd-DTPA. Eur
Radiol 10:1245-1252, 2000. Medline Similar articles
74. Montet X, Ivancevic MK, Belenger J, et al: Noninvasive measurement of absolute renal perfusion by contrast medium-
enhanced magnetic resonance imaging. Invest Radiol 38:584-592, 2003. Medline Similar articles
75. Prasad PV, Cannillo J, Chavez DR, et al: First-pass renal perfusion imaging using MS-325, an albumin-targeted MRI
contrast agent. Invest Radiol 34:566-571, 1999. Medline Similar articles
76. Aumann S, Schoenberg SO, Just A, et al: Quantification of renal perfusion using an intravascular contrast agent (part 1):
results in a canine model. Magn Reson Med 49:276-287, 2003.
77. Schoenberg SO, Aumann S, Just A, et al: Quantification of renal perfusion abnormalities using an intravascular contrast
agent (part 2): results in animals and humans with renal artery stenosis. Magn Reson Med 49:288-298, 2003.
78. Prasad PV, Kim D, Kaiser AM, et al: Noninvasive comprehensive characterization of renal artery stenosis by combination
of STAR angiography and EPISTAR perfusion imaging. Magn Reson Med 38:776-787, 1997. Medline Similar articles
79. Reichen J, Egger B, Ohara N, et al: Determinants of hepatic function in liver cirrhosis in the rat. Multivariate analysis. J Clin
Invest 82:2069-2076, 1988. Medline Similar articles
7 of 16 15-04-2010 20:17
80. Varin F, Huet PM: Hepatic microcirculation in the perfused cirrhotic rat liver. J Clin Invest 76:1904-1912, 1985. Medline
Similar articles
81. Huet PM, Goresky CA, Villeneuve JP, et al: Assessment of liver microcirculation in human cirrhosis. J Clin Invest
82. Reichen J: Assessment of hepatic function with xenobiotics. Semin Liver Dis 15:189-201, 1995. Medline
Similar articles
83. Keiding S, Engsted E, Ott P: Sorbitol as a test substance for measurement of liver plasma flow in humans. Hepatology
84. Iwasa M, Nakamura K, Nakagawa T, et al: Single photon emission computed tomography to determine effective hepatic
blood flow and intrahepatic shunting. Hepatology 21:359-365, 1995. Medline Similar articles
85. Ziegler SI, Haberkorn U, Byrne H, et al: Measurement of liver blood flow using oxygen-15 labelled water and dynamic
positron emission tomography: limitations of model description. Eur J Nucl Med 23:169-177, 1996. Medline
Similar articles
86. Leen E, Goldberg JA, Anderson JR, et al: Hepatic perfusion changes in patients with liver metastases: comparison with
those patients with cirrhosis. Gut 34:554-557, 1993. Medline Similar articles
87. Fong Y: Doppler perfusion index in colorectal cancer. Lancet 355:5-6, 2000. Medline Similar articles
88. Miles KA, Hayball M, Dixon AK: Colour perfusion imaging: a new application of computed tomography. Lancet
89. Bader TR, Herneth AM, Blaicher W, et al: Hepatic perfusion after liver transplantation: noninvasive measurement with
dynamic single-section CT. Radiology 209:129-134, 1998. Medline Similar articles
90. Leggett DA, Kelley BB, Bunce IH, Miles KA: Colorectal cancer: diagnostic potential of CT measurements of hepatic
perfusion and implications for contrast enhancement protocols. Radiology 205:716-720, 1997. Medline Similar articles
91. Van Beers BE, Leconte I, Materne R, et al: Hepatic perfusion parameters in chronic liver disease: dynamic CT
measurements correlated with disease severity. Am J Roentgenol 176:667-673, 2001.
92. Schwickert HC, Roberts TP, Shames DM, et al: Quantification of liver blood volume: comparison of ultra short TI inversion
recovery echo planar imaging (ULSTIR-EPI), with dynamic 3D-gradient recalled echo imaging. Magn Reson Med 34:845-852,
1995.
93. Blustajn J, Cuenod CA, Clement O, et al: Measurement of liver blood volume using a macromolecular MRI contrast agent at
equilibrium. Magn Reson Imaging 15:415-421, 1997. Medline Similar articles
94. Scharf J, Zapletal C, Hess T, et al: Assessment of hepatic perfusion in pigs by pharmacokinetic analysis of dynamic MR
images. J Magn Reson Imaging 9:568-572, 1999. Medline Similar articles
95. Sinoway LI, Hendrickson C, Davidson WR, Jr, et al: Characteristics of flow-mediated brachial artery vasodilation in human
subjects. Circ Res 64:32-42, 1989. Medline Similar articles
96. Lassen NA, Lindbjerg J, Munck O: Measurement of blood-flow through skeletal muscle by intramuscular injection of
xenon-133. Lancet 15:686-689, 1964.
97. Andersen P, Saltin B: Maximal perfusion of skeletal muscle in man. J Physiol 366:233-249, 1985. Medline
Similar articles
98. Guldvog I, Kjaernes M, Thoresen M, Walloe L: Blood flow in arteries determined transcutaneously by an ultrasonic Doppler
velocitymeter as compared to electromagnetic measurements on the exposed vesels. Acta Physiol Scand 109:211-216, 1980.
99. Larsson SE, Cai H, Zhang Q, et al: Measurement by laser-Doppler flowmetry of microcirculation in lower leg muscle at
different blood fluxes in relation to electromyographically determined contraction and accumulated fatigue. Eur J Appl Physiol
Occup Physiol 70:288-293, 1995. Medline Similar articles
100. Toussaint JF, Kwong KK, Mkparu FO, et al: Perfusion changes in human skeletal muscle during reactive hyperemia
measured by echo-planar imaging. Magn Reson Med 35:62-69, 1996. Medline Similar articles
101. Donahue KM, Van Kylen J, Guven S, et al: Simultaneous gradient-echo/spin-echo EPI of graded ischemia in human
skeletal muscle. J Magn Reson Imaging 8:1106-1113, 1998. Medline Similar articles
102. Belliveau JW, Kennedy DN, Jr, McKinstry RC, et al: Functional mapping of the human visual cortex by magnetic
resonance imaging. Science 254:716-719, 1991. Medline Similar articles
103. Deliganis AV, Maravilla KR, Heiman JR, et al: Female genitalia: dynamic MR imaging with use of MS-325 initial
experiences evaluating female sexual response. Radiology 225:791-799, 2002. Medline Similar articles
104. Heiman JR: Psychophysiological models of female sexual response. Int J Impot Res 10(Suppl 2):S94-97; discussion
S98-101, 1998.
page 454
page 455
105. Laan E, Everaerd W: Physiological measures of vaginal vasocongestion. Int J Impot Res 10(Suppl 2):S107-110;
discussion S124-125, 1998.
106. Hayes CE, Tsuruda JS, Mathis CM: Temporal lobes: surface MR coil phased-array imaging. Radiology 189:918-920,
8 of 16 15-04-2010 20:17
107. Wu EX, Wong KK, Andrassy M, Tang H: High-resolution in vivo CBV mapping with MRI in wild-type mice. Magn Reson
108. Niendorf ER, Santyr GE, Brazy PC, Grist TM: Measurement of Gd-DTPA dialysis clearance rates by using a look-locker
imaging technique. Magn Reson Med 36:571-578, 1996. Medline Similar articles
109. Chen Z, Prato FS, McKenzie C: T1 fast acquisition relaxation mapping (T1-FARM): an optimized reconstruction. IEEE
Trans Med Imaging 17:155-160, 1998. Medline Similar articles
110. Rusinek H, Lee VS, Johnson G: Optimal dose of Gd-DTPA in dynamic MR studies. Magn Reson Med 46:312-316, 2001.
111. Dumoulin CL, Buonocore MH, Opsahl LR, et al: Noninvasive measurement of renal hemodynamic functions using
gadolinium enhanced magnetic resonance imaging. Magn Reson Med 32:370-378, 1994. Medline Similar articles
112. Niendorf ER, Grist TM, Lee FT, Jr, et al: Rapid in vivo measurement of single-kidney extraction fraction and glomerular
filtration rate with MR imaging. Radiology 206:791-798, 1998. Medline Similar articles
113. Coulam CH, Lee JH, Wedding KL, et al: Noninvasive measurement of extraction fraction and single-kidney glomerular
filtration rate with MR imaging in swine with surgically created renal arterial stenoses. Radiology 223:76-82, 2002. Medline
Similar articles
114. Baumann D, Rudin M: Quantitative assessment of rat kidney function by measuring the clearance of the contrast agent
Gd(DOTA) using dynamic MRI. Magn Reson Imaging 18:587-595, 2000. Medline Similar articles
115. Laurent D, Poirier K, Wasvary J, Rudin M: Effect of essential hypertension on kidney function as measured in rat by
dynamic MRI. Magn Reson Med 47:127-134, 2002. Medline Similar articles
116. Smith AM, Materne R, Van Beers BE: Quantitative measurement of blood perfusion, GFR and arterial vascular fraction in
the kidney cortex. In: Proceedings of 9th Annual Meeting of International Society of Magnetic Resonance in Medicine, 2001.
117. Lee VS, Rusinek H, Kim S, et al: Analysis of dynamic three-dimensional (3D) MR renography: regional characterization
by multicompartmental modeling. In: Proceedings of 9th Annual Meeting of International Society of Magnetic Resonance in
Medicine, 2001.
118. Grenier N, Trillaud H, Combe C, et al: Diagnosis of renovascular hypertension: feasibility of captopril-sensitized dynamic
MR imaging and comparison with captopril scintigraphy. Am J Roentgenol 166:835-843, 1996.
119. Prasad PV, Goldfarb J, Sundaram C, et al: Captopril MR renography in a swine model: toward a comprehensive
evaluation of renal arterial stenosis. Radiology 217:813-818, 2000. Medline Similar articles
120. Pickering TG, Blumenfeld JD, Laragh JH: Renovascular hypertension and ischemic nephropathy. In Brenner BM (ed): The
Kidney. Philadelphia, WB Saunders, 1996, pp 2106-2125.
121. Lee VS, Rusinek H, Johnson G, et al: MR renography with low-dose gadopentetate dimeglumine: feasibility. Radiology
122. Mostafavi MR, Saltzman B, Prasad PV: Magnetic resonance imaging in the evaluation of ureteropelvic junction obstructed
kidney. Urology 50:601-602; discussion 602-603, 1997. Medline Similar articles
123. Rohrschneider WK, Becker K, Hoffend J, et al: Combined static-dynamic MR urography for the simultaneous evaluation of
morphology and function in urinary tract obstruction. II. Findings in experimentally induced ureteric stenosis. Pediatr Radiol
124. Rohrschneider WK, Hoffend J, Becker K, et al: Combined static-dynamic MR urography for the simultaneous evaluation of
morphology and function in urinary tract obstruction. I. Evaluation of the normal status in an animal model. Pediatr Radiol
125. Agildere AM, Tarhan NC, Bozdagi G, et al: Correlation of quantitative dynamic magnetic resonance imaging findings with
pathology results in renal transplants: a preliminary report. Transplant Proc 31:3312-3316, 1999. Medline
Similar articles
126. Dorsam J, Knopp MV, Carl S, et al: Ureteral complications after kidney transplantation-evaluation with functional magnetic
resonance urography. Transplant Proc 29:132-135, 1997. Medline Similar articles
127. Folkman J, Beckner K: Angiogenesis imaging. Acad Radiol 7:783-785, 2000. Medline Similar articles
128. Li WW: Tumor angiogenesis: molecular pathology, therapeutic targeting, and imaging. Acad Radiol 7:800-811, 2000.
129. Passe TJ, Bluemke DA, Siegelman SS: Tumor angiogenesis: tutorial on implications for imaging. Radiology
130. Neeman M: Preclinical MRI experience in imaging angiogenesis. Cancer Metastasis Rev 19:39-43, 2000. Medline
Similar articles
131. Neeman M, Provenzale JM, Dewhirst MW: Magnetic resonance imaging applications in the evaluation of tumor
angiogenesis. Semin Radiat Oncol 11:70-82, 2001. Medline Similar articles
132. Brasch R, Turetschek K: MRI characterization of tumors and grading angiogenesis using macromolecular contrast media:
status report. Eur J Radiol 34:148-155, 2000. Medline Similar articles
133. Gillies RJ, Bhujwalla ZM, Evelhoch J, et al: Applications of magnetic resonance in model systems: tumor biology and
physiology. Neoplasia 2:139-151, 2000. Medline Similar articles
9 of 16 15-04-2010 20:17
134. Bhujwalla ZM, Artemov D, Glockner J: Tumor angiogenesis, vascularization, and contrast-enhanced magnetic resonance
imaging. Top Magn Reson Imaging 10:92-103, 1999. Medline Similar articles
135. Crone C: The permeability of capillaries in various organs as determined by use of the 'indicator diffusion' method. Acta
Physiol Scand 58:292-305, 1963. Medline Similar articles
136. Yuh WT: An exciting and challenging role for the advanced contrast MR imaging. J Magn Reson Imaging 10:221-222,
137. Taylor JS, Reddick WE: Evolution from empirical dynamic contrast-enhanced magnetic resonance imaging to
pharmacokinetic MRI. Adv Drug Deliv Rev 41:91-110, 2000. Medline Similar articles
138. Daldrup HE, Shames DM, Husseini W, et al: Quantification of the extraction fraction for gadopentetate across breast
cancer capillaries. Magn Reson Med 40:537-543, 1998. Medline Similar articles
139. Taylor JS, Tofts PS, Port R, et al: MR imaging of tumor microcirculation: promise for the new millennium. J Magn Reson
140. Tofts PS, Brix G, Buckley DL, et al: Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of
a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 10:223-232, 1999. Medline
Similar articles
141. Su MY, Jao JC, Nalcioglu O: Measurement of vascular volume fraction and blood-tissue permeability constants with a
pharmacokinetic model: studies in rat muscle tumors with dynamic Gd-DTPA enhanced MRI. Magn Reson Med 32:714-724,
142. Degani H, Gusis V, Weinstein D, et al: Mapping pathophysiological features of breast tumors by MRI at high spatial
resolution. Nat Med 3:780-782, 1997. Medline Similar articles
143. Flickinger FW, Allison JD, Sherry RM, Wright JC: Differentiation of benign from malignant breast masses by time-intensity
evaluation of contrast enhanced MRI. Magn Reson Imaging 11:617-620, 1993. Medline Similar articles
144. Kaiser WA, Zeitler E: MR imaging of the breast: fast imaging sequences with and without Gd-DTPA. Preliminary
145. Gribbestad IS, Nilsen G, Fjosne HE, et al: Comparative signal intensity measurements in dynamic gadolinium-enhanced
MR mammography. J Magn Reson Imaging 4:477-480, 1994. Medline Similar articles
146. Tofts PS: Modeling tracer kinetics in dynamic Gd-DTPA MR imaging. J Magn Reson Imaging 7:91-101, 1997.
147. Tofts PS, Berkowitz B, Schnall MD: Quantitative analysis of dynamic Gd-DTPA enhancement in breast tumors using a
permeability model. Magn Reson Med 33:564-568, 1995. Medline Similar articles
148. den Boer JA, Hoenderop RK, Smink J, et al: Pharmacokinetic analysis of Gd-DTPA enhancement in dynamic three-
dimensional MRI of breast lesions. J Magn Reson Imaging 7:702-715, 1997. Medline Similar articles
149. van der Woude HJ, Verstraete KL, Hogendoorn PC, et al: Musculoskeletal tumors: does fast dynamic contrast-enhanced
subtraction MR imaging contribute to the characterization? Radiology 208:821-828, 1998. Medline Similar articles
150. Jager GJ, Ruijter ET, van de Kaa CA, et al: Dynamic TurboFLASH subtraction technique for contrast-enhanced MR
imaging of the prostate: correlation with histopathologic results. Radiology 203:645-652, 1997. Medline Similar articles
151. Liu PF, Krestin GP, Huch RA, et al: MRI of the uterus, uterine cervix, and vagina: diagnostic performance of dynamic
contrast-enhanced fast multiplanar gradient-echo imaging in comparison with fast spin-echo T2-weighted pulse imaging. Eur
152. Barentsz JO, Jager GJ, van Vierzen PB, et al: Staging urinary bladder cancer after transurethral biopsy: value of fast
dynamic contrast-enhanced MR imaging. Radiology 201:185-193, 1996. Medline Similar articles
153. Hawighorst H, Weikel W, Knapstein PG, et al: Angiogenic activity of cervical carcinoma: assessment by functional
magnetic resonance imaging-based parameters and a histomorphological approach in correlation with disease outcome. Clin
Cancer Res 4:2305-2312, 1998. Medline Similar articles
154. Barentsz JO, Berger-Hartog O, Witjes JA, et al: Evaluation of chemotherapy in advanced urinary bladder cancer with fast
dynamic contrast-enhanced MR imaging. Radiology 207:791-797, 1998. Medline Similar articles
155. Reddick WE, Taylor JS, Fletcher BD: Dynamic MR imaging (DEMRI) of microcirculation in bone sarcoma. J Magn Reson
156. van der Woude HJ, Bloem JL, Verstraete KL, et al: Osteosarcoma and Ewing's sarcoma after neoadjuvant
chemotherapy: value of dynamic MR imaging in detecting viable tumor before surgery. Am J Roentgenol 165:593-598, 1995.
157. Padhani AR, MacVicar AD, Gapinski CJ, et al: Effects of androgen deprivation on prostatic morphology and vascular
permeability evaluated with MR imaging. Radiology 218:365-374, 2001. Medline Similar articles
158. Knopp MV, Brix G, Junkermann HJ, Sinn HP: MR mammography with pharmacokinetic mapping for monitoring of breast
cancer treatment during neoadjuvant therapy. Magn Reson Imaging Clin N Am 2:633-658, 1994. Medline
Similar articles
159. George ML, Dzik-Jurasz AS, Padhani AR, et al: Non-invasive methods of assessing angiogenesis and their value in
predicting response to treatment in colorectal cancer. Br J Surg 88:1628-1636, 2001. Medline Similar articles
160. Devries AF, Griebel J, Kremser C, et al: Tumor microcirculation evaluated by dynamic magnetic resonance imaging
predicts therapy outcome for primary rectal carcinoma. Cancer Res 61:2513-2516, 2001. Medline Similar articles
10 of 16 15-04-2010 20:17
page 455
page 456
161. de Vries A, Griebel J, Kremser C, et al: Monitoring of tumor microcirculation during fractionated radiation therapy in
patients with rectal carcinoma: preliminary results and implications for therapy. Radiology 217:385-391, 2000. Medline
Similar articles
162. Mayr NA, Yuh WT, Arnholt JC, et al: Pixel analysis of MR perfusion imaging in predicting radiation therapy outcome in
cervical cancer. J Magn Reson Imaging 12:1027-1033, 2000. Medline Similar articles
163. Burn PR, McCall JM, Chinn RJ, et al: Uterine fibroleiomyoma: MR imaging appearances before and after embolization of
uterine arteries. Radiology 214:729-734, 2000. Medline Similar articles
164. Jha RC, Ascher SM, Imaoka I, Spies JB: Symptomatic fibroleiomyomata: MR imaging of the uterus before and after
uterine arterial embolization. Radiology 217:228-235, 2000. Medline Similar articles
165. Li W, Brophy DP, Chen Q, et al: Semiquantitative assessment of uterine perfusion using first pass dynamic contrast-
enhanced MR imaging for patients treated with uterine fibroid embolization. J Magn Reson Imaging 12:1004-1008, 2000.
166. Padhani AR, Dzik-Jurasz A: Perfusion MR imaging of extracranial tumor angiogenesis. Top Magn Reson Imaging
167. Sugahara T, Korogi Y, Kochi M, et al: Correlation of MR imaging-determined cerebral blood volume maps with histologic
and angiographic determination of vascularity of gliomas. Am J Roentgenol 171:1479-1486, 1998.
168. Aronen HJ, Gazit IE, Louis DN, et al: Cerebral blood volume maps of gliomas: comparison with tumor grade and
histologic findings. Radiology 191:41-51, 1994. Medline Similar articles
169. Knopp EA, Cha S, Johnson G, et al: Glial neoplasms: dynamic contrast-enhanced T2*-weighted MR imaging. Radiology
170. Bagley LJ, Grossman RI, Judy KD, et al: Gliomas: correlation of magnetic susceptibility artifact with histologic grade.
171. Maeda M, Itoh S, Kimura H, et al: Tumor vascularity in the brain: evaluation with dynamic susceptibility-contrast MR
172. Fisel CR, Ackerman JL, Buxton RB, et al: MR contrast due to microscopically heterogeneous magnetic susceptibility:
numerical simulations and applications to cerebral physiology. Magn Reson Med 17:336-347, 1991. Medline
Similar articles
173. Boxerman JL, Hamberg LM, Rosen BR, Weisskoff RM: MR contrast due to intravascular magnetic susceptibility
perturbations. Magn Reson Med 34:555-566.
174. Donahue KM, Krouwer HG, Rand SD, et al: Utility of simultaneously acquired gradient-echo and spin-echo cerebral blood
volume and morphology maps in brain tumor patients. Magn Reson Med 43:845-853, 2000. Medline Similar articles
175. van Dijke CF, Brasch RC, Roberts TP, et al: Mammary carcinoma model: correlation of macromolecular contrast-
enhanced MR imaging characterizations of tumor microvasculature and histologic capillary density. Radiology 198:813-818,
176. Turetschek K, Huber S, Floyd E, et al: MR imaging characterization of microvessels in experimental breast tumors by
using a particulate contrast agent with histopathologic correlation. Radiology 218:562-569, 2001. Medline
Similar articles
177. Pham CD, Roberts TP, van Bruggen N, et al: Magnetic resonance imaging detects suppression of tumor vascular
permeability after administration of antibody to vascular endothelial growth factor. Cancer Invest 16:225-230, 1998.
178. Bogdanov AA, Lewin M, Weissleder R: Approaches and agents for imaging the vascular system. Adv Drug Deliv Rev
179. Tjuvajev JG, Avril N, Oku T, et al: Imaging herpes virus thymidine kinase gene transfer and expression by positron
emission tomography. Cancer Res 58:4333-4341, 1998. Medline Similar articles
180. Weissleder R, Tung CH, Mahmood U, Bogdanov A, Jr: In vivo imaging of tumors with protease-activated near-infrared
fluorescent probes. Nat Biotechnol 17:375-378, 1999. Medline Similar articles
181. Benaron DA, Stevenson DK: Optical time-of-flight and absorbance imaging of biologic media. Science 259:1463-1466,
182. Contag PR, Olomu IN, Stevenson DK, Contag CH: Bioluminescent indicators in living mammals. Nat Med 4:245-247,
183. Smith BR, Johnson GA, Groman EV, Linney E: Magnetic resonance microscopy of mouse embryos. Proc Natl Acad Sci
U S A 91:3530-3533, 1994. Medline Similar articles
184. Brooks PC, Clark RA, Cheresh DA: Requirement of vascular integrin alpha v beta 3 for angiogenesis. Science
264:569-571, 1994.
185. Winter PM, Caruthers SD, Kassner A, et al: Molecular imaging of angiogenesis in nascent Vx-2 rabbit tumors using a
novel alpha(nu)beta3-targeted nanoparticle and 1.5 tesla magnetic resonance imaging. Cancer Res 63:5838-5843, 2003.
186. Hogemann-Savellano D, Bos E, Blondet C, et al: The transferrin receptor: a potential molecular imaging marker for human
11 of 16 15-04-2010 20:17
cancer. Neoplasia 5:495-506, 2003. Medline Similar articles

187. Buxton RB, Wong EC, Frank LR: Dynamics of blood flow and oxygenation changes during brain activation: the balloon
model. Magn Reson Med 39:855-864, 1998. Medline Similar articles
188. Brezis M, Rosen S: Hypoxia of the renal medulla-its implications for disease. N Engl J Med 332:647-655, 1995.
189. Epstein FH, Agmon Y, Brezis M: Physiology of renal hypoxia. Ann N Y Acad Sci 718:72-81, discussion 81-82, 1994.
190. Thadhani R, Pascual M, Bonventre JV: Acute renal failure. N Engl J Med 334:1448-1460.
191. Heyman SN, Fuchs S, Brezis M: The role of medullary ischemia in acute renal failure. N Horiz 3:597-607.
192. Brezis M, Epstein FH: Cellular mechanisms of acute ischemic injury in the kidney. Annu Rev Med 44:27-37, 1993.
193. Rosen S, Epstein FH, Brezis M: Determinants of intrarenal oxygenation: factors in acute renal failure. Ren Failure
14:321-325, 1992.
194. Brezis M, Dinour D, Greenfeld Z, Rosen S: Susceptibility of Henle's loop to hypoxic and toxic insults. Adv Nephrol Necker
Hosp 20:41-56, 1991. Medline Similar articles
195. Brezis M, Rosen SN, Epstein FH: The pathophysiological implications of medullary hypoxia. Am J Kidney Dis
196. Brezis M, Rosen S, Silva P, Epstein FH: Renal ischemia: a new perspective. Kidney Int 26:375-83, 1984. Medline
Similar articles
197. Brezis M, Rosen S, Silva P, Epstein FH: Selective anoxic injury to thick ascending limb: an anginal syndrome of the renal
medulla? Adv Exp Med Biol 180:239-249, 1984. Medline Similar articles
198. Johnson RJ, Herrera-Acosta J, Schreiner GF, Rodriguez-Iturbe B: Subtle acquired renal injury as a mechanism of
salt-sensitive hypertension. N Engl J Med 346:913-923, 2002. Medline Similar articles
199. Cowley AW, Jr, Mattson DL, Lu S, Roman RJ: The renal medulla and hypertension. Hypertension 25:663-673, 1995.
200. Cowley AW, Roman RJ, Fenoy FJ, Mattson DL: Effect of renal medullary circulation on arterial pressure. J Hypertens
10(Suppl):S187-193, 1992.
201. Cowley AW, Jr. Role of the renal medulla in volume and arterial pressure regulation. Am J Physiol 273:R1-15, 1997.
202. Ugurbil K, Hu X, Chen W, et al: Functional mapping in the human brain using high magnetic fields. Philos Trans R Soc
Lond B Biol Sci 354:1195-213, 1999. Medline Similar articles
203. Rajagopalan P, Krishnan KR, Passe TJ, Macfall JR: Magnetic resonance imaging using deoxyhemoglobin contrast
versus positron emission tomography in the assessment of brain function. Prog Neuropsychopharmacol Biol Psychiatry
204. Prasad PV, Edelman RR, Epstein FH: Noninvasive evaluation of intrarenal oxygenation with BOLD MRI. Circulation
205. Prasad PV, Chen Q, Goldfarb JW, et al: Breath-hold R2* mapping with a multiple gradient-recalled echo sequence:
application to the evaluation of intrarenal oxygenation. J Magn Reson Imaging 7:1163-1165, 1997. Medline
Similar articles
206. Abramovitch R, Frenkiel D, Neeman M: Analysis of subcutaneous angiogenesis by gradient echo magnetic resonance
207. van Zijl PC, Eleff SM, Ulatowski JA, et al: Quantitative assessment of blood flow, blood volume and blood oxygenation
effects in functional magnetic resonance imaging. Nat Med 4:159-167, 1998. Medline Similar articles
208. Robinson SP, Collingridge DR, Howe FA, et al: Tumour response to hypercapnia and hyperoxia monitored by FLOOD
magnetic resonance imaging. NMR Biomed 12:98-106, 1999. Medline Similar articles
209. Taylor NJ, Baddeley H, Goodchild KA, et al: BOLD MRI of human tumor oxygenation during carbogen breathing. J Magn
210. Li KC, Dalman RL, Wright GA: In vivo flow-independent T2 measurements of superior mesenteric vein blood in diagnosis
of chronic mesenteric ischemia: a preliminary evaluation. Acad Radiol 6:530-534, 1999. Medline Similar articles
211. Li KC, Dalman RL, Ch'en IY, et al: Chronic mesenteric ischemia: use of in vivo MR imaging measurements of blood
oxygen saturation in the superior mesenteric vein for diagnosis. Radiology 204:71-77, 1997. Medline Similar articles
212. Li KC, Wright GA, Pelc LR, et al: Oxygen saturation of blood in the superior mesenteric vein: in vivo verification of MR
imaging measurements in a canine model. Work in progress. Radiology 194:321-325, 1995. Medline Similar articles
213. Heiss SG, Li KC: Magnetic resonance angiography of mesenteric arteries. A review. Invest Radiol 33:670-81, 1998.
214. Linsenmeier RA, Braun RD, McRipley MA, et al: Retinal hypoxia in long-term diabetic cats. Invest Ophthalmol Vis Sci
215. Ito Y, Berkowitz BA: MR studies of retinal oxygenation. Vision Res 41:1307-1311, 2001. Medline Similar articles
12 of 16 15-04-2010 20:17
216. Berkowitz BA, Wilson CA: Quantitative mapping of ocular oxygenation using magnetic resonance imaging. Magn Reson
217. Hayes M, Taplin GV: Lung imaging with radioaerosols for the assessment of airway disease. Semin Nucl Med
3
218. Middleton H, Black RD, Saam B, et al: MR imaging with hyperpolarized He gas. Magn Reson Med 33:271-275, 1995.
219. Bachert P, Schad LR, Bock M, et al: Nuclear magnetic resonance imaging of airways in humans with use of
3
hyperpolarized He. Magn Reson Med 36:192-196, 1996.
129
220. Mugler JP, 3rd, Driehuys B, Brookeman JR, et al: MR imaging and spectroscopy using hyperpolarized Xe gas:
preliminary human results. Magn Reson Med 37:809-815, 1997.
221. Edelman RR, Hatabu H, Tadamura E, et al: Noninvasive assessment of regional ventilation in the human lung using
oxygen-enhanced magnetic resonance imaging. Nat Med 2:1236-1239, 1996. Medline Similar articles
222. Chen Q, Jakob PM, Griswold MA, et al: Oxygen enhanced MR ventilation imaging of the lung. MAGMA 7:153-161, 1998.
223. Chiarotti G, Cristiani G, Bliulotto L: Proton relaxation in pure liquids and in liquids containing paramagnetic gases in
solution. Nuovo Cimento 1:863-873, 1995.
page 456
page 457
224. Young IR, Clarke GJ, Bailes DR, et al: Enhancement of relaxation rate with paramagnetic contrast agents in NMR
imaging. J Comput Tomogr 5:543-547, 1981. Medline Similar articles
225. Mai VM, Chen Q, Bankier AA, Edelman RR: Multiple inversion recovery MR subtraction imaging of human ventilation from
inhalation of room air and pure oxygen. Magn Reson Med 43:913-916, 2000. Medline Similar articles
226. Loffler R, Muller CJ, Peller M, et al: Optimization and evaluation of the signal intensity change in multisection oxygen-
enhanced MR lung imaging. Magn Reson Med 43:860-866, 2000. Medline Similar articles
227. Mai VM, Tutton S, Prasad PV, et al: Computing oxygen-enhanced ventilation maps using correlation analysis. Magn
228. Mai VM, Liu B, Polzin JA, et al: Ventilation-perfusion ratio of signal intensity in human lung using oxygen-enhanced and
arterial spin labeling techniques. Magn Reson Med 48:341-350, 2002. Medline Similar articles
229. Mai VM, Bankier AA, Prasad PV, et al: MR ventilation-perfusion imaging of human lung using oxygen-enhanced and
arterial spin labeling techniques. J Magn Reson Imaging 14:574-579, 2001. Medline Similar articles
230. Chen Q, Levin DL, Kim D, et al; Pulmonary disorders: ventilation-perfusion MR imaging with animal models. Radiology
231. Hatabu H, Chen Q, Levin DL, et al: Ventilation-perfusion MR imaging of the lung. Magn Reson Imaging Clin N Am
7:379-392, x, 1999. Medline Similar articles
232. Nakagawa T, Sakuma H, Murashima S, et al: Pulmonary ventilation-perfusion MR imaging in clinical patients. J Magn
233. Pennell DJ: Ventricular volume and mass by CMR. J Cardiovasc Magn Reson 4:507-513, 2002. Medline
Similar articles
234. Quinones MA, Gaasch WH, Alexander JK: Influence of acute changes in preload, afterload, contractile state and heart
rate on ejection and isovolumic indices of myocardial contractility in man. Circulation 53:293-302, 1976. Medline
Similar articles
235. The world health report 2002: Reducing risks, promoting healthy life. World Health Organization, Geneva, 2002.
236. Ross J, Jr: Regional myocardial function and microvascular dysfunction. Does the alternate cascade represent ischemia?
Cardiologia 44:797-799, 1999. Medline Similar articles
237. Abraham TP, Nishimura RA: Myocardial strain: can we finally measure contractility? J Am Coll Cardiol 37:731-734, 2001.
238. Barkhausen J, Ruehm SG, Goyen M, et al: MR evaluation of ventricular function: true fast imaging with steady-state
precession versus fast low-angle shot cine MR imaging: feasibility study. Radiology 219:264-269, 2001. Medline
Similar articles
239. Plein S, Bloomer TN, Ridgway JP, et al: Steady-state free precession magnetic resonance imaging of the heart:
comparison with segmented k-space gradient-echo imaging. J Magn Reson Imaging 14:230-236, 2001. Medline
Similar articles
240. Li W, Stern JS, Mai VM, et al: MR assessment of left ventricular function: quantitative comparison of fast imaging
employing steady-state acquisition (FIESTA) with fast gradient echo cine technique. J Magn Reson Imaging 16:559-564,
241. Slavin GS, Saranathan M: FIESTA-ET: high-resolution cardiac imaging using echo-planar steady-state free precession.
242. Jung BA, Hennig J, Scheffler K: Single-breathhold 3D-trueFISP cine cardiac imaging. Magn Reson Med 48:921-925,
2002.
13 of 16 15-04-2010 20:17
243. Plein S, Smith WH, Ridgway JP, et al: Qualitative and quantitative analysis of regional left ventricular wall dynamics using
real-time magnetic resonance imaging: comparison with conventional breath-hold gradient echo acquisition in volunteers and
patients. J Magn Reson Imaging 14:23-30, 2001. Medline Similar articles
244. Alfakih K, Thiele H, Plein S, et al: Comparison of right ventricular volume measurement between segmented k-space
gradient-echo and steady-state free precession magnetic resonance imaging. J Magn Reson Imaging 16:253-258, 2002.
245. Spotnitz HM: Macro design, structure, and mechanics of the left ventricle. J Thorac Cardiovasc Surg 119:1053-1077,
246. Moore CC, McVeigh ER, Zerhouni EA: Quantitative tagged magnetic resonance imaging of the normal human left
ventricle. Top Magn Reson Imaging 11:359-371, 2000. Medline Similar articles
247. Rademakers FE, Rogers WJ, Guier WH, et al: Relation of regional cross-fiber shortening to wall thickening in the intact
heart. Three-dimensional strain analysis by NMR tagging. Circulation 89:1174-1182, 1994. Medline Similar articles
248. Zerhouni EA, Parish DM, Rogers WJ, et al: Human heart: tagging with MR imaging-a method for noninvasive assessment
of myocardial motion. Radiology 169:59-63, 1988. Medline Similar articles
249. Axel L, Dougherty L: MR imaging of motion with spatial modulation of magnetization. Radiology 171:841-845, 1989.
250. Mosher TJ, Smith MB: A DANTE tagging sequence for the evaluation of translational sample motion. Magn Reson Med
251. Fischer SE, McKinnon GC, Maier SE, Boesiger P: Improved myocardial tagging contrast. Magn Reson Med 30:191-200,
252. Stuber M, Spiegel MA, Fischer SE, et al: Single breath-hold slice-following CSPAMM myocardial tagging. MAGMA
253. McVeigh ER: MRI of myocardial function: motion tracking techniques. Magn Reson Imaging 14:137-150, 1996.
254. Osman NF, Kerwin WS, McVeigh ER, Prince JL: Cardiac motion tracking using CINE harmonic phase (HARP) magnetic
resonance imaging. Magn Reson Med 42:1048-1060, 1999. Medline Similar articles
255. Kraitchman DL, Sampath S, Castillo E, et al: Quantitative ischemia detection during cardiac magnetic resonance stress
testing by use of FastHARP. Circulation 107:2025-2030, 2003. Medline Similar articles
256. Tuch DS, Reese TG, Wiegell MR, et al: High angular resolution diffusion imaging reveals intravoxel white matter fiber
heterogeneity. Magn Reson Med 48:577-582, 2002. Medline Similar articles
257. Wedeen VJ, Reese TG, Napadow VJ, Gilbert RJ: Demonstration of primary and secondary muscle fiber architecture of
the bovine tongue by diffusion tensor magnetic resonance imaging. Biophys J 80:1024-1028, 2001. Medline
Similar articles
258. Wiegell MR, Larsson HB, Wedeen VJ: Fiber crossing in human brain depicted with diffusion tensor MR imaging.
259. Tseng WY, Reese TG, Weisskoff RM, et al: Myocardial fiber shortening in humans: initial results of MR imaging.
260. Reese TG, Weisskoff RM, Smith RN, et al: Imaging myocardial fiber architecture in vivo with magnetic resonance. Magn
261. Wedeen VJ, Tseng WY, Reese TG, et al: MRI finds normal myocardial shortening and reduced cross-fiber shortening in
hypertrophic cardiomyopathy. Paper presented at the 71st Scientific Session of the American Heart Association, Dallas, 1998.
262. Chen Q, Mai VM, Bankier AA, et al: Ultrafast MR grid-tagging sequence for assessment of local mechanical properties of
the lungs. Magn Reson Med 45:24-28, 2001. Medline Similar articles
263. Muthupillai R, Lomas DJ, Rossman PJ, et al: Magnetic resonance elastography by direct visualization of propagating
acoustic strain waves. Science 269:1854-1857, 1995. Medline Similar articles
264. Muthupillai R, Rossman PJ, Lomas DJ, et al: Magnetic resonance imaging of transverse acoustic strain waves. Magn
265. Maroudas A, Thomas H: A simple physicochemical micromethod for determining fixed anionic groups in connective
tissue. Biochim Biophys Acta 215:214-216, 1970. Medline Similar articles
266. Maroudas A: Physicochemical properties of cartilage in the light of ion exchange theory. Biophys J 8:575-595, 1968.
267. Lesperance LM, Gray ML, Burstein D: Determination of fixed charge density in cartilage using nuclear magnetic
resonance. J Orthop Res 10:1-13, 1992. Medline Similar articles
23
268. Shapiro EM, Borthakur A, Gougoutas A, Reddy R: Na MRI accurately measures fixed charge density in articular
cartilage. Magn Reson Med 47:284-291, 2002.
269. Bashir A: Sodium NMR relaxation parameters in cartilage: Implications for MR imaging. In Electrical Engineering and
Computer Science. Cambridge, MIT, 1995.
270. Smith HE, Mosher TJ, Dardzinski BJ, et al: Spatial variation in cartilage T2 of the knee. J Magn Reson Imaging 14:50-55,
14 of 16 15-04-2010 20:17

271. Mosher TJ, Dardzinski BJ, Smith MB: Human articular cartilage: influence of aging and early symptomatic degeneration
on the spatial variation of T2-preliminary findings at 3 T. Radiology 214:259-266, 2000.
272. Mosher TJ, Smith H, Dardzinski BJ, et al: MR imaging and T2 mapping of femoral cartilage: in vivo determination of the
magic angle effect. Am J Roentgenol 177:665-669, 2001.
273. Bashir A, Gray ML, Burstein D: Gd-DTPA2- as a measure of cartilage degradation. Magn Reson Med 36:665-673, 1996.
274. Bashir A, Gray ML, Hartke J, Burstein D: Nondestructive imaging of human cartilage glycosaminoglycan concentration by
275. Raghunand N, Howison C, Sherry AD, et al: Renal and systemic pH imaging by contrast-enhanced MRI. Magn Reson
276. Ward KM, Aletras AH, Balaban RS: A new class of contrast agents for MRI based on proton chemical exchange
dependent saturation transfer (CEST). J Magn Reson 143:79-87, 2000. Medline Similar articles
13
277. Golman K, Ardenkjaer-Larsen JH, Svensson J, et al: C-angiography. Acad Radiol 9(Suppl 2):S507-510, 2002.
31
278. Stubbs M, Bhujwalla ZM, Tozer GM, et al: An assessment of P MRS as a method of measuring pH in rat tumours. NMR
Biomed 5:351-359, 1992.
279. Raghunand N, Altbach MI, van Sluis R, et al: Plasmalemmal pH-gradients in drug-sensitive and drug-resistant MCF-7
31
human breast carcinoma xenografts measured by P magnetic resonance spectroscopy. Biochem Pharmacol 57:309-312,
1999.
280. Raghunand N, Gillies RJ: pH and chemotherapy. Novartis Found Symp 240:199-211; discussion 265-268, 2001.
281. Raghunand N, Gillies RJ: pH and drug resistance in tumors. Drug Resist Update 3:39-47, 2000.
282. Raghunand N, Martinez-Zaguilan R, Wright SH, Gillies RJ: pH and drug resistance. II. Turnover of acidic vesicles and
resistance to weakly basic chemotherapeutic drugs. Biochem Pharmacol. 57:1047-1058, 1999. Medline
Similar articles
283. Martinez-Zaguilan R, Raghunand N, Lynch RM, et al: pH and drug resistance. I. Functional expression of plasmalemmal
+
V-type H -ATPase in drug-resistant human breast carcinoma cell lines. Biochem Pharmacol 57:1037-1046, 1999. Medline
Similar articles
284. Raghunand N, He X, van Sluis R, et al: Enhancement of chemotherapy by manipulation of tumour pH. Br J Cancer
1
285. van Sluis R, Bhujwalla ZM, Raghunand N, et al: In vivo imaging of extracellular pH using H MRSI. Magn Reson Med
41:743-750, 1999.
286. Mori S, Eleff SM, Pilatus U, et al: Proton NMR spectroscopy of solvent-saturable resonances: a new approach to study pH
effects in situ. Magn Reson Med 40:36-42, 1998. Medline Similar articles
287. Aime S, Barge A, Delli Castelli D, et al: Paramagnetic lanthanide(III) complexes as pH-sensitive chemical exchange
saturation transfer (CEST) contrast agents for MRI applications. Magn Reson Med 47:639-648, 2002. Medline
Similar articles
page 457
page 458
288. Aime S, Delli Castelli D, Terreno E: Novel pH-reporter MRI contrast agents. Angew Chem Int Ed Engl 41:4334-4336,
289. Aime S, Delli Castelli D, Fedeli F, Terreno E: A paramagnetic MRI-CEST agent responsive to lactate concentration. J Am
Chem Soc 124:9364-9365, 2002. Medline Similar articles
290. Zhang S, Merritt M, Woessner DE, et al: PARACEST agents: modulating MRI contrast via water proton exchange. Acc
Chem Res 36:783-790, 2003. Medline Similar articles
291. Aime S, Calabi L, Biondi L, et al: Iopamidol: Exploring the potential use of a well-established x-ray contrast agent for MRI.
292. MacFall JR, Charles HC, Black RD, et al: Human lung air spaces: potential for MR imaging with hyperpolarized He-3.
293. Salerno M, Altes TA, Mugler JP, 3rd, et al: Hyperpolarized noble gas MR imaging of the lung: potential clinical
applications. Eur J Radiol 40:33-44, 2001.
294. Salerno M, de Lange EE, Altes TA, et al: Emphysema: hyperpolarized helium 3 diffusion MR imaging of the lungs
compared with spirometric indexes-initial experience. Radiology. 222:252-260, 2002.
295. Golman K, Axelsson O, Johannesson H, et al: Parahydrogen-induced polarization in imaging: subsecond (13)C
angiography. Magn Reson Med 46:1-5, 2001. Medline Similar articles
13
296. Svensson J, Mansson S, Johansson E, et al: Hyperpolarized C MR angiography using trueFISP. Magn Reson Med
50:256-262, 2003.
15 of 16 15-04-2010 20:17
297. Johansson E, Mansson S, Wirestam R, et al: Cerebral perfusion assessment by bolus tracking using hyperpolarized 13C.
Magn Reson Med 51:464-472, 2004.
298. Johansson E, Olsson LE, Mansson S, et al: Perfusion assessment with bolus differentiation: a technique applicable to
hyperpolarized tracers. Magn Reson Med 52:1043-1051, 2004. Medline Similar articles
299. Golman K, Ardenkjaer-Larsen JH, Petersson JS, et al: Molecular imaging with endogenous substances. Proc Natl Acad
Sci U S A 100:10435-10439, 2003. Medline Similar articles
300. Morishita S, Sumi M, Nishimura R, et al; Carbon-13 chemical shift imaging of [1-13C] glucose under metabolism in the
rat head in vivo. Radiat Med 10:94-100, 1992. Medline Similar articles
301. Sonnewald U, Gribbestad IS, Westergaard N, et al: Nuclear magnetic resonance spectroscopy: biochemical evaluation of
brain function in vivo and in vitro. Neurotoxicology 15:579-590, 1994. Medline Similar articles
© 2010 Elsevier
16 of 16 15-04-2010 20:17
AGNETIC ESONANCE PECTROSCOPY

Chris Boesch
The aim of this chapter is an introduction into the basic vocabulary used in magnetic resonance spectroscopy
(MRS), in particular for non-spectroscopists and beginners in the field. It shall explain the basic methodological
principles of clinical MRS and show the many traps in acquisition and interpretation of MR spectra. The
chapter also aims to demonstrate the great variety of MRS with multiple nuclei, many different metabolites,
and various organs as illustrated in Figures 17-1 and 17-2.
Clinical MRS is often seen as an add-on to magnetic resonance imaging (MRI). This may be partially correct
for the diagnostic routine and for very specific examinations; however, there are many reasons why this limited
view is misleading. A responsible radiologist would never describe radiological findings either in X-ray, CT, or
in MRI, without sufficient knowledge of anatomy, pathology, and some basic principles of the respective
method. In the same way, spectroscopic data requires understanding of the basic principles of MRS, patho-
physiology, and biochemistry. Just three examples of potential malpractice: 1. Incorrect prescription of an
MRS voxel can conceal pathology and lead to a false negative result; 2. Unrecognized eddy current artifacts
may reduce or raise metabolite signals below or above normal values and thus result in false positive findings;
3. Motion of the patient between prescans, reference scans, and final spectrum can jeopardize all automatic
adjustments, leading to an incorrect, probably also automated interpretation of a ruined spectrum. Such
negative consequences for the patient can be prevented if the necessary knowledge of the method is available
and applied by the responsible staff.
page 459
page 460
Add to lightbox
Figure 17-1 31P-MR spectra of different organs: brain, heartmuscle, liver, and skeletal muscle. The spectra
have been acquired with different techniques and show clear variations in the metabolite concentrations, as is
expected since the metabolism of the various organs is different. A typical example is the resonance of
phosphocreatine (0 ppm by definition), which is very strong in skeletal muscle and absent in liver tissue.
A clinical MRS examination can be divided into several parts: 1. Selection of the specific sequence and
prescription of localization and parameters. This step is crucial and influences the outcome of an exam even
more than may be the case in morphological imaging. Different volume selection methods with their pros and
cons will be described in the following paragraphs. 2. Adjustments and acquisition of the spectra is often
1 of 2 15-04-2010 20:18
automated and rather robust; however, all automatic routines have been optimized by the manufacturers for a
specific purpose-e.g., for brain cortex or prostate. As soon as the acquisition changes slightly (e.g., a brain
spectrum is to be acquired at the skull base instead of the cortex), or, in particular, if a spectrum is acquired in
an organ other than that for which the manufacturer has optimized the pulse sequence and the prescan (e.g.,
in skeletal muscle), automatic routines may fail or at least lead to suboptimal results. Under these
circumstances, it is necessary to readjust the prescan manually, or at least to be aware that a suboptimal
prescan can jeopardize the results. 3. In addition to the final spectrum, clinical MRS uses reference scans to
adjust eddy current effects, phasing, quantitation, etc. These procedures are also quite reliable; however,
basic knowledge of the process helps to estimate and monitor, for example, effects of patient motion between
reference scan and spectrum. 4. The interpretation of a clinical spectrum is definitely more than just measuring
peak heights with a ruler. It begins with the recognition of artifacts in a spectrum that could lead to erroneous
findings. While it is self evident that a radiologist needs to identify artifacts in an MR image, it is even more
important to be aware of-often hidden-artifacts in an MR spectrum. Automated spectral fitting is an enormous
help in clinical MRS, however, the limitations of these procedures are such that a patient's diagnosis should
never rely on bare numbers only-the original spectra should always be consulted. 5. Obviously, clinical MRS
goes far beyond understanding the spectroscopic basics-it includes knowledge of specific findings, and
awareness of the sensitivity and specificity of the method to detect a specific pathology. This is not the target
of this chapter; however, many textbooks and reviews are available for further studies. 1-64
Sophisticated MRS for spectroscopists will not be covered in detail in this chapter either. Editing techniques,
polarization transfer, magnetization transfer, diffusion spectroscopy, sophisticated pulse shapes, echo-planar
chemical shift imaging (CSI), isotope enrichment, multidimensional MRS, etc. are methods with a great
potential. While they are briefly mentioned in this chapter, a number of textbooks and reviews are available
that cover these promising topics and provide a broad theoretical background of nuclear magnetic resonance
65-94
(NMR).
To summarize, this chapter aims to provide the methodological basis of spectroscopic examinations. It
explains basic terms and principles, as well as limitations, artifacts, and potential misinterpretations in MRS. In
addition, it shall give an impression of the wide range of potential MRS applications beyond N-acetyl-aspartate
(NAA) and protons. An important intention of this chapter is also to motivate clinicians to look into the
methodological aspects of MRS examinations and to recognize the potential of this method.
© 2010 Elsevier
2 of 2 15-04-2010 20:18
BASICS OF NUCLEAR MAGNETIC RESONANCE

page 460
page 461
Add to lightbox
Figure 17-2 The versatility of MRS. 1H-, 13C-, and 31P-MR spectra of human skeletal muscle reveal
complementary information on the metabolism of the organ. While 13C-MR spectra (glycogen) and
1
H-MR spectra (IMCL372) give access to the major long-term energy stores of human muscle, 31P-MR
spectra show high-energy phosphates (creatine-phosphate; ATP) that are used in the short-term energy
balance of skeletal muscle. In addition, 31P- and 1H-MR spectra can include information on the pH
1
value of the tissue, while deoxymyoglobin in H-MR spectra shows the oxygenation level of the muscle.
The spectra inserts for deoxymyoglobin and carnosine are scaled arbitrarily. (EMCL, extramyocellular
lipids; IMCL, intramyocellular lipids; PDE, phosphodiesters; Pi, inorganic phosphate; PME,
phosphomonoesters; TMA, trimethylammonium groups.
Many phenomena of MRI and MRS can be explained on the basis of the Larmor equation (also called
gyromagnetic ratio):
Equation 17-1 describes the fact that a spin with a gyromagnetic ratio γ in a magnetic field of strength B
can receive and emit radio waves of the frequency ν, the so-called "resonance frequency." In the
following, components and consequences of this relatively simple equation are discussed:
1. Nuclei in the tissues of our body can become radio transmitters and receivers if they are placed
in a magnetic field. For typical field strength used in medical applications, the frequency ν is in the
order of 100 MHz, i.e., in the typical range at which radio stations are broadcasting (Fig.17-3).
2. The frequency ν is dependent on the strength of the magnetic field B and a constant number that
is called gyromagnetic ratio γ.
1 of 6 15-04-2010 20:21
3. The gyromagnetic ratio γ is unique for each specific isotope (Table 17-1). Isotopes of a given
element have the same number of nuclear protons but differing numbers of neutrons. The values
of γ are sufficiently different that the radio waves can be separated easily (see Fig. 17-3).
4. Variations of the magnetic field strength result in variations of the resonance frequency ν. It will
be explained in the following how this can be used to separate signals spatially (spatial
localization) and chemically (spectral selection).
Table 17-1 lists characteristic parameters of selected nuclei and explains the enormous differences in
the sensitivity of different nuclei and metabolites. All isotopes in Table 17-1 are stable, i.e., they do not
decay and do not emit radioactive radiation. While MRI is mainly using the signal from 1H (hydrogen with
1 proton and no neutrons, thus also called "protons"), MR spectroscopy is also observing other nuclei,
such as 13C (carbon-13 with 6 protons and 7 neutrons) or 31P (phosphorus-31 with 15 protons and 16
neutrons). Despite the fact that the signal acquisition of these nuclei is based on the same physical
effect, their measurement requires very specific and adapted techniques. For example, it will be shown
1
in a later paragraph why H-MRS needs water suppression and uses almost exclusively gradient-based
volume selection, while C-MRS and 31P-MRS often apply pulse-and-acquire techniques with
13
simultaneous irradiation of the protons, so-called decoupling.

page 461
page 462
Add to lightbox
Figure 17-3 The spectrum of electromagnetic waves. A, Penetration depth of electromagnetic waves in
biological tissue. Electromagnetic waves at very low frequencies (left) and at very high frequencies
(right) can penetrate the body such that information about the interior of our body can be obtained in two
"biological windows." A third, limited, access to the interior in the infrared region can be used to
measure blood oxygenation. B, Typical representatives of the electromagnetic spectrum are radio
waves used for broadcasting, microwaves for cooking and RADAR, infrared-, visible-, and ultraviolet-
light, X-rays used in radiology, and γ-radiation used in nuclear medicine. Electromagnetic waves have
an increasing specific energy with increasing frequency, i.e., the specific energy of radio waves is
unable to break chemical bonds specifically (they are non-ionizing), while the energy of X-rays and
γ-rays is sufficient to break a chemical bond, even at very low levels of totally absorbed energy. C, The
2 of 6 15-04-2010 20:21
Larmor equation has attractive consequences-the MR effect allows tissue in our body to become a
radio receiver and transmitter if it is placed in a strong magnetic field. MR signals can be found in the
very common range of radio signals that are used for broadcasting. This illustrates how the resonance
frequencies of different nuclei are clearly separated, thus allowing an unequivocal observation. D, MR
spectroscopy resolves small variations of the resonance frequency of one isotope caused by the
chemical environment in the molecules. The resulting spectra cover a very small range of frequencies in
the order of parts-per-million (ppm). (Adapted from Boesch C: Molecular aspects of magnetic
resonance imaging and spectroscopy. Mol Aspects Med 20:185-318, 1999. © 1999, with permission
from Elsevier.)
Table 17-1. Characteristic Magnetic Resonance Parameters of Selected Nuclei

Inherent
sensitivity
at Typical
Frequency constant concentrations
at 1.5 field Natural of nuclei in Typical
Spin tesla (relative abundance biological overall
1
Nucleus/Isotope Name number (MHz) to H) (%) tissue (mM/L)* sensitivity*
1 Hydrogen 1/2 1 100 In water 1
H 63.87 99.985
(protons) => MRI
0.02 In 2 × 10-4
metabolites
2
H Deuterium 1 9.8 0.00965 0.015 100 In water 1 × 10-6
0.02 In 3 × 10-10
metabolites
12 Carbon 0 no NMR 0 100 In adipose 0
C 98.89
signal tissue
13 Carbon 1/2 0.0159 100 In adipose 2 × 10-4
C 16.06 1.108
tissue
0.05 In -7
1 × 10
metabolites
14
N Nitrogen 1 4.61 0.00101 99.63 2 2 × 10-5
15 Nitrogen 1/2 0.00104 -7
N 6.46 0.37 2 1 × 10
16 Oxygen 0 no NMR 0 0
O 99.96 50
signal
17
O Oxygen 5/2 8.66 0.0291 0.037 50 6 × 10-6
19
F Fluorine 1/2 60.08 0.833 100 0.001 1 × 10-5
23 Sodium 3/2 0.0925 -5
Na 16.89 100 0.05 5 × 10
31
P Phosphorus 1/2 25.85 0.0663 100 0.1 5 × 10-5
35
Cl Chlorine 3/2 6.26 0.0047 75.53 0.03 2 × 10-6
39 Potassium 3/2 0.00051 -7
K 2.98 93.08 0.05 3 × 10
page 462
page 463
*The "typical concentrations of the nuclei" and the "typical overall sensitivity" are rough estimations to
illustrate the order of magnitude only. The "typical overall sensitivity" for nuclei and metabolites that are
typically observed by in vivo MR spectroscopy is drastically lower than the sensitivity of water in 1H-MRI
as it is shown in the top line.
3 of 6 15-04-2010 20:21
The spin number in Table 17-1 will not be explained in detail; it should just point toward additional
complexity of some nuclei with spin numbers above ½ that leads to complex spectral and relaxation
behaviors. It also shows that some isotopes have no spin and thus cannot be observed by nuclear
magnetic resonance at all-e.g., the most abundant carbon isotope 12C. For such elements, it is
necessary to select another isotope e.g. 13C that amounts to about 1% of all naturally occurring carbon
in our body. Unfortunately, the much lower natural abundance results in a reduced overall sensitivity. On
the other hand, a natural abundance far below 100% represents an opportunity since molecules can be
labeled by the rare isotope and the label can be followed specifically along metabolic steps in vivo, e.g.,
13 95-102
infused C-labeled glucose or acetate can be observed in the tricarboxylic acid cycle.
The column "Frequency at 1.5 tesla" in Table 17-1 and Fig. 17-3 illustrates how different nuclei emit
radio waves at completely different frequencies and can thus be distinguished easily. Using the Larmor
equation (Eq. 17-1), it is straightforward to calculate resonance frequencies at other field
strengths-e.g., 340 MHz for 1H and 85 MHz for 13C at 8 Tesla, a magnetic field strength that is currently
the upper limit for experimental whole-body magnets. Higher resonance frequencies-either due to a
stronger magnetic field or a larger γ (Table 17-1)-result in a higher inherent sensitivity, generated by the
induction of currents in the receiver antenna.79,103-111
Typical concentrations of metabolites in human tissue are listed for illustration only; other metabolites
may have much lower concentrations. Just as an example: while major metabolites in the brain are in
the order of 10 mmol/L, the concentration of an important metabolite, phenylalanine, is in the order of
300 μmol/L,112 thus requiring optimized and adapted acquisition strategies.113 A potential advantage of
19
low natural concentrations can be found for F where anesthesia agents can be observed without
background signal from naturally occurring 19F in human tissue87,114-117
All the factors in Table 17-1 contribute to the typical overall sensitivity of a specific nucleus and
1
metabolite concentration shown in the rightmost column. If an overall sensitivity of 1 is assigned to H in
water, one finds outrageous differences in the overall sensitivity of other nuclei and other metabolites,
going down by up to ten orders of magnitude-e.g., for deuterium in typical metabolites of the human
body. While such a low overall sensitivity is prohibitive for the acquisition of a classic MR spectrum, it
does not imply that low-sensitivity nuclei are of no value for MRS or MRI. Using isotope labeling and
high-concentration substances-e.g. deuterated water-low-sensitivity nuclei can provide significant
information about the tissue. The trend towards higher static magnetic field strengths as mentioned
above will additionally help to overcome some of the difficulties with low-sensitivity nuclei in the near
future.79,107,108,111,118-128
Table 17-1 also illustrates an inherent disadvantage of MRS versus MRI. While MRI typically observes
1
H in water, with a total concentration of approximately 100 M, MRS receives signals from metabolites
with much lower concentrations. Even if both modalities observed the same nucleus 1H, Figure 17-4
illustrates the difference in signal strength of MRI and MRS; in order to obtain similar signal amplitude,
MRS would need to acquire signals of typical metabolites from a much larger volume than MRI that
receives signal from 1H in highly concentrated water. In other words, the spatial resolution of
(metabolite-) MRS is inherently lower than that of (water-) MRI.
4 of 6 15-04-2010 20:21
Add to lightbox
page 463
page 464
Figure 17-4 The sensitivity difference of MRI and MRS. MRI (left) with typical water concentrations of
110 Molar, and MRS (right), which observes metabolites in much lower concentrations. In order to
obtain the same signal intensity, MRS would have to collect signals from a much larger voxel (cubic
volume). It is often overlooked that the volume increases in the third power, i.e., while the total volume for
metabolites has to be about 5000-fold that for water, the voxel length is only about 17 times longer. On
the other hand, a reduction of the voxel length by a factor of two reduces the volume and the signal
intensity by almost an order of magnitude (23 = 8), i.e., a slight increase in spatial resolution in all three
directions decreases the signal intensity dramatically.
While the value of γ for various isotopes leads to very distinct resonance frequencies, two additional
effects result in much smaller, yet essential, variations of the magnetic field strength. These two effects
are an immediate consequence of the Larmor equation
and are in fact the basics for MRI as well as for MRS. It shows that any change of the magnetic field
strength B + ∆B leads to a variation of the resonance frequency ν + ∆ν that can be detected. Variations
of the magnetic field strength B can be the result of two major mechanisms (Fig. 17-5):
1. Additional currents in specifically designed (orthogonal) coils lead to spatial variations of the
magnetic field strength, to so-called gradients. If such gradients are switched on, the resonance
frequency is dependent on the location of the spin, an effect that can be used for spatial
encoding. MRI and gradient-localized MRS use this effect of gradients for the generation of
images, for the selection of slices, for the definition of a sensitive volume (voxel) in MRS, and for
other effects. Figure 17-6 shows that in a homogenous field all spins in a sample resonate at the
same frequency, while, after the application of a gradient, spins in the right tube resonate at a
higher frequency and become distinguishable from the spins in the left tube.
2. Chemical bonds in a molecule (i.e., the glue of chemical substances) consist of electrons that are
shared between the participating nuclei. These electrons shield the magnetic field such that the
nuclei experience a weaker magnetic field strength than if they were fully exposed to the applied
field. However, the distribution of the electrons is unequal, resulting in variations of the resonance
frequency according to the Larmor equation as discussed in the next section.
5 of 6 15-04-2010 20:21
Add to lightbox
Figure 17-5 Two mechanisms that change the magnetic field strength spatially or locally. While
technically generated gradients (upper row) vary the magnetic field strength by several mT (10 mT/m is
a moderate gradient in up-to-date MR systems), spectroscopy must deal with much smaller variations
of the magnetic field by the chemical environment in a molecule, typically in the range of several parts-
per-million (ppm). Since the offset of spectra is arbitrarily set to a reference substance, 1.50 000 00
tesla is set to this position in the spectrum for didactic purposes.
© 2010 Elsevier
6 of 6 15-04-2010 20:21
CHEMICAL ENVIRONMENT, CHEMICAL SHIFT

Figure 17-7 illustrates how 1H-nuclei in a molecule experience different magnetic field strengths
depending on their chemical environment. Since the oxygen atom to the left attracts the shared
electrons much more than the carbon and hydrogen atoms to the right (this attraction is called electron
negativity), the 1H nucleus bound to the C=O group (left side) is much less shielded than the 1H
nucleus within the CH3 group on the right. The Larmor equation can be adapted such that a "shielding
constant" σ describes the effect of the shielding electrons:
In other words, if the electrons shield a nucleus effectively-i.e., if σ is (relatively) large-the resonance
frequency ν will decrease accordingly. On the other hand, if a nucleus is bare of surrounding electrons
due to a highly electronegative chemical partner, σ will be small and the resonance frequency ν
proportionately higher.
Add to lightbox
Figure 17-6 The effect of gradients-spatial variations of the magnetic field strength. In a magnetic field
of homogeneous strength (left column), all spins in the sample resonate at the same frequency and are
indistinguishable. If a field gradient is applied (right column), i.e., if the magnetic field strength
increases along one spatial direction, the resonance frequencies become distinguishable and a
projection of the sample is obtained. This is the original idea that led finally to MRI and to the Nobel
Prizes for Lauterbur181 and Sir Peter Mansfield386 in 2003. In turn, if a gradient is applied to the
sample as shown to the right, a pulse with a limited bandwidth (i.e., with a limited range of frequencies)
excites just a slice of the sample. This slice selective excitation is a well-known technique in MRI and
MRS. (Courtesy of M. Ith)
According to this description, the unshielded 1H bound to C=O will have a higher resonance frequency
than the 1H within the CH3 group in Figure 17-7. These variations of the resonance frequency based on
the chemical environment are called chemical shifts.
page 464
page 465
1 of 7 15-04-2010 20:23
Add to lightbox
Figure 17-7 1H-nuclei in a molecule experience different magnetic field strengths, depending on their
chemical environment. Since the oxygen atom to the left attracts electrons much more than the carbon
1 1
to the right, the H nucleus bound to the C=O group is much less shielded by electrons than the H
1
nuclei within the CH3 group on the right. According to the Larmor equation, the unshielded H bound to
1
C=O will have a higher resonance frequency than the H within the CH3 methyl group. The terms
"up-field" and "down-field" are not in contradiction to Figure 17-5 since they are not addressing the
field strength experienced by the nuclei but are historically justified by the techniques that have been
applied these days (sweeping fields).
2 of 7 15-04-2010 20:23
Add to lightbox
page 465
page 466
Figure 17-8 A hypothetical spectrum to illustrate the vocabulary used in MRS (bottom row). The
chemical shift is a result of the chemical environment and is measured relative to a common reference
substance. It is expressed in parts-per-million (ppm) of the resonance frequency, a unit that is
independent of the magnetic field strength. Under relaxed conditions, the area of a resonance line is
proportional to the number of observed nuclei, i.e., to the concentration of a substance. The line width
of a signal depends on T2*, which depends on the transversal relaxation time T2 of the chemical group
and on the homogeneity of the magnetic field ("shim"). Coupling of resonance lines leads to split
"multiplets," depending on the coupling partner. In the upper row, the acquisition of an MR spectrum is
explained, mainly for didactic and historical reasons. While continuous wave acquisition is more
intuitive and was used in the beginning of nuclear magnetic resonance (NMR), almost every spectrum
nowadays is acquired by Fourier techniques.180
Figure 17-8 shows a hypothetical spectrum to illustrate the vocabulary used in MRS. The chemical shift
is a result of the chemical environment and is characteristic for a specific type of chemical substance.
However, alterations of the magnetic field strength by the chemical environment are very small (see
also Fig. 17-5), typically in the range of one part per million (ppm). Since it would be inconvenient to
measure the resonance frequency in absolute terms-e.g., 63.866224 MHz as shown in Fig. 17-5-the
difference of the resonance frequency (the chemical shift), is indicated in ppm from a reference
substance at 0 ppm (see Fig. 17-8). Since the chemical shift of a specific substance is well defined
under identical conditions (i.e., without variations of temperature, pH, and/or ionic strength of the
solution), the position on the chemical shift axis can be used to identify a chemical group or at least to
narrow the list of potential candidates. The relative unit "ppm" has the additional advantage that
chemical shifts expressed in ppm are field independent, for example, the methyl group in NAA is found
at 2.02 ppm independent of the field strength at which the spectrum has been measured.
Under relaxed conditions, the area of a resonance line is proportional to the number of observed nuclei,
i.e., to the concentration of a substance. The line width of a signal depends on T2*, which itself
depends on the transversal relaxation time T2 of the chemical group and on the homogeneity of the
3 of 7 15-04-2010 20:23
magnetic field (shim). In other words, the amplitude of the signals may change as result of an
increasing line width and is, therefore, not proportional to the concentration of a metabolite.
While the ppm scale is extremely accurate, MR spectra traditionally have no y-axis, unlike other
x-y-plots. In principle, one could add a voltage scale to it, however, MR signals detected from the
sample and amplified by many stages of electronic equipment have an arbitrary amplification factor,
and it requires careful quantitation procedures to translate this arbitrary amplification into absolute
values. While only a few MR imaging procedures require absolute intensities, and high-resolution NMR
spectra deal much more with couplings, cross peaks, etc, in vivo spectroscopy would benefit from
absolute values of signal intensities. For quite a while, ratios of signal areas from different metabolites
have been used to overcome this problem. However, in recent years, many successful attempts to
obtain absolute values have been made128-151 in order to make results comparable inter-individually,
between patients and healthy volunteers, and between different publications and methods. Such an
absolute quantitation goes far beyond adding a voltage scale to the amplitude of a spectrum for
several reasons: 1. the area and not the amplitude is proportional to the concentration of a metabolite;
2. not all resonance lines would require the same scaling, since multiple nuclei in one metabolite can
contribute to a peak, e.g., three nuclei in a methyl group; and 3. resonance lines depend on many other
parameters such as relaxation times. Absolute quantitation will be discussed later in this chapter.
Add to lightbox
Figure 17-9 A spectrum of ethanol in vitro. In the lower row, obtained by a short-echo point-resolved
1
spectroscopy (PRESS) sequence at 1.5 tesla, three H-groups in ethanol (CH3-CH2-OH) can be
distinguished and the splitting of the resonance lines is illustrated, a result of the so-called spin-spin
coupling. The inserts demonstrate heteronuclear spin-spin coupling between 13C nuclei and the
observed protons. The comparison of the two spectra demonstrates the effect of increasing strength
and better homogeneity of the magnetic field. The upper row represents the quality of spectra that
387
have been obtained at much lower field strength and field quality when Arnold and Liddel and
388
Ramsey observed the phenomenon of the chemical shift for the first time in history. For practical
reasons, the spectrum in the upper row has been obtained by mathematical treatment (apodization) of
4 of 7 15-04-2010 20:23
the lower spectrum in this example, however, lower field strength or homogeneity (insufficient
shimming) would result in the same spectrum. (Adapted from Boesch C: Molecular aspects of
magnetic resonance imaging and spectroscopy. Mol Aspects Med 20:185-318, 1999. © 1999, with
permission from Elsevier.)
Some resonance lines, such as the methyl group in ethanol (Fig. 17-9), are split and appear as
1
multiplets due to so-called spin-spin coupling. Coupling can be explained as follows: if a H nucleus has
a coupling partner of spin ½, this coupling partner can have a spin in either the up or down state. While
the first state decreases the field strength at the location of the observed 1H, the later increases the
field. Since the probability of an up spin and down spin are almost the same, we can expect that the
1
resonance frequency of 50% of the observed H are slightly shifted to the right while the other 50%
are slightly shifted to the left, resulting in a doublet. We can now extrapolate these combinations to two
coupling partners, resulting in a triplet of 1:2:1 as it can be seen at 1 ppm in Figure 17-9, or to more
coupling partners as seen in Figure 17-8. Figure 17-9 illustrates in addition heteronuclear coupling with
the insert at 0 and 2 ppm, showing small mirror images of the larger triplet. The origin of these
satellites75 can be explained on the basis of Table 17-1, where we can see that 1 in 100 carbon nuclei
12 13
is not a C without a spin, but a C with spin ½. In other words, 1 in 100 methyl groups is connected
to a 13C with spin 1/2, resulting in the very same splitting as we have seen between protons, i.e., in the
homonuclear case. However, since only 1% of the methyl groups experiences this splitting, 99% show
13
the typical triplet and the two satellites represent those methyl groups that are bound to a C atom.
This fact can be used for heteronuclear editing of 1H spectra.96,97,152-156
page 466
page 467
Add to lightbox
1 31 13
Figure 17-10 H-, P- and C-MR spectra of human skeletal muscle. The same scaling of the
chemical shift axis was used for all nuclei, indicating the enormous differences in the chemical shift
dispersion, i.e., the range of chemical shifts covered by biological molecules. This range is typically 10
1 31 13 1
ppm for H, about 25 ppm for P, and up to 200 ppm for C-spectra. In H-MRS, deoxymyoglobin at
78 ppm is an exception, since it is shifted far outside the common spectrum by paramagnetic effects
and so is visible, whereas it would be covered by other resonances without this effect.
Figure 17-10 illustrates the range that is typically covered by different nuclei. While the major
5 of 7 15-04-2010 20:23
metabolites in a 1H spectrum cover a range of about 10 ppm, 31

P metabolites extend over a range of
13
about 25 ppm, and C metabolites over about 200 ppm.
The chemical shift dispersion of the various nuclei plays a major role in volume selection and in
acquisition techniques, as will be shown later. Chemical shift artifacts, displacements of selected
voxels, bandwidth of radiofrequency pulses, and requirements for the gradient strength are the major
negative consequences of larger chemical shift dispersion. On the other hand, smaller chemical shift
dispersion leads to a crowded spectrum where almost only singlets are observable without
considerable overlap from other resonances.
1
A few metabolites in a H spectrum, such as aromatic and exchangeable protons around 7 to 10
ppm112,113,157-162 and deoxymyoglobin at 78 ppm163-167 are outside the small ppm range mentioned
above, however, these metabolites are not yet observed in routine MRS examinations.
Add to lightbox
Figure 17-11 The spectral changes that occur if two species of molecules (A and B) are in exchange.
Depending on the exchange rate, either two distinct lines are visible (slow exchange) or an average of
the two positions is visible (fast exchange). pH titration is one case of fast exchange, as shown to the
right, where the resonance line of inorganic phosphate (Pi) shifts its position with the intracellular pH
value. It represents an average position of the acidic and basic forms of the phosphate ion, since the
two species exchange rapidly such that the MR experiment just can observe an average position. This
titration curve can be used to determine the intracellular pH value noninvasively. Other mechanisms for
titration of resonance lines are ionic strength of the solution, e.g., ATP-resonances that are dependent
2+
on the Mg concentration.
It has already been mentioned that the chemical shift of a molecule is extremely well defined, such that
the position of a resonance in a spectrum can be used to identify a compound. However, while this is
true for stable conditions-i.e., defined pH, temperature, molecular shape etc.-the chemical shift can be
altered by different conditions, in particular, if two or more species of molecules exchange rapidly (Fig.
17-11). If nuclei exist in two molecules that interconvert after a certain time, the resonance frequency
of these nuclei depends on the exchange rate, i.e., on the speed at which the two molecules exchange.
If two species exchange slowly compared with the timescale of an MR experiment, the MR spectrum
reveals just the relative amount of the two substances (see the left column of Figure 17-11), i.e., in MR
terms, these are two distinct substances with two different resonance positions. If the exchange is fast
compared with an MR experiment, the two separate lines disappear (see the middle column of Figure
17-11). All nuclei exchange fast enough to experience both states, just with varying probability,
depending on the relative fraction of the two states. This results in one line with constant amplitude, but
6 of 7 15-04-2010 20:23
with moving position depending on the relative fraction of the two states. As soon as the MR timescale
and exchange rate are comparable, the spectra become more complicated than described above.
Specialist literature is recommended for further details.75,81
page 467
page 468
An important and very practical example of fast exchange in in vivo spectroscopy is the exchange of
acidic and basic forms of molecules. As shown in Figure 17-11, the resonance shift of such pairs
- 2-
reveals the pH value of the environment, e.g., the pair H2PO4 and HPO4 contributes to the inorganic
31
phosphate signal in P-MRS and allows a very accurate measurement of the intracellular pH. With a
pH of 6.77, the resonance peak of inorganic phosphate shifts from 3.23 ppm to 5.70 ppm. Therefore,
the chemical shift difference between phosphocreatine and inorganic phosphate (see Fig. 17-10)
allows the calculation of the pH with an accuracy of about 0.05 units. 168-171 Similarly, pH dependent
1
resonance shifts can be observed in H-MRS of skeletal muscle, where the protons in the histidine
residue of carnosine titrate, revealing the intracellular pH.159-161
Other applications of exchange mechanisms are the Mg2+-dependent shift of ATP in 31P-MRS171-174
or-on a much faster timescale and using other techniques-the measurement of the enzymatic flux
175-179
through the creatine phosphokinase reaction.
© 2010 Elsevier
7 of 7 15-04-2010 20:23
DATA ACQUISITION: TIME DOMAIN VERSUS FREQUENCY DOMAIN

Intuitively, one would expect that spectra be obtained by sweeping the frequency range like a radio
receiver as shown in the right part of Figure 17-8. This so-called continuous wave (CW) technique
180
was used for more than a decade before Ernst and Anderson suggested the introduction of Fourier
transform (FT) techniques into NMR. For this seminal idea and in particular for the consequential
development to multiple dimensions, including "Fourier imaging," Richard Ernst received the Nobel Prize
for Chemistry in 1991. Fourier techniques acquire data in the time domain-i.e., the precession of the
magnetization in a sample is observed over a certain time and all contributing frequencies are sorted
out by the Fourier transformation. Since the whole range of frequencies is acquired simultaneously, this
technique is much faster and much more powerful than the historical and intuitive CW technique. This
transition from CW to FT techniques found its analog situation in imaging, when the original projection
181 182
reconstruction was replaced by the currently-used Fourier imaging.
Add to lightbox
Figure 17-12 The relationship of the "time domain" and the "frequency domain" in MR spectroscopy.
While data is acquired in the time domain by recording the signal emitted by the sample, Fourier
transformation sorts out all contributing frequencies and displays the occurrence of the various
frequencies in a spectrum. The top row shows a hypothetical signal with a single frequency of infinite
duration, resulting in an infinitely sharp line in the spectrum. The following rows illustrate how variations
in the time domain change the spectrum in the frequency domain. A short signal in the time domain
corresponds to a broad line in the frequency domain, while a longer decay better defines the
frequencies, i.e., results in a sharper line.
As shown in Figure 17-12, Fourier transformation relates a time domain with a frequency domain-i.e.,
it connects data acquired in a coil with data presentation in a spectrum. Figure 17-12 also
demonstrates how variations or manipulations of the acquired data influence the spectrum. An infinite
radio wave of constant frequency and amplitude would result in exactly one infinitely narrow line in the
1 of 9 15-04-2010 20:25
spectrum. This does not occur in MR since the free induction decay (FID) has a limited lifetime,
resulting in finite line width of the resonance line in the spectrum. Now, if the FID is shortened by an
inhomogeneous magnetic field (insufficient shimming) or by multiplication with a decaying function
(apodization), the FID gets shorter and the resonance line broader since more frequencies are
present. The fourth row in Figure 17-12 shows a synthetic FID of multiple lines and its corresponding
spectrum. While it is helpful to recognize the time domain primarily as data that is acquired in the coil
and digitized in analog-digital converters, it is important to expand and to generalize this view. Fourier
transformation and mathematical treatment in the time domain can smooth data or increase resolution
independent of the original data acquisition. This is a very common process and is widely used, from
astronomy and geology to the restoration of old movies. In MRS, understanding the duality of time and
frequency domain is important for the design of pulse sequences, quantitation of spectra, and for the
data handling as shown in Figure 17-13. An acquired FID can be multiplied by various mathematical
functions in order to enhance either the first part that is dominated mainly by broad lines with high
signal-to-noise or to augment the later part that is dominated by the sharper lines and by noise. The
middle row shows an FT of the untreated signal. In comparison, multiplication with an exponential
function, called apodization, leads to a suppression of the noise and to a preference of broad lines-i.e.,
the spectrum gets a better signal-to-noise ratio (SNR) at the cost of broader lines and a decreased
resolution. On the other hand, multiplication with a function that suppresses the first data points in the
time domain results in a resolution enhancement at the cost of a reduced SNR.
page 468
page 469
Add to lightbox
Figure 17-13 The effect of "apodization"-a mathematical treatment of MR signals that are typically
acquired in the "time-domain" (left) and represent a measurable voltage induced in a radiofrequency
coil by the precession of the macroscopic magnetization. A Fourier transformation sorts out the
frequencies that contribute to the time-domain signal, producing a spectrum in the frequency domain
(right column). Apodization changes a spectrum specifically: broad resonances with high signal-
to-noise ratios contribute mostly to the first part of a free induction decay, while the end is dominated
by smaller lines and by noise. Multiplication with appropriate functions can enhance either resolution or
signal-to-noise, i.e., sensitivity as it is demonstrated in the inserts.
2 of 9 15-04-2010 20:25
Add to lightbox
Figure 17-14 A synthetic data set (left column) and an in vivo brain spectrum (right column) with
different phase settings (0 = appropriate setting, 22.5, and 45 degrees). The raw data for the
respective spectra are identical, i.e., just the phasing is different. The figure illustrates that the line
height obviously is not an appropriate measurement for concentrations or ratios. It also demonstrates
that the baseline is a crucial factor for the determination of resonance areas. While it is easy to phase
isolated lines correctly, it is much more complicated in a crowded in vivo spectrum. Current software
on commercial scanners acquires reference scans with unsuppressed water or uses fitting algorithms
to phase the spectra quite reliably. However, these algorithms can fail in moving organs (e.g., heart,
liver), if the patient moves during the scan, or if artifacts interfere with the algorithms.
NMR signals are radio waves that are characterized by amplitude, frequency, and phase. The later is a
physical property that typically is difficult to understand wherever it occurs in MR, independent of MRI
or MRS. It describes the relative timing of two oscillations, similar to two runners in a stadium who run
the same radius (= same amplitude) with the same speed (= same frequency), however, several
meters apart (= different phase). While this explanation is correct in principle, it has limited value to
understand the effect of the phase in MRI or MRS. The phase of a spectrum can be described in a
more visual way, as seen in Figure 17-14, which shows how the phase defines shape and symmetry of
the resonance lines. In the left column of Figure 17-14, a simulated spectrum is shown with an
appropriate setting of the phase in the lowermost row. If other values are used for the phase, the lines
become distorted and unsymmetrical. This is obvious in isolated lines, however, not in crowded spectra
like the brain spectra on the right that are adjusted in the same way as the synthetic spectra on the
left. On the other hand, the wrong phasing can change metabolite peaks and baseline considerably,
such that other amplitudes and areas are reported. In modern MR systems, the phase of 1H-spectra is
adjusted by the acquisition of an additional spectrum where the water signal is used to adjust the
phase. In addition, up-to-date quantitation algorithms can consider the phasing as well. While all these
are relatively robust procedures, they can fail if the reference scan and spectrum are not acquired
under identical conditions, for example, if the patient moves during the examination. This can result in a
wrong phasing of the metabolite and in subsequent errors in the automatic quantitation. In addition,
spectra of X-nuclei have no "water signal" that can be used to adjust the phase. Amazingly, spectra
with badly adjusted phases do appear in publications from time to time, leading to a mistrust of the
overall quality of the spectroscopic data therein.
3 of 9 15-04-2010 20:25
Methods that observe the signal immediately following the excitation by a radiofrequency pulse are
called pulse-and-acquire techniques. As is shown below, these techniques have specific advantages
for X-nuclei, that is, for non-proton nuclei with low sensitivity, large chemical shift dispersion, short
relaxation transversal times, and other reasons. Pulse-and-acquire techniques observe the FID that is
generated by any pulse that is not a multiple of 180 degrees as shown in Figure 17-15.
page 469
page 470
Add to lightbox
Figure 17-15 All the detectable signals (free induction decays, spin-echoes, and stimulated echoes)
that are produced by three subsequent radiofrequency (RF) pulses. Hahn described these signals in a
183
seminal paper in 1950. If the three pulses are combined with gradients, i.e., with variations of the
magnetic field strength, the three spatial dimensions can be used to select a specific volume in space,
a so-called voxel. Every radiofrequency pulse that does not have a pulse angle of exactly 0 or 180
degrees generates a free induction decay (FID). Two pulses can, in addition, be source for a
spin-echo (SE). Three pulses can either re-form a spin-echo or generate a stimulated echo (STE). In
volume selection sequences, it is crucial to destroy all unwanted signal combinations by so-called
crusher gradients.
In contrast to the acquisition of X-nuclei signals, most clinical 1H-MRS examinations apply
gradient-based volume selection methods nowadays. Figure 17-15 lists all detectable signals that are
produced by three subsequent radio pulses and that can be used to select signals from a well-defined
volume. The number of three is in fact not arbitrary since the combination of three pulses with
gradients, i.e., with variations of the magnetic field strength, can be used to select the three spatial
dimensions of a desired volume in space, a so-called voxel (see below). In a seminal publication in
1950,183 Hahn described all signals generated by three pulses, including FID, spin-echoes, and
stimulated echoes. Every pulse can be the source of an FID unless it is a perfect multiple of 180
degrees, which is technically very difficult or even impossible in volume selection sequences with
modulated pulses. Two pulses can generate a spin-echo as shown in Figure 17-16. An additional pulse
can refocus the spin-echo of two pulses, leading to another spin-echo that is created by all three
pulses. The "stimulated echo" is another echo that is formed after three pulses. It uses the
magnetization present in the z-direction after the second pulse and leads to an echo with maximally
50% amplitude if all three pulses are perfectly at 90 degrees. For the selection of a sensitive volume, it
is crucial to avoid all unwanted signal contributions by so-called crusher gradients, i.e., by gradients
that dephase all signals but rephase only those that are generated by the correct pulses.
The fact that magnetization dephases with time is a well-known mechanism in MRI as well as MRS.
4 of 9 15-04-2010 20:25
Figure 17-17 demonstrates the effect of increased echo times (TE). Unlike MRI, where the contrast is
mainly based on a specific effect of the relaxation times T1 and T2, MRS aims at the concentrations of
metabolites-relative or absolute-and any influence of the relaxation times should be avoided. Since
concentrations are related to the signal area in the fully relaxed longitudinal state (i.e., repetition of the
experiments is very long; TR [Gt ] T1) and without influences of the transversal relaxation rate T2,
sequences with short TE are preferred (i.e., TE [Lt ] T2). In addition, short TE sequences reveal much
more information as seen in Figure 17-17. Acquisition of short TE spectra was not feasible until about
a decade ago, when gradients systems improved.184-187 This historical development is one reason why
spectra are still acquired at long TE. On the other hand, while the acquisition of short TE spectra is
possible on up-to-date MR systems now, Figure 17-17 shows that there are also disadvantages of
short TE spectra. Such spectra are crowded, and it is difficult to determine the baseline. In turn,
spectra at long TE reveal just a few singlets, which can be quantified easily. However, the use of short
echo time spectra is now state-of-the-art and modern techniques of spectra deconvolution and fitting
135,145,148,188-197
help to overcome the separation of crowded spectra.
page 470
page 471
Add to lightbox
Figure 17-16 The formation of multiple spin-echoes and the signal decay in a spin-echo experiment.
Following a 90-degree pulse (A), the transverse magnetization begins to dephase, i.e., depending on
the exact resonance frequency, some spins are faster while others are slower. A 180-degree pulse
around the axis x' mirrors the spins (B), which can rephase in (C), i.e., slower spins are in advance
while faster spins are behind. If all spins experience the same magnetic field strength over the time
period A-D, the signal would be fully recovered. However, stochastic (arbitrary) processes lead to
fluctuations of the local magnetic field in a molecule that cannot be compensated for and lead to a
remaining dephasing and signal reduction in D. The lower trace illustrates recoverable and
unrecoverable dephasing-dephasing that can be reversed leads to a T2* decay and to the formation of
an echo, while dephasing related to stochastic processes cannot be recovered and leads to the T2
decay of an echo train. (Courtesy of M. Ith)
5 of 9 15-04-2010 20:25
Add to lightbox
Figure 17-17 Spectra of a human brain at different echo times (TE). The spectrum at TE 20 ms
contains numerous resonances and abundant information, however, this is at the cost of decreased
signal separation and a badly defined baseline. At long TE, the spectrum is clear, the baseline is flat,
and an interpretation is easier, but at the cost of reduced information content.
Add to lightbox
page 471
page 472
1
Figure 17-18 The decomposition of a cerebral H-MR spectrum into baseline (BL) and metabolites of
interest (MOI) based on saturation recovery data.198 The macromolecule baseline in 1H-MRS alters
the determination of metabolite concentrations considerably. (Courtesy of L. Hofmann and R. Kreis)
6 of 9 15-04-2010 20:25
Add to lightbox
Figure 17-19 Artifacts in the spectrum that result from eddy currents in the MR system (coil, cover,
shields). Since the unsuppressed water signal experiences the same effects, an appropriate
correction of the metabolite signals with the phase information of the water resonance is possible and
robust. However, this requires that the reference scan and final spectrum be acquired under the same
conditions-patient motion between or during the measurements can jeopardize the correction and lead
to unpredictable and concealed artifacts. (Courtesy of R. Kreis).
The separation of the baseline from the metabolite signals is a demanding process that influences
spectral quantitation considerably.135,198-201 Figure 17-18 shows one attempt to separate the signals
from tissue components with short T2 that contribute to the broad baseline. Based on a T1 separation,
198
the baseline can be measured in a series of experiments. Other attempts use a mathematical
135,201
deconvolution of baseline and metabolite signals or apply inversion recovery sequences.199,200 An
accurate determination of the baseline is mandatory, yet often underestimated. An inappropriate
determination of the baseline can easily change absolute quantities and ratios to an extent that
questions the findings of a paper or a report.
Figure 17-19 shows how eddy currents can destroy spectral quality. Eddy currents are induced by
switched gradients in the structures of an MR magnet, in particular in main coil and dewar, surviving
the intended effect of the gradient and introducing instabilities of the field strength. Eddy currents in
modern MR systems are drastically reduced by engineering solutions, 184-187 and the remaining effects
can be corrected for,202-210 for example by a separate acquisition of the water signal. Since the decay
of the water signal is known and phase shifts (see above) during the decay can be attributed to eddy
currents, a successful correction of the water signal (left column in Figure 17-19) can be transferred to
the metabolite signals. Similar to other automatic correction procedures in MRS, this is a robust
procedure; however, motion of the patient during the reference scan can lead to erroneous corrections
and a spectroscopist needs to recognize such artifacts.
7 of 9 15-04-2010 20:25
Add to lightbox
Figure 17-20 This shows the enormous difference in amplitude between the water signal (which is
mainly used in MRI) and the signals from metabolites, which are about 10,000 times weaker. The
1
water signal is typically suppressed in H-MR spectra, since it would either produce a steep baseline
or even exceed the resolution of older receivers/digitizers.
1
H-MRS has many advantages over MRS of X-nuclei, in particular because of the high inherent
sensitivity and the abundant information content of the spectra. One disadvantage, however, is the fact
1
that two major metabolites of the human body contribute with very strong signals to H-MR spectra:
water and fat. Figure 17-20 illustrates how to get rid of these two enormous signals in an MR
spectrum of the brain. Since the fat signal mainly originates from the subcutaneous fat in brain MRS,
an accurate selection of the sensitive volume can reduce the fat signal considerably. The sequence has
to minimize the strong water signal, which hides the signals from metabolites with concentrations that
are about 10,000 times smaller.211-216 Since the water signal would either produce a steep baseline or
exceed the resolution of the electronics of receiver/digitizer, the water signal is typically suppressed or,
1
alternatively, not excited in H-MR spectra. Since digitizers with higher resolution are available, and
because water suppression saturates also exchanging protons, promising attempts to measure 1H
162,217,218
spectra without water suppression have been published.
The signal-to-noise-ratio (SNR) is one of the most important numbers in MRS. As seen in Table 17-1,
MR spectroscopy suffers from low sensitivity and low concentrations compared with other
spectroscopy methods, such as optical spectroscopy, and also as compared with MRI. MR
spectroscopy has to overcome the problem of low SNR by several means. One is averaging, i.e., the
repetition of the same experiment N times. While MRI uses N of 1, 2 or 4, MRS may repeat an
13
identical experiment 128 times or-in C-MRS-N may be between 1000 and 10,000. This averaging of
the signal with multiple acquisitions is at the cost of a longer examination time. Since SNR is
proportional to the square root of the number of acquisitions (Eq. 17-4), it may soon become
impossible to raise the number of averages in clinical examinations:
Equation 17-5 shows how decreasing the voxel dimensions can easily lead to unfeasible numbers of
scans, i.e., to prolonged examinations:
In other words: in order to have the same SNR, one would need to acquire 64 times longer if the length
of the voxel was halved in every direction. This explains why resolution in MRS is always bargaining
against SNR.
page 472
page 473
Since whole-body magnets at higher field strengths have become available during recent
years,79,107,108,111,118-128 MRS will benefit considerably from increased SNR and better spectral
8 of 9 15-04-2010 20:25
resolution without prohibitively long acquisition times. Other not yet widely used methods to overcome
the limited sensitivity of MRS use the transfer of polarization between high sensitivity nuclei and those
with low sensitivity.71,219-234
© 2010 Elsevier
9 of 9 15-04-2010 20:25
VOLUME SELECTION
A fundamental difference between high-resolution NMR and in-vivo MRS is given by the fact that MRS
needs volume selection components in every acquisition sequence. This has been a major research
topic in the last two decades, and numerous suggestions for such sequences have been
made.90,235-279 Meanwhile, three principal types of volume selection methods have been established
and are used, depending on the specific question, organ, nucleus, and metabolite:
1. Pulse-and-acquire techniques for X-nuclei in combination with surface coils

2. Gradient-based methods with intersecting slices for clinical single voxel spectroscopy such as
point-resolved spectroscopy (PRESS), stimulated echo aquisition mode (STEAM), and image-
selected in-vivo spectroscopy (ISIS)
3. Spectroscopic imaging (CSI, chemical shift imaging) using phase gradients to produce
metabolite maps.
Table 2 summarizes a few volume selection methods that are or have been used in vivo. Several
245 247
methods, such as depth-resolved surface-coil spectroscopy (DRESS), OSIRIS, and PRESS-CSI,
have not been listed separately since they represent combinations of the methods indicated, in part,
they will be discussed in the text below.
The pulse-and-acquire technique, using a surface coil for transmission and detection, was the first
suggestion for selecting signal from a specific volume.235 Easy implementation and high SNR were
particularly important at the beginning of in vivo MRS. Since then, gradient-based techniques have
almost completely replaced pulse-and-acquire for most clinical applications of MRS. However, pulse-
and-acquire still is very important for X-nuclei that suffer from low sensitivity (13C- and 31P-MRS)
and/or metabolites with very short T2 (e.g. 1H-deoxymyoglobin163-167). Unlike MRI, where surface coils
are combined with body-coil transmission, MRS often uses transmission/reception surface coils, in
particular for X-nuclei where no specific body coil is available and coils are often home-built. Figure
17-21 illustrates the complicated shape of the selected volume if surface coils are used for
transmission.71,79,235-238 The inhomogeneous sensitivity profile makes quantitation very difficult and the
preference of structures close to the surface favors contamination-for example, glycogen of abdominal
skeletal muscle and lipids from subcutaneous fat contribute strongly to 13C-MR spectra of liver. On the
other hand, pulse-and-acquire techniques are at present indispensable for 13C-MR spectra of skeletal
muscle and liver.280-291 One of the most important metabolites-glycogen-has an extremely short T2 and
any current echo-based method (see below) would reduce the signal intensity by order of magnitudes.
In addition, 13C-MR spectra have a very large chemical shift dispersion (see Fig. 17-10), and,
therefore, amplify all chemical shift-dependent problems, including pulse width and shift displacement.
Adiabatic pulses238,239,292-296 make the excitation angle largely independent from the amplitude of the
applied radiofrequency field. While this helps to overcome inhomogeneity problems of surface coil
transmission, at least partially, the sensitivity profile of a surface coil remains inhomogeneous and
complicates absolute quantitation of metabolite concentrations considerably. Nevertheless, pulse-
and-acquire sequences with surface coils remain indispensable for MR spectroscopy of X-nuclei.
Topical magnetic resonance (TMR)241 and rotating frame spectroscopy242 have not been applied by
many groups; however, they are historically and didactically interesting. Topical magnetic resonance
selects the sensitive volume by shaping the static magnetic field such that only a limited volume is
homogeneous enough to produce a measurable signal-i.e., fulfils the resonance condition-while the rest
of the tissue is dephased by the inhomogeneous magnetic field.
Rotating frame spectroscopy is based on the effect of increasing radiofrequency amplitudes and
242-244
stepwise increasing flip angles. The measured signal is then deconvoluted by a Fourier
transformation. While this method has been abandoned due to some inflexibility, the basic idea of using
the inhomogeneity of the radiofrequency field for spatial encoding is attractive. Even if principle
differences exist between rotating frame spectroscopy and parallel imaging techniques in MRI and
1 of 8 15-04-2010 20:26
MRS,270,297,298 the success of these methods may motivate a revival of rotating frame spectroscopy.
The introduction of very strong, fast, highly linear, and stable gradients by the manufacturers184-187
promoted the use of gradient-based volume selection techniques in MRS, such as STEAM,253
PRESS,255 CSI,90,248,249 DRESS,245 and ISIS.246 Gradient stability and eddy currents (see Fig. 17-19)
limited the application of gradients in spectroscopy for almost a decade since spectra require a much
better defined magnetic field-temporarily and spatially-than common MRI sequences. Critical imaging
sequences such as echo-planar imaging and phase-sensitive angiography promoted the development
of shielded gradients.184,185
page 473
page 474
Table 17-2. Various Volume Selection Methods for Magnetic Resonance

Spectroscopy
Method* Principles Variatiable Advantages Disadvantages Comments Refs
Pulse- Restricted Radiofrequency No additional Complicated Historically 235
and-Acquire sensitive soft- and/or excitation; important; still

volume of hardware; complicated used for
surface coils no time quantitation X-nuclei or
delay (short short
T2 T2-components
components)
Topical Limited Static magnetic No switching Inflexible; Historically 241
Magnetic volume with field gradients complicated important; no

Resonance homogeneous necessary shape longer used
(TMR) magnetic field
Rotating Increasing Radiofrequency No switching Limited Historically 242
Frame pulse angles gradients flexibility; one important; no

Spectroscopy and necessary dimension only; longer used
subsequent complex shape;
Fourier limited
transformation resolution
Image- Add/subtract Gradient-based Flexibility, Complicated Mainly used in 246
scans with slice selection short T2 add/subtract 31
Selected in P MRS
vivo different components scheme; small
Spectroscopy combinations differences of
(ISIS) of 180-degree very large
pulses signals
Point Three pulses Gradient-based Flexibility, Loss of signal Major method 255
double slice selection high signal with echo time 1
Resolved in H-MRS;
Spectroscopy spin-echo yield TE higher signal
(PRESS) encode three (quantitation; yield destroyed
slices very short if imperfect
T2-components) pulses are
used
Stimulated Three pulses Gradient-based Flexibility, Loss of signal Major method 253
Echo double slice selection robustness with echo time in 1H-MRS;

Acquisition stimulated TE less
Mode echo encode (quantitation; susceptibility to
(STEAM) three slices very short pulse
T2-components) imperfections
than PRESS
2 of 8 15-04-2010 20:26
Chemical Phase- Gradient-based Flexibility, Point-spread- Major method 248

,
Shift Imaging encoding of phase- simultaneous function; long in 1H- and 249
(MRSI, CSI) spatial encoding acquisition acquisition X-MRS for
dimension; of multiple times; large metabolite
generation of spectra data sets maps
images
page 474
page 475
*Combinations of these methods are discussed in the text.
Add to lightbox
Figure 17-21 The sensitive volume of a square-shaped, transmit/receive surface coil. The upper row
shows a calculated sensitivity profile perpendicular (A) and parallel (B) to the plane of the coil. The
lower row illustrates a typical situation in liver spectroscopy where the rendered liver parenchyma (C)
is overlaid by an approximate sensitivity distribution (D). This combination of complicated sensitivity
profile and anatomical situation makes absolute quantitation of surface coil data extremely difficult. In
addition, signals from the surface (e.g., abdominal muscle) are very prominent, while signals from
deeper lying structures are discriminated. (Courtesy of B. Jung, J. Slotboom, R. Kreis)
Image-selected in vivo spectroscopy246-unlike echo-based sequences such as PRESS and

STEAM-keeps the magnetization in the z-direction until a read-out pulse flips the remaining
magnetization into the x-y-plane (observation plane). Therefore, just longitudinal (T1) processes
influence the result, not the transversal (T2) decay, which has no influence during the ISIS sequence.
This makes the method particularly favorable for long T1- and short T2-metabolites as they are
observed in 31P-MRS. Since ISIS is an add/subtract method that acquires the individual signals from
the whole volume, multiple differences of very large signals define the final spectrum, making the
method susceptible to changes between the individual acquisitions. OSIRIS247 overcame this problem
by additional saturation of the outer volume.
Figure 17-22 explains the basic principles of echo- and gradient-based techniques such as PRESS, 255
253
and STEAM. In order to select a cubic volume (a voxel), three orthogonal slices are selected with
the voxel at the intersection. The selection of a slice in MRS is the same as slice selection in MRI (see
Fig. 17-6): if a gradient is applied, a pulse with a limited bandwidth will excite only a slice of spins in
3 of 8 15-04-2010 20:26
the sample. As can be seen in Figure 17-15, a double spin-echo or a stimulated echo results from
three subsequent radio frequency pulses: 90-180-180 degrees for the double spin-echo and 90-90-90
degrees for the stimulated echo. If every pulse is used to select one spatial direction, a first pulse
selects a slice, the second a bar and, finally, the third one selects a voxel (Fig. 17-22). Details of the
two specific sequences PRESS and STEAM are shown in Figures 17-23 and 17-24, respectively.
Add to lightbox
page 475
page 476
Figure 17-22 The selection of a voxel at the intersection of three planes in a dual spin-echo (PRESS)
experiment. If the two 180-degree pulses were replaced by 90-degree pulses and the timing was
adjusted accordingly, the same scheme would be valid for the volume selection in a STEAM sequence
as well. (Courtesy of M. Ith)
Add to lightbox
255
Figure 17-23 Details from the point-resolved spectroscopy (PRESS) sequence, including water
and outer volume suppression prior to the PRESS sequence. Symmetrical "crusher" gradients on both
sides of the 180-degree pulses dephase unwanted signals that are generated by imperfect
radiofrequency (RF) pulses (as shown in Figure 17-15). A series of water suppression (WS) pulses
prior to the sequence reduce the water signal by several orders of magnitude while the outer volume
suppression (OVS) pulses additionally restrict the borders of the voxel in all six directions (left-right,
anterior-posterior, superior-inferior).
4 of 8 15-04-2010 20:26
PRESS255 observes a double spin-echo (see Fig. 17-15) from three pulses that define three
orthogonal slices (see Fig. 17-23). Since the magnetization is always in the transversal plane, T2
processes are active during the entire echo time. As Figure 17-15 illustrates, three pulses generate up
to eight FID and echoes. Except for the double echo SEαβγ, all other signals in Figure 17-15 are not
correctly spatially encoded in a PRESS sequence. For example, FIDγ results from one whole slice in
the z-direction while SEβγ represents a bar that is produced by the intersection of a slice in the y- and
z-directions. It is obvious that a mixture of such signals necessarily leads to artifacts that depend on
tissue far from the selected voxel. So-called crusher gradients on both sides of the 180-degree pulses
are applied to get rid of these spurious signals, since only signals that experience the first and the
symmetrical crusher will be dephased and rephased to the same extent. Because the complete
PRESS sequence should be as short as possible, water suppression and outer volume suppression
are placed prior to the first 90-degree pulse. Water suppression is achieved by a series of spectrally
selective pulses that suppress the water resonance by several orders of magnitude (see Fig. 17-20
and paragraph above). Outer volume suppression supports the definition of the voxel. Prior to the
PRESS sequence, slices adjacent to the voxel in all six directions (left-right, anterior-posterior,
superior-inferior) are applied and the signals are dephased by the gradients as indicated. This makes
the borders of the voxel steeper than the following selective pulse could achieve.
STEAM,253 VOSY,254 and VEST252 observe a stimulated echo (see Fig. 17-15) from three pulses that
define three orthogonal slices (see Fig. 17-24), analogous to the PRESS sequence. Since only
magnetization in the z-direction is refocused by the third pulse in the stimulated echo, transversal
magnetization between the second and third pulse (period TM) is lost and, therefore, transversal
relaxation is not effective in TM. This helps to shorten the total TE of the sequence since TM is only
influenced by T1 mechanisms. In addition, water suppression pulses and gradients can be placed
either prior to the sequence or in the period TM.
Since the maximum amplitude of a stimulated echo is limited to one half of a spin-echo, one would
expect an advantage in SNR for PRESS. However, this advantage can be partially lost by imperfect
180-degree pulses; STEAM is more robust since it needs 90-degree pulses only. Modern coils and
radiofrequency amplifiers, as well as sophisticated pulse shapes, produce better defined 180-degree
pulses such that PRESS increasingly benefits from the inherently better SNR.
Add to lightbox
253
Figure 17-24 Details from the STEAM sequence, including water and outer volume suppression.
Outer volume suppression (OVS) pulses additionally restrict the borders of the voxel in all six
directions (left-right, anterior-posterior, superior-inferior). In contrast to the PRESS sequence (Figure
17-23), water suppression (WS) can be placed either before or between the 90-degree pulses.
Symmetrical "crusher" gradients on both sides of the second and third 90-degree pulses dephase
unwanted signals that are generated by imperfect radiofrequency (RF) pulses (as shown in Figure
17-15). During the TM period, i.e., the time between the second and third 90-degree pulse, the
magnetization remains in the z-direction, thus making T2-processes during this period ineffective.
5 of 8 15-04-2010 20:26
Add to lightbox
page 476
page 477
Figure 17-25 An illustration of the chemical shift displacement of single-voxel techniques-depending
on their chemical shift, metabolites in the same spectrum do not have their origin in identical voxels. If
the red cube represents N-acetyl-aspartate (NAA), the green shows the sensitive volume of creatine,
while the blue one would correspond to the selected voxel of myo-inositol. Increasing gradient strength
during slice selection can reduce this effect, however, this requires a larger bandwidth and thus
typically more power of the selective pulse.
Chemical-shift artifacts are clearly recognizable in an MR image, while chemical-shift displacements

are often neglected in MR spectra since they are not visible but result in variations of line intensities.
Figure 17-25 illustrates how slices select a slightly different region for metabolites with different
chemical shift. Since slice selection is used to define the sensitive volume in all directions, this effect
has to be considered in all three dimensions. An estimation for a 1.5 T system gives an impression of
the effect. If a gradient strength of 1.0 mT/m is used for slice selection in a PRESS sequence, the
pulse bandwidth for a 10 mm voxel is 680 Hz and the displacement between NAA (2 ppm), creatine (3
ppm), and myo-inositol (4 ppm) is 1 mm and 2 mm, respectively. If the aromatic region (8 ppm) and
lactate (1 ppm) are included, the effect gets even larger and the displacement becomes a
considerable fraction of the voxel size. Since the displacement occurs in all three dimensions (see Fig.
17-25), a 1 mm displacement in every direction results in a final 72% overlap of the cubes for NAA and
myo-inositol in single voxel spectroscopy. This must be considered in clinical exams, in particular in
examinations of focal lesions and if ratios are built, since these ratios may represent spatial diversity
rather than specific pathology. Prior to a spectroscopic imaging sequence (see below), often a PRESS
or STEAM box is selected for a restriction of the volume. Since this box typically is much larger than
the volume in single voxel spectroscopy, the gradients are smaller and, thus, the chemical-shift
displacement can be dramatic at the edges of the box.
Some methods combine the volume selection by a surface coil with gradient-based techniques (e.g.,
DRESS245). While a slice is selected parallel to the coil in the depth, the inhomogeneous sensitivity of
the surface coil limits the sensitive volume in the two other directions. This method has been used in
64
particular for X-nuclei, but has not found wide propagation.
6 of 8 15-04-2010 20:26
Add to lightbox
Figure 17-26 To the left, the radiofrequency, gradients, and acquisition window of an original
chemical-shift imaging (CSI) sequence, i.e., with selection of a slice prior to the CSI phase-
248
encoding. The shaded gradients are increased stepwise throughout an experiment. The
disadvantage of this type of sequence is a delayed data acquisition and, subsequently, a baseline roll.
Without slice selection, a 4-dimensional data set with less time delay of the acquisition can be
acquired,249 however, since the whole volume is excited, very large fields-of-view have to be selected
to prevent in-folding. To the right, a combination of a point-resolved spectroscopy (PRESS) box
selection with subsequent phase-encoding in the Y- and Z-directions is shown.
A completely different approach for volume selections is made in CSI or "spectroscopic imaging,"
where gradients are used to encode one or more spatial dimensions by the phase of the signals. In the
248,249
early days of CSI, a phase-encoding gradient was applied on the FID generated over the whole
sample or from a preselected slice.248 Since an immediate reading of the FID causes phase
249
90
problems, newer CSI sequences combine selection of a PRESS or STEAM box with additional
phase-encoding gradients (Fig. 17-26). Chemical shift imaging data sets can be displayed in different
ways, either as metabolite maps or as a series of spectra (Fig. 17-27). Both presentations have
specific advantages. Metabolite maps are very popular in clinical examinations since they resemble
classical presentation of MRI data. However, they have the disadvantage that artifacts are hardly
recognized, which is much easier in arrays of whole spectra (see Fig. 17-27). Two important
advantages of CSI result from the phase-encoding: 1. metabolites experience no chemical shift
displacement in the phase direction (just at the border of the selected box due to the slice selection);
and 2. voxels can be shifted in fractions of voxel dimensions if the data set is transformed back into the
time domain followed by a phase shift. An inherent problem of CSI data acquisition is the limited size of
the matrix. Fourier transformation with an insufficient number of elements (e.g., below 64 or 128),
cannot perfectly assign all signals to the correct voxel due to the limited data set. The distribution of a
point-like signal source by this mechanism is called point-spread function71,77,94,299 and describes how
the Fourier transformation of a limited data set leads to bleeding of signal from one voxel to another.
This smearing effect can be reduced by spatial apodization functions, but only at the cost of a reduced
spatial resolution, similar to the sensitivity enhancement with reduced spectral resolution shown in
Figure 17-13. Appropriate arrangement of the phase steps (i.e., nonuniform sampling of the k-space)
256-261
can avoid the reduction of the spatial resolution partially. The acquisition of a spectroscopic
image needs considerable time and several methodologies offer ways to overcome this and other
disadvantages: multi-slice CSI,262,263 echo-planar spectroscopic imaging,90,250,263-267 spatial-
251,268,269 270 271-274
spectroscopic selective pulses, parallel spectroscopic imaging, Hadamard encoding,
275-279
and the observation of a reduced number of resonances.
The choice of using single voxel techniques or spectroscopic imaging should be defined by the clinical
question, since both techniques are complementary. While CSI data sets are most appropriate for
7 of 8 15-04-2010 20:26
focal lesions where the important finding is an inhomogeneous spatial distribution of metabolites, single
voxel acquisition can be beneficial for diffuse lesions and metabolic disorders where no spatial
inhomogeneity is expected.
© 2010 Elsevier
8 of 8 15-04-2010 20:26
QUANTITATION
There are two principally different views of clinical MRS: the approach of "pattern recognition," where
a spectrum is analyzed as kind of an image, versus the concept of MRS as kind of a "clinical chemistry
in vivo."
page 477
page 478
Add to lightbox
Figure 17-27 A chemical-shift imaging (CSI) data set from a brain examination of a ring-enhancing
lesion. Prior to the CSI phase-encoding in anterior/posterior and right/left direction, a point-resolved
spectroscopy (PRESS) box (green) was selected to limit the field-of-view and to avoid contamination
by large signals from subcutaneous fat. Two different ways of data display are possible: in the upper
row, three metabolite images for choline, creatine, and N-acetyl-aspartate (NAA) are shown. In the
lower row, single spectra from 9 different sites in the lesion are depicted. This shows clearly the
absence of NAA in the cyst (spectrum #5) and increased choline in spectra #2 and #8. However,
artifacts from in-folding lipid signals cannot be separated in the metabolite maps; they are visible in
the single spectra, e.g., in spectra #4 to #6. On the other hand, metabolite maps give a better
impression of the homogeneity and relevance of the data set, e.g., enhanced choline signals are
clearly depicted in the rim of the cyst. Therefore, both methods of data display reveal specific
information and should be looked at.
In the pattern recognition approach, one determines sensitivity and specificity of changes in the MR
spectral pattern for various pathologies, independent of the biochemical nature of metabolites that
contribute to the spectrum. In most cases, ratios of two or more signal intensities are sufficient to
provide interpretable numbers. This pragmatic and fast approach can be quite successful in clinical
diagnosis; however, it is dependent on a specific acquisition technique, MR system, and setting. In
addition, ratios can be subject to many influences; for example, an increase of a metabolite in the
numerator of the ratio can lead to the same number as a decrease of the metabolite in the
denominator, or variations of the relaxation times can dominate concentration changes.
Representatives of this view use MRS typically as an add-on to a general MRI examination.
The clinical chemistry approach quantifies metabolites as is done in reports from clinical chemistry
1 of 6 15-04-2010 20:28
labs, e.g., for g/L hemoglobin. This approach needs to overcome some principal problems of in-vivo
MRS, and the standardization process may introduce additional errors; however, it has the advantages
that single metabolites can be compared against a normal range, that results should be independent of
a specific sequence or MR system, and that the numbers are comparable with other methods, such as
biopsy and chemical analysis. Representatives of the clinical chemistry approach often have to
integrate different clinical modalities and need to compare MRS with wet chemistry.
page 478
page 479
Add to lightbox
Figure 17-28 Variations in the visibility of chemical groups in a molecule, depending on the
environment. The left column shows a brain spectrum before and 90 min after an alcoholic drink. The
right column shows difference spectra between spectra obtained before and after ethanol (EtOH)
intake (top, middle), and an in vitro spectrum of ethanol (bottom). In the in vitro spectrum, the area of
the CH2 signal correctly represents about two thirds of the CH3 signal. However, in vivo, the CH2
signal in both volunteers is reduced, compared with the CH3 group, i.e., the CH2 group is only partially
visible. This reduced visibility can be an effect of limited mobility, for example, in a membrane
environment. Such a reduced visibility has to be taken into account if a resonance is used for absolute
quantitation. (Adapted from Kreis R, Hofmann L, Boesch C: Visibility of the methylene signals of
ethanol in 1H-MR spectra of the human brain and implications for model fitting. Proc Intl Soc Magn
Reson Med 8:426, 2000.)
Why is it so difficult to define absolute concentrations in MRS, and why are ratios of signal amplitudes
so popular in clinical MRS? It has been mentioned earlier (see Fig. 17-8 and corresponding text) that
the amplification of MR signals is inherently arbitrary. Voltages induced in the detecting coil depend on
the excitation pulses, coil geometry and loading, quality factor, and many other parameters that may
change from examination to examination. Amplification and digitization of the detected voltages
introduce additional factors that may be variable from patient to patient. Other factors that may change
the signal of a metabolite-even if the concentration remains constant-are relaxation times T1 and T2
(see Fig. 17-17), nuclear Overhauser effect (NOE), diffusion, magnetization transfer, changes of voxel
shape and size, chemical-shift displacements, and numerous other mechanisms. While the relaxation
times are intentionally used to increase contrast in MR images, MR spectra should be acquired under
relaxed conditions in order to obtain a measure for the concentration of the metabolites. It is also
evident that the multiplicities of lines (e.g., doublets) and the number of nuclei per resonance line (3
nuclei per molecule in a methyl group) must be taken into account for the calculation of absolute
2 of 6 15-04-2010 20:28
concentrations.
It may not be obvious at first glance, but it is mandatory to first determine the visibility of a specific
metabolite in a spectrum for the determination of absolute concentrations. Figure 17-28 illustrates how
two protons in the same molecule300 show differences in their visibility.301 While the CH3 group in
ethanol is clearly detectable, the CH2 group is only partially visible in vivo. The spectrum in solution,
presented in the lower trace, shows resonance lines of the two groups in ethanol as expected from
stoichiometry. In vivo, however, the area of the CH2 resonance of ethanol is clearly reduced (upper
trace). Such a reduction of the visibility can result from immobilization-e.g., in a membrane. 300 For most
metabolites observed in clinical MRS, restricted visibility has not been an issue so far, however,
magnetization transfer from the water-suppression pulse can influence various metabolites, including
exchangeable protons,162 and even major metabolite signals in the brain.302 Another important
metabolite, glycogen, was supposed to be partially invisible due to the large mass of the molecule.
Careful experiments have shown that glycogen is 100% visible in 13C-MRS, despite its size.303 The
visibility of some metabolites such as creatine in skeletal muscle304 and fluoxetine in brain tissue305 is
still under investigation.
The area (and not the amplitude) is proportional to the concentration of metabolites under specific
conditions (see Fig. 17-8). It is amazing how difficult the correct determination of the area of a
resonance line in a crowded spectrum can be. While a correct adjustment of the phase (see Fig.
17-14) is relatively easy, as long as an appropriate reference scan is available, the determination of
the line width and/or of the baseline (see Fig. 17-18) is much more difficult and the algorithms to do
that are not very robust.
page 479
page 480
Add to lightbox
3 of 6 15-04-2010 20:28
Add to lightbox
Figure 17-29 This demonstrates fitting of a brain spectrum in LCModel,135 where an experimental
spectrum is approximated by a linear combination of metabolite spectra that have been acquired in
solution. Left and middle columns, the metabolites that contribute to the fitted spectrum in the right
column. Bottom right, the residual difference shows almost no difference between the fit and the
experimental spectrum. (Chtot, total choline; Cr, creatine; mI, myo-inositol; NAA, N-acetyl-aspartate.)
(Courtesy of L. Hofmann, R. Kreis, M. Ith.)
Numerous fitting algorithms135,145,148,188-201 have been suggested to analyze clinical spectra more or
less automatically. Different approaches can be distinguished: 1. fitting in the time domain versus fitting
in the frequency domain or a combination of both; and 2. using prior knowledge (i.e., known chemical
shifts, relative intensities, coupling constants) to approximate the experimental spectrum versus linear
combination of spectra from metabolites that have been acquired in solution. While an extended
analysis of fitting algorithms would go far beyond the scope of this chapter, Figure 17-29 shows one
example of a fitting algorithm, "LCModel."135 Solution spectra that have been acquired under identical
conditions are combined such that the sum fits the experimental spectrum. Based on a reference scan
without water saturation,132,133 absolute concentrations of up to about 30 metabolites can be
estimated, depending on the basis set of solution spectra. In the original basis set, metabolites such as
alanine, aspartic acid, tri-methyl-ammonium compounds, creatines, γ-aminobutyric acid (GABA),
glucose , glutamine, glutamate, lactate, myo-inositol, scyllo-inositol, NAA, N-acetyl-aspartyl-
glutamate, and taurine have been included. As soon as other metabolites occur (e.g., with pathology,
113,306-309
applied pharmaceuticals, brain decomposition) additional solution spectra can be added.
The described "LCModel"135 is just one example of a whole series of fitting algorithms with various
acronyms, such as TDFDFIT, MRUI, VARPRO, AMARES, and LPSVD.145,148,188-201 Following an
appropriate determination of resonance areas, the signal intensities must be converted into absolute
4 of 6 15-04-2010 20:28
measures such as mmol/kg wet weight. Several approaches have been suggested, 128-144,285,310-312
1 13
either based on internal reference substances such as water in H-spectra or creatine in C-MRS, or
on external references that contain solutions with known concentrations.
Figure 17-17 illustrates spectra after various TE delays. Baseline and definition of the lines are
improving with increasing TE. In turn, the SNR decreases and slight variations of the T2-values may
lead to dramatic changes in signal amplitude. Therefore, short TE spectra are preferred for absolute
quantitation, since it would be very time consuming to measure T2 of all metabolites in every
examination, which would be necessary to correct for individual T2 changes.
The value of MRS as a diagnostic tool depends strongly on the reproducibility of the individual
spectrum. Figure 17-30 demonstrates the high reproducibility of up-to-date MR spectra with 10
142-144,313-316
repeated spectra in one volunteer. In fact, baseline determination and line-width effects
are more prominent sources of errors than the reproducibility of the data acquisition.317
Table 17-3 illustrates the accuracy of absolute concentrations determined by different authors and MR
systems. In weighted averages, Kreis129 determined CV of 0.7% to 2% for the major metabolites,
numbers that are surprisingly good for biological measurements.
page 480
page 481
Add to lightbox
Figure 17-30 This gives an idea of the excellent reproducibility of 1H-MR spectra in human brain that is
possible in up-to-date MR systems. Ten spectra from the same volunteer are shown, together with the
average spectrum in the middle. (Ch, choline; Cr, creatine; mI, myo-inositol; NAA, N-acetyl-aspartate.)
(Courtesy of P. Maloca, C. Fusch, R. Kreis.)
Table 17-3. The Content of Commonly Observed Metabolites in Gray Matter,

Determined by Quantitative 1 H-Magnetic Resonance Spectroscopy. A Literature
5 of 6 15-04-2010 20:28
Survey Illustrating the Accuracy of Quantitative 1H-MRS in the Brain

N-acetyl-aspartate Creatine Choline myo-Inositol
CV (% CV (% CV (% CV (%
Mean of Mean of Mean of Mean of
Area N (mmol/kgww) mean) (mmol/kgww) mean) (mmol/kgww) mean) (mmol/kgww) mean) Refs
Occipital 10 9.24 6.1 8.04 5.8 1.43 16.1 7.26 17.6 366
Insular 10 11.80 14.2 8.92 7.4 2.27 10.7 8.11 20.1 367
Occipital 10 11.13 4.2 9.00 3.5 1.71 6.9 6.56 8.6 367
Occipital 10 9.98 7.5 7.98 5.8 0.99 14.1 5.11 14.1 368
Occipital 10 9.45 7.7 7.55 5.4 0.94 14.9 4.83 13.9 368
Occipital 10 11.49 9.7 9.22 13.0 1.15 20.0 5.88 16.3 368
Parietal 15 11.70 18.8 8.20 17.1 1.40 21.4 6.20 17.7 369
Parietal 16 9.90 19.4 8.17 18.8 1.63 35.3 370
Unweighted
Average
(mean ±
SEM) 10.59 3.5 8.39 2.5 1.44 10.7 6.28 7.0
Weighted
Average
(mean ±
SEM) 10.32 0.9 8.36 0.7 1.35 1.5 5.92 1.6
CV, coefficient of varience; SEM, standard error of the mean.

Adapted from Kreis R: Quantitative localized 1H MR spectroscopy for clinical use. Prog NMR
Spectroscopy 31:155-195, 1997.
© 2010 Elsevier
6 of 6 15-04-2010 20:28
MRS BEYOND N-ACETYL-ASPARTATE

This section aims to illustrate the wide range of in-vivo MRS applications that go beyond the observation of
an NAA/creatine-ratio in 1H spectra of the brain. It will also give an impression of what could be possible in
future.
Every organ and every nucleus generates particular methodological problems and requires specific
techniques as is illustrated in Figs. 17-1 and 17-2. In turn, all these spectra provide information on different
metabolites and present findings on numerous biochemical pathways and pathologies. Therefore, MRS
examinations can differ considerably, depending on the question, the metabolites, and the technique. For
13 1
example, the quantitation of liver glycogen by C-MRS has to consider low SNR, short T2, H-decoupling,
motion of a relatively homogeneous organ, complex sensitive volume of a surface coil, and contamination by
1
the glycogen signal from the abdominal muscles; while the observation of NAA in a brain H-MR spectrum
can be acquired in an inhomogeneous, motionless organ, with high SNR, long T2, and in a well-defined voxel.
This wide range of MRS applications may be an advantage in the long run, but, for the time being, it
overextends the methodological capacity of many MR groups and MR systems. In addition, many clinicians
are also hesitating to invest a lot into the implementation of different MRS techniques, in particular of
X-nuclei, since the numerous MRS applications may be very successful in research, but are not yet ready as
a diagnostic tool in an individual patient.
31
Figure 17-1 shows P-MR spectra of brain, heart muscle, liver, and skeletal muscle acquired with different
techniques. The brain spectrum has been obtained with an ISIS sequence, heart muscle and liver data result
from a one-dimensional CSI-experiment, and the skeletal muscle spectrum is the product of a pulse-
and-acquire sequence with surface coil localization. The metabolite concentrations of the various organs
show clear differences-as is expected since their metabolism is different. A typical example is the resonance
of phosphocreatine (0 ppm by definition), which is very strong in skeletal muscle and absent in liver tissue.
1 13 31
Figure 17-2 illustrates spectra of human skeletal muscle obtained by H-, C-, and P-MRS. While the
1
H-MR spectrum has been acquired with a gradient-based PRESS sequence, water suppression and outer-
13 31
volume suppression, the C- and P-spectra were obtained by a pulse-and-acquire technique using
13
double-tuned surface coils for transmission and signal reception. During acquisition of the C-MR spectrum,
1 1 13
H have been irradiated, in addition, to decouple the J-coupling between H and C in the glycogen
C1-nucleus (Figure 17-31).
One example illustrates in more detail how specific problems have to be solved for a particular nucleus.
Looking at Figs. 17-10 and 17-25, it is evident why gradient-based techniques are unfavorable for 13C-MRS.
13 1
Since the chemical shift range of C-MR spectra is much larger than for H, either very strong gradients
would result in an enormous bandwidth of the slice selective pulse or the chemical-shift dependent
displacement would be disastrous. To give a numerical example, a gradient strength of 5 mT/m would require
13
a bandwidth of 2130 Hz for a 4 cm voxel length if C was directly selected by slice excitation. In turn, the
chemical-shift difference between creatine (157 ppm) and glycogen-C1 (100 ppm; see Fig. 17-31) of 910 Hz
would lead to a chemical shift displacement of 1.7 cm. In other words, the voxels for creatine and
glycogen-C1 would overlap by only 19% if this displacement were considered in all three spatial dimensions.
For the whole chemical-shift range of about 200 ppm (e.g., for the chemical shift difference between the
methyl carbons in fat and the carboxyl group in acetone), the displacement would increase to 6 cm. Under
these circumstances, the voxels selected for the carboxyl groups and for the methyl carbons would not even
13
touch each other. Since gradient-based localization is obviously problematic in C-MRS, decoupling (see
Fig. 17-31) is often successfully combined with pulse-and-acquire techniques, as shown in Figure 17-32
291
where the signal of glycogen-C1 is followed at rest and during replenishment after a depleting workload.
Since surface coil excitation makes absolute quantitation difficult, the signal of creatine is used as an internal
standard in this example.310,311
31
Metabolite changes that can be observed by P-MRS of skeletal muscle are on a much shorter time scale
than changes of glycogen as shown in Figure 17-32. Since subcutaneous fat contains almost no metabolite
that would contaminate the 31P-MRS spectrum, pulse-and-acquire techniques are suitable for the observation
of metabolite changes under workload (Fig. 17-33). The high SNR of surface coils allows a high temporal
resolution that is necessary to determine pH changes and recovery of phosphocreatine during and after
1 of 21 15-04-2010 20:29
workload.318-324
Add to lightbox
page 482
page 483
13
Figure 17-31 A C-MR spectrum of skeletal muscle (upper trace) without and (lower trace) with
1
H-decoupling and nuclear Overhauser effect (NOE) buildup. Both mechanisms help to increase the signal-
to-noise ratio. (Adapted from Boesch C, Kreis R: Imaging and spectroscopy of muscle, In Young IR, Grant
DM, Harris RK (eds): Methods in Biomedical Magnetic Resonance Imaging and Spectroscopy
(Encyclopedia of Nuclear Magnetic Resonance). Chichester UK: John Wiley & Sons, 2000, pp
1307-1316.)
Add to lightbox
Figure 17-32 The usage and replenishment of glycogen in human skeletal muscle, observed by natural-
abundance 13C-MRS. The resonance of creatine is used for relative quantitation311,312 since it can be
assumed that creatine remains relatively constant. (Adapted from Rotman S, Slotboom J, Kreis R, et al:
Muscle glycogen recovery after exercise measured by 13C-MRS in humans: Effect of nutritional
solutions. MAGMA 11:114-121, 2000.)
Many comprehensive lists of observable metabolites for the most common nuclei have been published in the
past.71,85,325,326 Table 17-4 represents an extract of such a table85 and illustrates the large number of
metabolites that can be observed by 1H-MRS, going far beyond the acquisition of NAA, creatine, and choline.
Multinuclear MRS can observe many other metabolites that are crucial for human metabolism.
2 of 21 15-04-2010 20:29
Add to lightbox
31
Figure 17-33 Changes in the P-MR spectrum during workload and recovery. For moderate workload,
recovery can be described by a mono-exponential function with a time constant that is representative for the
oxidative capacity of the muscle.323 (ATP, adenosine triphosphate; PCr, phosphocreatine; Pi, inorganic
phosphate.)
1
Figure 17-34 shows the aromatic region in the H-spectrum of human brain and skeletal muscle-a spectral
range that is often neglected, but which contains information on clinically highly relevant metabolites such as
phenylalanine (Phe). Blood levels of Phe are extremely elevated in untreated phenylketonuria (PKU), an
inherited disorder of amino-acid metabolism112 with profound consequences for brain development. It has
been shown that cerebral Phe can be detected noninvasively by 1H-MRS113,157,158 (see Fig. 17-34).
Difference 1H-MR spectra help with the separation of the Phe resonance from the other aromatic amino
acids and allow a determination of blood-brain barrier kinetics of this metabolite in humans. 113 Figure 17-34
shows an additional metabolite with resonances that contribute to the aromatic region-the histidine residue in
carnosine shows two resonances, at 7 and 8 ppm, in a 1H-MR spectrum of skeletal muscle.159-161 Since the
chemical-shift positions of the two resonances are pH dependent, carnosine can be used for pH
31
measurements, similar to inorganic phosphate in the P-spectrum (see Fig. 17-11).
The chemical shift of myoglobin (Mb) resonances is strongly dependent on the oxygenation state of the
163-167,327 1
molecule. The signals from oxy-myoglobin are in the normal frequency range of a H MR spectrum
and, therefore, are overlapped by many other resonances at 1.5 T. It has been reported that the resonance
327
of the methyl group of Val-E11 can be observed at 2.8 ppm under oxygenated conditions. While this
resonance disappears with deoxygenation, the F8 proximal histidyl N-proton shifts to 78 ppm (see Fig.
17-10), a region that is free from overlapping resonances, even at moderate field strength. Despite very
short T2 relaxation times, this histidine resonance can be observed reliably with appropriate techniques, such
163-167
as pulse-and-acquire and surface coil detection.
page 483
page 484
Table 17-4. Selected in 1H-MR Spectra in vivo

Chemical
[ppm] Metabolites group Abbreviation Comments References
0.95 Fatty-acyl-chains -CH3 Intramyocellular 371 372
,
lipids (IMCL)
3 of 21 15-04-2010 20:29
1.12 Propan-1,2-diol Exogenous, 307

neonatal brain
1.15 Fatty-acyl-chains -CH3 Extramyocellular 371-374
lipids (EMCL),
shift orientation
dependent
1.20 Ethanol -CH3 Exogenous, 309
brain
1.30 Lactate -CH3 Lac Underneath 373 375-377
,
lipids, long TE
or editing
1.48 Alanine -CH3 Ala In brain; after 377 378
,
voluntary
hyperventilation
1.85 Acetate -CH3 Ac Brain, 373 376 379
, ,
cancerous liver
([commat]1.94
ppm)
2.00 N-acetyl-aspartate -CH3 NAA Brain 373 375 379
, ,
2.05 N-acetyl-aspartyl-glutamate -CH3 NAAG Brain (white 380
matter)
2.10-2.35 Glutamate β-CH2 Glu Brain 373 377
,
γ-CH2
2.10-2.45 Glutamine β-CH2 Gln Brain 373 377
,
γ-CH2
2.13 Acetyl-carnitine -CH3 In muscle during 381
or after
strenuous
exercise
2.22 Aceto-acetate -CH3 Brain, diabetic 308 309
,
keto-acidosis
2.25 Gamma-aminobutyric acid γ-CH2 GABA Brain 373
2.60 N-acetyl-aspartate β-CH2 NAA Brain 373 377 379

, ,
3.00 Creatine/phosphocreatine N-CH3 Cr/PCr Brain, muscle 371 373 375
, , ,
377 379
,
3.03 Glutathione Brain, weak 308
contribution to
Cr/PCr peak
3.20 Choline, N-(CH3)3 Cho, GPCh Brain, muscle 371 373 377
, , ,
phosphocholine/glycerophosphocholine 379
3.21 Carnitine Car Brain, muscle 308 382

,
3.30 Taurine N-CH2,S-CH2 Tau N-CH2 373 377 379
, ,
[commat]3.2
ppm & S-CH2
[commat]3.4
ppm collapse at
low field
strength to a
singlet
3.35 Inositol (scyllo-) ring H1-H6 scyllo-Ins Brain (mainly 383
gray matter)
4 of 21 15-04-2010 20:29
3.40 Glucose C4α,β, C5β Glc Difference 377 384

,
hyper-
euglycemia in
brain
3.50 Glycine -CH2 Gly Visible at long 373
TE, else below
inositol
3.50 Inositol (myo-) ring H1, H3, myo-Ins Brain, not visible 373 377 383
, ,
H4, H6 at long TE (270
ms)
3.75 Glutamate α-CH2 Glu Brain 373 377
,
3.75 Glutamine α-CH2 Gln Brain 373 377
,
3.80 Glucose C5α, C6α Glc Difference 377 384
,
hyper-
euglycemia in
brain
3.90 Creatine/phosphocreatine N-CH2 Cr/PCr Brain, muscle 371 373 375
, , ,
377
4.06 Inositol (-myo) H2 myo-Ins Brain 377
4.77 Water H2O Chemical shift 371 379

,
temperature
dependent
5.40 Glycogen C1-proton Liver (rat) 385
5.52 Fatty-acyl-chains -C*H=C*H- All tissues 371 374 379

, ,
7.05 Carnosine C4 proton Car Muscle 371 382
,
His
7.32 Phenylalanine ring protons Phe Brain, 113 157 379
, ,
cancerous liver
8.05 Carnosine C2 proton Muscle 371 382
,
His
79.00 Myoglobin (deoxygenated) Nδ proton of deoxy-Mb Visible in the 163
F8 proximal deoxygenated
His form under
ischemic
conditions in
muscle
Adapted from Boesch C: Molecular aspects of magnetic resonance imaging and spectroscopy. Mol Aspects
Med 20:185-318, 1999. © 1999, with permission from Elsevier.
These examples are far from being a complete list-the intention is just to illustrate the versatility and the
potential of MRS. In turn, it also demonstrates that various techniques are necessary to obtain optimal
results. The future will show if this is justified for research applications only (where it has already been
proven) or if this may also be feasible for clinical diagnostics in individual patients.
page 484
page 485
5 of 21 15-04-2010 20:29
Add to lightbox
Figure 17-34 Spectra of the human brain (A-C) and human skeletal muscle (D). The left part of the spectrum
(i.e., the aromatic region left of the water resonance) typically is not shown, however, it contains important
information on relevant metabolites, e.g., carnosine (in muscle; D) and phenylalanine (in brain; A-C).
Spectrum A is the sum of spectra acquired in healthy volunteers, while spectrum B is the sum of spectra
obtained in several phenylketonuria patients. Trace C represents the difference between traces A and B,
and clearly shows the phenylalanine resonance.
MRS offers a series of different editing mechanisms, which can be used to select specific, otherwise hidden,
metabolite signals in a MR spectrum. The examples given here serve as a demonstration of the versatility of
MRS only; further details would go beyond the scope of this chapter. Difference spectra (see Fig. 17-34),
either between healthy volunteers and patients, or between the same individual before and after an
113,328
intervention, can be seen understood as a straightforward and simple form of spectral editing. It
requires a considerable degree of reproducibility; however, this can be achieved in modern MR systems as
shown in Figure 17-30. Metabolites such as lactate, GABA, glutamate/glutamine, or glutathione are hidden in
1
crowded regions of the H-MR spectrum. Homo- or heteronuclear coupling with other nuclei (see Figs.17-8
and 17-9) can be used to edit spectra such that only the desired resonances survive the sequence. These
methods include add/subtract schemes with selective irradiation on a J-coupled partner in the molecule,
multi-quantum filters, and polarization transfer.96,152-156 While edited spectra are extremely powerful for
specific questions, they suffer from the fact that they need at least the same amount of time as unedited
spectra, but just measure a few preselected metabolites.
A chemical alternative and supplementary method to spectral editing of the natural abundance of metabolites
is isotope labeling, which can increase SNR dramatically. Metabolites enriched in a rare isotope can be
followed specifically (some in small animal or isolated organ studies particularly for financial reasons), e.g.,
13 13 13 13 13
infused [1- C]-glucose, [1,2- C2]-glucose, [2- C]-acetate, [3- C]-pyruvate, [1- C]-ribose, and other
95-102,329-332
stable isotopes A further expansion of this technique is so-called "isotopomer analysis" where the
13 13
coupling pattern of C- C bonds can be analyzed to study coexisting metabolite pathways
specifically.333-343
6 of 21 15-04-2010 20:29
Add to lightbox
Figure 17-35 An example of two-dimensional spectroscopy in human skeletal muscle. The second
dimension in the two-dimensional plot demonstrates the coupling of the two lines in Cr2, a finding that was
unexpected since it results from dipolar coupling and not of the common J-coupling. The figure shall illustrate
the versatility and potential of modern MRS that has access to many methods from high-resolution NMR that
are not yet often used in vivo. Cr2, creative-CH2; Cr3, creative-CH3; X3, tentative assignment to taurine;
TMA, trimethylammonium groups. (Adapted from Kreis R, Boesch C: Spatially localized, one- and
two-dimensional NMR spectroscopy and in vivo application to human muscle. J Magn Reson Series B
113:103-118, 1996.)
Multidimensional MRS (Fig. 17-35) is a method that is widely used and very successful in high-resolution
NMR,75,80-82 but is not yet in widespread use in vivo. "Multidimensional" in this context does not mean multiple
dimensions in space, but multiple parameters, such as multiple chemical shifts of different nuclei, or chemical
shifts combined with coupling constants as shown in Figure 17-35. Despite the enormous potential of an
entire pool of methods from high-resolution NMR, some problems limit the applications in vivo so far. In vivo,
the sequences have to be combined with volume selection, which adds an additional complication. Typically,
multidimensional sequences require long acquisition times since one dimension has to be stepped through,
something that is limited by natural time constraints in vivo. Technical requirements, such as field
homogeneity or pulse power constraints, result in further limitations. Nevertheless, multidimensional MR
344-351
methods are attractive for specific questions.
MR spectroscopy is inherently a molecular method85 that is able to describe characteristics of metabolites on

a molecular level. Magnetization-transfer experiments are well known from imaging where the spatial
neighborhood of water and fixed molecules is detected by magnetization-transfer techniques. In MRS,
magnetization transfer is monitored between different pools of metabolites, thus allowing a noninvasive
investigation of tissue at the molecular level.302,305,352-359
Another molecular mechanism that can be investigated by MR is molecular diffusion. The random motion of
the nuclei in a gradient field de-phases the magnetization and can be used to characterize the amount, the
84,360-365
direction, and the spatial limitations of diffusion. These mechanisms allow probing of cellular
dimensions and metabolite mobility.
page 485
page 486
Currently, MR systems with high and ultra-high magnetic field strength are installed worldwide,79,107,118-128
driven mainly by brain activity studies (fMRI), angiography of smaller vessels, and higher SNR for various
7 of 21 15-04-2010 20:29
MRI applications. MRS benefits indirectly from these technical developments with higher SNR and better
spectral resolution. While a series of research institutions are mainly solving the many technical and
methodological problems that are posed by field strengths at 7 T and above, 3 T has become the standard
for high-field in-vivo MR and numerous systems are currently installed. Animal systems around 10 T and
122
above demonstrate the enormous potential of these strong fields for MR spectroscopy. The well-known
advantages of higher field strength are an improved spectral resolution and increased SNR, both serious
limitations of MRS at lower field strengths. On the other hand, T2 relaxation times of some metabolites seem
to decrease with higher field strength, which counteracts the improved spectral resolution somewhat. 123 In
addition, limitations of the power deposition restrict the application of 1H-decoupling and fast imaging
sequences. Problems at the higher 1H-frequency (see Table 17-1) such as power deposition, penetration
depth, phase distortion, and diamagnetic resonances are much less prominent at lower frequencies.
Therefore, X-nuclei such as 13C, 31P, 23Na, 35Cl, 39K (see Table 17-1) will benefit most from the higher field
strength without the disadvantages of the higher radiofrequency. It can be expected that these applications
will be promoted by ultra-high fields in the near future.
Sophisticated methods mentioned in the last paragraph are not yet installed in routine clinical MR systems.
On the other hand, MR systems ordered today will serve for the next 5 years or more. During that period,
multinuclear MR and some of the more sophisticated techniques may become routine and an up-to-date MR
system should be able to be upgraded in the future. In-vivo MR spectroscopy is technically and
methodologically very demanding, which may partially explain why MRS is not yet widely applied in the
clinical routine. On the other hand, the technological achievements of the most recent years, in particular, the
increasing number of ultra-high field whole-body systems, are very beneficial for MRS. The technical
development has now reached a threshold that is necessary for many MRS applications in vivo.
REFERENCES
1. Danielsen ER, Ross B: Magnetic Resonance Spectroscopy Diagnosis of Neurological Disease. New York: Marcel Dekker, 1999.
2. Gupta RK, Lufkin RB: MR Imaging and Spectroscopy of Central Nervous System Infection. New York: Plenum Press, 2001.
3. Kim EE, Jackson EF, Bandettini PA, et al: Molecular Imaging in Oncology: PET, MRI, and MRS. Berlin: Springer Verlag, 1999.
4. Tofts P: Quantitative MRI of the Brain: Measuring Changes Caused by Disease. Chichester, UK: John Wiley & Sons, 2003.
5. Filippi M, Arnold DL, Comi G: Magnetic Resonance Spectroscopy in Multiple Sclerosis. Milan: Springer Verlag Telos, 2001.
6. Young IR, Grant DM, Harris RK: Methods in Biomedical Magnetic Resonance Imaging and Spectroscopy. Chichester, UK: John
Wiley & Sons, 2000.
7. Mukherji SK: Clinical Applications of Magnetic Resonance Spectroscopy. New York: John Wiley & Sons, 1998.
8. Higgins CB, Ingwall JS, Pohost GM: Current and Future Applications of Magnetic Resonance in Cardiovascular Disease. Armonk,
NY: Futura, 1998.
9. Salibi N, Brown MA: Clinical MR Spectroscopy: First Pinciples. New York: Wiley-Liss, 1997.
10. Nasrallah HE, Pettegrew JW: NMR Spectroscopy in Psychiatric Disorders. Arlington VA: American Psychiatric Publishing, 1995.
11. Gillies RJ: NMR in Physiology and Biomedicine. San Diego: Academic Press, 1994.
12. Boesch C, Kreis R: Imaging and spectroscopy of muscle, In Young IR, Grant DM, Harris RK (eds): Methods in Biomedical
Magnetic Resonance Imaging and Spectroscopy (Encyclopedia of Nuclear Magnetic Resonance). Chichester UK: John Wiley & Sons,
2000, pp. 1307-1316.
13. Rinck PA: Magnetic Resonance in Medicine. The Basic Textbook of the European Magnetic Resonance Forum. Oxford: Blackwell
Scientific, 1993.
14. Howe FA, Opstad KS: 1H MR spectroscopy of brain tumours and masses. NMR Biomed 16:123-131, 2003.
15. Novotny EJ, Jr., Fulbright RK, Pearl PL, et al: Magnetic resonance spectroscopy of neurotransmitters in human brain. Ann Neurol
54:S25-S31, 2003.
16. Gulati S, Shah T, Menon S, et al: Magnetic resonance spectroscopy in pediatric neurology. Indian J Pediatr 70:317-325, 2003.
17. Briellmann RS, Pell GS, Wellard RM, et al: MR imaging of epilepsy: state of the art at 1.5 T and potential of 3 T. Epileptic Disord
5:3-20, 2003.
18. Nelson SJ: Multivoxel magnetic resonance spectroscopy of brain tumors. Mol Cancer Ther 2:497-507, 2003. Medline
Similar articles
19. Port RE, Wolf W: Noninvasive methods to study drug distribution. Invest New Drugs 21:157-168, 2003. Medline
Similar articles
20. Vaidya SJ, Payne GS, Leach MO, et al: Potential role of magnetic resonance spectroscopy in assessment of tumour response in
childhood cancer. Eur J Cancer 39:728-735, 2003. Medline Similar articles
21. Shulman RG, Hyder F, Rothman DL: Biophysical basis of brain activity: implications for neuroimaging. Q Rev Biophys 35:287-325,
8 of 21 15-04-2010 20:29
22. Kwock L, Smith JK, Castillo M, et al: Clinical applications of proton MR spectroscopy in oncology. Technol Cancer Res Treat
23. Katz-Brull R, Lavin PT, Lenkinski RE: Clinical utility of proton magnetic resonance spectroscopy in characterizing breast lesions. J
Natl Cancer Inst 94:1197-1203, 2002. Medline Similar articles
24. Del Sole A, Gambini A, Falini A, et al: In vivo neurochemistry with emission tomography and magnetic resonance spectroscopy:
clinical applications. Eur Radiol 12:2582-2599, 2002. Medline Similar articles
25. Adusumilli S, Pretorius ES: Magnetic resonance imaging of prostate cancer. Semin Urol Oncol 20:192-210, 2002. Medline
Similar articles
26. Malhi GS, Valenzuela M, Wen W, et al: Magnetic resonance spectroscopy and its applications in psychiatry. Aust N Z J Psychiatry
27. Shen J, Rothman DL: Magnetic resonance spectroscopic approaches to studying neuronal: glial interactions. Biol Psychiatry
28. Lyoo IK, Renshaw PF: Magnetic resonance spectroscopy: current and future applications in psychiatric research. Biol Psychiatry
29. Kuzniecky RI, Knowlton RC: Neuroimaging of epilepsy. Semin Neurol 22:279-288, 2002. Medline Similar articles
30. Drost DJ, Riddle WR, Clarke GD: Proton magnetic resonance spectroscopy in the brain: report of AAPM MR Task Group #9. Med
Phys 29:2177-2197, 2002. Medline Similar articles
31. Avison MJ, Nath A, Berger JR: Understanding pathogenesis and treatment of HIV dementia: a role for magnetic resonance?
Trends Neurosci 25:468-473, 2002. Medline Similar articles
32. von Kienlin M, Beer M, Greiser A, et al: Advances in human cardiac 31P-MR spectroscopy: SLOOP and clinical applications. J
33. Taylor-Robinson SD: Applications of magnetic resonance spectroscopy to chronic liver disease. Clin Med 1:54-60, 2001.
34. Zakian KL, Koutcher JA, Ballon D, et al: Developments in nuclear magnetic resonance imaging and spectroscopy: application to
radiation oncology. Semin Radiat Oncol 11:3-15, 2001. Medline Similar articles
35. Filippi M: In-vivo tissue characterization of multiple sclerosis and other white matter diseases using magnetic resonance based
techniques. J Neurol 248:1019-1029, 2001. Medline Similar articles
36. Ross B, Bluml S: Magnetic resonance spectroscopy of the human brain. Anat Rec 265:54-84, 2001. Medline
Similar articles
37. Cecil KM, Jones BV: Magnetic resonance spectroscopy of the pediatric brain. Top Magn Reson Imaging 12:435-452, 2001.
38. Huppi PS: MR imaging and spectroscopy of brain development. Magn Reson Imaging Clin N Am 9:1-17, vii, 2001. Medline
Similar articles
39. Hunter JV, Wang ZJ: MR spectroscopy in pediatric neuroradiology. Magn Reson Imaging Clin N Am 9:165-89, ix, 2001.
40. Burtscher IM, Holtas S: Proton magnetic resonance spectroscopy in brain tumours: clinical applications. Neuroradiology
41. Roach M, III, Kurhanewicz J, Carroll P: Spectroscopy in prostate cancer: hope or hype? Oncology (Huntingt) 15:1399-1410, 2001.
42. Evelhoch JL, Gillies RJ, Karczmar GS, et al: Applications of magnetic resonance in model systems: cancer therapeutics.
43. Gillies RJ, Bhujwalla ZM, Evelhoch J, et al: Applications of magnetic resonance in model systems: tumor biology and physiology.
44. Inder TE, Huppi PS: In vivo studies of brain development by magnetic resonance techniques. Ment Retard Dev Disabil Res Rev
45. Koutcher JA, Zakian K, Hricak H: Magnetic resonance spectroscopic studies of the prostate. Mol Urol 4:143-152, 2000.
46. Lee PL, Gonzalez RG: Magnetic resonance spectroscopy of brain tumors. Curr Opin Oncol 12:199-204, 2000. Medline
Similar articles
47. Brooks DJ: Morphological and functional imaging studies on the diagnosis and progression of Parkinson's disease. J Neurol 247
Suppl 2:II11-II18, 2000.
page 486
page 487
48. Argov Z, Arnold DL: MR spectroscopy and imaging in metabolic myopathies. Neurol Clin 18:35-52, 2000. Medline
Similar articles
49. Burlina AP, Aureli T, Bracco F, et al: MR spectroscopy: a powerful tool for investigating brain function and neurological diseases.
Neurochem Res 25:1365-1372, 2000. Medline Similar articles
50. Pomper MG, Port JD: New techniques in MR imaging of brain tumors. Magn Reson Imaging Clin N Am 8:691-713, 2000.
51. Maheshwari SR, Fatterpekar GM, Castillo M, et al: Proton MR spectroscopy of the brain. Semin Ultrasound CT MR 21:434-451,
52. Stefano ND, Narayanan S, Matthews PM, et al: Proton MR spectroscopy to assess axonal damage in multiple sclerosis and other
9 of 21 15-04-2010 20:29
white matter disorders. J Neurovirol 6 Suppl 2:S121-S129, 2000.

53. Beckmann N, Hof RP, Rudin M: The role of magnetic resonance imaging and spectroscopy in transplantation: from animal models
to man. NMR Biomed 13:329-348, 2000. Medline Similar articles
54. Stubbs M: Application of magnetic resonance techniques for imaging tumour physiology. Acta Oncol 38:845-853, 1999.
55. Rudin M, Beckmann N, Porszasz R, et al: In vivo magnetic resonance methods in pharmaceutical research: current status and
perspectives. NMR Biomed 12:69-97, 1999. Medline Similar articles
56. Shevell MI, Ashwal S, Novotny E: Proton magnetic resonance spectroscopy: clinical applications in children with nervous system
diseases. Semin Pediatr Neurol 6:68-77, 1999. Medline Similar articles
57. Achten E: Aspects of proton MR spectroscopy in the seizure patient. Neuroimaging Clin N Am 8:849-862, 1998. Medline
Similar articles
58. Bell JD, Bhakoo KK: Metabolic changes underlying 31P MR spectral alterations in human hepatic tumours. NMR Biomed
11:354-359, 1998.
59. Castillo M, Kwock L: Proton MR spectroscopy of common brain tumors. Neuroimaging Clin N Am 8:733-752, 1998. Medline
Similar articles
60. Cecil KM, Lenkinski RE: Proton MR spectroscopy in inflammatory and infectious brain disorders. Neuroimaging Clin N Am
61. Ricci PE, Jr.: Proton MR spectroscopy in ischemic stroke and other vascular disorders. Neuroimaging Clin N Am 8:881-900, 1998.
62. Matthews PM, Taivassalo T: Applications of magnetic resonance spectroscopy to diagnosis and monitoring of mitochondrial
disease. Ital J Neurol Sci 18:341-351, 1997. Medline Similar articles
63. Henriksen O: MR spectroscopy in clinical research. Acta Radiol 35:96-116, 1994. Medline Similar articles
64. Bottomley PA: MR spectroscopy of the human heart: The status and the challenges. Radiology 191:593-612, 1994. Medline
Similar articles
65. Freeman R: Magnetic Resonance in Chemistry and Medicine. Oxford: Oxford Press, 2003.
66. Lambert JB, Mazzola EP: Nuclear Magnetic Resonance Spectroscopy: An Introduction to Principles, Applications, and
Experimental Methods. Upper Saddle River, NJ: Prentice Hall, 2003.
67. Gadian DG: NMR and Its Applications to Living Systems. Oxford: Oxford University Press, 2002.
68. Levitt MH: Spin Dynamics: Basics of Nuclear Magnetic Resonance. New York: John Wiley & Sons, 2001.
69. Kuperman V: Magnetic Resonance Imaging: Physical Principles and Applications. San Diego: Academic Press, 2000.
70. Hore PJ, Jones JA, Wimperis S: NMR: The Toolkit. Oxford: Oxford University Press, 2000.
71. De Graaf RA: In vivo NMR Spectroscopy: Principles and Techniques. Chichester UK: John Wiley & Sons, 1999.
72. Haacke EM, Brown RW, Thompson MR, et al: Magnetic Resonance Imaging: Physical Principles and Sequence Design. New
York: John Wiley & Sons, 1999.
73. Macomber RS: A Complete Introduction to Modern NMR Spectroscopy. Philadelphia: Wiley-Interscience, 1998.
74. Friebolin H: Basic One- and Two-Dimensional NMR Spectroscopy. Chichester UK: John Wiley & Sons, 1998.
75. Günther H: NMR Spectroscopy. Basic Principles, Concepts, and Applications in Chemistry. Chichester UK: John Wiley & Sons,
1994.
76. Sanders JKM, Hunter BK: Modern NMR Spectroscopy. Oxford: Oxford University Press, 1993.
77. Brigham EO: The Fast Fourier Transform and Applications. Englewood Cliffs, NJ: Prentice Hall, 1988.
78. Slichter CP: Principles of Magnetic Resonance. Berlin: Springer-Verlag, 1990.
79. Chen CN, Hoult DI: Biomedical Magnetic Resonance Technology. Bristol: Adam Hilger, 1989.
80. Freeman R: A Handbook of Nuclear Magnetic Resonance. Harlow, UK: Longman Scientific & Technical, 1987.
81. Derome AE: Modern NMR Techniques for Chemistry Research. Oxford: Pergamon Press, 1987.
82. Ernst RR, Bodenhausen G, Wokaun A: Principles of Nuclear Magnetic Resonance in One and Two Dimensions. Oxford: Clarendon
Press, 1987.
83. Abragam A: The Principle of Nuclear Magnetism. Oxford: Oxford University Press, 1961.
84. Nicolay K, Braun KP, Graaf RA, et al: Diffusion NMR spectroscopy. NMR Biomed 14:94-111, 2001. Medline
Similar articles
85. Boesch C: Molecular aspects of magnetic resonance imaging and spectroscopy. Mol Aspects Med 20:185-318, 1999.
86. Kwock L: Localized MR spectroscopy: basic principles. Neuroimaging Clin N Am 8:713-731, 1998. Medline Similar articles
87. Bachert P: Pharmakokinetics using fluorine NMR in vivo. Prog NMR Spectroscopy 33:1-56, 1998.
88. Freeman DM, Hurd R: Decoupling: theory and practice. II. State of the art: in vivo applications of decoupling. NMR Biomed
89. Freeman R, Kupce E: Decoupling: theory and practice. I. Current methods and recent concepts. NMR Biomed 10:372-380, 1997.
90. Pohmann R, von Kienlin M, Haase A: Theoretical evaluation and comparison of fast chemical shift imaging methods. J Magn
10 of 21 15-04-2010 20:29
Reson 129:145-160, 1997. Medline Similar articles

91. Cox IJ: Development and applications of in vivo clinical magnetic resonance spectroscopy. Prog Biophys Mol Biol 65:45-81, 1996.
92. Cousins JP: Clinical MR spectroscopy: fundamentals, current applications, and future potential. Am J Roentgenol 164:1337-1347,
1995.
93. Nicolay K, van der Toorn A, Dijkhuizen RM: In vivo diffusion spectroscopy. An Overview. NMR Biomed 8:365-374, 1995.
94. Bottomley PA, Hardy CJ, Roemer PB, et al: Problems and expediencies in human 31P spectroscopy. The definition of localized
volumes, dealing with saturation and the technique-dependence of quantification. NMR Biomed 2:284-289, 1989.
13
95. Beckmann N, Turkalj I, Seelig J, et al: C NMR for the assessment of human brain glucose metabolism in vivo. Biochemistry
96. De Graaf RA, Mason GF, Patel AB, et al: In vivo 1H-[13C]-NMR spectroscopy of cerebral metabolism. NMR Biomed 16:339-357,
2003.
97. Dijkhuizen RM, De Graaf RA, Garwood M, et al: Spatial assessment of the dynamics of lactate formation in focal ischemic rat
brain. J Cereb Blood Flow Metab 19:376-379, 1999. Medline Similar articles
98. Gruetter R, Adriany G, Choi IY, et al: Localized in vivo 13C NMR spectroscopy of the brain. NMR Biomed 16:313-338, 2003.
13
99. Morris P, Bachelard H: Reflections on the application of C-MRS to research on brain metabolism. NMR Biomed 16:303-312,
2003.
100. Ross B, Lin A, Harris K, et al: Clinical experience with 13C MRS in vivo. NMR Biomed 16:358-369, 2003.
101. Shen J, Petersen KF, Behar KL, et al: Determination of the rate of the glutamate/glutamine cycle in the human brain by in vivo 13C
NMR. Proc Natl Acad Sci USA 96:8235-8240, 1999.
102. vanZijl PCM, Chesnick AS, DesPres D, et al: In vivo proton spectroscopy and spectroscopic imaging of (1-13C)-glucose and its
metabolic products. Magn Reson Med 30:544-551, 1993. Medline Similar articles
103. Callaghan PT, Eccles CD: Sensitivity and resolution in NMR imaging. J Magn Reson 71:426-445, 1987.
104. Maudsley AA: Sensitivity in Fourier imaging, J Magn Reson 68:363-366, 1986.
105. Macovski A: Noise in MRI. Magn Reson Med 36:494-497, 1996. Medline Similar articles
106. Edelstein WA, Glover GH, Hardy CJ, et al: The intrinsic signal-to-noise ratio in NMR imaging. Magn Reson Med 3:604-618, 1986.
107. Chen CN, Sank VJ, Cohen SM, et al: The field dependence of NMR imaging. I. Laboratory assessment of signal-to-noise ratio
and power deposition. Magn Reson Med 3:722-729, 1986. Medline Similar articles
108. Collins CM, Smith MB: Signal-to-noise ratio and absorbed power as functions of main magnetic field strength, and definition of
"90 degrees " RF pulse for the head in the birdcage coil. Magn Reson Med 45:684-691, 2001. Medline Similar articles
109. Ohliger MA, Grant AK, Sodickson DK: Ultimate intrinsic signal-to-noise ratio for parallel MRI: electromagnetic field
considerations. Magn Reson Med 50:1018-1030, 2003. Medline Similar articles
110. Murphy BW, Carson PL, Ellis JH, et al: Signal-to-noise measures for magnetic resonance imagers. Magn Reson Imaging
111. Kowalski ME, Jin JM, Chen J: Computation of the signal-to-noise ratio of high-frequency magnetic resonance imagers. IEEE
Trans Biomed Eng 47:1525-1533, 2000. Medline Similar articles
112. Pietz J, Kreis R, Rupp A, et al: Large neutral amino acids block phenylalanine transport into brain tissue in patients with
phenylketonuria. J Clin Invest 103:1169-1178, 1999. Medline Similar articles
113. Kreis R, Pietz J, Penzien J, et al: Identification and quantitation of phenylalanine in the brain of patients with phenyketonuria by
means of localized in vivo 1H magnetic resonance spectroscopy. J Magn Reson Series B 107:242-251, 1995.
114. Venkatasubramanian PN, Shen YJ, Wyrwicz AM: Characterization of the cerebral distribution of general anesthetics in vivo by
two-dimensional 19F chemical shift imaging. Magn Reson Med 35:626-630, 1996.
115. Menon DK, Lockwood GG, Peden CJ, et al: In vivo fluorine-19 magnetic resonance spectroscopy of cerebral halothane in
postoperative patients: preliminary results. Magn Reson Med 30:680-684, 1993. Medline Similar articles
116. Preece NE, Challands J, Williams SCR: 19F NMR studies of enflurane elimination and metabolism. NMR Biomed 5:101-106,
1992.
117. Lockwood GG, Dob DP, Bryant DJ, et al: Magnetic resonance spectroscopy of isoflurane kinetics in humans. Part II: Functional
localization. Br J Anaesth 79:586-589, 1997. Medline Similar articles
118. Langkowski JH, Wieland J, Bomsdorf H, et al: Pre-operative localized in vivo proton spectroscopy in cerebral tumors at 4.0
Tesla-First results. Magn Reson Imaging 7:547-555, 1989.
119. Barfuss H, Fischer H, Hentschel D, et al: In vivo magnetic resonance imaging and spectroscopy with a 4T whole-body magnet.
NMR Biomed 3:31-45, 1990.
120. Middleton DA, Hockings PD, Glen S, et al: Image directed proton spectroscopy of gerbil brain at 7 tesla. NMR Biomed
8:118-126, 1995.
121. Hetherington HP, Pan JW, Chu WJ, et al: Biological and clinical MRS at ultra-high field. NMR Biomed 10:360-371, 1997.
11 of 21 15-04-2010 20:29
page 487
page 488
1
122. Gruetter R, Weisdorf SA, Rajanayagan V, et al: Resolution improvements in in vivo H NMR spectra with increased magnetic field
strength. J Magn Reson 135:260-264, 1998.
123. Mlynarik V, Gruber S, Moser E: Proton T (1) and T (2) relaxation times of human brain metabolites at 3 Tesla. NMR Biomed
14:325-331, 2001.
124. Tkac I, Andersen P, Adriany G, et al: In vivo 1H NMR spectroscopy of the human brain at 7 T. Magn Reson Med 46:451-456,
2001.
125. Tomasi DG, Ernst T: Echo planar imaging at 4 Tesla with minimum acoustic noise. J Magn Reson Imaging 18:128-130, 2003.
126. Ugurbil K, Adriany G, Andersen P, et al: Ultrahigh field magnetic resonance imaging and spectroscopy. Magn Reson Imaging
127. Burgess RE, Yu Y, Abduljalil AM, et al: High signal-to-noise FLASH imaging at 8 Tesla. Magn Reson Imaging 17:1099-1103,
1999.
128. Bartha R, Drost DJ, Menon RS, et al: Comparison of the quantification precision of human short echo time (1)H spectroscopy at
1.5 and 4.0 Tesla. Magn Reson Med 44:185-192, 2000.
129. Kreis R: Quantitative localized 1H MR spectroscopy for clinical use. Prog NMR Spectroscopy 31:155-195, 1997.
130. Alger JR, Symko SC, Bizzi A, et al: Absolute quantitation of short TE brain 1H-MR spectra and spectroscopic imaging data. J
Comput Assist Tomogr 17:191-199, 1993.
131. Barker PB, Soher BJ, Blackband SJ, et al: Quantitation of proton NMR spectra of the human brain using tissue water as an
internal concentration reference. NMR Biomed 6:89-94, 1993. Medline Similar articles
132. Ernst T, Kreis R, Ross BD: Absolute quantitation of water and metabolites in the human brain. I. Compartments and water. J
Magn Reson Series B 102:1-8, 1993.
133. Kreis R, Ernst T, Ross BD: Absolute quantitation of water and metabolites in the human brain. II. Metabolite concentrations. J
134. Michaelis T, Merboldt KD, Bruhn H, et al: Absolute concentrations of metabolites in the adult human brain in vivo: Quantification of
localized proton MR spectra. Radiology 187:219-227, 1993. Medline Similar articles
135. Provencher SW: Estimation of metabolite concentration from localized in vivo proton NMR spectra. Magn Reson Med
136. Henriksen O: In vivo quantitation of metabolite concentrations in the brain by means of proton MRS. NMR Biomed 8:139-148,
137. Bovee WMMJ, deBeer R, vanOrmondt D, et al: Quantitation in whole body MRS, In Grant DM, Harris RK (eds): Encyclopedia of
Nuclear Magnetic Resonance. Chichester, UK: Wiley, 1996, pp. 3900-3906.
138. Soher BJ, van Zijl PC, Duyn JH, et al: Quantitative proton MR spectroscopic imaging of the human brain. Magn Reson Med
139. Barantin L, Le Pape A, Akoka S: A new method for absolute quantitation of MRS metabolites. Magn Reson Med 38:179-182,
140. Keevil SF, Barbiroli B, Brooks JC, et al: Absolute metabolite quantification by in vivo NMR spectroscopy: II. A multicentre trial of
protocols for in vivo localised proton studies of human brain. Magn Reson Imaging 16:1093-1106, 1998. Medline
Similar articles
141. Podo F, Henriksen O, Bovee WM, et al: Absolute metabolite quantification by in vivo NMR spectroscopy: I. Introduction, objectives
and activities of a concerted action in biomedical research. Magn Reson Imaging 16:1085-1092, 1998. Medline
Similar articles
142. Schirmer T, Auer DP: On the reliability of quantitative clinical magnetic resonance spectroscopy of the human brain. NMR Biomed
143. Wiedermann D, Schuff N, Matson GB, et al: Short echo time multislice proton magnetic resonance spectroscopic imaging in
human brain: metabolite distributions and reliability. Magn Reson Imaging 19:1073-1080, 2001. Medline Similar articles
144. Hofmann L, Slotboom J, Jung B, et al: Quantitative 1H-magnetic resonance spectroscopy of human brain: Influence of
composition and parameterization of the basis set in linear combination model-fitting. Magn Reson Med 48:440-453, 2002.
145. Vanhamme L, Van Huffel S, Van Hecke P, et al: Time-domain quantification of series of biomedical magnetic resonance
spectroscopy signals. J Magn Reson 140:120-130, 1999. Medline Similar articles
146. Hajek M, Burian M, Dezortova M: Application of LCModel for quality control and quantitative in vivo 1H MR spectroscopy by short
echo time STEAM sequence. MAGMA 10:6-17, 2000.
147. Cavassila S, Deval S, Huegen C, et al: Cramer-Rao bounds: an evaluation tool for quantitation. NMR Biomed 14:278-283, 2001.
148. Ebel A, Soher BJ, Maudsley AA: Assessment of 3D proton MR echo-planar spectroscopic imaging using automated spectral
analysis. Magn Reson Med 46:1072-1078, 2001.
149. Keevil SF, Newbold MC: The performance of volume selection sequences for in vivo NMR spectroscopy: implications for
quantitative MRS. Magn Reson Imaging 19:1217-1226, 2001. Medline Similar articles
150. Mierisova S, Ala-Korpela M: MR spectroscopy quantitation: a review of frequency domain methods. NMR Biomed 14:247-259,
151. Vanhamme L, Sundin T, Hecke PV, et al: MR spectroscopy quantitation: a review of time-domain methods. NMR Biomed
12 of 21 15-04-2010 20:29
152. Chen W, Adriany G, Zhu XH, et al: Detecting natural abundance carbon signal of NAA metabolite within 12-cm3 localized volume
of human brain using 1H-13C NMR spectroscopy. Magn Reson Med 40:180-184, 1998.
153. Fitzpatrick SM, Hetherington HP, Behar KL, et al: The flux from glucose to glutamate in the rat brain in vivo as determined by
1
H-observed, 13C-edited NMR spectroscopy. J Cereb Blood Flow Metab 10:170-179, 1990.
154. Pan JW, Mason GF, Vaughan JT, et al: 13C editing of glutamate in human brain using J-refocused coherence transfer
spectroscopy at 4.1 T. Magn Reson Med 37:355-358, 1997. Medline Similar articles
1
155. Pfeuffer J, Tkac I, Choi IY, et al: Localized in vivo H NMR detection of neurotransmitter labeling in rat brain during infusion of
[1-13C] D-glucose. Magn Reson Med 41:1077-1083, 1999.
156. Rothman DL, Behar KL, Hetherington HP, et al: 1H-Observe/13C-decouple spectroscopic measurements of lactate and
glutamate in rat brain in vivo. Proc Natl Acad Sci USA 82:1633-1637, 1985. Medline Similar articles
157. Novotny EJ, Avison MJ, Herschkowitz N, et al: In vivo measurement of phenylalanine in human brain by proton nuclear magnetic
resonance spectroscopy. Pediatr Res 37:244-249, 1995. Medline Similar articles
158. Weglage J, Moller HE, Wiedermann D, et al: In vivo NMR spectroscopy in patients with phenylketonuria: clinical significance of
interindividual differences in brain phenylalanine concentrations. J Inherit Metab Dis 21:81-82, 1998. Medline Similar articles
159. Bruhn H, Frahm J, Gyngell ML, et al: Localized proton NMR spectroscopy using stimulated echoes: Applications to human
skeletal muscle in vivo. Magn Reson Med 17:82-94, 1991. Medline Similar articles
160. Damon BM, Hsu AC, Stark HJ, et al: The carnosine C-2 proton's chemical shift reports intracellular pH in oxidative and glycolytic
muscle fibers. Magn Reson Med 49:233-240, 2003. Medline Similar articles
161. Pan JW, Hamm JR, Hetherington HP, et al: Correlation of lactate and pH in human skeletal muscle after exercise by 1H NMR.
Magn Reson Med 20:57-65, 1991.
162. Mori S, Eleff SM, Pilatus U, et al: Proton NMR spectroscopy of solvent-saturable resonances: a new approach to study pH effects
in situ. Magn Reson Med 40:36-42, 1998. Medline Similar articles
163. Wang Z, Noyszewski EA, Leigh JS: In vivo MRS measurement of deoxymyoglobin in human forearms. Magn Reson Med
164. Brillault-Salvat C, Giacomini E, Jouvensal L, et al: Simultaneous determination of muscle perfusion and oxygenation by
interleaved NMR plethysmography and deoxymyoglobin spectroscopy. NMR Biomed 10:315-323, 1997. Medline
Similar articles
165. Chen W, Cho Y, Merkle H, et al: In vitro and in vivo studies of 1H NMR visibility to detect deoxyhemoglobin and deoxymyoglobin
signals in myocardium. Magn Reson Med 42:1-5, 1999.
166. Mole PA, Chung Y, Tran TK, et al: Myoglobin desaturation with exercise intensity in human gastrocnemius muscle. Am J Physiol
277:R173-R180, 1999.
167. Kreis R, Brügger K, Skjelsvik C, et al: Quantitative 1H magnetic resonance spectroscopy of myoglobin de- and re-oxygenation in
skeletal muscle: Reproducibility and effects of localization and disease. Magn Reson Med 46:240-248, 2001.
168. Farrow NA, Richards JH, Ross BD: pH measurements in vivo in whole body systems. In Grant DM, Harris RK (eds): Encyclopedia
of Nuclear Magnetic Resonance. Chichester, UK: Wiley, 1996, pp. 3562-3565.
169. Petroff OA, Prichard JW: Cerebral pH by NMR. Lancet 2:105-106, 1983. Medline Similar articles
31
170. Moon RB, Richards JH: Determination of intracellular pH by P magnetic resonance. J Biol Chem 248:7276-7278, 1973.
171. Barker PB, Butterworth EJ, Boska MD, et al: Magnesium and pH imaging of the human brain at 3.0 Tesla. Magn Reson Med
41:400-406, 1999.
172. Golding EM, Golding RM: Interpretation of 31P MRS spectra in determining intracellular free magnesium and potassium ion
concentrations. Magn Reson Med 33:467-474, 1995.
173. Gupta RK, Moore RD: 31P NMR studies of intracellular free Mg2+ in intact frog skeletal muscle. J Biol Chem 255:3987-3993,
1980.
174. Helpern JA, Vande Linde AM, Welch KM, et al: Acute elevation and recovery of intracellular [Mg2+] following human focal cerebral
ischemia. Neurology 43:1577-1581, 1993. Medline Similar articles
175. Bottomley PA, Ouwerkerk R, Lee RF, et al: Four-angle saturation transfer (FAST) method for measuring creatine kinase reaction
rates in vivo. Magn Reson Med 47:850-863, 2002. Medline Similar articles
176. Koretsky AP, Wang S, Klein MP, et al: 31P NMR saturation transfer measurements of phosphorus exchange reactions in rat heart
and kidney in situ. Biochemistry 25:77-84, 1986. Medline Similar articles
177. Le Rumeur E, Le Tallec N, Kernec F, et al: Kinetics of ATP to ADP beta-phosphoryl conversion in contracting skeletal muscle by
in vivo 31P NMR magnetization transfer. NMR Biomed 10:67-72, 1997.
178. Rees D, Smith MB, Harley J, et al: In vivo functioning of creatine phosphokinase in human forearm muscle, studied by 31P NMR
saturation transfer. Magn Reson Med 9:39-52, 1989.
31
179. van Dorsten FA, Nederhoff MG, Nicolay K, et al: P NMR studies of creatine kinase flux in M-creatine kinase-deficient mouse
heart. Am J Physiol 275:H1191-H1199, 1998.
180. Ernst RR, Anderson WA: Application of Fourier transform spectroscopy to magnetic resonance. Rev Sci Instrum 37:93-102,
13 of 21 15-04-2010 20:29
1966.
242:190-191, 1973.
182. Kumar A, Welti D, Ernst RR: NMR fourier zeugmatography. J Magn Reson 18:69-83, 1975.
184. Mansfield P, Chapman B: Active magnetic screening of gradient coils in NMR imaging. J Magn Reson 66:573-576, 1986.
185. Mansfield P, Chapman B: Multishield active magnetic screening of coil structures in NMR. J Magn Reson 72:211-223, 1987.
page 488
page 489
186. Harvey PR, Katznelson E: Modular gradient coil: A new concept in high-performance whole-body gradient coil design. Magn
187. Seeger U, Klose U, Seitz D, et al: Proton spectroscopy of human brain with very short echo time using high gradient amplitudes.
188. Barkhuijsen H, deBeer R, Bovee WMMJ, et al: Application of Linear Prediction and Singular Value Decomposition (LPSVD) to
Determine NMR Frequencies and Intensities from the FID. Magn Reson Med 2:86-89, 1985. Medline Similar articles
189. Overloop K, Van Hecke P, Vanstapel F, et al: Evaluation of signal processing methods for the quantification of a multi-exponential
signal: the glycogen 13C-1 NMR signal. NMR Biomed 9:315-321, 1996.
190. Vanhamme L, vandenBoogaart A, Van Huffel S: Improved method for accurate and efficient quantification of MRS data with use
of prior knowledge. J Magn Reson 129:35-43, 1997. Medline Similar articles
191. Mierisova S, van den Boogaart A, Tkac I, et al: New approach for quantitation of short echo time in vivo 1H MR spectra of brain
using AMARES. NMR Biomed 11:32-39, 1998.
192. Slotboom J, Boesch C, Kreis R: Versatile frequency domain fitting using time domain models and prior knowledge. Magn Reson
193. Soher BJ, Young K, Govindaraju V, et al: Automated spectral analysis III: application to in vivo proton MR spectroscopy and
spectroscopic imaging. Magn Reson Med 40:822-831, 1998. Medline Similar articles
194. Young K, Soher BJ, Maudsley AA: Automated spectral analysis II: application of wavelet shrinkage for characterization of
non-parameterized signals. Magn Reson Med 40:816-821, 1998. Medline Similar articles
195. Young K, Govindaraju V, Soher BJ, et al: Automated spectral analysis I: formation of a priori information by spectral simulation.
196. Naressi A, Couturier C, Castang I, et al: Java-based graphical user interface for MRUI, a software package for quantitation of in
vivo/medical magnetic resonance spectroscopy signals. Comput Biol Med 31:269-286, 2001. Medline Similar articles
197. Romano R, Motta A, Camassa S, et al: A new time-domain frequency-selective quantification algorithm. J Magn Reson
198. Hofmann L, Slotboom J, Boesch C, et al: Characterization of the macromolecule baseline in localized 1H-MR spectra of human
brain. Magn Reson Med 46:855-863, 2001.
1
199. Behar KL, Rothman DL, Spencer DD, et al: Analysis of macromolecule resonances in H NMR spectra of human brain. Magn
Reson Med 32:294-302, 1994.
200. Mader I, Seeger U, Karitzky J, et al: Proton magnetic resonance spectroscopy with metabolite nulling reveals regional differences
of macromolecules in normal human brain. J Magn Reson Imaging 16:538-546, 2002. Medline Similar articles
201. Soher BJ, Young K, Maudsley AA: Representation of strong baseline contributions in 1H MR spectra. Magn Reson Med
45:966-972, 2001.
202. Terpstra M, Andersen PM, Gruetter R: Localized eddy current compensation using quantitative field mapping. J Magn Reson
203. Lin C, Wendt RE, III, Evans HJ, et al: Eddy current correction in volume-localized MR spectroscopy. J Magn Reson Imaging
204. Roebuck JR, Hearshen DO, O'Donnell M, et al: Correction of phase effects produced by eddy currents in solvent suppressed
1
H-CSI. Magn Reson Med 30:277-282, 1993.
205. Riddle WR, Gibbs SJ, Willcott MR: Removing effects of eddy currents in proton MR spectroscopy. Med Phys 19:501-509, 1992.
206. Boesch C, Gruetter R, Martin E: Temporal and spatial analysis of fields generated by eddy currents in superconducting magnets:
Optimization of corrections and quantitative characterization of magnet / gradient system. Magn Reson Med 20:268-284, 1991.
207. Klose U: In vivo proton spectroscopy in presence of eddy currents. Magn Reson Med 14:26-30, 1990. Medline
Similar articles
208. Kim YS, Cho ZH: Eddy-current-compensated field-inhomogeneity mapping in NMR imaging. J Magn Reson 78:459-471, 1988.
209. Malko JA, Nelson RC: Controlled eddy currents: Applications to MR imaging. J Comput Assist Tomogr 11:1044-1049, 1987.
210. Helms G, Piringer A: Restoration of motion-related signal loss and line-shape deterioration of proton MR spectra using the
residual water as intrinsic reference. Magn Reson Med 46:395-400, 2001. Medline Similar articles
211. Starcuk Z, Jr., Starcuk Z, Mlynarik V, et al: Low-power water suppression by hyperbolic secant pulses with controlled offsets and
14 of 21 15-04-2010 20:29
delays (WASHCODE). J Magn Reson 152:168-178, 2001. Medline Similar articles

212. Males RG, Vigneron DB, Star-Lack J, et al: Clinical application of BASING and spectral/spatial water and lipid suppression
pulses for prostate cancer staging and localization by in vivo 3D 1H magnetic resonance spectroscopic imaging. Magn Reson Med
43:17-22, 2000.
213. De Graaf RA, Luo Y, Garwood M, et al: B1-insensitive, single-shot localization and water suppression. J Magn Reson Series B
113:35-45, 1996.
214. Ernst T, Hennig J: Improved water suppression for localized in vivo 1H spectroscopy. J Magn Reson Series B 106:181-186,
1995.
215. Ogg RJ, Kingsley PB, Taylor JS: WET, a T1- and B1-insensitive water-suppression method for in vivo localized 1H NMR
spectroscopy. J Magn Reson B 104:1-10, 1994.
1
216. Haase A, Frahm J, Hanicke W, et al: H NMR chemical shift selective (CHESS) imaging. Phys Med Biol 30:341-344, 1985.
217. Hurd RE, Gurr D, Sailasuta N: Proton spectroscopy without water suppression: the oversampled J-resolved experiment. Magn
218. Clayton DB, Elliott MA, Leigh JS, et al: 1H spectroscopy without solvent suppression: characterization of signal modulations at
short echo times. J Magn Reson 153:203-209, 2001. Medline Similar articles
13
219. Bendall MR, den Hollander JA, Arias-Mendoza F, et al: Application of multipulse NMR to observe C-labeled metabolites in
biological systems. Magn Reson Med 2:56-64, 1985.
13
220. Norris DG, Schuff N, Leibfritz D: INEPT enhanced C spectroscopy using double-tuned surface coils. J Magn Reson
78:362-366, 1988.
221. Mueller S, Beckmann N: 13C spectroscopic imaging. A simple approach to in vivo 13C investigations. Magn Reson Med
12:400-406, 1989.
222. Bomsdorf H, Röschmann P, Wieland J: Sensitivity enhancement in whole-body natural abundance 13C spectroscopy using
13
C/1H double-resonance techniques at 4 tesla. Magn Reson Med 22:10-22, 1991.
223. Saner M, McKinnon G, Boesiger P: Glycogen detection by in vivo 13C NMR: A comparison of proton decoupling and polarization
transfer. Magn Reson Med 28:65-73, 1992.
224. Ende G, Bachert P: Dynamic 13C-1H nuclear polarization of lipid methylene resonances applied to broadband-decoupled in vivo
13
C MR spectroscopy of human breast and calf tissue. Magn Reson Med 30:415-423, 1993.
225. Artemov D, Bhujwalla ZM, Glickson JD: In vivo selective measurements of [lcub ]1-13C[rcub ]-glucose metabolism in tumors by
heteronuclear cross polarization. Magn Reson Med 33:151-155, 1995. Medline Similar articles
226. Gruetter R, Adriany G, Merkle H, et al: Broadband decoupled, 1H-localized 13C MRS of the human brain at 4 Tesla. Magn Reson
Med 36:659-664, 1996.
227. Gonen O, Mohebbi A, Stoyanova R, et al: In vivo phosphorus polarization transfer and decoupling from protons in three-
dimensional localized nuclear magnetic resonance spectroscopy of human brain. Magn Reson Med 37:301-306, 1997. Medline
Similar articles
228. Bachelard H: Landmarks in the application of 13C-magnetic resonance spectroscopy to studies of neuronal/glial relationships.
Dev Neurosci 20:277-288, 1998.
229. Lee SM, Tzou DL, Chiou GJY, et al: An alternative scheme for 13C chemical-shift imaging via inverse detection of protons through
an MILS (or inverse SLIM) technique. J Magn Reson 130:102-110, 1998.
230. van den Bergh AJ, van den Boogert HJ, Heerschap A: Heteronuclear cross polarization for enhanced sensitivity of in vivo 13C MR
spectroscopy on a clinical 1.5 T MR system. J Magn Reson 135:93-98, 1998.
231. Watanabe H, Ishihara Y, Okamoto K, et al: In vivo 3D localized 13C spectroscopy using modified INEPT and DEPT. J Magn
Reson 134:214-222, 1998.
232. Payne GS, Leach MO: Surface-coil polarization transfer for monitoring tissue metabolism in vivo. Magn Reson Med 43:510-516,
233. Jung WI, Bunse M, Lutz O: Quantitative evaluation of the lactate signal loss and its spatial dependence in press localized (1)H
NMR spectroscopy. J Magn Reson 152:203-213, 2001. Medline Similar articles
234. Pan JW, De Graaf RA, Petersen KF, et al: [2,4-13 C2 ]-beta-Hydroxybutyrate metabolism in human brain. J Cereb Blood Flow
Metab 22:890-898, 2002. Medline Similar articles
235. Ackerman JJH, Grove TH, Wong GG, et al: Mapping of metabolites in whole animals by 31P NMR using surface coils. Nature
283:167-170, 1980.
236. Evelhoch JL, Crowley MG, Ackerman JJH: Signal-to-noise optimization and observed volume localization with circular surface
coils. J Magn Reson 56:110-124, 1984.
237. Haase A, Haenicke W, Frahm J: The influence of experimental parameters in surface-coil NMR. J Magn Reson 56:401-412,
1984.
238. Garwood M, Ugurbil K, Rath AR, et al: Magnetic resonance imaging with adiabatic pulses using a single surface coil for RF
transmission and signal detection. Magn Reson Med 9:25-34, 1989. Medline Similar articles
15 of 21 15-04-2010 20:29
239. Bendall MR, Pegg DT: Uniform sample excitation with surface coils for in vivo spectroscopy by adiabatic rapid half passage. J
Magn Reson 67:376-381, 1986.
240. Brindle KM, Smith MB, Rajagopalan B, et al: Spectral editing in 31-P NMR spectra of human brain. J Magn Reson 61:559-563,
1985.
241. Gordon RE, Hanley PE, Shaw D, et al: Localization of metabolites in animals using 31-P topical magnetic resonance. Nature
287:736-738, 1980.
242. Hoult DI: Rotating frame zeugmatography. J Magn Reson 33:183-197, 1979.
243. Blackledge MJ, Styles P, Radda GK: Rotating-frame depth selection and its application to the study of human organs. J Magn
Reson 71:246-258, 1987.
244. Cox SJ, Styles P: Towards biochemical imaging. J Magn Reson 40:209-212, 1980.
245. Bottomley PA, Foster TB, Darrow RD: Depth-resolved surface-coil spectroscopy (DRESS) for in vivo 1H, 31P, and 13C NMR. J
Magn Reson 59:338-342, 1984.
246. Ordidge RJ, Connelly A, Lohmann JAB: Image-selected in vivo spectroscopy (ISIS). A new technique for spatially selective NMR
spectroscopy. J Magn Reson 66:283-294, 1986.
247. Connelly A, Counsell C, Lohman JAB, et al: Outer volume suppressed image related in vivo spectroscopy (OSIRIS), a
high-sensitivity localization technique. J Magn Reson 78:519-525, 1988.
248. Brown TR, Kincaid BM, Ugurbil K: NMR chemical shift imaging in three dimensions. Proc Natl Acad Sci USA 79:3523-3526,
249. Maudsley AA, Hilal SK, Perman WH, et al: Spatially resolved high resolution spectroscopy by 'four-dimensional' NMR. J Magn
Reson 51:147-152, 1983.
250. Mansfield P: Spatial mapping of the chemical shift in NMR. Magn Reson Med 1:370-386, 1984. Medline Similar articles
page 489
page 490
251. Doddrell DM, Brooks WM, Bulsing JM, et al: Spatial and chemical-shift-encoded excitation. SPACE, a new technique for volume-
selected NMR spectroscopy. J Magn Reson 68:367-372, 1986.
252. Granot J: Selected volume excitation using stimulated echoes (VEST). Applications to spatially localized spectroscopy and
imaging. J Magn Reson 70:488-492, 1986.
253. Frahm J, Merboldt KD, Haenicke W: Localized proton spectroscopy using stimulated echoes. J Magn Reson 72:502-508, 1987.
254. Kimmich R, Hoepfel D: Volume-selective multipulse spin-echo spectroscopy. J Magn Reson 72:379-384, 1987.
255. Bottomley, PA: Selective volume method for performing localized NMR spectroscopy. US Patent 4,480,228, 1984.
256. Brooker HR, Mareci TH, Mao JT: Selective Fourier transform localization. Magn Reson Med 5:417-433, 1987. Medline
Similar articles
257. Kuhn B, Dreher W, Norris DG, et al: Fast proton spectroscopic imaging employing k-space weighting achieved by variable
repetition times. Magn Reson Med 35:457-464, 1996. Medline Similar articles
258. Hugg JW, Maudsley AA, Weiner MW, et al: Comparison of k-space sampling schemes for multidimensional MR spectroscopic
259. Loffler R, Sauter R, Kolem H, et al: Localized spectroscopy from anatomically matched compartments: improved sensitivity and
localization for cardiac 31P MRS in humans. J Magn Reson 134:287-299, 1998.
260. Adalsteinsson E, Star-Lack J, Meyer CH, et al: Reduced spatial side lobes in chemical-shift imaging. Magn Reson Med
261. Pohmann R, von Kienlin M: Accurate phosphorus metabolite images of the human heart by 3D acquisition-weighted CSI. Magn
Reson Med 45:817-826, 2001.
262. Duyn JH, Moonen CT: Fast proton spectroscopic imaging of human brain using multiple spin-echoes. Magn Reson Med
263. Duyn JH: High-speed interlaced spin-echo magnetic resonance imaging. Magn Reson Med 43:905-908, 2000. Medline
Similar articles
264. Adalsteinsson E, Irarrazabal P, Topp S, et al: Volumetric spectroscopic imaging with spiral-based k-space trajectories. Magn
265. Posse S, Tedeschi G, Risinger R, et al: High speed 1H spectroscopic imaging in human brain by echo-planar spatial-spectral
encoding. Magn Reson Med 33:34-40, 1995.
266. Adalsteinsson E, Irarrazabal P, Spielman DM, et al: Three-dimensional spectroscopic imaging with time-varying gradients. Magn
267. Posse S, DeCarli C, LeBihan D: Three-dimensional echo-planar MR spectroscopic imaging at short echo times in the human
brain. Radiology 192:733-738, 1994. Medline Similar articles
268. Morrell G, Macovski A: Three-dimensional spectral-spatial excitation. Magn Reson Med 37:378-386, 1997. Medline
Similar articles
269. Spielman D, Meyer C, Macovski A, et al: 1H spectroscopic imaging using a spectral-spatial excitation pulse. Magn Reson Med
270. Dydak U, Pruessmann KP, Weiger M, et al: Parallel spectroscopic imaging with spin-echo trains. Magn Reson Med 50:196-200,
16 of 21 15-04-2010 20:29
271. Goelman G, Walter G, Leigh JS: Hadamard spectroscopic imaging technique as applied to study human calf muscles. Magn
272. Dreher W, Leibfritz D: Double-echo multislice proton spectroscopic imaging using Hadamard slice encoding. Magn Reson Med
1
273. Gonen O, Arias-Mendoza F, Goelman G: 3D localized in vivo H spectroscopy of human brain by using a hybrid of 1D-Hadamard
with 2D-chemical shift imaging. Magn Reson Med 37:644-650, 1997.
274. Goelman G: Hadamard encoding with surface coils for high SNR MR spectroscopy. Magn Reson Imaging 17:777-781, 1999.
275. Speck O, Scheffler K, Hennig J: Fast 31P chemical shift imaging using SSFP methods. Magn Reson Med 48:633-639, 2002.
276. Dreher W, Leibfritz D: Fast proton spectroscopic imaging with high signal-to-noise ratio: spectroscopic RARE. Magn Reson Med
277. Dreher W, Leibfritz D: Improved proton spectroscopic U-FLARE imaging for the detection of coupled resonances in the rat brain
in vivo. Magn Reson Imaging 17:611-621, 1999. Medline Similar articles
278. Ernst T, Lee JH, Ross BD: Direct 31P imaging in human limb and brain. J Comput Assist Tomogr 17:673-680, 1993.
279. Norris DG, Dreher W: Fast proton spectroscopic imaging using the sliced k-space method. Magn Reson Med 30:641-645, 1993.
280. Boesch C, Ith M, Jung B, et al: Effect of oral D-tagatose on liver volume and hepatic glycogen accumulation in healthy male
volunteers. Regul Toxicol Pharmacol 33:257-267, 2001. Medline Similar articles
281. Casey A, Mann R, Banister K, et al: Effect of carbohydrate ingestion on glycogen resynthesis in human liver and skeletal muscle,
measured by (13)C MRS. Am J Physiol 278:E65-E75, 2000.
282. Roser W, Beckmann N, Wiesmann U, et al: Absolute quantitation of the hepatic glycogen content in a patient with glycogen
storage disease by 13C Magnetic Resonance Spectroscopy. Magn Reson Imaging 14:1217-1220, 1996.
283. Fried R, Beckmann N, Keller U, et al: Early glycogenolysis and late glycogenesis in human liver after intravenous administration of
galactose: Non-invasive determinations by 13C MRS and proton MRI in combination with a stable isotope tracer. Am J Physiol
270:G14-G19, 1996.
284. Shulman RG, Rothman DL, Price TB: Nuclear magnetic resonance studies of muscle and applications to exercise and diabetes.
Diabetes 45:S93-S98, 1996.
285. Gruetter R, Magnusson I, Rothman DL, et al: Validation of 13C NMR measurements of liver glycogen in vivo. Magn Reson Med
31:583-588, 1994.
286. Borgs M, VanHecke P, Overloop K, et al: In situ 13C NMR quantification of hepatic glycogen. NMR Biomed 6:371-376, 1993.
287. Beckmann N, Fried R, Turkalj I, et al: Noninvasive observation of hepatic glycogen formation in man by 13C MRS after oral and
intravenous glucose administration. Magn Reson Med 29:583-590, 1993.
288. Van Cauteren M, Miot F, Segebarth CM, et al: Excitation characteristics of adiabatic half-passage RF pulses used in surface coil
MR spectroscopy. Application to 13C detection of glycogen in the rat liver. Phys Med Biol 37:1055-1064, 1992.
289. Heerschap A, Luyten PR, vanderHeyden JI, et al: Broadband proton decoupled natural abundance 13C NMR spectroscopy of
humans at 1.5 T. NMR Biomed 2:124-132, 1989.
290. Jue T, Rothman DL, Tavitian BA, et al: Natural-abundance C-13 NMR study of glycogen repletion in human liver and muscle. Proc
Natl Acad Sci USA 86:1439-1442, 1989. Medline Similar articles
291. Rotman S, Slotboom J, Kreis R, et al: Muscle glycogen recovery after exercise measured by 13C-MRS in humans: Effect of
nutritional solutions. MAGMA 11:114-121, 2000.
292. Garwood M, DelaBarre L: The return of the frequency sweep: designing adiabatic pulses for contemporary NMR. J Magn Reson
293. De Graaf RA, Nicolay K: Adiabatic rf pulses: Applications to in vivo NMR. Concepts Magn Reson 9:247-268, 1997.
294. Conolly S, Glover G, Nishimura D, et al: A reduced power selective adiabatic spin-echo pulse sequence. Magn Reson Med
295. Ugurbil K, Garwood M, Bendall M: Amplitude- and frequency-modulated pulses to achieve 90 plane rotations with
inhomogeneous B1 fields. J Magn Reson 72:177-185, 1987.
296. Silver MS, Joseph RI, Hoult DI: Highly selective Pi/2 and Pi pulse generation. J Magn Reson 59:347-351, 1984.
297. Sodickson DK, Manning WJ: Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil
arrays. Magn Reson Med 38:591-603, 1997. Medline Similar articles
298. Pruessmann KP, Weiger M, Scheidegger MB, et al: SENSE: sensitivity encoding for fast MRI. Magn Reson Med 42:952-962,
299. Murphy-Boesch J, Jiang H, Stoyanova R, et al: Quantification of phosphorus metabolites from chemical shift imaging spectra with
corrections for point spread effects and B1 inhomogeneity. Magn Reson Med 39:429-438, 1998. Medline Similar articles
300. Meyerhoff DJ, Rooney WD, Tokumitsu T, et al: Evidence of multiple ethanol pools in the brain: an in vivo proton magnetization
transfer study. Alcohol Clin Exp Res 20:1283-1288, 1996. Medline Similar articles
301. Kreis R, Hofmann L, Boesch C: Visibility of the methylene signals of ethanol in 1H-MR spectra of the human brain and
implications for model fitting. Proc Intl Soc Magn Reson Med 8:426, 2000.
17 of 21 15-04-2010 20:29
302. Leibfritz D, Dreher W: Magnetization transfer MRS, NMR Biomed 14:65-76, 2001. Medline Similar articles
13
303. Gruetter R, Prolla TA, Shulman RG: C NMR visibility of rabbit muscle glycogen in vivo. Magn Reson Med 20:327-332, 1991.
304. Kreis R, Jung B, Slotboom J, et al: Effect of exercise on the creatine resonances in 1H MR spectra of human skeletal muscle. J
Magn Reson 137:350-357, 1999.
305. Strauss WL, Dager SR: Magnetization transfer of fluoxetine in the human brain using fluorine magnetic resonance spectroscopy.
Biol Psychiatry 49:798-802, 2001. Medline Similar articles
306. Ith M, Bigler P, Scheurer E, et al: Observation and identification of metabolites emerging during postmortem decomposition of
brain tissue by means of in situ 1H-magnetic resonance spectroscopy. Magn Reson Med 48:915-920, 2002.
307. Cady EB, Lorek A, Penrice J, et al: Detection of propan-1,2-diol in neonatal brain by in vivo proton magnetic resonance
spectroscopy. Magn Reson Med 32:764-767, 1994. Medline Similar articles
308. Kreis R, Ross BD: Cerebral metabolic disturbances in patients with subacute and chronic diabetes mellitus: Detection with proton
MR spectroscopy. Radiology 184:123-130, 1992. Medline Similar articles
309. Ross B, Kreis R, Ernst T: Clinical tools for the 90s: magnetic resonance spectroscopy and metabolite imaging. Eur J Radiol
14:128-140, 1992.
310. Ith M, Jung B, Zehnder M, et al: Absolute quantitation of glycogen by means of 13C-MRS: A comparison of two different
approaches. Proc Intl Soc Magn Reson Med 11:261, 2003.
311. Jehenson P, Duboc D, Bloch G, et al: Diagnosis of muscular glycogenosis by in vivo natural abundance 13C NMR spectroscopy.
Neuromuscul Disord 1:99-101, 1991.
312. Ith M, Jung B, Zehnder M, et al: Evaluation of two different methods for absolute quantitation of glycogen in 13C-MR spectra.
MAGMA 16 (Suppl 1):S174-S175, 2003.
313. Li BS, Babb JS, Soher BJ, et al: Reproducibility of 3D proton spectroscopy in the human brain. Magn Reson Med 47:439-446,
2002.
314. Chard DT, McLean MA, Parker GJ, et al: Reproducibility of in vivo metabolite quantification with proton magnetic resonance
spectroscopic imaging. J Magn Reson Imaging 15:219-225, 2002. Medline Similar articles
315. Brooks WM, Friedman SD, Stidley CA: Reproducibility of 1H-MRS in vivo. Magn Reson Med 41:193-197, 1999.
316. Marshall I, Wardlaw J, Cannon J, et al: Reproducibility of metabolite peak areas in 1H MRS of brain. Magn Reson Imaging
14:281-292, 1996.
page 490
page 491
317. Kreis R, Boesch C: Bad spectra can be better than good spectra. Proc Intl Soc Magn Reson Med 11:264, 2003.
318. Kushmerick MJ, Meyer RA, Brown TR: Regulation of oxygen consumption in fast- and slow-twitch muscle. Am J Physiol
263:C598-C606, 1992.
319. Kemp GJ, Taylor DJ, Radda GK: Control of phosphocreatine resynthesis during recovery from exercise in human skelatal muscle.
NMR Biomed 6:66-72, 1993. Medline Similar articles
320. Quistorff B, Johansen L, Sahlin K: Absence of phosphocreatine resynthesis in human calf muscle during ischaemic recovery.
Biochem J 291:681-686, 1993. Medline Similar articles
321. McCully KK, Iotti S, Kendrick K, et al: Simultaneous in vivo measurements of HbO2 saturation and PCr kinetics after exercise in
normal humans. J Appl Physiol 77:5-10, 1994. Medline Similar articles
322. Boska MD, Nelson JA, Sripathi N, et al: 31P MRS studies of exercising human muscle at high temporal resolution. Magn Reson
323. Kreis R, Kamber M, Koster M, et al: Creatine supplementation-part II: in vivo magnetic resonance spectroscopy. Med Sci Sports
Exerc 31:1770-1777, 1999. Medline Similar articles
324. Conley KE, Jubrias SA, Esselman PC: Oxidative capacity and ageing in human muscle. J Physiol 526:203-210, 2000.
325. Fan TWM: Metabolite profiling by one- and two-dimensional analysis of complex mixtures. Prog NMR Spectroscopy 28:161-219,
1996.
326. Govindaraju V, Young K, Maudsley AA: Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed
327. Kreutzer U, Wang DS, Jue T: Observing the 1H NMR signal of the myoglobin Val-E11 in myocardium. An index of cellular
oxygenation. Proc Natl Acad Sci USA 89:4731-4733, 1992.
328. Kreis R, Pfenninger J, Herschkowitz N, et al: In vivo proton magnetic resonance spectroscopy in a case of Reye's syndrome.
Intensive Care Med 21:266-268, 1995. Medline Similar articles
329. Kunnecke B, Seelig J: Glycogen metabolism as detected by in vivo and in vitro 13C-NMR spectroscopy using (1,2-13C2)glucose
as substrate. Biochim Biophys Acta 1095:103-113, 1991.
330. Kanamori K, Ross BD: In vivo activity of glutaminase in the brain of hyperammonaemic rats measured by 15N nuclear magnetic
resonance. Biochem J 305 (Pt 1):329-336, 1995.
331. Gruetter R, Seaquist ER, Ugurbil K: A mathematical model of compartmentalized neurotransmitter metabolism in the human
brain. Am J Physiol Endocrinol Metab 281:E100-E112, 2001.
332. Lebon V, Petersen KF, Cline GW, et al: Astroglial contribution to brain energy metabolism in humans revealed by 13C nuclear
18 of 21 15-04-2010 20:29
magnetic resonance spectroscopy: elucidation of the dominant pathway for neurotransmitter glutamate repletion and measurement of
astrocytic oxidative metabolism. J Neurosci 22:1523-1531, 2002.
13
333. Malloy CR, Sherry AD, Jeffrey FM: Analysis of tricarboxylic acid cycle of the heart using C isotope isomers. Am J Physiol
259:H987-H995, 1990.
13 13
334. Cerdan S, Kunnecke B, Seelig J: Cerebral metabolism of [1,2- C2]acetate as detected by in vivo and in vitro C NMR. J Biol
Chem 265:12916-12926, 1990.
335. Jeffrey FM, Rajagopal A, Malloy CR, et al: 13C-NMR: a simple yet comprehensive method for analysis of intermediary
metabolism. Trends Biochem Sci 16:5-10, 1991. Medline Similar articles
336. Lewandowski ED, Doumen C, White LT, et al: Multiplet structure of 13C NMR signal from glutamate and direct detection of
tricarboxylic acid (TCA) cycle intermediates. Magn Reson Med 35:149-154, 1996.
337. Jucker BM, Rennings AJ, Cline GW, et al: In vivo NMR investigation of intramuscular glucose metabolism in conscious rats. Am
J Physiol 273:E139-E148, 1997.
338. Bertocci LA, Jones JG, Malloy CR, et al: Oxidation of lactate and acetate in rat skeletal muscle: analysis by 13C-nuclear magnetic
resonance spectroscopy. J Appl Physiol 83:32-39, 1997.
339. Previs SF, Cline GW, Shulman GI: A critical evaluation of mass isotopomer distribution analysis of gluconeogenesis in vivo. Am J
Physiol 277:E154-E160, 1999.
340. Carvalho RA, Babcock EE, Jeffrey FM, et al: Multiple bond 13C-13C spin-spin coupling provides complementary information in a
13
C NMR isotopomer analysis of glutamate. Magn Reson Med 42:197-200, 1999.
341. Lapidot A, Haber S: Effect of acute insulin-induced hypoglycemia on fetal versus adult brain fuel utilization, assessed by (13)C
MRS isotopomer analysis of [U-(13)C]glucose metabolites. Dev Neurosci 22:444-455, 2000. Medline Similar articles
342. Jones JG, Solomon MA, Cole SM, et al: An integrated (2)H and (13)C NMR study of gluconeogenesis and TCA cycle flux in
humans. Am J Physiol Endocrinol Metab 281:E848-E856, 2001.
343. Lapidot A, Gopher A: Cerebral metabolic compartmentation. Estimation of glucose flux via pyruvate carboxylase/pyruvate
dehydrogenase by 13C NMR isotopomer analysis of D-[U-13C]glucose metabolites. J Biol Chem 269:27198-27208, 1994.
344. Behar KL, Ogino T: Assignment of resonances in the 1H spectrum of rat brain by two-dimensional shift correlated and J-resolved
NMR spectroscopy. Magn Reson Med 17:285-303, 1991.
345. Brereton IM, Galloway GJ, Rose SE, et al: Localized two-dimensional shift correlated spectroscopy in humans at 2 Tesla. Magn
Reson Med 32:251-257, 1994.
346. Ryner LN, Sorenson JA, Thomas MA: Localized 2D J-resolved 1H MR spectroscopy: strong coupling effects in vitro and in vivo.
347. Kreis R, Boesch C: Spatially localized, one- and two-dimensional NMR spectroscopy and in vivo application to human muscle. J
348. Dreher W, Leibfritz D: Detection of homonuclear decoupled in vivo proton NMR spectra using constant time chemical shift
encoding: CT-PRESS. Magn Reson Imaging 17:141-150, 1999. Medline Similar articles
349. Adalsteinsson E, Spielman DM: Spatially resolved two-dimensional spectroscopy. Magn Reson Med 41:8-12, 1999. Medline
Similar articles
350. Thomas MA, Yue K, Binesh N, et al: Localized two-dimensional shift correlated MR spectroscopy of human brain. Magn Reson
351. Thomas MA, Hattori N, Umeda M, et al: Evaluation of two-dimensional L-COSY and JPRESS using a 3 T MRI scanner: from
phantoms to human brain in vivo. NMR Biomed 16:245-251, 2003.
352. Renema WK, Klomp DW, Philippens ME, et al: Magnetization transfer effect on the creatine methyl resonance studied by CW
off-resonance irradiation in human skeletal muscle on a clinical MR system. Magn Reson Med 50:468-473, 2003. Medline
Similar articles
353. Kruiskamp MJ, De Graaf RA, Van der Grond J, et al: Magnetic coupling between water and creatine protons in human brain and
skeletal muscle, as measured using inversion transfer 1H-MRS. NMR Biomed 14:1-4, 2001.
354. De Graaf RA, van Kranenburg A, Nicolay K: Off-resonance metabolite magnetization transfer measurements on rat brain in situ.
355. Kruiskamp MJ, De Graaf RA, van Vliet G, et al: Magnetic coupling of creatine/phosphocreatine protons in rat skeletal muscle, as
studied by 1H-magnetization transfer MRS. Magn Reson Med 42:665-672, 1999.
356. Luo Y, Rydzewski J, De Graaf RA, et al: In vivo observation of lactate methyl proton magnetization transfer in rat C6 glioma. Magn
357. Meyerhoff DJ: Proton magnetization transfer of metabolites in human brain. Magn Reson Med 42:417-420, 1999. Medline
Similar articles
358. Roell SA, Dreher W, Leibfritz D: Combining CW and pulsed saturation allows in vivo quantitation of magnetization transfer
observed for total creatine by 1H-NMR-spectroscopy of rat brain. Magn Reson Med 42:222-227, 1999.
359. Roell SA, Dreher W, Busch E, et al: Magnetization transfer attenuates metabolite signals in tumorous and contralateral animal
brain: in vivo observations by proton NMR spectroscopy. Magn Reson Med 39:742-748, 1998. Medline Similar articles
360. van der Toorn A, Dijkhuizen RM, Tulleken CA, et al: Diffusion of metabolites in normal and ischemic rat brain measured by
localized 1H MRS. Magn Reson Med 36:914-922, 1996.
19 of 21 15-04-2010 20:29
361. Assaf Y, Cohen Y: In vivo and in vitro bi-exponential diffusion of N-acetyl aspartate (NAA) in rat brain: a potential structural probe?
NMR Biomed 11:67-74, 1998. Medline Similar articles
362. Pfeuffer J, Tkac I, Gruetter R: Extracellular-intracellular distribution of glucose and lactate in the rat brain assessed noninvasively
1
by diffusion-weighted H nuclear magnetic resonance spectroscopy in vivo. J Cereb Blood Flow Metab 20:736-746, 2000.
363. De Graaf RA, van Kranenburg A, Nicolay K: In vivo (31)P-NMR diffusion spectroscopy of ATP and phosphocreatine in rat skeletal
muscle. Biophys J 78:1657-1664, 2000. Medline Similar articles
364. Boesch C, Kreis R: Dipolar coupling and ordering effects observed in magnetic resonance spectra of skeletal muscle. NMR
365. Kaminogo M, Ishimaru H, Morikawa M, et al: Proton MR spectroscopy and diffusion-weighted MR imaging for the diagnosis of
intracranial tuberculomas. Report of two cases. Neurol Res 24:537-543, 2002. Medline Similar articles
366. Kreis R, Arcinue E, Ernst T, et al: Hypoxic encephalopathy after near-drowning studied by quantitative 1H-magnetic resonance
spectroscopy. Metabolic changes and their prognostic value. J Clin Invest 97:1142-1154, 1996.
367. Kreis R, Fusch C, Maloca P, et al: Supposed pathology may be individuality: Interindividual and regional differences of brain
metabolite concentrations determined by 1H MRS. SMR Annual Meeting 2:45, 1994.
368. Soher BJ, Hurd RE, Sailasuta N, et al: Quantitation of automated single-voxel proton MRS using cerebral water as an internal
reference. Magn Reson Med 36:335-339, 1996. Medline Similar articles
369. Husted CA, Duijn JH, Matson GB, et al: Molar quantitation of in vivo proton metabolites in human brain with 3D magnetic
resonance spectroscopic imaging. Magn Reson Imaging 12:661-667, 1994.
370. Ala-Korpela M, Usenius JP, Keisala J, et al: Quantification of metabolites from single-voxel in vivo 1H NMR data of normal human
brain by means of time-domain data analysis. MAGMA 3:129-136, 1995.
371. Schick F, Eismann B, Jung WI, et al: Comparison of localized proton NMR signals of skeletal muscle and fat tissue in vivo: Two
lipid compartments in muscle tissue. Magn Reson Med 29:158-167, 1993. Medline Similar articles
372. Boesch C, Slotboom H, Hoppeler H, et al: In vivo determination of intra-myocellular lipids in human muscle by means of localized
1
H-MR-spectroscopy. Magn Reson Med 37:484-493, 1997.
373. Frahm J, Bruhn H, Gyngell ML, et al: Localized high-resolution proton NMR spectroscopy using stimulated echoes: Initial
applications to human brain in vivo. Magn Reson Med 9:79-93, 1989. Medline Similar articles
374. Desmoulin F, Seelig J: A homonuclear shift correlated and spatially localized spectroscopy using stimulated echoes. Magn
page 491
page 492
1
375. Hanstock CC, Rothman DL, Prichard JW, et al: Spatially localized H NMR spectra of metabolites in the human brain. Proc Natl
Acad Sci USA 85:1821-1825, 1988.
376. Bruhn H, Frahm J, Gyngell ML, et al: Cerebral metabolism in man after acute stroke: New observations using localized proton
NMR spectroscopy. Magn Reson Med 9:126-131, 1989. Medline Similar articles
377. Michaelis T, Merboldt KD, Haenicke W, et al: On the identification of cerebral metabolites in localized 1H NMR spectra of human
brain in vivo. NMR Biomed 4:90-98, 1991.
378. vanRijen PC, Luyten PR, BerkelbachvanderSprenkel JW, et al: 1H and 31P NMR measurement of cerebral lactate, high-energy
phosphate levels, and pH in humans during voluntary hyperventilation; Associated EEG, capnographic, and Doppler findings. Magn
Reson Med 10:182-193, 1989.
379. Barany M, Spigos DG, Mok E, et al: High resolution proton magnetic resonance spectroscopy of human brain and liver. Magn
380. Frahm J, Michaelis T, Merboldt K, et al: On the N-acetyl methyl resonance in localized 1H NMR spectra of human brain in vivo.
NMR Biomed 4:201-204, 1991.
1
381. Kreis R, Jung B, Rotman S, et al: Non-invasive observation of acetyl-group buffering by H-MR spectroscopy in exercising human
muscle. NMR Biomed 12:471-476, 1999.
382. Frahm J, Michaelis T, Merboldt KD, et al: Localized NMR spectroscopy in vivo: Progress and problems. NMR Biomed 2:188-195,
383. Michaelis T, Helms G, Merboldt KD, et al: Identification of Scyllo-inositol in proton NMR spectra of human brain in vivo. NMR
384. Gruetter R, Rothman DL, Novotny EJ, et al: Detection and assignment of the glucose signal in 1H NMR difference spectra of the
human brain. Magn Reson Med 27:183-188, 1992.
385. Chen W, Avison MJ, Bloch G, et al: Proton NMR observation of glycogen in vivo. Magn Reson Med 31:576-579, 1994.
386. Mansfield P, Grannell PK: NMR 'diffraction' in solids. J Phys C: Solid State 6:L422-L426, 1973.
387. Arnold JT, Dharmatti SS, Packard ME: Chemical effects on nuclear induction signals from organic compounds. J Chem Phys
19:507, 1951.
388. Liddel U, Ramsey NF: Temperature dependent magnetic shielding in ethyl alcohol. J Chem Phys 19:1608, 1951.
20 of 21 15-04-2010 20:29
© 2010 Elsevier
21 of 21 15-04-2010 20:29
IGH IELD AGNETIC ESONANCE MAGING

Robert E. Lenkinski
Over the past five decades there has been a consistent trend in nuclear magnetic resonance (NMR) to
move to increasingly higher static fields. The major motivations for this trend come from expectations
of increased signal-to-noise ratios (SNR) in imaging studies and increased SNR and chemical shift
dispersion in NMR spectroscopy. Even in the NMR regime, where the samples are generally small (0.5
to 10 mL) and much less conductive than tissue, there have been intense debates surrounding whether
the practical gains made by moving to higher fields can approach those predicted by theory. In human
MR imaging a similar trend to higher field (i.e., from 1.5 to 3-4 T and even to 7T systems) has
occurred, albeit more slowly. This trend has generated its own debates regarding the relative wisdom
of performing MR imaging and MR spectroscopy studies at these higher fields.
Many of these issues involved in higher field imaging, such as radiofrequency (RF) homogeneity, power
deposition (specific absorption rate, SAR), static field homogeneity, inter alia, have been reviewed
recently by Norris,1 Takahashi et al,2 and Ugurbil et al.3 However, the major focus of these discussions
has been primarily on MR imaging and MR spectroscopy applications in the brain. In this chapter we
will review some of these issues and extend these discussions to using higher field MR systems for
applications to image organs outside the brain. The motivation for providing this discussion is the
increasing availability of high-field MR scanners (3-4 T) equipped with whole-body RF coils. Some of
these scanners have received FDA clearance for clinical whole-body MR imaging (see
http://www.FDA.gov/cdrh/index.html). Our own experience, gathered over the past 2 years in
whole-body applications at 3 T, forms the basis for optimism about the future of whole-body imaging at
this field strength.
In the very early days of MR imaging there were a number of papers that discussed the factors that
might limit the field strengths for useful MR imaging. Bottomley & Andrew suggested that, based on the
interactions of RF with tissue, 20 MHz (~0.5 T) would be the upper limit because of so-called RF
penetration effects.115 In fact, the issue was not RF penetration but rather the phase shift in the RF
across the body part being imaged caused by the electrical properties of tissue. Hoult and Lauterbur4
and Mansfield and Morris5 provided scientific arguments that 10 MHz (~0.25 T) would be the effective
limit. Subsequently, Hoult suggested that there might be an optimum field strength for imaging the head
which would be around 0.7 T and that imaging the torso might be optimum at about half of the field
strength (0.35 T).6,7 More recently, Hoult8 has provided an excellent explanation for why these
seemingly pessimistic limits were set and, even more importantly, of how technical advances provided
the basis for imaging at 1.5 T (63 MHz) which has become the most widely accepted field for
diagnostic human studies. These technical advances were a result of the development of RF coils with
distributed capacitance (see Alderman and Grant9) and the implementation of imaging sequences using
symmetric spin-echoes and magnitude reconstructions.10
page 493
page 494
Although it is tempting to make light of these early pessimistic predictions about field strengths, it is
more useful to recognize that they were based on seemingly sound theoretical principles. These
principles and considerations were correct but limited by the context in which they were developed.
What changed was that innovative researchers made advances that either directly addressed these
issues or found alternative imaging approaches that made these considerations irrelevant. The current
discussions on the disadvantages of high-field whole-body imaging may be similar to those that
occurred 20 years ago concerning field strengths, in that they are based on our present knowledge
and approaches rather than anticipating the development of new technologies that will address current
concerns.
© 2010 Elsevier
1 of 1 15-04-2010 20:31
ADVANCES IN MAGNET DESIGN

All the early high-field magnets were quite large and were based on conventional unshielded designs.
These magnets presented siting problems associated with their relatively large fringe fields. The
unwieldy size of these cryostats also was not conducive to patient studies because of a higher
potential for claustrophobic refusal and limited access to the subject within the bore tube of the
magnet. Although there have not been any major "breakthroughs" in magnet technology over the past
decade, there have been incremental improvements which have made 3-4 T magnets much more
suitable for use as clinical scanners. Most of the current 3 T magnets are based on self-shielded
designs that reduce the fringe fields to those comparable to unshielded 1.5 T magnets. In addition,
there has been the development of "short-bore" 3 T magnets (currently offered by all the major
imaging companies) which have a 60 cm diameter clear bore and a much smaller "footprint " than the
older "long-bore" designs (see, for example, references 11 and 12). These magnets have very similar
dimensions to 1.5 T short-bore magnets. Their smaller size, manageable fringe fields, and patient
friendliness make them easier to use as clinical scanners. An additional improvement in magnet
technology has been the development of magnet designs with helium cold-heads that reduce helium
consumption and remove the need for the use of liquid nitrogen as a secondary coolant. This advance
has reduced the physical dimensions of the cryostat and has reduced the operating costs of modern
magnets.
© 2010 Elsevier
1 of 1 15-04-2010 20:32
THE FIELD DEPENDENCE OF RADIOFREQUENCY INHOMOGENEITIES

The state of the art in terms of theoretical models and experimental results was reviewed in 2000 by
Hoult.8 Briefly, several groups, including Hoult's, have modeled the B1 field inside different models of
human tissue at various field strengths.8,13-36 The reason for carrying out these kinds of calculations is
to develop expressions for the spatial variation of the B1 field, the power deposition, and the intrinsic
SNR of the study, all as a function of the applied magnetic field.
At 1.5 T, Glover et al21 showed that there can be significant B1 artifacts associated with RF
inhomogeneities in the body that were minimized using circularly polarized RF excitation. At 3-4 T
several groups have shown by calculations8,13,15,16,33,37 and experiments (see, for example, reference
33) that there is so-called "center brightening" or "field focusing" in images of the human brain. This
effect results from the fact that because of tissue conductivity, the B1 field in the center of the brain is
higher than at the periphery even though the driving field would be uniform in a non-conducting medium.
Theoretically, this effect is expected to be more pronounced at higher fields, such as 7-8 T, which has
been experimentally verified.33 Alsop38 has shown that it is possible to minimize this effect with a novel
spiral-based RF volume coil design at higher static fields.
At a given field strength and conductivity, this effect should also scale with the size of the object being
imaged. Theory would predict that the effect might be more pronounced in the body than in the brain.
The B1 field in the body, using a surface coil, has also been modeled at 3-4 T17,30,31,35 and
experimentally verified for both surface coils35 and a body coil transmitter.39 We have measured the B1
field generated by a whole-body RF transmitter at 3 T on a scanner equipped with a prototype
high-pass birdcage body coil (57 cm diameter, 53 cm in length).40 B1 mapping was performed with a
modified single-shot fast spin-echo (SSFSE) sequence. A field of view of 48 cm, a slice thickness of 5
mm, a slice spacing of 40 cm, and a matrix size of 128 × 96 were selected. The refocusing RF pulse
train was tailored to reduced flip angles in order to minimize RF power deposition. 41 This method for
reducing RF power deposition will be discussed in a following section. Examples of human images
obtained at 3 T using this method will also be shown.
Ten images were obtained with the excitation pulse amplitude stepped from near zero to approximately
120°. Five slices were obtained with a TR of 4 s. In a silicone oil phantom the B1 maps shown in Figure
18-1 indicate excellent uniformity of the RF field. The silicone oil phantom has minimal dielectric and
resistive properties. In contrast, the B1 variation in volunteers, shown in Figure 18-2, was much
greater. The B1 is highest near the skin and then drops rapidly with increasing depth. This decrease
slows and then reverses such that the center of the abdomen has higher B1 than the surroundings. The
B1 variation from the maximum to the minimum is approximately a factor of two. This corresponds to a
variation of ±33% around the mean. These effects can be explained by recognizing that for cylindrical
objects, dielectric effects tend to produce "center-brightened" B1 profiles, while resistive effects cause
a decrease in B1 with depth into the object. The in vivo B1 pattern observed is consistent with a
40
balance of the two effects producing slight center brightening and edge brightening in the B1 profile.
page 494
page 495
Add to lightbox
1 of 2 15-04-2010 20:33
Add to lightbox
Figure 18-1 B1 maps in the coronal plane from the silicon oil phantom (A) and the corresponding
SSFSE images (B). The color code is linear for B1.
39
The B1 field in a saline phantom at 1.5 T and 3 T (taken from Greenman et al ) is shown in Figure
18-3A. Note the center brightening in the 3 T image. The variation across the slice is shown in Figure
18-3B. The much larger variation in B1 observed at 3 T is consistent with theoretical predictions.
Examples of B1 field profiles, taken in an oblique, short-axis view of the heart at 1.5 and 3 T, are
39
shown in Figures 18-4A and 18-4B (taken from Greenman et al ). These images and profiles indicate
a larger variation in the B1 field in from the edge to the center of the chest which is consistent with the
phantom studies and theory.
The sensitivity (SNR) of the MR experiment should increase linearly with the field strength.20,30,31
However, as indicated above, as the field strength increases the variation in B1 increases, causing the
SNR to be spatially dependent. For example, as shown in Figure 18-5, Vaughn et al33 reported an
improvement in SNR at 7 T compared to 4 T of about 1.5 at the edges of the brain compared to 2.0 at
35
the center of the brain. Wen et al found that the SNR improved at slightly less than a linear
dependence in the heart when comparing 1.5 T, 3 T, and 4 T images of the same volunteer. Our own
experience has been that the SNR improvement is also slightly less than a factor of two when
comparing 1.5 T and 3 T body coil images of normal volunteers. However, this deviation from theory
may be caused by practical issues such as differences in coil loading, differences in setting the correct
90° flip angle, and differences in noise figures between the two systems.
The power deposition (SAR) should vary as the square of the field to about 4.7 T. Above this field, two
groups33,42 have reported that the SAR increases less steeply than the square law dependence in the
human brain. The observations of Vaughn et al33 are consistent with the theoretical treatment of Hoult.8
For whole-body imaging this means that 3 T should require about four times higher SAR than 1.5 T, all
other parameters being equal. Many multislice, multiecho imaging sequences, such as fast spin-echo,
are at or near the maximum SAR guidelines approved by the FDA even at 1.5 T. Alsop41 has shown
that the SAR of a RARE-based FSE sequence can be reduced by using low-angle instead of 180°
refocusing pulses. For example, using 90° refocusing pulses will reduce the SAR by a factor of four
with a reduction of SNR of less than 20%. Using this modification in FSE sequences, comparable
spatial coverage can be maintained at 3 T as is used at 1.5 T without exceeding the SAR guidelines.
The effective TE must be lengthened in these reduced flip angle sequences in order to preserve T2
contrast at 3 T. As illustrated by the examples that will be presented, we have made this reduced flip
angle sequence a standard part of many of our body imaging protocols at 3 T.
© 2010 Elsevier
2 of 2 15-04-2010 20:33
THE FIELD DEPENDENCE OF THE RELAXATION TIMES OF HUMAN TISSUES

page 495
page 496
Add to lightbox
Add to lightbox
1 of 4 15-04-2010 20:36
Add to lightbox
page 496
page 497
Figure 18-2 A, B1 maps in the coronal plane through the abdomen of a normal volunteer (top) and the
corresponding SSFSE images (bottom). The color code is linear for B1. B, B1 maps in the axial plane
through the abdomen of a normal volunteer (top) and the corresponding SSFSE images (bottom). The
color code is linear for B1. C, Coronal image obtained through the pelvis of a normal volunteer using
the body coil.
Add to lightbox
2 of 4 15-04-2010 20:36
Add to lightbox
Add to lightbox
Figure 18-3 B1 maps of a cylindrical phantom containing an aqueous solution of 50 mM NaCl. As
opposed to the B1 field at 1.5 T (A), the greater nonuniformity of the B1 field at 3 T (B) is apparent
from the bright center and dark edges. C, B1 field values along the dashed and solid lines that are
drawn on the phantom images in (A) and (B) acquired at 1.5 T and 3 T, respectively. The variation in
B1 field value is substantially greater at 3 T due to the shorter RF wavelength. (Reproduced with
permission from reference 39. Copyright © 2003 Wiley-Liss, Inc., A Wiley Company.)
The relaxation properties of various tissues43 and their dependence on the field (up to 100 MHz) have
been reported in the literature. More recently, Fischer et al44 reported a formula that fits the T1s of
gray and white matter as a function of the field strength. The common assessments of high-field MR,
particularly in the brain, were that the T1s of gray and white matter would not only lengthen but would
also converge. These predictions implied that the contrast between gray and white matter would
decrease and that image acquisition times would increase. Both of these predictions created a
relatively pessimistic view about the merits of high-field imaging for applications in the brain. The T1s of
gray and white matter have been reported at both 3 T45 and 4 T.46-48 In spite of the initial pessimistic
predictions, there have been excellent images with good contrast between gray and white matter
reported at fields as high as 7-8 T (see reference 3). The T2s of brain tissue have been reported to be
3 of 4 15-04-2010 20:36
slightly shorter at high field as compared to 1.5 T. In his recent review, Norris1 has discussed the
mechanism of T2 shortening at higher fields. He called this mechanism "dynamic averaging". The T2
shortening results from water diffusion through microscopic Bo gradients in tissue and theory predicts a
Bo2 dependence for this effect. At 7T this mechanism has been shown to dominate T2 relaxation of
water in brain tissue.49
The relaxation times of various tissues in the body have been reported by de Bazelaire et al.50 These
authors found that the T1 relaxation times were generally longer and T2 relaxation times were
generally shorter at 3 T than at 1.5 T. The magnitude of change was found to vary considerably in
different tissues.
© 2010 Elsevier
4 of 4 15-04-2010 20:36
THE IMPACT OF HIGH FIELD ON CONTRAST AGENT RELAXIVITIES AND

THEIR CLINICAL EFFICACY
The theory for determining the relaxivities of gadolinium-based contrast agents has been reviewed by
Caravan et al (see, for example, references 51 and 52). Theory predicts that for small gadolinium
chelates that have inner sphere waters of hydration, the relaxivities are relatively independent of field.
The low molecular weight gadolinium-based contrast agents (such as Magnevist, Prohance, and
Omniscan) are used clinically primarily as T1-shortening agents. We have already discussed the fact
that at higher fields, the T1s of tissues should get longer. On this basis alone, if, as theory predicts,
the relaxivities of the agents are unaffected by the field strength, the effects of the administration of an
equal dose of contrast agent should be more pronounced at higher field.
page 497
page 498
Add to lightbox
Add to lightbox
Add to lightbox
1 of 2 15-04-2010 20:39
Figure 18-4 B1 maps across an oblique short axis view of the heart of the same normal volunteer at
1.5 T (A) and 3 T (B). In (B) there is a substantial variation in field intensity across the heart while the
field in (A) is relatively constant across the heart. The B1 fields in images (A) and (B) are drawn in (C).
Pixel 0 corresponds to the anterior end of the lines and pixel 72 corresponds to the posterior end of
the lines. The B1 value varies in the anterior/posterior direction by nearly 100% at 3 T while the 1.5 T
B1 value varies by only about 10% and is much more flat. (Reproduced with permission from
reference 39. Copyright © 2003 Wiley-Liss, Inc., A Wiley Company.)
A small number of studies have compared the results of contrast-enhanced studies at 1.5 T and 3
53-57 54
T. All these studies have been performed in the brain and spine. Bernstein et al have reported
the R1 relaxivity of gadoteridol (Prohance) in saline to be 4.74 ± 0.05 mM/s at 1.5 T and 4.43 ± 0.03
mM/s at 3 T, in keeping with the theoretical predictions. They also found that contrast-enhanced MR
53,56,57
angiography of the cervical spine showed promise at 3 T. The group in Vienna have reported
improved visualization of both primary and metastatic brain tumors on contrast-enhanced MR studies
conducted at 3 T, as compared to 1.5 T and 3 T studies. Further work needs to be performed to
assess the relative advantages of contrast-enhanced studies in other parts of the body at 3 T and
higher.
© 2010 Elsevier
2 of 2 15-04-2010 20:39
ANATOMIC AND CLINICAL NEUROIMAGING AT HIGH FIELD

Based on the considerations of SAR and the changes in relaxation time, it is clear that there needs to
be optimization of pulse sequences in order to realize the potential advantages of higher field imaging.
Some of these points have been discussed previously. They have also been recently reviewed by
Frayne et al55 for neuroimaging at 3 T. Once the challenges of high field are met, it is clear that the
SNR gains afforded by higher fields can improve anatomic imaging of the brain. The development of
whole-body RF transmit coils at 3 T and 4 T has created an opportunity to employ multiple coil arrays
at the higher fields in a similar fashion to the manner in which these kinds of coils are routinely used at
1.5 T. This includes applications of parallel imaging (see, for example, references 58-63).
An example of the quality of images obtained at 3 T using a prototype 4 element receiver coil is shown
in Figure 18-6. The spatial resolution obtained in these images is superior to comparable images
obtained at 1.5 T. The FSE image shown in Figure 18-6 was acquired using the reduced flip angle
refocusing sequence discussed previously.41
There has been at least one study aimed at determining the clinical advantages of 3 T scanning for
multiple sclerosis (MS). This group used identical acquisition conditions to image 25 subjects with MS
at 1.5 T and 3 T using FSE and (T1-weighted) SPGR with and without gadolinium contrast injections.
Even using scan protocols optimized for 1.5 T, the 3 T scans showed a 21% increase in the number of
contrast-enhancing lesions detected, a 30% increase in enhancing lesion volume, and a 10% increase
64
in total lesion volume.
65-67 68
Higher fields have proven to be useful for MR angiography and MR venography. For MR
angiography this is, in large part, due to the fact that the improved SNR can be utilized to provide
images with higher spatial resolution, allowing for the visualization of smaller vessels at the higher field.
The increased sensitivity to susceptibility at higher fields improves the MR venograms.
© 2010 Elsevier
1 of 1 15-04-2010 20:39
PERFUSION AND DIFFUSION-WEIGHTED MAGNETIC RESONANCE OF THE BRAIN AT

HIGH FIELD
Wang et al have shown that the increase in the T1 of blood at higher fields enhances the measurement of brain
69
perfusion using arterial spin labeling. The improvement is linear with fields up to about 4 T, above which T2*
shortening begins to reduce the improvements.
page 498
page 499
Add to lightbox
Add to lightbox
1 of 2 15-04-2010 20:42
Add to lightbox
Figure 18-5 A, The SNR measured from fully relaxed gradient-echo images acquired at 4 T (left) and 7 T (center)
using a TEM coil. The average SNR ratio at each location is shown on the 7 T images under the SNR measurement
in each region. The calculated SNR improvement based on a Maxwell model is shown on the right. Note the
excellent agreement between the experimental and measured values of the SNR ratio. (Reproduced with
permission from reference 33.) B, Sagittal 7 T image (TSE, 11 echoes, 7 min exam, 20 cm FOV, 512 × 512, 0.4
mm × 0.4 mm, 3 mm thick slices) shows superior SNR (by a factor of approximately 2.2) compared to the 3 T
image (C), acquired with identical parameters. Magnetic susceptibility artifact is more apparent along the anterior
aspect of the frontal lobes on the higher field image. (Courtesy of Dr Gregory Sorenson, Massachusetts General
Hospital.)
Diffusion-tensor MR imaging was compared at 1.5 T and 3 T by Hunsche et al.70 When the SNR was sufficient,
these authors found no differences in fractional anisotropy. The SNR was found to be 40% higher at 3 T, leading to
higher resolution without introduction of noise-related errors. The authors noted that some of the gains came at the
cost of increased geometric distortions caused by 3 T magnetic field inhomogeneities. This observation has been
made for many echoplanar imaging (EPI) sequences at 3 T. However, there have been at least two different
non-EPI based methods suggested for use in cases where B0 inhomogeneities cause artifacts at high field. The first
is Propeller, an FSE-based sequence that has been employed with good results for diffusion-weighted imaging
(DWI).71,72 We have been using line scan DWI (LSDWI)73-78 at 3 T. An example of LSDWI in the optic chiasm at 3
T is shown in Figure 18-7. Conventional EPI-based DWI was unsuccessful in this region. Note the relatively
distortionless images, even in the region where B0 inhomogeneities would make EPI useless. We suggest that
these two approaches, or others like them, may gain increasing use at higher field both within and outside the brain.
© 2010 Elsevier
2 of 2 15-04-2010 20:42
HIGH-FIELD FUNCTIONAL MAGNETIC RESONANCE IMAGING

The applications of high field to fMRI have been reviewed and discussed in detail by Norris1 and
3
Ugurbil et al. This application was one of the major driving forces in moving to higher field. Kruger et
al79 have shown clear SNR advantages in blood oxygen level-dependent (BOLD) imaging at 3 T as
80-84
compared to 1.5 T. BOLD-based fMRI images have been obtained at 7 T in humans. These 7 T
fMRI studies also show clear advantages over 4 T (and, by implication, lower field) images, provided
that care is taken to optimize the acquisition protocols. Higher field (9.4 T) fMRI studies have been
carried out in rats85 and also show improvements related to the higher field strength. These results
confirm the view that higher fields have become and will continue to be the preferred platforms for
activation studies in humans.
page 499
page 500
Add to lightbox
Figure 18-6 An axial T2-weighted reduced flip angle FSE image obtained using a four element head
coil at 3 T. The acquisition parameters were: TR 4000, TE 104.9, 512 × 256, 24 cm FOV, 16 echo
train length, 3 mm slice/skip 0, 31.2 kHz bandwidth, 2 NEX, 22 slices and total scan time of 4 min 32
s. Note the excellent visualization of the optic chiasm and other small anatomic structures.
© 2010 Elsevier
1 of 1 15-04-2010 20:43
MUSCULOSKELETAL IMAGING AT HIGH FIELD

Many of the technical issues concerning B1 inhomogeneities and SAR considerations should have little
impact on musculoskeletal imaging at high field strengths. MR imaging of the knee and wrist should
also benefit from the fact that excellent designs for transmit/receive coils for imaging these body parts
have existed at 1.5 T for many years. Examples of 3 T images of the knee and wrist are shown in
Figures 18-8 and 18-9. Note the excellent resolution and SNR supported by the 3 T. MR imaging of the
shoulder should present somewhat greater technical challenges in that this application requires
off-center fields of view and is usually performed with receive-only surface coils. An example of a 3 T
image of the shoulder is shown in Figure 18-10.
© 2010 Elsevier
1 of 1 15-04-2010 20:44
MAGNETIC RESONANCE IMAGING OF THE BODY AT HIGH FIELD
Add to lightbox
Figure 18-7 An example of diffusion-weighted images obtained on a normal volunteer acquired with
the line scan diffusion method at 3 T using a four element head coil. Oblique images were acquired
with 2 mm slice thickness, FOV 20 × 5 cm, TR/TE=97/74 ms, four excitations, 256 × 256, 3.9 kHz
2 2
bandwidth and two b-factors: 5 s/mm and 1000 s/mm . The optic chiasm is designated with the
arrows shown on each figure. The orientation of the diffusion tensor is indicated by the blue lines in the
figures on the right. Note that there are negligible effects of susceptibility even close to the sinuses.
40
As discussed earlier, we had a 3 T with a prototype whole-body RF coil. Many of the pessimistic
views about imaging the brain at 3 T and higher fields, such as T1 lengthening and convergence of
tissues, B1 inhomogeneities and SAR considerations, were voiced in even stronger terms concerning
the acquisition of high-quality images of the torso, abdomen, and pelvis at these field strengths. As
shown earlier, for cylindrical objects like the body, dielectric effects tend to produce "center-
brightened" B1 profiles, while resistive effects cause a decrease in B1 with depth into the object. These
two effects balance each other, resulting in the kinds of images shown in Figure 18-2C.
page 500
page 501
1 of 10 15-04-2010 20:47
Add to lightbox
Add to lightbox
Figure 18-8 A, A T1-weighted FSE coronal image of the knee of a normal volunteer at 3 T. The
acquisition parameters were: TR 700, TE 20, echo train length 2, 16 cm FOV, 2 mm slice/skip 0, 256
× 224, 31.25 kHz bandwidth. B, Sagittal gradient-echo image at 3 T.
2 of 10 15-04-2010 20:47
Add to lightbox
Add to lightbox
page 501
page 502
Figure 18-9 A, T1-weighted FSE coronal image of the wrist of a normal volunteer at 3 T. The
acquisition parameters were: TR 700, TE 20, echo train length 2, 8 cm FOV, 2 mm slice/skip 0, 256 ×
224, 31.25 kHz bandwidth and 2 averages. B, A coronal gradient-echo image of the wrist of a normal
volunteer at 3 T. The acquisition parameters were: TR 450, TE 15, 10° flip angle, 8 cm FOV, 2 mm
slice/skip 0, 256 × 256, 31.25 kHz bandwidth and 2 averages.
3 of 10 15-04-2010 20:47
Add to lightbox
Figure 18-10 A coronal T1-weighted FSE image of the shoulder of a normal volunteer taken at 3 T.
The acquisition parameters were: TR 700, TE 20, echo train length 2, 16 cm FOV, 3 mm slice/skip 1,
256 × 256, 31.25 kHz bandwidth and 2 averages.
Add to lightbox
Figure 18-11 Axial single-shot FSE images of the abdomen of a normal volunteer taken at 1.5 and 3
T. The top images were acquired with the body coil as the receiver. The bottom images were acquired
with a four element torso array. Note the SNR improvement at 3 T.
4 of 10 15-04-2010 20:47
Add to lightbox
Figure 18-12 A coronal T2-weighted FSE image of the pelvis of a normal volunteer with an incidental
finding of a bone island. These images were acquired using the reduced flip angle FSE sequence
described in the text for reduced SAR using the body coil as a receiver. The acquisition parameters
were: TR 4000, TE 105, 16 echo train length, 40 cm FOV, 4 mm slice/skip 2, 256 × 256, 16 kHz
bandwidth and 2 excitations.
Add to lightbox
page 502
page 503
5 of 10 15-04-2010 20:47
Figure 18-13 An axial T1-weighted FSE image of the head and spine of a normal volunteer acquired
with the body coil. The acquisition parameters were: TR 700, TE 20, 2 echo train length, 40 cm FOV, 4
mm slice/skip 2, 256 × 256, 16 kHz bandwidth and 2 excitations.
Add to lightbox
Figure 18-14 An axial T2-weighted FSE image of the prostate in a patient with biopsy-proven cancer.
These images were acquired using the reduced flip angle FSE sequence described in the text for
reduced SAR using the four element external torso array. The acquisition parameters were: TR 4750,
TE 149.7, 16 echo train length, 16 cm FOV, 4 mm slice/skip 1, 256 × 256, 16 kHz bandwidth, 5
excitations, 18 slices and 6 minutes and 20 seconds. Note the excellent visualization of the zonal
architecture in the gland.
Our own initial measurements of SNR at 3 T in body-sized phantoms indicated that the average
improvement in SNR at 3 T was about 1.8 times that of 1.5 T. This increase in SNR is almost the gain
achieved by using local surface coils for reception. As a basis for comparison, we show the single-shot
FSE images (see Fig. 18-11) obtained on the same normal volunteer with the body coil and a torso
array at 1.5 and 3 T. Although the B1 inhomogeneities are higher in the 3 T images obtained with the
body coil alone, the SNR is also higher and, as expected, approaches the SNR achieved with the torso
array at 1.5 T. The images obtained with a torso array at 3 T show even higher SNR, as expected.
Examples of T1-weighted images obtained using an FSE sequence are shown in Figure 18-12 (coronal
image of a female volunteer) and Figure 18-13 (sagittal image of the brain and spine of a normal
volunteer). Note that these images were obtained within the FDA guidelines for SAR using the reduced
flip angle FSE sequence described earlier.41 These images illustrate that it is possible to obtain usable
images with the body coil alone. Potential applications include whole-body screening and moving-table
studies.
We have recently completed a study that qualitatively compared the image quality of torso phased
array 3 T imaging of the prostate with that of endorectal 1.5 T imaging.86 The image quality of
T2-weighted images acquired with a torso phased array at 3 T was comparable to that acquired with
an endorectal coil at 1.5 T. Examples of 3 T images of the prostate are shown in Figures 18-14 and
18-15. Note the excellent contrast and resolution in these images which indicate that 3 T MR imaging
with an external coil array may be a viable option for clinical imaging of the prostate.
We have also completed a pilot study using endorectal coil MR of the prostate at 3 T.87 As expected,
6 of 10 15-04-2010 20:47
the SNR is even higher than that obtained with the external torso array, which supports higher
resolution acquisitions. Representative T2-weighted FSE images are shown in Figure 18-16. Note that
the 1.5 mm slice thickness of these axial images permits multiplanar reconstruction into both coronal
and sagittal scan planes, leading to an overall time saving in acquisition.
Add to lightbox
Figure 18-15 A coronal T2-weighted FSE image of the prostate in the same patient as shown in Fig.
18-14. These images were acquired using the reduced flip angle FSE sequence described in the text
for reduced SAR using the four element external torso array. The acquisition parameters were: TR
4750, TE 151,16 echo train length, 16 cm FOV, 4 mm slice/skip 1, 256 ×k192, 16 Hz bandwidth, 5
excitations, 17 slices and 4 minutes and 45 seconds. Note the excellent visualization of the seminal
vesicles.
An example of a sagittal FSE image of the female pelvis of a normal female volunteer acquired with the
external torso array is shown in Figure 18-17. Note the excellent depiction of the zonal anatomy of the
uterus on this image. This image quality indicates the potential of 3 T imaging of the female pelvis.
We have also been evaluating the role that 3 T can play in imaging the liver. In particular, we have
been exploring single breath-hold strategies. One area of interest is the application of MR to study
88-90
fatty liver disease. In this context we have been assessing the three-point Dixon method for
obtaining fat and water images of the liver at 3 T in a single breath-hold. An example of the results
achievable with the torso array at 3 T is shown in Figure 18-18. The images are free of any respiratory
artifact and show good resolution and SNR, indicating the potential clinical utility of this approach. Note
that at 1.5 T comparable images might take almost four times longer to acquire, leading to
unacceptably long breath-holds.
These examples illustrate that, although there may be concerns regarding the technical aspects of
performing body imaging at 3 T and higher, the development of whole-body RF coils, together with
strategies to reduce and manage SAR, have made body applications feasible at 3 T. As indicated by
the examples shown here, there are clear advantages at 3 T for imaging some parts of the body. In
some areas like cardiac imaging, the theoretical SNR advantages have not been fully realized. 35,39
Advances involving strategies to minimize the effects of inhomogeneous RF will probably have some
impact on improving this application.
page 503
page 504
7 of 10 15-04-2010 20:47
Add to lightbox
Add to lightbox
Figure 18-16 A, An axial T2-weighted FSE image of a slice of the prostate acquired at 3 T using an
endorectal coil combined with an external torso array. The acquisition parameters were: TR 6200, TE
160, 1.5 mm slice/skip 0, 320 × 192, 18 echo train length, 12 cm FOV and 4 excitations. B, A coronal
image obtained from the multiplanar reformatting of the axial data. (Reprinted with permission from
reference 87. © 2004, with permission from Association of University Radiologists.)
8 of 10 15-04-2010 20:47
Add to lightbox
page 504
page 505
Figure 18-17 A sagittal T2-weighted FSE image of the pelvis of a normal young female volunteer.
These images were acquired using the reduced flip angle FSE sequence described in the text for
reduced SAR using the four element external torso array. The acquisition parameters were: TR 4750,
TE 149.7, 16 echo train length, 16 cm FOV, 4 mm slice/skip 1, 256 × 256, 16 kHz bandwidth, 4
excitations, 18 slices and 4 minutes and 32 seconds. Note the excellent visualization of the zonal
architecture in the uterus.
Add to lightbox
Add to lightbox
Add to lightbox
9 of 10 15-04-2010 20:47
Figure 18-18 Axial images of the liver of a normal volunteer taken with the three-point Dixon method at
3 T. The images were acquired in a single breath-hold using the four element torso array. The
acquisition parameters were: TR 900, 7 mm slice/skip 30, 512 × 256, 16 kHz bandwidth, 1 excitation
and 7 slices in 24 seconds. The combined image is shown on the top. The water image is shown on
the left and the fat image is shown on the right.
© 2010 Elsevier
10 of 10 15-04-2010 20:47
PROTON MAGNETIC RESONANCE SPECTROSCOPY AT HIGH FIELD

Human Brain
Besides fMRI, proton and other MRS applications have always provided the impetus to move to higher
fields. However, at this point there appears to be some confusion in the literature regarding the
advantages of high field for proton MRS studies of human brain. The Minnesota group have shown
excellent spectra from human brain at 7 T. In these spectra the line width of the methyl resonance of
creatine was found to be 9.5 Hz, as compared to 5.5 Hz at 4 T.91 As was the case in MR imaging, the
line widths at the higher fields have a significant contribution from diffusion of the metabolites through
microscopic susceptibility gradients. As pointed out by these authors, shimming at the higher field is
crucial to the acquisition of high-quality spectra.
Several investigators have shown that the SNR gains at 3 T and 4 T are relatively modest compared
with those predicted by theory. Bartha et al92 acquired multiple stimulated-echo acquisition mode
(STEAM) spectra (TE = 20 ms, volume = 8 cm3) in a single individual at 1.5 T and 4 T to compare the
precision of metabolite quantification using automated software that incorporated field strength-specific
prior knowledge. At 4 T, the SNR based on peak height increased about 80%, while the line widths
increased about 50%. This SNR increase improved the precision of quantification of metabolites by
about 40%. Barker et al93 compared single-voxel proton spectra of the human brain recorded in five
subjects at both 1.5 T and 3 T using the STEAM pulse sequence and data acquisition parameters that
were closely matched between the two field strengths. Spectra recorded in the white matter of the
centrum semiovale and in phantoms were compared in terms of resolution and SNR, and T2 were
estimated at both field strengths. Spectra at 3 T demonstrated only a 20% improvement in sensitivity,
compared to 1.5 T at short echo times (TE = 20 ms). Although spectra in phantoms demonstrated
significantly improved resolution at 3 T, as compared to 1.5 T, the in vivo spectra showed almost a
twofold increase in line width at 3 T. Based on a comparison of multivoxel studies at 1.5 and 3 T,
Gonen et al94 showed that the expected gains in SNR (23-46%) and spectral resolution were less than
theoretically predicted. However, even these modest improvements led to more reliable peak area
estimation and an H-1-MRS acquisition approximately 50% shorter at 3 T versus 1.5 T.
Li et al95 have shown why the gains at higher field are not as large as expected. Their explanation
deals with the contribution to the line widths from diffusion of the metabolites through microscopic
susceptibility gradients. In principle, the increase in line width may be reduced by decreasing the voxel
95
size in the spectral acquisition. This has been experimentally verified by Li et al and is shown in
Figure 18-19. Note the reduction in line width at the smaller voxel size. Since the SNR (as determined
by the peak height) scales inversely with the line width for a given peak area, these authors have
pointed out that the SNR decreases much less than expected as the voxel size is decreased and the
contribution of dynamic averaging to the line width is reduced. Thus, for higher field proton MRS,
optimal SNR gains may be achievable at smaller voxel sizes than are currently being employed at 1.5
T.
Body
We have recently shown that it is possible to perform single breath-hold proton MRS in the abdomen
96
and thorax at 3 T. Breath-holding either eliminated or markedly reduced phase and frequency shifts
and outer voxel contamination that were associated with the motion of the abdomen and the thorax
during breathing. Our method involves collecting each spectrum as a separate frame. We partially
suppress the water resonance in order to leave a sufficiently large water resonance to ascertain
whether there are any phase, frequency or line width changes during each breath-hold. The spectra
are then individually phased and summed for each breath-hold. Multiple summed breath-holds can also
be obtained in this manner.
Spectra of renal cell carcinoma metastases in the abdomen and thorax were obtained utilizing multiple
breath-hold averaging. An example of the results obtained at 3 T for a tumor is shown in Figure 18-20.
Breath-holding was found to be essential for spectroscopic investigation of the thorax. An example of a
1 of 3 15-04-2010 20:48
spectrum obtained from a tumor in the thorax is shown in Figure 18-21. These spectra exhibited a
resonance at 3.2 ppm, attributed to the trimethylamine moiety of choline metabolites. The results of
this study suggest a practical strategy for implementation of proton MRS in the body in regions where
respiratory motion is a potential difficulty.
page 505
page 506
Add to lightbox
Add to lightbox
Add to lightbox
Figure 18-19 A, An axial slice taken from a normal volunteer at 3 T. B and C, The real part of the
spectrum taken from the overlapping regions indicated on the image at 1.5 cm and 0.75 linear
dimension. D and E, NAA maps constructed from the data collected at the two different spatial
2 of 3 15-04-2010 20:48
resolutions: 1.5 and 0.75 cm respectively. (Reproduced with permission from reference 94.)
© 2010 Elsevier
3 of 3 15-04-2010 20:48
MULTINUCLEAR IMAGING AND SPECTROSCOPY AT HIGH FIELD

Historically, there has been a great deal of interest in performing spectroscopic or imaging studies on
NMR active nuclei other than protons such as P-31, C-13, Na-23, and O-17. These nuclei all have
much lower intrinsic sensitivities than protons and could, in principle, benefit from higher fields. The use
of high fields for multinuclear studies has been reviewed by Ugurbil et al.3 The major motivation for
studying these nuclei is that each of them can provide different and important physiologic and
metabolic information about the tissue being sampled (for a review of the uses of these nuclei in the
brain, see reference 97). Rather than provide an exhaustive review of multinuclear MR, we will focus
on current or potential applications of multinuclear MR at high fields.
Wei et al have shown that P-31 MRS of the human brain at 7 T can be performed with relatively small
voxels and in under 10 minutes of acquisition time.98-100 Several groups have reported P-31 MR
imaging approaches (rather than MR spectroscopic studies), some of which have been implemented at
high field.101-104 These methods produce images of the compounds being studied. An example of a
P-31 image of the foot of a normal volunteer obtained at 3 T in 4 minutes is shown in Figure 18-22.
Note that at 1.5 T this acquisition would take about four times longer. We are currently evaluating this
approach to determine muscle viability in the feet of patients with diabetes.
The utility of C-13 studies in humans has been reviewed by Gruetter et al105 and Shulman et al.105-107
Many of these studies are limited by SNR and thus can and will benefit from higher field strengths. One
caveat in these studies is that the SAR requirements for proton decoupling increase at higher field and
care must be taken to stay below the FDA guidelines for SAR.
There has also been some renewed interest in imaging Na-23. Thulborn et al108 have shown that Na-23
109-111
images are useful in assessing stroke in humans. Sandstede et al have shown that Na-23
images can be clinically useful in detecting myocardial infarctions even at 1.5 T. A number of Na-23
imaging sequences have been recently reported.112-114 These sequences have been implemented at 3
112 113
T and 4 T. An example of a Na-23 image, obtained from the heart of a normal volunteer at 4 T, is
shown in Figure 18-23. Note the acquisition time was about 8 minutes. At 1.5 T this image would take
about four times longer to acquire. This is about the scan time reported by Sanstede. 109 This
comparison again illustrates the SNR advantages of higher fields.
page 506
page 507
1 of 3 15-04-2010 20:49
Add to lightbox
Add to lightbox
Add to lightbox
2 of 3 15-04-2010 20:49
Add to lightbox
Figure 18-20 A, A single-shot FSE image of the abdomen recorded in a breath-hold at end expiration.
A large renal cell carcinoma metastasis in the right adrenal gland is demonstrated. The MRS voxel
(square in white, 2 × 2 × 2 cm3) was localized in the center of the tumor. B-D. The proton spectrum at
the location demonstrated in (A) acquired with one, four, and eight breath-holds, (8, 32, and 64 frames,
respectively). The total scan time for eight breath-holds was 2.1 min. The large signal at 4.7 ppm is
due to partially suppressed water. The signal at 3.2 ppm was assigned to the trimethylamine moiety
(TMA) that is common to the choline-containing compounds. The line width of the TMA signal in this
example was 15 Hz. The signal at ~1.3 ppm was assigned to lipids. Breath-hold data were summed
in post-processing. Note the improvement in the SNR with the increase in the number of frames.
(Reproduced with permission from reference 96.)
© 2010 Elsevier
3 of 3 15-04-2010 20:49
CONCLUSION
At the conclusion of his excellent review on high-field imaging, David Norris writes that at the 2003
meeting of the European Society of Magnetic Resonance in Medicine held in Cannes, "the audience
was convinced that technical innovation would finally bring such (high field) systems to the mainstream
of practice".1 At the time, Norris was probably referring to 3 T and 4 T systems. More properly, we
should refer to these scanners as "intermediate field" systems. At the International Society of Magnetic
Resonance in Medicine (ISMRM) meeting held in Kyoto in 2004, General Electric Medical Systems
estimated that about 25% of new scanner sales in the United States would be 3 T systems. This
follows their announcement made in Toronto at the 2003 ISMRM meeting that 3T would become the
premier platform for MR imaging.
As noted in the introduction of this chapter, there has always been a debate in the MR community
concerning the relative merits of higher fields. This debate should remind us of the tension that has
always existed between pioneers and settlers. Pioneers are always looking over the horizon to explore
new spaces. Settlers want stability and order. In the MR community, the pioneers are currently
working on systems ranging from 7 T to 9.4 T while the settlers have begun to use 3 T and 4 T
systems.
page 507
page 508
Add to lightbox
1 of 8 15-04-2010 20:51
Add to lightbox
Figure 18-21 A, A fast gradient-echo image of the thorax recorded in a breath-hold at end expiration.
A large renal cell carcinoma metastasis in the mediastinum is demonstrated. The MRS voxel (outlined
3
area, 2 × 2 × 2cm ) was localized in the center of the tumor. B, The proton spectra at the location
demonstrated in (A) acquired during free breathing (left) and in four breath-holds (right), respectively. A
total of 32 scans were recorded for each spectrum (1.07 min). Breath-hold data were summed in
post-processing. The large signal at 4.7 ppm is due to partially suppressed water. The signal at 3.2
ppm (arrow) was assigned to the trimethylamine moiety (TMA) that is common to the choline-
containing compounds. The line width of the TMA signal in this example was 25 Hz. The signal at ~1.3
ppm was assigned to lipids. Note the improvement in the SNR of both the TMA and the lipid signals in
multiple breath-holds compared to free breathing. (Reproduced with permission from reference 96.)
Add to lightbox
Add to lightbox
page 508
page 509
Figure 18-22 Axial images through the foot of a normal volunteer obtained at 3 T using a doubly tuned
2 of 8 15-04-2010 20:51
birdcage coil. The image on the left is a T1-weighted proton MR image. The image on the right is a
P-31 (total phosphorus) image acquired in 4 minutes at high spatial resolution. The acquisition details
103
are provided in references 102 and . Note the presence of a vial of reference sample that can be
used as a standard for determining concentrations.
Add to lightbox
Add to lightbox
Figure 18-23 Axial MR images of the heart of a normal volunteer acquired at 4 T using a 5 inch circular
surface coil for transmit and receive. The image on the right is a proton image showing the chest wall
and myocardium. The image on the left is a Na-23 image obtained using the method described in
reference 113 in 10 minutes with no cardiac or respiratory gating. The myocardium can be visualized.
The bright signal intensity just below the surface of the chest arises from cartilage in the ribs.
3 of 8 15-04-2010 20:51
REFERENCES
1. Norris DG: High field human imaging. J Magn Reson Imaging 18:519-529, 2003. Medline Similar articles
2. Takahashi M, Uematsu H, Hatabu H: MR imaging at high magnetic fields. Eur J Radiol 46:45-52, 2003. Medline
Similar articles
3. Ugurbil K, Adriany G, Andersen P, et al: Ultrahigh field magnetic resonance imaging and spectroscopy. Magn Reson
4. Hoult DI, Lauterbur PC: Sensitivity of the zeugmatographic experiment involving human samples. J Magn Reson 34:425-433,
1979.
5. Mansfield P, Morris PG: NMR imaging in biomedicine. New York: Academic Press, 1982.
6. Chen CN, Sank VJ, Cohen SM, Hoult DI: The field-dependence of NMR imaging. 1. Laboratory assessment of signal-
to-noise ratio and power deposition. Magn Reson Med 3:722-729, 1986.
7. Hoult DI, Chen CN, Sank V: The field-dependence of NMR imaging. 2. Arguments concerning an optimal field-strength.
Magn Reson Med 3:730-746, 1986.
8. Hoult DI: Sensitivity and power deposition in a high-field imaging experiment. J Magn Reson Imaging 12:46-67, 2000.
9. Alderman DW, Grant DM: Efficient decoupler coil design which reduces heating in conductive samples in superconducting
spectrometers. J Magn Reson 36:447-451, 1979.
10. Bydder GM, Steiner RE, Young IR, et al: Clinical NMR imaging of the brain-140 cases. Am J Roentgenol 139:215-236,
1982.
11. Zhao H, Crozier S, Doddrell DM: Compact clinical MRI magnet design using a multi-layer current density approach. Magn
12. Mulder GBJ, Peeren GN: MRI apparatus having a short uniform field magnet with an internal space. New York: US Philips
Corporation, 2001.
13. Alecci M, Collins CM, Smith MB, Jezzard P: Radio frequency magnetic field mapping of a 3 tesla birdcage coil:
experimental and theoretical dependence on sample properties. Magn Reson Med 46:379-385, 2001.
14. Baertlein BA, Ozbay O, Ibrahim T, et al: Theoretical model for an MRI radio frequency resonator. IEEE Trans Biomed Eng
15. Collins CM, Li SZ, Smith MB: SAR and B-1 field distributions in a heterogeneous human head model within a birdcage
coil. Magn Reson Med 40:847-856, 1998. Medline Similar articles
16. Collins CM, Smith MB: Signal-to-noise ratio and absorbed power as functions of main magnetic field strength, and
definition of "90 degrees" RF pulse for the head in the birdcage coil. Magn Reson Med 45:684-691, 2001. Medline
Similar articles
17. Collins CM, Smith MB: Calculations of B-1 distribution, SNR, and SAR for a surface coil adjacent to an anatomically
accurate human body model. Magn Reson Med 45:692-699, 2001. Medline Similar articles
18. Collins CM, Yang QX, Wang JH, et al: Different excitation and reception distributions with a single-loop transmit-receive
surface coil near a head-sized spherical phantom at 300 MHz. Magn Reson Med 47:1026-1028, 2002.
19. De Zanche N, Allen PS: Sensitivity calculations and comparisons for shielded elliptical and circular birdcage coils. Magn
20. Edelstein WA, Glover GH, Hardy CJ, Redington RW: The intrinsic signal-to-noise ratio in NMR imaging. Magn Reson Med
21. Glover GH, Hayes CE, Pelc NJ, et al: Comparison of linear and circular-polarization for magnetic-resonance imaging. J
Magn Reson 64:255-270, 1985.
22. Ibrahim TS, Lee R, Baertlein BA, et al: Application of finite difference time domain method for the design of birdcage RF
head coils using multi-port excitations. Magn Reson Imaging 18:733-742, 2000. Medline Similar articles
23. Ibrahim TS, Lee R, Baertlein BA, et al: Computational analysis of the high pass birdcage resonator: finite difference time
domain simulations for high-field MRI. Magn Reson Imaging 18:835-843, 2000. Medline Similar articles
24. Ibrahim TS, Lee R, Baertlein BA, Robitaille PM: B1 field homogeneity and SAR calculations for the birdcage coil. Phys
Med Biol 46:609-619, 2001. Medline Similar articles
1
25. Ibrahim TS, Lee R, Abduljalil AM, et al: Dielectric resonances and B field inhomogeneity in UHFMRI: computational
analysis and experimental findings. Magn Reson Imaging 19:219-226, 2001. Medline Similar articles
26. Ibrahim TS, Abduljalil AM, Baertlein BA, et al: Analysis of B1 field profiles and SAR values for multi-strut transverse
electromagnetic RF coils in high field MRI applications. Phys Med Biol 46:2545-2555, 2001. Medline Similar articles
27. Kangarlu A, Baertlein BA, Lee R, et al: Dielectric resonance phenomena in ultra high field MRI. J Comput Assist Tomogr
28. Keltner JR, Carlson JW, Roos MS, et al: Electromagnetic-fields of surface coil in vivo NMR at high-frequencies. Magn
4 of 8 15-04-2010 20:51
29. Kowalski ME, Jin JM, Chen J: Computation of the signal-to-noise ratio of high-frequency magnetic resonance imagers.
IEEE Trans Biomed Eng 47:1525-1533, 2000. Medline Similar articles
30. Ocali O, Atalar E: Ultimate intrinsic signal-to-noise ratio in MRI. Magn Reson Med 39:462-473, 1998. Medline
Similar articles
31. Singerman RW, Denison TJ, Wen H, Balaban RS: Simulation of B-1 field distribution and intrinsic signal-to-noise in
cardiac MRI as a function of static magnetic field. J Magn Reson 125:72-83, 1997. Medline Similar articles
page 509
page 510
32. Strilka RJ, Li SZ, Martin JT, et al: A numerical study of radiofrequency deposition in a spherical phantom using surface
coils. Magn Reson Imaging 16:787-798, 1998. Medline Similar articles
33. Vaughan JT, Garwood M, Collins CM, et al: 7T vs. 4T: RF power, homogeneity, and signal-to-noise comparison in head
images. Magn Reson Med 46:24-30, 2001. Medline Similar articles
34. Wang JH, Yang QX, Zhang XL, et al: Polarization of the RF field in a human head at high field: a study with a quadrature
surface coil at 7.0 T. Magn Reson Med 48:362-369, 2002.
35. Wen H, Denison TJ, Singerman RW, Balaban RS: The intrinsic signal-to-noise ratio in human cardiac imaging at 1.5, 3,
and 4 T. J Magn Reson 125:65-71, 1997.
36. Yang QX, Wang JH, Zhang XL, et al: Analysis of wave behavior in Lossy dielectric samples at high field. Magn Reson Med
37. Barfuss H, Fischer H, Hentschel D, et al: In vivo magnetic resonance imaging and spectroscopy of humans with a 4 T
whole-body magnet. NMR Biomed 3:31-45, 1990.
38. Alsop DC, Connick TJ, Mizsei G: A spiral volume coil for improved RF field homogeneity at high static magnetic field
strength. Magn Reson Med 40:49-54, 1998. Medline Similar articles
39. Greenman RL, Shirosky JE, Mulkern RV, Rofsky NM: Double inversion black-blood fast spin-echo imaging of the human
heart: a comparison between 1.5T and 3.0T. J Magn Reson Imaging 17:648-655, 2003.
40. Alsop DC, Watkins RD, Greenman RL, et al: In-vivo mapping of B1 uniformity produced by a whole body 3T RF coil.
Proceedings of the International Society of Magnetic Resonance in Medicine. Glasgow, Scotland, 1094, 2001.
41. Alsop DC: The sensitivity of low flip angle RARE imaging. Magn Reson Med 37:176-184, 1997.
42. Robitaille PM, Abduljalil AM, Kangarlu A, et al: Human magnetic resonance imaging at 8 T. NMR Biomed 11:263-265,
1998.
43. Bottomley PA, Foster TH, Argersinger RE, Pfeifer LM: A review of normal tissue hydrogen NMR relaxation times and
relaxation mechanisms from 1-100 MHz: dependence on tissue type, NMR frequency, temperature, species, excision, and
age. Med Phys 11:425-448, 1984.
44. Fischer HW, Rinck PA, van Haverbeke Y, Muller RN: Nuclear relaxation of human brain gray and white matter: analysis of
field dependence and implications for MRI. Magn Reson Med 16:317-334, 1990. Medline Similar articles
45. Wansapura JP, Holland SK, Dunn RS, Ball WS Jr: NMR relaxation times in the human brain at 3 Tesla. J Magn Reson
Imaging 9:531-538, 1999.
46. Bomsdorf H, Helzel T, Kunz D, et al: Spectroscopy and imaging with a 4 tesla whole-body MR system. NMR Biomed
1:151-158, 1988.
47. Kim SG, Hu XP, Ugurbil K: Accurate T1 determination from inversion-recovery images-applications to human brain at 4T.
Magn Reson Med 31:445-449, 1994.
48. Jezzard P, Duewell S, Balaban RS: MR relaxation times in human brain: measurement at 4 T. Radiology 199:773-779,
1996.
49. Michaeli S, Garwood M, Zhu XH, et al: Proton T2 relaxation study of water, N-acetylaspartate, and creatine in human brain
using Hahn and Carr-Purcell spin echoes at 4T and 7T. Magn Reson Med 47:629-633, 2002.
50. de Bazelaire CMJ, Duhamel GD, Rofsky NM, Alsop DC: MR imaging relaxation times of abdominal and pelvic tissues
measured in vivo at 3 T: preliminary results. Radiology 230:652-659, 2004.
51. Caravan P, Ellison JJ, McMurry TJ, Lauffer RB: Gadolinium(III) chelates as MRI contrast agents: structure, dynamics, and
applications. Chem Rev 99:2293-2352, 1999. Medline Similar articles
52. Zhou X, Caravan P, Clarkson RB, Westlund PO: On the philosophy of optimizing contrast agents. An analysis of 1H NMRD
profiles and ESR lineshapes of the Gd(III)complex MS-325+HSA. J Magn Reson 167:147-160, 2004.
53. Ba-Ssalamah A, Nobauer-Huhmann IM, Pinker K, et al: Effect of contrast dose and field strength in the magnetic resonance
detection of brain metastases. Invest Radiol 38:415-422, 2003. Medline Similar articles
54. Bernstein MA, Huston J III, Lin C, et al: High-resolution intracranial and cervical MRA at 3.OT: technical considerations and
initial experience. Magn Reson Med 46:955-962, 2001.
55. Frayne R, Goodyear BG, Dickhoff P, et al: Magnetic resonance imaging at 3 Tesla: challenges and advantages in clinical
neurological imaging. Invest Radiol 38:385-402, 2003.
56. Nobauer-Huhmann IM, Ba-Ssalamah A, Mlynarik V, et al: Magnetic resonance imaging contrast enhancement of brain
tumors at 3 tesla versus 1.5 tesla. Invest Radiol 37:114-119, 2002.
57. Trattnig S, Ba-Ssalamah A, Nobauer-Huhmann IM, et al: Use of contrast agent in high-field MRI (3 T). Radiologe 44:56-64,
5 of 8 15-04-2010 20:51

58. Preibisch C, Pilatus U, Bunke R, et al: Functional MRI using sensitivity-encoded echo planar imaging (SENSE-EPI).
59. Heidemann RM, Ozsarlak O, Parizel PM, et al: A brief review of parallel magnetic resonance imaging. Eur Radiol
60. van den Brink JS, Watanabe Y, Kuhl CK, et al: Implications of SENSE MR in routine clinical practice. Eur J Radiol 46:3-27,
61. de Zwart JA, van Gelderen P, Kellman P, Duyn JH: Application of sensitivity-encoded echo-planar imaging for blood
oxygen level-dependent functional brain imaging. Magn Reson Med 48:1011-1020, 2002. Medline Similar articles
62. Sodickson DK, McKenzie CA, Ohliger MA, et al: Recent advances in image reconstruction, coil sensitivity calibration, and
coil array design for SMASH and generalized parallel MRI. Magn Reson Materials Phys Biol Med 13:158-163, 2002.
63. Ra JB, Rim CY: Fast imaging using subencoding data sets from multiple detectors. Magn Reson Med 30:142-145, 1993.
64. Sicotte NL, Voskuhl RR, Bouvier S, et al: Comparison of multiple sclerosis lesions at 1.5 and 3 Tesla. Invest Radiol
38:423-427, 2003.
65. Willinek WA, Born M, Simon B, et al: Time-of-flight MR angiography: comparison of 3.0-T imaging and 1.5-T imaging-initial
experience. Radiology 229:913-920, 2003.
66. Willinek WA, Gieseke J, von Falkenhausen M, et al: Sensitivity encoding (SENSE) for high spatial resolution time-of-flight
MR angiography of the intracranial arteries at 3 T. Rofo-Fortschritte Auf Dem Gebiet Der Rontgenstrahlen Und Der
Bildgebenden Verfahren 176:21-26, 2004.
67. Al-Kwifi O, Emery DJ, Wilman AH: Vessel contrast at three Tesla in time-of-flight magnetic resonance angiography of the
intracranial and carotid arteries. Magn Reson Imaging 20:181-187, 2002. Medline Similar articles
68. Reichenbach JR, Barth M, Haacke EM, et al: High-resolution MR venography at 3 Tesla. J Comput Assist Tomogr
24:949-957, 2000.
69. Wang JJ, Alsop DC, Li L, et al: Comparison of quantitative perfusion imaging using arterial spin labeling at 1.5 and 4.0
tesla. Magn Reson Med 48:242-254, 2002.
70. Hunsche S, Moseley ME, Stoeter P, Hedehus M: Diffusion-tensor MR imaging at 1.5 and 3 T: initial observations.
Radiology 221:550-556, 2001.
71. Forbes KPN, Pipe JG, Bird CR, Heiserman JE: PROPELLER MRI: clinical testing of a novel technique for quantification
and compensation of head motion. J Magn Reson Imaging 14:215-222, 2001. Medline Similar articles
72. Pipe JG, Farthing VG, Forbes KP: Multishot diffusion-weighted FSE using PROPELLER MRI. Magn Reson Med 47:42-52,
73. Gudbjartsson H, Maier SE, Mulkern RV, et al: Line scan diffusion imaging. Magn Reson Med 36:509-519, 1996.
74. Gudbjartsson H, Maier SE, Jolesz FA: Double line scan diffusion imaging. Magn Reson Med 38:101-109, 1997.
75. Maier SE, Gudbjartsson H, Patz S, et al: Line scan diffusion imaging: characterization in healthy subjects and stroke
patients. Am J Roentgenol 171:85-93, 1998.
76. Mamata H, Mamata Y, Westin CF, et al: High-resolution line scan diffusion tensor MR imaging of white matter fiber tract
anatomy. Am J Neuroradiol 23:67-75, 2002. Medline Similar articles
77. Bammer R, Herneth AM, Maier SE, et al: Line scan diffusion imaging of the spine. Am J Neuroradiol 24:5-12, 2003.
78. Kubicki M, Maier SE, Westin CF, et al: Comparison of single-shot echo-planar and line scan protocols for diffusion tensor-
imaging. Acad Radiol 11:224-232, 2004. Medline Similar articles
79. Kruger G, Kastrup A, Glover GH: Neuroimaging at 1.5 T and 3 T: comparison of oxygenation-sensitive magnetic resonance
imaging. Magn Reson Med 45:595-604, 2001.
80. Yacoub E, Duong TQ, van de Moortele PF, et al: Spin-echo fMRI in humans using high spatial resolutions and high
magnetic fields. Magn Reson Med 49:655-664, 2003. Medline Similar articles
81. Duong TQ, Yacoub E, Adriany G, et al: Microvascular BOLD contribution at 4 and 7 T in the human brain: gradient-echo
and spin-echo fMRI with suppression of blood effects. Magn Reson Med 49:1019-1027, 2003.
82. Duong TQ, Yacoub E, Adriany G, et al: High-resolution, spin-echo BOLD, and CBF AM at 4 and 7 T. Magn Reson Med
48:589-593, 2002.
83. Pfeuffer J, van de Moortele PF, Yacoub E, et al: Zoomed functional imaging in the human brain at 7 Tesla with
simultaneous high spatial and high temporal resolution. Neuroimage 17:272-286, 2002.
84. Yacoub E, Shmuel A, Pfeuffer J, et al: Imaging brain function in humans at 7 Tesla. Magn Reson Med 45:588-594, 2001.
85. Silva AC, Koretsky AP: Laminar specificity of functional MRI onset times during somatosensory stimulation in rat. Proc Natl
Acad Sci USA 99:15182-15187, 2002. Medline Similar articles
86. Sosna J, Rofsky NM, Pedrosa I, et al: MR imaging of the prostate at 3-tesla: comparison of an external phased array coil to
6 of 8 15-04-2010 20:51
imaging with an endorectal coil at 1.5 tesla. Acad Radiol 11:857-862, 2004.
87. Bloch BN, Rofsky NM, Baroni RH, et al: 3 Tesla magnetic resonance imaging (MRI) of the prostate with combined pelvic
phased array and endorectal coils-initial experience. Acad Radiol 11:863-867, 2004. Medline Similar articles
88. Heiken JP, Lee JK, Dixon WT: Fatty infiltration of the liver: evaluation by proton spectroscopic imaging. Radiology
90. Lee JK, Dixon WT, Ling D, et al: Fatty infiltration of the liver: demonstration by proton spectroscopic imaging. Preliminary
91. Tkac I, Andersen P, Adriany G, et al: In vivo H-1 NMR spectroscopy of the human brain at 7 T. Magn Reson Med
46:451-456, 2001.
92. Bartha R, Drost DJ, Menon RS, Williamson PC: Comparison of the quantification precision of human short echo time H-1
spectroscopy at 1.5 and 4.0 Tesla. Magn Reson Med 44:185-192, 2000.
93. Barker PB, Hearshen DO, Boska MD: Single-voxel proton MRS of the human brain at 1.5T and 3.0T. Magn Reson Med
45:765-769, 2001.
94. Gonen O, Gruber S, Li BSY, et al: Multivoxel 3D proton spectroscopy in the brain at 1.5 versus 3 T: signal-to-noise ratio and
resolution comparison. Am J Neuroradiol 22:1727-1731, 2001.
95. Li BSY, Regal J, Gonen O: SNR versus resolution in 3D H-1 MRS of the human brain at high magnetic fields. Magn Reson
Med 46:1049-1053, 2001.
page 510
page 511
96. Katz-Brull R, Rofsky NM, Lenkinski RE: Breathhold abdominal and thoracic proton MR spectroscopy at 3T. Magn Reson
Med 50:461-467, 2003.
97. Ross B, Bluml S: Magnetic resonance spectroscopy of the human brain. Anat Rec 265:54-84, 2001. Medline
Similar articles
98. Lei H, Ugurbil K, Chen W: Measurement of unidirectional P-i to ATP flux in human visual cortex at 7 T by using in vivo P-31
magnetic resonance spectroscopy. Proc Natl Acad Sci USA 100:14409-14414, 2003.
99. Lei H, Zhu XH, Zhang XL, et al: P-31-P-31 coupling and ATP T-2 measurement in human brain at 7T. Magn Reson Med
50:656-658, 2003.
100. Lei H, Zhu XH, Zhang XL, et al: In vivo P-31 magnetic resonance spectroscopy of human brain at 7 T: an initial
experience. Magn Reson Med 49:199-205, 2003.
101. Speck O, Scheffler K, Hennig J: Fast P-31 chemical shift imaging using SSFP methods. Magn Reson Med 48:633-639,
102. Greenman RL, Elliott MA, Vandenborne K, et al: Fast imaging of phosphocreatine using a RARE pulse sequence. Magn
103. Greenman RL, Axel L, Ferrari VA, Lenkinski RE: Fast imaging of phosphocreatine in the normal human myocardium
using a three-dimensional RARE pulse sequence at 4 tesla. J Magn Reson Imaging 15:467-472, 2002.
104. Ernst T, Lee JH, Ross BD: Direct P-31 imaging in human limb and brain. J Comput Assist Tomogr 17:673-680, 1993.
105. Gruetter R, Adriany G, Choi IY, et al: Localized in vivo C-13 NMR spectroscopy of the brain. NMR Biomed 16:313-338,
106. Shulman RG, Hyder F, Rothman DL: Biophysical basis of brain activity: implications for neuroimaging. Q Rev Biophys
107. Shulman RG, Rothman DL: C-13 NMR of intermediary metabolism: implications for systemic physiology. Annu Rev
Physiol 63:15-48, 2001. Medline Similar articles
108. Thulborn KR, Gindin TS, Davis D, Erb P: Comprehensive MR imaging protocol for stroke management: tissue sodium
concentration as a measure of tissue viability in nonhuman primate studies and in clinical studies. Radiology 213:156-166,
109. Sandstede JJW, Pabst T, Beer M, et al: Assessment of myocardial infarction in humans with Na-23 MR imaging:
comparison with cine MR imaging and delayed contrast enhancement. Radiology 221:222-228, 2001. Medline
Similar articles
110. Pabst T, Sandstede J, Beer M, et al: Sodium T-2* relaxation times in human heart muscle. J Magn Reson Imaging
111. Pabst T, Sandstede J, Beer M, et al: Evaluation of sodium T-1 relaxation times in human heart. J Magn Reson Imaging
112. Boada FE, Gillen JS, Shen GX, et al: Fast three dimensional sodium imaging. Magn Reson Med 37:706-715, 1997.
113. Clayton DB, Lenkinski RE: MR imaging of sodium in the human brain with a fast three-dimensional gradient-recalled-echo
sequence at 4 T. Acad Radiol 10:358-365, 2003.
114. Kohler S, Preibisch C, Nittka M, Haase A: Fast three-dimensional sodium imaging of human brain. Magn Reson
7 of 8 15-04-2010 20:51
Materials Phys Biol Med 13:63-69, 2001.

115. Bottomley PA, Andrew ER: RF magnetic field penetration, phase-shift and power dissipation in biological tissue-
implications for NMR imaging. Phys Med Biol 23:630-643, 1978. Medline Similar articles
© 2010 Elsevier
8 of 8 15-04-2010 20:51
NTERVENTIONAL AND NTRAOPERATIVE AGNETIC

ESONANCE MAGING
John A. Carrino
Ferenc A. Jolesz
INTRODUCTION
The major goal of interventional MR is to provide dynamic image guidance for surgical and
percutaneous interventional procedures. Open surgeries and various minimally invasive procedures
have specific and somewhat unique additional requirements for imaging apart from but related to the
original diagnostic task. Image guidance, unlike diagnostic imaging, does not require high specificity
since in most cases the specific histopathologic diagnosis has already been established. Image
guidance needs more dynamic and adaptive imaging methods, balancing a flexible compromise
between temporal, spatial, and contrast resolution.
Image guidance is a complex task that involves localization, targeting, navigation, monitoring, and
control. Localization is a combination of the best diagnostic imaging with the best anatomic definition of
the operational volume. Accurate targeting or trajectory optimization can be improved by preprocedural
surgical planning. Seamless navigation relies on the ability to provide operator-defined and
task-oriented interactive imaging. Monitoring of a therapeutic effect requires dynamic imaging.
Quantitative imaging and closed loop control of energy deposition may best achieve adequate control
of certain therapies such as thermal ablation.
There are numerous reasons why MRI is best suitable for image guidance, the main one being
excellent and complex tissue characterization. The multiparametric and flexible image contrast that
provides not only characterization of diseased tissue but the definition of the related anatomy and
some functionally relevant tissue parameters (flow, perfusion, diffusion, tissue temperature) makes
MRI the best imaging method to provide intraprocedural guidance. Additional features, like the lack of
ionizing radiation, interactive multiplanar and/or volumetric imaging and the possibility of real-time or
near real-time image updating, make MRI the best imaging modality for integration with therapy.
There are some limitations for interventional MR, generally revolving around device and equipment
compatibility in the magnetic and radiofrequency environment. Most of the safety and compatibility-
related issues have been addressed and resolved.
The development of the intraoperative and interventional MR imager represents an important example
of creative vision and interdisciplinary teamwork. The result is a remarkable tool for a wide variety of
procedures that potentially encompass nearly the entire application gamut of minimally invasive surgical
and percutaneous radiology efforts. For certain surgical domains, the development of the
intraoperative MR imager marks a significant advance similar to the introduction of the operating
microscope.
© 2010 Elsevier
1 of 1 15-04-2010 20:52
TECHNICAL ASPECTS
Overview
page 512
page 513
There are a number of technical concerns to consider when putting an interventional magnet into
operation. Among the most pertinent are: the configuration and field strength of the magnets (defining
a compromise between access and signal-to-noise ratio-SNR), the safety and compatibility of the
devices and instruments that will be used in or near the magnetic field, the spatial accuracy of imaging
for localization and targeting, the optimal use of the imaging hardware and software (the dynamic
range of gradients, the limitation and availability of pulse sequences, radiofrequency coils) and the level
of integration with other guidance methods for planning, navigation, and 3D tracking.
The Magnet
Magnet Design
In terms of selecting an existing or designing and developing a new magnet for interventions or
intraoperative guidance, there are trade-offs with respect to magnet configuration, field strength, and
gradient strength. The issue of field strength is relatively simple. The higher the field, the higher the
SNR which can be used to improve spatial and temporal resolution and can make techniques like
temperature or flow-sensitive imaging, functional brain MRI, diffusion imaging or MR spectroscopy
more useful. Therefore, for intraoperative or interventional procedures the higher field systems are
preferable. However, there are reasons why most of the interventional and intraoperative systems are
operating in low- or mid-field strength, the main ones being access and cost.
The bore configuration trade-offs are access to the patient versus field strength and field homogeneity.
The cylindrical configuration of conventional high-field MR imaging systems using a superconducting
magnet excludes direct contact with the patient but provides high image quality and temporal
resolution. The lack of direct access prohibits real-time intraoperative imaging for open surgeries and
some minimally invasive procedures. Nevertheless, infrequent updates can be obtained during the open
procedures when the patient is moved in and out of the magnet and some catheter-based applications
and most of the thermal ablations can be monitored in real time within the closed bore of the high-field
magnet. The most obvious solution to overcome the access-related problems is to image the patient
within an open configuration magnet that allows different degrees of access to the patient, depending
on the design and procedure type. Open configuration, low-field, and mid-field MRI systems have been
developed primarily as the result of resistive magnet design and without much consideration of
potential interventional use. Nevertheless, they allow limited access to the patient and some direct
control of interventional, mostly percutaneous procedures so that it is feasible to adapt commercial
systems for interventional use. Other field strength trade-offs are safety (better at low field) versus
image quality and speed (better at high field). There are also other considerations in terms of
maintaining image quality, such as having to use a higher dose of intravenous contrast at lower field
strengths to produce enough conspicuity of enhancing lesions (particularly when spectral fat
suppression techniques are not available).1
1 of 15 15-04-2010 20:58
Add to lightbox
Figure 19-1 Magnet bore configurations for interventional MRI. There are three major types of
1
interventional magnet designs currently used for image-guided therapy. The "clam shell" (lower left)
configuration has a horizontally opened gap and is an adaptation of the routine open low-field MR
imager.2 The "cylinder" type (top middle) configuration is an adaptation of a short-bore high-field MR
imager but in general does not allow for direct patient access for the majority of procedures.3 The
"double donut" (lower right) configuration has a vertically opened gap specifically designed for
interventional work.
There are three main magnet bore configurations that are currently used for interventional MRI (Fig.
19-1). The "clam shell" configuration has a horizontally opened gap and is basically an adaptation and
slight modification to the routine open low-field MRI. The "cylinder" type configuration is an adaptation
of a short-bore high-field MR imager (some with a flared opening to allow an operator to reach in) but
in general does not allow for direct patient access for the majority of procedures. Therefore, imaging
and intervention are "uncoupled". That is, manipulation of devices or surgical work is done outside the
bore and then the patient must be moved into the magnet for imaging. This interventional mode has
been referred to as the "in/out" paradigm (for lack of a better moniker). Full surgical procedures may
be performed in the weak fringe fields surrounding an MRI system, using standard operating room
equipment. This approach can offer a cost advantage over fitting an entire operative suite with
"MRI-compatible" surgical equipment.
The third configuration is a mid-field (0.5 T) magnet specifically designed for intervention, which has
two cylinders separated by a gap for access, referred to as the "double donut" (SIGNA SP, GE
Medical Systems, Milwaukee, WI).2 Unlike a horizontal gap magnet, the vertical gap configuration is
well suited for interventional work because it allows the operator full access to the exposed anatomy of
the patient. The operator can perform various procedures while standing or sitting between the two
static magnet "donuts" of the system. Working in this mode does not require the patient to be moved
and has been referred to as the "stationary" paradigm. The patient in the magnet can be erect, sitting
or lying either parallel or perpendicular to the bore axis.
page 513
page 514
3
A wide variety of procedures may be performed on closed magnets but certain procedures that
require a particular position or minimal patient movement will be compromised or not even feasible. For
the clam shell configuration the main magnetic field is oriented vertically (perpendicular to the floor)
while for the cylinder and the double donut types the main magnetic field is oriented horizontally
(parallel to the floor), which has some implications for imaging. Each of these, with the exception of the
"double donut", is available from various manufacturers (Phillips, Siemens, General Electric).
In terms of intraoperative MR imaging (predominantly for neurosurgery) other configuration options

also exist. A movable MRI magnet can be placed in an operating room without affecting established
neurosurgical procedure with field strengths from 0.12 T (Odin Medical Technologies, Newton,
2 of 15 15-04-2010 20:58
Massachusetts) to 1.5 T (IMRIS Inc, Winnipeg, Canada). For the low-field mobile option, a foldable
local shield can be wheeled into the operating suite and unfolded during the imaging session. For the
high-field option, there is active shielding of the magnet but the operating room also requires
radiofrequency (RF) shielding. Thus far, a vertical configuration represents the best solution for
continuous intraoperative imaging. Various flat magnet design solutions are suggested as the future of
the ideal "stationary" paradigm.4
MR Safety and Compatibility

The MR safety profile of devices may be considered on three levels: safe in the magnet room, safe
around the patient, and safe to use in the patient. Many devices are magnet room safe if the forces on
the device are low and the device can work in the magnetic field. "Safe around the patient" refers to
devices that do not distort or degrade the image nor arc during imaging. Magnetic radiofrequency
fields applied during MRI may induce heating in devices made from conductive materials. Therefore,
"safe to use in the patient" refers to no local heating from radiofrequency energy deposition. An
additional feature is that both the device and the target will be visible when imaged and there are no
significant artifacts.
Since many objects are not MR room safe, it is imperative to test all objects (usually with a strong
hand-held magnet) prior to bringing them into the room. All unsafe objects need to be kept out of the
MRI area entirely to avoid any potential mishaps, although this is more difficult in an interventional
environment than a purely diagnostic imaging environment. Of course, a training program should be in
place to educate all personnel involved in procedures about magnet safety. MRI compatibility issues
are related to ferromagnetism, electric interference, image distortion, and device heating from
radiofrequency energy. Instruments may cause artifacts, leading to image distortion. The size of the
artifact depends mostly on the following attributes: size of instrument, orientation of the device to the
main magnetic field, static magnetic field strength, pulse sequence, and frequency-encoding gradient
direction.5
Radiofrequency energy deposition is a relevant consideration because needles, wires, and other
devices can act as radiofrequency antennae. The resultant heating at the tips of guidewires, catheters,
and other wire-shaped devices is an important safety issue. This effect is field strength dependent and
changes with the square of the field strength. Therefore, radiofrequency energy deposition at 1.5 T is
ninefold greater than at 0.5 T. Needles tend to heat more easily at higher radiofrequencies (higher
tesla levels necessitate shorter wavelength radiofrequencies). Some of the variables that affect the
relative magnitude of this heating have been identified but predictors of the absolute amount of heating
to formulate safety margins do not presently exist. The current safety regulation for local
radiofrequency heating, developed for externally positioned RF coils, is probably not suitable for
internal RF coils. Experimental results show an increase in coupling when moving a guidewire toward
the wall of an external RF transmit coil, documented by increased temperature of a saline solution in
6
close proximity to the tip of the guidewire. The coupling of the wire not only presents a potential
hazard to the patient but also interferes with the visualization of the wire. A safe alternative would be to
use non-conducting guidewires such as optical fibers7 for endovascular MR-guided procedures.
Magnetic Resonance Imaging

Localization
Spatial accuracy is extremely important for correct targeting and avoiding complications. Image-guided
8
procedures are fundamentally using frameless stereotactic techniques that demand exact localization.
The accuracy of localization depends upon good homogeneity of the main magnetic field and the
gradient field. Geometric distortion occurs because of "shift artifacts" and "warp artifacts".
Local frequency changes cause "shift artifacts" analogous to chemical shift. Shift artifact displaces
pixels in the frequency direction. The sources of shift artifact are: main magnetic field heterogeneity,
patient-induced magnetic susceptibility, materials/tissue interfaces (devices, air, etc.), and different
tissues/water in the same voxel. It can be seen on imaging as a shift at the surface of the air and the
3 of 15 15-04-2010 20:58
person or object. The effect of the frequency direction on needle artifact is important. In terms of
minimizing "shift artifact", the frequency direction should be orthogonal to the long axis of the needle to
provide the most accurate tip position. The type of pulse sequences used should be spin-echo and fast
spin-echo and not gradient-echo, to minimize the artifact. Other parameter modifications include
widening the bandwidth (allows more hertz per pixel), using a lower field strength magnet, and using a
lower spatial resolution matrix.
page 514
page 515
Non-linear gradients distort image scaling, causing "warp artifacts". This is magnet system dependent
and independent of patient, procedure or devices. Gradients that are weaker than the linear field
stretch the image while gradients stronger than the linear field shrink the image. This phenomenon can
also occur in the section selection direction. Therefore an ideal image (without geometric distortion) is
created at the linear gradient field showing actual positions. The strategies to minimize gradient warp
artifact are moving the area of interest toward the isocenter, using lower gradient field strength (thicker
section, narrower bandwidth, larger field of view), and using self-referenced guidance techniques
(described under Guidance Methods section below).
The image distortions mostly influence targeting when the targeting device and the target are in
different locations within the heterogeneous portion of the field (an accurate trajectory cannot be
defined). However, when they are at the same location or very close to each other, their actual
positions can be established even on the distorted images because the target and the targeting device
are both distorted in the same way at the same location.
Dynamic MRI
Interventional MR poses some of the greatest demands on the temporal and spatial resolution of
imaging technology because of a combination of factors: the changing orientation of the imaging plane;
surgical and therapeutic events occurring within the field of view (FOV); physiologic motion near and
within the FOV; intensity changes in the FOV due to introduction of contrast agents, deposition of heat
or other therapy effects; and the need for 3D volume information to safely and effectively apply,
monitor, and control therapies. MR-guided interventions require imaging methods that prescribe,
acquire, reconstruct, and display a series of images dynamically. Dynamic imaging differs from
standard MR imaging in that continually updating or reacquiring image data forms a large number of
images successively and rapidly.
Dynamic MRI methods are classified in two major categories: non-adaptive and adaptive. Nonadaptive
dynamic MRI methods perform a monitoring task, which is not influenced by the changes or events
within the imaging volume. These include four general categories of methods, all designed with the
basic goal of increasing the temporal resolution:
1. various fast pulse sequences (gradient-echo or echo-planar techniques)

2. reduced k-space sampling without using prior knowledge
3. reduced k-space sampling exploiting prior knowledge
4. reduced k-space sampling using model-based reconstruction.
Fast pulse sequences attempt to rapidly manipulate magnetic field gradients and RF pulses to acquire
as much spatially encoded information as possible in a short period, with sensitivity to different physical
properties (T1 and T2 relaxation, susceptibility, diffusion, chemical shift). Fast pulse sequence methods
may suffer from low tissue contrast or reduced spatial resolution and all diminish SNR. Images
acquired using reduced k-space sampling may also suffer from reduced spatial resolution, artifacts or
low tissue contrast or may be restricted to a smaller FOV.
MR fluoroscopy is a method for high-speed MR image acquisition with the goals of short acquisition
time per image (500 ms or less), high image rate (10 images or more per second), and high-speed
image reconstruction (150 ms or less from data acquisition to image display). 9 This may be achieved
by a variety of acquisition techniques and is an active area of ongoing development.10,11 MR
4 of 15 15-04-2010 20:58
fluoroscopic image data can be acquired with a limited flip angle pulse sequence with reduced TRs and
fewer phase encodes per image. The sequences can be applied continuously with images formed by
updating one set of data with data from the most recently taken measurements. In addition, a sliding-
window reconstruction technique together with radial scanning allows updating of the images at a
frame rate of 16 images per second. A specially designed backprojector is needed to perform the
real-time reconstruction of the image data. Also, a stream of images from a single receiver channel
can be reconstructed at multiple FOVs within each image frame. Alternatively, or in addition, when
multiple receiver channels are available, image streams from each channel can be independently
reconstructed at multiple FOVs. Artifacts from the motion became less evident on images as
progressively shorter acquisition times are used. The advent of ultra-fast imaging techniques has
extended the utility of MRI from a static and purely diagnostic status to an imaging modality ideally
suited for a number of therapeutic applications.
Adaptive methods have the potential to provide the spatial and temporal resolution that will be required
by interventional MR for true "real-time" MRI. Adaptive MRI methods are distinguished as those in
which the image data acquisition strategy is modified dynamically depending on information obtained
from the current and most recently acquired images (a context-sensitive technique). Adaptive imaging
methods can be implemented by modifying the spatial encoding of the FOV, by changing the algorithm
through which the spatial-encoding order is enacted during the pulse sequence or by a combination of
these two. When spatial encoding is optimized to the best estimate of the current FOV, then additional
computations are necessary before each image acquisition. If the algorithm by which the spatial-
encoding order is enacted during the pulse sequence is dynamically modified, one or more acquired
data sets need to be analyzed immediately after acquisition. Adaptive methods can be implemented in
the context of fast pulse sequences such as echo-planar or fast gradient-echo pulse sequences, which
in present use do not exploit known redundancy in data acquisition.
page 515
page 516
12
Three dynamically adaptive MRI methods are wavelet transform-encoded MRI, singular value
decomposition-encoded MRI,13 and adaptive Fourier transform-encoded MRI. Wavelet transform and
singular value decomposition encoding techniques are fundamentally different from standard Fourier
transform MRI because they use spatially selective RF excitations to encode data in an alternative
orthogonal basis set. In both methods, prior image information is exploited to reduce redundancy in
data acquisition. The multiresolution structure of the wavelet transform-encoded data can adaptively
locate and selectively resolve regions of change in an image. Singular value decomposition-encoded
MRI provides near-optimal spatial encoding and is suitable for a multiresolution adaptation as well as
keyhole updating method analogous to that employed in keyhole Fourier MRI. Adaptive Fourier
transform-encoded MRI differs from standard Fourier MRI in that the necessity of reacquiring data in
k-space is determined dynamically, minimizing oversampling.
RF Encoding
The standard practice in most MRI-guided therapy methods is to employ some form of RF
manipulation such as section selection and k-space sampling schemes for MRI encoding. However,
non-Fourier spatial encoding methodology exploits the already well-developed field of RF pulse
engineering to provide an additional flexibility that can enhance MR-guided therapy applications.
There are a number of specific advantages of RF encoding, including: flexibility to design encoding
functions that are optimized for a given application, ability to employ multiple resolutions within the
imaging volume, point spread functions free of ringing artifact, and robust reduced FOV imaging. A
further advantage is that RF encoding can both enhance and be enhanced by parallel imaging
methodologies. It is anticipated that the advantages of RF encoding will be exploited through the
development of hybrid imaging sequences. The combination of RF encoding with Fourier-encoded MRI
will make these methods effective as means for improving MRI-guided therapy applications.
Multiresolution and Altered FOV Techniques

Multiple resolutions along phase and slice-select dimensions (MURPS) enables dynamic imaging of
5 of 15 15-04-2010 20:58
focal changes using a graded, multiresolution approach. 14 With MURPS, it is possible to make a direct
trade-off between lower spatial and/or temporal resolution in parts of the volume for better spatial and
temporal resolution in other parts. Multiple FOV imaging is a technique useful for interventional
applications. A stream of images from a single receiver channel can be reconstructed at multiple FOVs
within each image frame. When multiple receiver channels are available, image streams from each
channel can be independently reconstructed at multiple FOVs. Very narrow FOV images are acquired
on MR receiver channels that collect guidewire or catheter data. These very narrow FOV images
provide very high frame rate continuous, real-time imaging of the interventional devices (25 fps). Large
FOV images are formed from receiver channels that collect anatomic data from standard imaging
surface coils and simultaneously provide a dynamic, frequently updated roadmap. These multiple FOV
15
images are displayed together, improving visualization of interventional device placement. Another
proposed reduced FOV approach16 uses the combination of 2D spatially selective excitation17 and the
18 19,20
UNFOLD technique. Reduced FOV approaches can help to avoid aliasing artifacts.
Parallel Imaging
Current developments of various parallel imaging techniques are extremely promising for interventional
applications. Despite the many advances in ultra-fast MRI, there is still a great need for additional
increases in the speed of image acquisition. Two main parallel imaging techniques, simultaneous
acquisition of spatial harmonics (SMASH)21 and sensitivity encoding (SENSE),22 have been described
in the literature and have shown image acceleration factors of up to eight. In both techniques, however,
the SNR was shown to be inversely proportional to the square root of the acceleration factor, when a
regular sampling of k-space is performed.
The SMASH technique is restricted to regular sampling schemes, which limits the choice of the imaging
planes and constrains the SNR to be inversely proportional to the square root of the acceleration
factor. By contrast, the SENSE technique allows for a more general k-space sampling trajectory as
well as for a flexible coil positioning. In its practical application, however, SENSE requires a regular
sampling of k-space and achieves a SNR similar to that of SMASH. When non-regular sampling is used
in SENSE, reconstruction becomes computationally expensive and slow, thereby making it impractical.
Recently, new parallel imaging techniques have been developed such as sensitivity profiles from an
array of coils for encoding and reconstruction in parallel (SPACE RIP)23 that would allow for an
24
acceleration of image acquisition, without a significant loss in the SNR. SPACE RIP, moreover,
allows for the flexible placement of the receiver coils around the object of interest, as well as for a
flexible choice of k-space lines; it is amenable to parallel reconstruction, making real-time image
rendering possible. In addition, the SPACE RIP technique is compatible with selective excitation
approaches, where the RF excitation is manipulated and used as a symbiotic element to improve
parallel imaging strategies. Because of the flexibility in sampling k-space using the SPACE RIP
approach, the center of k-space can be sampled more densely so that the majority of the image
energy can be preserved in order to minimize SNR loss due to sub-sampling. These features make
SPACE RIP exceptionally attractive for interventional MRI applications.
Coils
In conventional clinical MRI systems, the patient is surrounded not only by the magnetic gradient coils
but also by the head or body radiofrequency transmit-receive coil. For interventional applications,
traditional coils are often unsuitable because they do not allow access by the operator to the region of
interest. Therefore, coils must be designed with a sufficient space (gap or opening) available to
perform percutaneous and surgical procedures. The requirement for full access prohibits the use of
transmit RF coils that surround the operational volume. Transmit-receive surface coils were used in the
GE SIGNA SP open system with limited success.
page 516
page 517
Various solutions for limited access transmit RF coils have been introduced.25 Flexible RF coils can
6 of 15 15-04-2010 20:58
adapt to the shape of the anatomic region under investigation. They are integrated into the surgical
drape, can be sterilized, and allow full access to the image volume. Optimal coil design for a particular
anatomic region can significantly improve image quality; therefore, individual coil development is
strategically important for interventional MR. Optimization of the flexible RF coils for the target organ,
or only one part of it, can provide a high-resolution roadmap for the interventional process. The
introduction of parallel imaging for interventional MRI applications will require the development of
special multi-coils or multipurpose multi-coil systems.
Planning
Intraprocedural imaging, in addition to the diagnostic task of discriminating abnormalities, also furnishes
the operator with updates on intraprocedural changes and how they may modify preintervention data.
A robust surgical guidance and visualization system should integrate capabilities for data analysis and
on-line interventional guidance into the setting of interventional MR. Various preprocedure scans can be
fused and automatically aligned with the operating field of the interventional MR system.
Both presurgical and intraoperative data may be segmented to generate three-dimensional surface
models of key anatomic and functional structures. Models are combined in a three-dimensional scene
along with reformatted sections that are driven by a tracked surgical device. Thus, preprocedure data
augment interventional imaging to expedite tissue characterization and precise localization and
targeting. As the procedure progresses and anatomic changes subsequently reduce the relevance of
preoperative data, interventional data are refreshed for software navigation in true real time. Dynamic
MRI methods and navigational aids provide the fundamental imaging tools for surgical navigation and
therapy targeting and monitoring in an interventional MRI system.
Improvements in surgical strategies are directly related to optimizing access routes and controlling the
spatial extent of destructive energies used for therapy. Various devices employed in MRI-guided
interventions (including surgical robots, computer-assisted interventional tools, energy delivery devices)
are greatly dependent on image-derived information for feedback control of their actions.
The computer methods required for 3D planning and guidance are generally performed in the following
sequence:
image segmentation
rendering
registration of imaged anatomy with the patient's anatomy
display, and
interaction of the operator with the displayed information.
The segmentation of the images is a computational process initiated by identification or selection of

different tissue characteristics (may be manual, automated or a combination, semi-automated).
Segmentation is followed by interactive identification of anatomic regions performed manually or
automatically by warping acquired image data to a generic anatomic atlas. The 3D rendering of the
data and creation of a computer display representation of the anatomy follow. After these initial steps
of segmentation and rendering, 3D images must be accurately matched with the orientation of
intraoperative images on a computer display or projected onto the patient's corresponding anatomic
region to be used for localization (augmented reality). Registration involves the accurate determination
of the geometric transformation to match images correctly to the patient or patient's images in order to
provide an exact roadmap for procedures.
One of the tools available to accomplish visualization, registration, segmentation, and quantification of
medical data is the 3D Slicer (Fig. 19-2). The 3D Slicer uniquely integrates several facets of image-
guided medicine into a single environment. It provides capabilities for automatic registration (aligning
data sets), semi-automatic segmentation (extracting structures such as vessels and tumors from the
data), generation of 3D surface models (for viewing the segmented structures), 3D visualization, and
quantitative analysis (measuring distances, angles, surface areas, and volumes) of various medical
26
scans. The 3D Slicer has been integrated with an open MR scanner to augment intraoperative
7 of 15 15-04-2010 20:58
imaging with a full array of preoperative data. The same analysis previously reserved for preoperative
data can now be applied to exploring the anatomic changes as the surgery progresses. Surgical
instruments are tracked and used to drive the location of reformatted slices. Real-time scans are
visualized as slices in the same 3D view along with the preoperative slices and surface models.
The 3D Slicer is an open source development project begun at the MIT Artificial Intelligence Laboratory
and the Surgical Planning Laboratory (SPL) at Brigham and Women's Hospital, a teaching affiliate of
Harvard Medical School. It is now in active research use at institutions around the world with
contributing developers sponsored by a variety of governmental, commercial, and institutional funding
sources. This software platform of the 3D Slicer is based on an open source approach and the
application can be downloaded from the website ( http://www.slicer.org/).
Navigation
page 517
page 518
Add to lightbox
Add to lightbox
8 of 15 15-04-2010 20:58
Figure 19-2 Surgical planning. 3D Slicer surgical navigation software is used to guide biopsy
approach to the target with two-dimensional and three-dimensional display of preloaded images. The
software is integrated into the intraoperative MR imager to achieve online transfer of the images. An
MR-compatible, "surgical assist" robot that works co-operatively within the intraoperative MRI system
is shown (A). It is designed to carry and manipulate a needle into a designated position and direction.
To overcome certain obstacles, the robot is mounted above the operator's head. From this ceiling
vantage point, two long, rigid arms, equipped with a needle holder at the end of the arms, reach to the
surgical field. Surgical planning representation using the 3D Slicer application (B) shows the pelvic
MRI with the prostate modeled in green and the preoperatively planned target and skin entry points for
the robot to track. The robot, control software, and the clinical implementation are the results of an
international collaboration among the Surgical Planning Laboratory (SPL) at Brigham and Women's
Hospital (Boston, MA), Computer Integrated Surgical Systems and Technology (CISST) at Johns
Hopkins University (Baltimore, MD) and the Surgical Assist Technology Group in the National Institute
of Advanced Industrial Science & Technology (Tsukuba, Japan). (Images courtesy of Steve Pieper
PhD, Simon DiMaio PhD and Ron Kikinis MD, Surgical Planning Laboratory, Harvard Medical
School, Brigham and Women's Hospital.)
Navigation is essentially interactive imaging controlled by the operator. Navigation tools enhance the
operator's ability to perform an intervention in 3D by intraprocedural use of interactive imaging in which
image planes are referenced to the patient's anatomy. Tracking sensors (optical, electromagnetic or
ultrasonic) define the position of instruments ("pointers") within the operational volume of the real
patient and computer-generated displays of specially processed 3D images demonstrate the same
position at the image-based "virtual patient". Navigational tools can identify points, trajectories, and
planes on the images and can relate them to target points, trajectories or cross-sectional anatomy.
Tracking sensors can work as a 3D computer mouse and can be used for real-time planning of a
procedure. For targeting, the entry point can be marked out on the skin surface of the patient and the
target point on the image; alternative trajectories can be displayed by moving the tracked instruments
on the patient; the trajectories (needle path, cut planes) can be displayed on multiplanar cross-
sectional images.
MRI is ideal for navigation through the use of real-time or near real-time interactive plane selection in
any 2D plane or using preacquired isotropic volumetric 3D images from which any plane can be
reconstructed in real time. A realistic 3D environment is created for the physician, containing the
patient's relevant anatomy that can be reoriented, redrawn, specially displayed, and interactively
manipulated to explore and rehearse an interventional procedure.
Guidance Methods
Guidance for procedures may be accomplished by sequential but static localization or tracking.
Tracking is the process by which objects are dynamically localized in the patient's co-ordinate system.
Surgical or interventional tools (needles, forceps, knives) can be tracked within the image space and
their position can serve as the reference for acquiring various image planes, choosing view angles or
performing not only real but also virtual manipulations.
There are three main guidance methods: externally referenced, self-referenced, and MR tracking.
Externally referenced (frameless stereotaxy) methods include optical guidance and radiofrequency
guidance but are prone to distortion errors. Self-referenced methods include using anatomic landmarks
and fiducial markers and tend to minimize but not eliminate distortion errors. MR tracking uses the MRI
hardware for guidance and is considered "distortion free".
Externally Referenced
Frameless stereotaxy is an externally referenced method for localization and targeting. Optical tracking
is performed by attaching two flashing, light-emitting diodes (LEDs) to an instrument (the needle
holder) surface. These LEDs produce invisible infrared beams detected by three sensors
(approximately 1 meter from the LEDs) and localized by simple triangulation. By this tracking method,
the interventional MRI console computer has continuous information about the instrument tip position
and about alignment of the instrument's axis within the space. This optical tracking concept can be
used in the open magnet for interactive image plane acquisition, allowing tracking of any tool or
instrument if a direct optical link can be established between the LEDs and the sensors located above
9 of 15 15-04-2010 20:58
the imaging field. The position can be updated approximately six times per second. However, the
device must remain in clear view of the camera and the operator.
page 518
page 519
Add to lightbox
Add to lightbox
Figure 19-3 Externally referenced guidance method: Frameless stereotaxy. A tool or instrument, in this
case a biopsy needle holder, is localized by two flashing, light-emitting diodes (LEDs) attached to its
surface using simple triangulation (PIXSYS Inc, Boulder, CO). This can be used for interactive image
plane acquisition. In one method, images can be generated centered on the position of the instrument
tip in the conventional axial-coronal-sagittal planes. In another method, a tool can be used in the
fashion of an ultrasound transducer and images can be generated along the axis in any orthogonal
10 of 15 15-04-2010 20:58
plane about the optically tracked tool. One-step localization and targeting using image guidance is the
essential feature of frameless stereotaxy. The near real-time imaging during needle advancement
offers an opportunity for trajectory correction and depth measurement. This tracking method is
appropriate only for rigid instruments, which are at least partially outside the body where the
instrument-fixed LEDs can be tracked by the sensors. Sagittal (A) and axial (B) gradient-echo MR
images (TR 20/TE 9.8, flip angle 90°) of the scapula are centered at the needle tip. The needle icon
(dashed line) is superimposed on the image during the intervention and indicates the current location
of the needle (longer lines with narrow spacing), while the continued trajectory of the needle is
displayed as the widely spaced dotted white line. This case illustrates aspiration of a spinogelnoid
notch cyst that was causing an entrapment neuropathy. (Images courtesy of Carl S Winalski MD,
Harvard Medical School, Brigham and Women's Hospital.)
In one method, images can be generated centered on the position of the instrument tip in the
conventional axial-coronal-sagittal plane. In another method, a tool can be used as an ultrasound
transducer and images can be generated along the axis of the optically tracked tool. An advantage in
this application is that image planes that are defined by the instrument axis can be changed without
moving the tool itself. Positioning the optically tracked tools and selecting the related image planes
provide an interactive way to localize targets, define trajectories, and review alternative access routes.
This one-step localization and targeting using image guidance is the essential feature of frameless
stereotaxy. There is no need for several steps to register images with the patient's anatomy and define
trajectories in order to perform a biopsy. In this simplified process, one interactively scans through the
image volume, using the instrument and attached LEDs, until the appropriate image or trajectory is
found. When the instrument is a biopsy needle holder, for example, target localization is immediately
followed by advancement of the needle, which is also continuously visible in images. Because the
needle holder is localized during biopsy, the images will be in the same plane as the needle (although
some distortion due to bending of the biopsy needle is possible). Because the optics are independent
of the MR image, this process may cause distortion errors (Fig. 19-3).
A device localized by an external radiofrequency transmitter performs radiofrequency tracking. The

device may be within or outside the patient. The object absorbing RF energy may interfere with device
localization. Because the RF field and MR image distortion are independent, this uncoupling could
potentially cause errors.
Self-Referenced
page 519
page 520
Add to lightbox
Figure 19-4 Self-referenced guidance method. Axial T2-weighted fast spin-echo image (TR 2000, TE
108, ETL 12) just above the knee joint shows a lobulated, relatively low signal intensity lesion in the
popliteal fossa (arrow). Self-referenced techniques are defined by placing MR-visible markers close to
a target. In this case, the lesion is being localized by the operator's finger (arrow); user exposure is not
an issue in the MR environment. The trajectory is calculated from the position of the target relative to
the marker (finger tip location). Percutaneous core needle biopsy showed this lesion to be pigmented
villonodular synovitis. (Image courtesy of Philipp K Lang MD, Harvard Medical School, Brigham and
11 of 15 15-04-2010 20:58
Women's Hospital.)
Self-referenced guidance includes techniques as simple as placing a finger over a palpable abnormality
and imaging at that site, showing the relationship of the operator's finger to the lesion and therefore
facilitating a puncture site entry (simple but elegant mechanism for musculoskeletal biopsies). Fiducial
markers are another mechanism for self-referenced localization and consist of MR-visible markers
placed close to a target. The trajectory is calculated from the position of the target relative to the
markers. Shift and warp artifacts are similar for markers and target. Therefore, the impact of distortion
is dependent on the distance of the markers from the target and this works best for more superficial
lesions that are located closer to the fiducial markers (Fig. 19-4).
MR Tracking
The optical tracking method described earlier is appropriate only for rigid instruments that are at least
partially outside the body where the instrument-fixed LED can be tracked by the sensors. Tracking the
tip of flexible devices (catheters, drains, flexible endoscopes) is not possible using optical methods.
MR tracking is performed by MR receive coils located on or within the tracking device. A small wire
loop attached to the tip of the catheter or other flexible device acts as an RF receiver coil and picks up
the spatially specific RF of surrounding tissue during successive orthogonal readout magnetic field
gradient pulses. Advancement of the device tip can then be displayed within a previously acquired
image plane by this MRI tip-tracking method. Also, the tip position can be used to dynamically select
the image plane. The same imaging parameters are used to localize the device in the image and thus
these techniques are inherently self-referencing.
The MR image shows the true device position. Localization is interwoven with image sequencing and
thus rapid position updates are needed (80 per second). This can theoretically work with any pulse
sequence. During MR tracking procedures, signals are generated throughout the patient using a large
transmit coil but are detected with small receive coils. When a magnetic field gradient is applied, the
frequency of the signal depends directly on its position along that gradient. A typical receive-only coil
only detects MR signal from a small volume, thus resulting in data characterized by a narrow frequency
range that indicates the position of the coil.
However, determining the coil location is complicated by a number of factors. SNR is a function of coil
and gradient orientation, availability of signal-producing material, and losses in the coil and cable.
Furthermore, at certain orientations the signal from a solenoid coil is dual peaked and its width may
vary from a single to several frequency intervals. Another commonly observed problem is coupling of
tracking coils to surface, body, and other tracking coils. Finally, susceptibility effects present in the
frequency-encoded data need to be removed when calculating the 3D coil position.
Real-time MRI-based tracking of catheter and probe positions involves either active or passive
methods. In most passive methods an image is obtained and the catheter or probe position is
determined with the aid of susceptibility artifacts caused by material of the probe itself.27 In the active
28
methods a receiver is built into the tip of a catheter. The position of the tip along one dimension is
determined from the frequency of the signal in the small receiver and the strength of the gradient field.
Hybrid methods have been proposed with simultaneous tracking using the small receiver coil and
continuous imaging for visualization.29 Active catheter tracking has been a subject of research for more
than a decade; most studies have focused on projecting the magnitude sensitivity pattern of small
micro-coils onto three orthogonal axes. The location of the micro-coil can then be determined by
30
identifying the peaks of the three projections.
31
Both passive and active MR tracking methods have been tested in vitro and in vivo in animal models
and in a small number of human subjects for catheters introduced to the superior mesenteric artery in
order to perform MR imaging during arterial portography.32 Similar techniques can be applied to needle
tip tracking for biopsies and the percutaneous insertion of devices other than in the vascular
33,34
system. With active tip tracking, the guidewire in the MR environment can act potentially as an
12 of 15 15-04-2010 20:58
antenna and cause considerable heating adjacent to tissues.35 A single one-element micro-coil
incorporated into an interventional device has proven to be effective in some applications but it can only
supply tip position information. Device orientation in addition to tip position is important information
needed to define device location. Multiple positions on the device must be tracked to determine its
orientation. A novel single micro-coil design with three separate winding elements that provides both
36
the device orientation and tip position is being testing and is likely to be more useful. Definition of MR
scan planes, by using the device orientation and the target tissue location, permits automatic tracking
of the insertion of the device (Fig. 19-5).
page 520
page 521
Add to lightbox
Figure 19-5 MR tracking guidance method. Projective two-dimensional phase-contrast MR angiogram
of a pig highlighting the arteries in the neck. The star icon illustrates the location of the tip of an active
MR-tracking guidewire that has been placed into the carotid artery. The square, circle, and triangle
icons illustrate the locations of three coils on the distal end of a 5 Fr catheter. MR tracking locations
were determined in three dimensions and in real time with no perceptible latency. (Image courtesy of
John Pile-Spellman, Columbia Presbyterian Hospital, and Charles L Dumolin PhD, MRI and
Image Guided Therapy Program, General Electric Company.)
A complementary aspect to catheter use in the MR environment would be not only the ability to track
but to manipulate and guide. By exploiting the high magnetic field environment of the MRI scanner, it is
technically feasible to develop a catheter whose tip can be remotely oriented within the magnetic field
by applying a DC current to a coil wound around the catheter tip to generate a magnetic moment and
consequent deflection (including achieving arbitrary 3-dimensional deflections). The feasibility of using
such a system in a phantom maze has been demonstrated.37 This approach facilitates the guidance of
remotely steered devices to sites of clinical interest.
Multimodality Systems
Although MRI is the most versatile and multifaceted image guidance modality, several different medical
imaging modalities can be used to guide procedures. The properties of each modality are different and
therefore trade-offs exist. A preferred approach might be to provide simultaneous access to multiple
13 of 15 15-04-2010 20:58
modalities in a single integrated interventional system, allowing the operator to optimally use the
strengths of each as they are needed in isolation, combination or succession.
XMR
X-ray fluoroscopy and MRI have complementary capabilities and this combination could be extremely
powerful. X-ray fluoroscopy is the oldest and most commonly used modality for image-guided
procedures and many operators are familiar with using it. It offers superb temporal and spatial
resolution and high contrast for certain structures (metallic devices, bone, gas) although the soft-tissue
contrast is poor without an opacifying agent. The 2D projection format of X-ray fluoroscopy is often
viewed as a weakness compared to tomographic imaging, but it can be a great advantage when one
needs to monitor a procedure which is performed within a thick volume (catheter tracking, vasculature
opacification, cement injection into a bone). Tomographic imaging can be cumbersome for volumetric
monitoring and insufficient if limited to a few sections. Combining the two modalities of X-ray and MRI,
either into a single hybrid system or closely co-located devices, an operator would be able to exploit
the strengths of each as needed, during the same procedure. These types of multimodality systems
may find a significant role in many interventional radiology and angiography procedures in the near
future and may also serve as a bridging technology until true "MR-fluoroscopy" techniques are widely
available.
Both imaging modalities can be implemented in a hybrid system combining the advantages of each.
However, certain restrictions are imposed from one modality on the other. For the reception of MR
signals, an RF coil has to be placed on or around the patient near the region of interest. X-ray
fluoroscopic images are less useful if X-ray attenuating materials, located between the X-ray source
and detector, obscure patient anatomy. Unfortunately, many materials used in conventional RF coils
are highly X-ray attenuating. Removing the RF coil when switching from MR to X-ray image acquisition
disrupts the procedure and is impractical for coils positioned underneath the patient. The use of X-ray
compatible RF coils in a hybrid system allows for X-ray fluoroscopic image acquisition with minimal or
no impact by the RF coil without compromising the MR image quality.
page 521
page 522
38-40
Recently, X-ray/MRI hybrid systems that provide co-located (actually closely located) access to
both X-ray fluoroscopy and MR during a single procedure have been described. They consist of
separate MRI and X-ray systems with a coupled patient transport mechanism. This design strategy
allows the use of standard X-ray components (image intensifiers and rotating anode X-ray tubes) and
provides minimal impact of one imaging system on the other. As a result, use of these systems
requires moving the patient (and the table) tens of feet between two separate gantries. Moving the
patient is a somewhat time-consuming and cumbersome approach, leading to potential risk and loss of
image registration. It would also invariably lead to less than optimal use of the modalities, since one
may be reluctant to move the patient repeatedly. Nonetheless, this paradigm may be useful for certain
applications but a paramount consideration for patient safety and expediency is to have a single table
that is used for both modalities. This is not commercially available yet but other prototypes are
available using an intermediate strategy of transferring patients between the MRI and angiographic
units via a carbon fiber tabletop; this maneuver can be accomplished within 10 seconds. 41
An open MRI configuration allows space to accommodate an integrated X-ray detector. The unique
design of the vertically open "double donut" interventional magnet (SIGNA SP, GE Medical Systems,
Milwaukee, WI) provides a near ideal space for this unit without substantially compromising access to
the patient. Integrating an X-ray fluoroscopy system into the bore of an interventional MRI unit creates
a truly hybrid system, allowing access to both MRI and X-ray fluoroscopy without having to move the
patient.42,43 Preliminary work demonstrates that such systems are technically feasible which motivates
operators for further development of the system and expanded clinical use.44 Preliminary work on
thoracic targeting and performing TIP procedures is also promising.45 Some believe that development
of a truly hybrid system with minimal positioning restrictions is paramount to permit a robust array of
interventional and intraoperative surgical procedures. The integrated nature of the system could be
14 of 15 15-04-2010 20:58
especially beneficial when X-ray and MR guidance techniques are used iteratively (Fig. 19-6).
Image Fusion and Interventional MRI

The patient's anatomy can change during surgery through unavoidable shifts and deformations.
Intraoperative imaging can update the distorted anatomy and the surgeon can use these pertinent data
for localization and navigation. This anatomic update can be obtained by most of the interventional MRI
systems that are at low or mid-field. However, imaging methods that require high SNR cannot be
applied intraoperatively. Because of this limitation, preoperatively acquired MRA, fMRI, DTI and MRS
data that are necessary for image-guided procedures have to be registered to the intraprocedural
images (Fig. 19-7). This image fusion can be obtained by rigid registration methods but in the case of
intraoperative deformations, this will result in misregistration.
Currently, non-rigid registration methods are under development to use preoperatively obtained
high-field image data for correct targeting during procedures like brain surgery or prostate biopsy
performed in a mid-field open magnet.46-49 Multimodality fusion can also be used intraoperatively when
the source of preoperative information is from an imaging modality other than MRI (SPECT, PET) or it
is generated by spatially referenced electrophysiologic data (brain or cardiac electrophysiology).
© 2010 Elsevier
15 of 15 15-04-2010 20:58
Overview
Add to lightbox
Figure 19-6 XMR: multimodality imaging guidance for procedures. A hybrid system capable of MR
and X-ray imaging of the same field of view without patient movement is shown, thus extending the
2
versatility of the stationary paradigm. A digital flat-panel X-ray system (1024 array of 200 μm pixels,
30 frames per second) is integrated into a vertically open interventional 0.5 T magnet (SIGNA SP, GE
Medical Systems, Milwaukee, WI). The X-ray tube is placed on the top (vertical arrow) with a flat panel
digital detector placed at the bottom (horizontal arrow) of the gap. The X-ray source was adapted from
that of a bone densitometry system (DPX, Lunar, Madison, WI) and included a small, fixed-anode,
X-ray tube insert (Brand X-ray, Addison, IL) in which most of the magnetic components had been
replaced with non-magnetic alternatives. The X-ray flat panel detector (prototype of the Revolution
detector, GE Medical Systems) contains a CsI (Tl) phosphor layer. The system combines the strengths
of MRI (soft-tissue contrast, arbitrary plane selection) with those of X-ray fluoroscopy (high-resolution
real-time projection images, clear portrayal of bony structures) and allows switching between the
imaging modalities without moving the patient. The integrated nature of the system could be especially
beneficial when X-ray and MR image guidance are used iteratively. Another possible configuration
(not shown) is placing a conventional fluoroscopy C-arm unit outside the 5 gauss line of a routine
magnet and using an "in/out" paradigm. (Image courtesy of Nobert J Pelc PhD and Arundhuti
Ganguly PhD, Stanford University.)
Since their introduction more than a decade ago, interventional MRI methods have already affected the
fields of radiology, radiotherapy, and surgery. Almost every type of diagnostic procedure and several
types of therapy have been performed using MRI guidance to varying degrees, depending on the need,
local expertise, and type of interventional magnet deployed. Biopsies have been accomplished in
virtually every organ system.
The major, already proven therapeutic domains are brain tumor resection, various thermal ablations,
and brachytherapy. However, interest in cardiovascular applications has recently increased because of
improved systems. The application of intraoperative image guidance for monitoring and controlling
open surgeries, endoscopic procedures, thermal ablations, high- and low-dose brachytherapy, and
targeted drug delivery can merge with and enhance the burgeoning field of minimally invasive
therapies.
page 522
page 523
1 of 26 15-04-2010 21:03
Add to lightbox
Figure 19-7 Multimodality image fusion: ganglioglioma. This case illustrates the use of multimodality
image fusion with ventricles in blue, neoplasm in green, functional MRI (fMRI) in purple, diffusion tensor
(DT) tractography in yellow. 3D models of the intracranial arteries (internal cerebral artery, middle
cerebral artery), derived from MR angiography in red, are all displayed superimposed on a 3D-SPGR
image using the Slicer software application. The medial portion of the tumor is infiltrating the
corticospinal tract region. (Images courtesy of Ion-Florin Talos MD, Surgical Planning Laboratory,
Harvard Medical School, Brigham and Women's Hospital.)
In the surgery domain, direct visibility of the exposed anatomy is restricted to the surfaces. Surgeons,
with or without the use of microscopes or endoscopes, cannot see beyond and below the surfaces
because of the reflection of visible light. Using only direct visualization, the margins of most malignant
tumors are invisible for the surgeons while MRI can provide high sensitivity tumor definition within the
limitation of spatial and contrast resolution. To overcome the limitations of human vision, imaging
modalities have been used to help surgeons to expand their vision beyond the surfaces and exposed
layers of their operational volume and to make diseased tissue more distinguishable from normal. In
addition, imaging modalities can depict structural and functional properties of tissues well beyond the
specificity and sensitivity of the human eye. Intraoperative MRI not only improves surgical localization
and targeting; it also improves intraoperative navigation by using interactive multiplanar imaging. The
superiority over conventional approaches encourages rapid adoption in the wider clinical community
where these systems are available. The promise of image-guided techniques to deliver higher quality
treatment with less insult to the patient, among other advantages, has spawned numerous
investigations, although many are preliminary.
Neurologic Surgery
Intraoperative MRI has been developed and implemented for multiple neurosurgical procedures,
including brain biopsies, craniotomies for resection of mass lesions, cyst drainages, thermal ablations
for brain tumors, functional neurosurgery, and a variety of other CNS-related procedures. Among
intraoperative imaging methods, MRI offers the best tissue characterization for the brain, which is
essential in defining tumor boundaries and depicting functional brain anatomy. Intraoperative imaging
also allows the practitioner to update the approach to intracranial lesions continuously and adjust for
brain shift. Serial intraoperative MR imaging guidance based upon frequent volumetric image updates
offers surgeons several advantages over previous intraoperative guidance systems using a single
image database generated from preoperative imaging. Intraoperative imaging and navigation can be
fully integrated with various MRI scan configurations, dynamic MR imaging methods, surgical tools,
2 of 26 15-04-2010 21:03
robotic devices, and thermal ablation systems such as MRI-guided focused ultrasound. 50
Stereotactic procedures employing frame-based systems and utilizing preoperative MR or CT have

several shortcomings such as long procedure time, patient discomfort and transport, poor fail-safe
capabilities and targeting inaccuracies due to brain shift. Conducting all procedural steps in an
interventional MRI has the potential to alleviate some of these deficiencies. With an intraoperative MRI
system, important anatomic, functional, and vascular structures can be successfully avoided;
boundaries of low- or high-grade malignant brain tumors can be more accurately defined, and foci of
possible higher grade within these lesions can be identified; foci of high-grade astrocytomas can be
differentiated from radiated brain; hyperacute hemorrhage or infarction during and after procedures
can be determined and the possible communication of cystic collections with CSF can be ascertained.
Intraoperative MR imaging techniques have the potential to greatly improve the stereotactic methods
used for functional neurosurgery. No longer are neurosurgeons and patients always constrained by
uncomfortable head frames and conventional stereotaxy. Accuracy and complication avoidance are
improved by intraoperative imaging. Safety of operative machinery and equipment in an MR imaging
operative suite is attainable, even with deep brain stimulating electrodes for epilepsy. Intraoperative
MR-guided neurosurgery represents a natural progression from frame-based to frameless stereotactic
techniques.51
page 523
page 524
Maximal surgical excision is the best current treatment option for low-grade brain tumors since
complete excision can reduce the risk of progression to higher grade, reduce the incidence of seizures,
and render patients eligible for adjuvant therapies. 52 Intraoperative MRI is helpful for navigation as well
as demonstrating the tumor margins to achieve a complete and safe resection of intracranial lesions
located in or near eloquent brain areas. It enables an image-based functional monitoring of the brain,
which is critical for motor, sensory or language function. However, the visual distinction of these lesions
from normal brain is difficult. Determining the optimal limits of the resection at the margins can be
uncertain.53 Therefore, the surgeon is faced with the difficult decision of how far to carry the surgical
resection knowing that the risk of permanent neurologic deficit must be weighed against the goal of
complete resection.
MRI has proved to be the most promising imaging method for neurosurgical guidance. It can detect
intracranial lesions with superior sensitivity; locate "eloquent" cortical areas (functional MRI) and white
matter fiber tracts (diffusion tensor imaging), and it provides high-quality visualization of intracranial
vasculature. But many current intraoperative "open" type MRI scanners suffer from significant
limitations. Due to low magnetic field strength and low gradients, not all of the necessary pulse
sequences can be utilized in an open scanner. Even if this were not the case, scanning times for
diffusion tensor imaging (DTI) and functional MRI (fMRI) would not be compatible with the time
constraints imposed by the surgical procedure. Thus there has been an interest in using higher field
strength MRI (1.5 or 3 T) for resection of brain tumors. Intraoperative functional techniques such as
MR spectroscopy, functional MRI, MR angiography and venography, and diffusion-weighted imaging,
which have become routine at some high-field MR units, can significantly influence surgical decision
making54 (see Fig. 19-7).
The potential complications associated with intraoperative MR-guided neurosurgery are similar in
incidence to those seen in the conventional neurosurgical operating room. However, these are
immediately recognized in an intraoperative MRI environment. Untoward events associated with
performing surgery in an MR environment are uncommon. Therefore, complications related to the
surgical procedure are reduced and the risks of neurologic deterioration due to tumor removal and
postoperative complications are minimized.55
Intraoperative deformations are a major issue. Mechanical factors, physiologic motions, and
pathophysiologic processes such as edema or hemorrhage typically cause these displacements. The
unwanted movement of tissues and the reduction or swelling of tissue volumes by the advancing
3 of 26 15-04-2010 21:03
surgery can be so substantial that preoperatively acquired images become virtually useless. Brain shift
is a continuous dynamic process that evolves differently in distinct brain regions.56 Therefore, only
serial imaging or continuous data acquisition can provide consistently accurate image guidance.
Fortunately, intraoperative imaging, which updates the original preoperative (baseline) model, provides
an effective solution.
In pediatric populations, interventional MRI has an expanding role in minimally invasive neurosurgery. In
a preliminary investigation it has been found to be safe, reliable, and useful for frameless stereotactic
procedures such as tumor biopsies, cyst aspirations, and catheter placements. There were no
hemorrhagic, neurologic or infectious complications in one prospective series.57 Neuroendoscopic
treatment of hydrocephalic children performed with an interventional MRI introduces new imaging
features that, in combination with endoscopy, are particularly valuable for performing these
operations.58 Real-time production of MR images in three dimensions during the procedure facilitates
endoscope guidance. Intraoperative changes such as brain shift, effects of perforation, and drainage
of cysts are shown during the procedure. Patency of cysts or fluid compartments inside the ventricular
system can be assessed by intraoperative intracyst injection of diluted gadolinium. Data confirm that
this technology lowers the incidence of and extends the duration to tumor recurrence in children. The
posterior fossa in a child poses a considerable challenge to the neurosurgeon. MRI-guided surgery
allows for real-time interaction between imaging and the neurosurgeon, with higher resolution than
other modalities, and results in improved resection imaging and real-time needle guidance in tumor
operations.59
Some key questions are: does intraoperative MRI improve outcomes and is there a favorable
cost-benefit for using this technology? Interventional MRI guidance is an effective tool for a more
extensive tumor excision compared with conventional neurosurgery without an increased risk to the
patient but its use lengthens the average time for a craniotomy. Preliminary reports concerning the
efficacy and usefulness of MRI-guided navigational tools for the performance of neurosurgery are
encouraging and have shown that the more extensive removal of glioblastomas afforded by
intraoperative MRI leads to significantly prolonged patient survival compared with conventional
60
surgery. Further outcomes analysis must be performed, however, to determine whether these new
techniques significantly decrease overall long-term morbidity or increase survival in those patients who
have low-grade astrocytomas.
Although cost analyses are few, there is some information supporting MR-guided neurosurgery. A
retrospective comparison of the costs and benefits of brain tumor resection in the conventional
operating room with an interventional magnetic resonance operating suite was performed.
Comparisons were made for adults and children for length of stay, hospital charges and payments,
hospital total direct and indirect costs, readmission rates, repeat resection interval, and net health
outcome. The data from this analysis suggest that MR-guided surgery improves net health outcomes
61
by reduced length of stay, reduced readmission rates, and reduced hospital charges and costs.
Overall, the advantages of intraoperative MRI provide a level of comfort to the surgeon and a
presumptive margin of safety to the patient that is unattainable during conventional surgical approaches
and, given the choice, some neurosurgeons would prefer to operate in the interventional magnet.
Intraoperative MR imaging has been successfully implemented for a variety of intracranial procedures
and provides continuous visual feedback, which can be helpful in all stages of neurosurgical intervention
without affecting the duration of the procedure or the incidence of complications.62 As minimally
invasive techniques are developed to treat brain lesions (e.g., focused ultrasound), noninvasive brain
mapping techniques will be required to guide the deployment of these techniques in individual patients.
page 524
page 525
Other Intraoperative Applications

Intraoperative MRI has already demonstrated usefulness in neurosurgery; efforts to use image-guided
therapy in the form of MRI in general surgery, however, are in pilot stages. One of the more promising
4 of 26 15-04-2010 21:03
applications of intraoperative MRI was demonstrated when the feasibility of MRI-guided lumpectomy
for breast cancer was shown.63 Similarly to glioma, the extent of breast cancer cannot be defined by
the surgeon using visual observation of the exposed tissue. Using MRI to define tumor margins has
great potential to improve the successful debulking of breast cancer. It has also been shown that a
64
surgically relatively difficult task (surgery for anal fistula) is feasible using MRI guidance. MRI
demonstrates the exact anatomy of the tracts and abscesses, and confirms that all have been
adequately treated. It may become particularly useful in surgery for recurrent and complex anal
fistulae, and may lead to fewer recurrences.
Diagnostic and Therapeutic Percutaneous Non-Vascular Procedures

Prostate
Many interventional MRI applications have been performed in the abdomen and pelvis. Biopsies are
indicated in patients with lesions not visible by other modalities (US or CT imaging) or that may be
difficult to access (hepatic dome or adrenal) without complex angulation. While US has been the
standard modality for prostate biopsies, MRI has been establishing a role in certain contexts: men in
whom prostate-specific antigen is elevated with a negative US-guided biopsy and a lesion visible on
MRI or men who have had rectal surgery (since US is done as a transrectal technique).
MRI-guided biopsy of the prostate may be accomplished in two ways:
1. in a closed bore system using a transgluteal approach

2. open horizontal system using a perineal approach.65
It has been demonstrated that preoperatively obtained high-field prostate images that demonstrate the
prostate cancer can be used to guide biopsies performed at an open mid-field system where the
image quality is sufficient to depict the anatomy of the prostate but cannot demonstrate the tumor.
Since the position of the patient is different during the biopsy procedure, non-rigid registration between
the high-field and low-field images is needed for accurate targeting. 66
Musculoskeletal
MRI-guided interventional procedures involving bone, soft-tissue, intervertebral discs, and joints are
safe and sufficiently rapid for use in clinical practice.67 There are several reports that show diagnostic
yield similar to other modalities.68-74 In terms of musculoskeletal lesions, there are several categories
of MRI interventions that the authors consider useful and advantageous for diagnostic procedures that
relate largely to biopsies. The categories of usefulness include: lesion not visible by other modalities;
collection or lesion adjacent to hardware; and targeting a specific portion of a lesion (based on
functional imaging or contrast enhancement). On the therapeutic side, the utility of MRI for
musculoskeletal applications is evident for cyst aspiration, intra-operative guidance or needle
localization and tumor ablation. MRI guidance has also been used for a host of spine injections (Fig.
19-8) and pain management procedures such as discography,75 periradicular nerve blocks,76
sympathetic blocks,77 sacroiliac joint injections,78 celiac plexus blocks79 and facet joint neurotomies
using cryotherapy.
Breast
MRI mammography is increasingly used to detect or characterize breast lesions. An extension of this is
to use MR as guidance for tissue-sparing histologic sampling of these lesions, particularly for lesions
that are occult on other imaging modalities. Techniques are available that allow preoperative hook wire
localization or core biopsy under MR guidance. For a variety of reasons closed architecture high-field
strength magnets are most widely used, thus necessitating the "in/out" paradigm (or reaching into the
80
bore when possible). Robotic guides are being developed for such biopsies.
The following approaches have been described: MR-guided freehand localization in supine position;
stereotactic localization in a supine position; and, most frequently used, localization in the prone
5 of 26 15-04-2010 21:03
position by means of a compression device that immobilizes the breast to prevent tissue shift during
intervention. There is a growing experience with interventions on open magnets. Recently,
percutaneous vacuum biopsy of enhancing breast lesions has become possible under MR guidance.
The system allows accurate and safe access to lesions in any location of the breast and direct
monitoring of representative excision by imaging and allows reliable histologic evaluation of lesions
smaller than 10 mm.81 MR-guided vacuum biopsy also allows reliable histologic sampling of small
82
contrast-enhancing lesions that are not visible by any other modality.
Thoracoabdominal Cavity
Drainage of fluid collections is a novel application for interventional MRI in the thoracoabdominal organs
or cavity. The multiplanar imaging capability of MR is particularly helpful for fluid collections in the
subphrenic location83 and has also been used for percutaneous cholecystostomy84 and nephrostomy.85
Tumor Ablations
page 525
page 526
Add to lightbox
Add to lightbox
6 of 26 15-04-2010 21:03
Add to lightbox
Add to lightbox
Figure 19-8 Neurosurgery: brain tumor resection. Preoperative axial T2-weighted FSE image (A)
shows a large astrocytoma (WHO grade II/IV) in the left cerebral hemisphere. Postoperative axial
T2-weighted FSE image (B) shows a resection cavity with air-fluid level. Residual tumor present
anteriorly (arrow) is adjacent to an eloquent brain region (motor and sensory strip). The presurgical (C)
and post-surgical (D) 3D images show this relationship (ventricles in blue, neoplasm in green).
(Images courtesy of Ion-Florin Talos MD, Surgical Planning Laboratory, Harvard Medical School,
Brigham and Women's Hospital.)
Tumor ablation is the direct application of chemical or thermal therapies to a specific focal tumor (or
tumors) in an attempt to achieve eradication or substantial tumor destruction. The methods of tumor
ablation most commonly used in current practice can be divided into two main categories: chemical
ablation and thermal ablation. Thermal ablation includes energy sources that destroy a tumor by using
thermal energy, with either heat (e.g., radiofrequency, laser, microwave and focused ultrasound) or
cold (cryoablation). Imaging is used in five separate and distinct ways: planning, targeting, monitoring,
7 of 26 15-04-2010 21:03
controlling, and assessing treatment response. 86
Thermal Ablation
page 526
page 527
Thermal treatment of neoplasms or neoplastic-like processes may become an effective, minimally

invasive, therapeutic tool given proper guidance of the thermal deposition. Proper guidance implies that
the spatial extent of the diseased tissue must be clearly identified and this predefined tissue volume
brought to an appropriate temperature for an appropriate length of time to insure effective cell
destruction while minimizing collateral cell damage to adjacent, healthy tissue. An advantage can be
gained if the thermal treatment can be combined with an imaging modality capable of monitoring in 3D
the tissue temperature changes which occur during the heating or freezing process in order to
maximize damage to target tissue while minimizing damage to healthy tissue or adjacent critical or
vulnerable structures (e.g., nerves). Clearly, there is an advantage if the thermal treatment can be
combined with an imaging modality that is capable of monitoring tissue temperature changes;
temperature monitoring then guides the thermal process to selectively damage targeted tissues.
MRI is well suited to thermal ablation because of the ability to perform thermometry for heating
procedures and the distinct conspicuity of the cryoablation zone ("iceball") for freezing procedures.
Control is necessary to confine thermal damage to targeted tissues. Control of the ablation is available
by real-time visualization of the thermal effects; the most important effect is temperature change.
Magnetic Resonance Thermometry

The roles of MR-guided thermal ablation are, first, to restrict energy deposition to the target tissue by
demonstrating transient temperature elevations compared to the surrounding normal tissue and,
second, to signal the irreversible phase transition (cell necrosis) within the target volume. Various
magnetic resonance techniques for monitoring tissue temperature are available and collectively this
field is known as MR thermometry.
Despite progress, significant work remains to be done in the development of practical MRI techniques
to monitor and improve the efficacy of various thermal treatments. The basic premise of MR
thermometry is that the temperature dependence of MR tissue characterization parameters, such as
tissue water relaxation times, diffusion coefficients, and resonant frequencies, can be exploited for
tissue temperature monitoring. MRI-related tissue characteristics (T1 relaxation, diffusion, and
chemical shift) are sensitive to temperature changes and have been applied for both qualitative and
quantitative thermal mapping.87-91
The relative benefits and limitations of four potential MR temperature probes have been evaluated.92,93
These probes are: water T1 relaxation, water diffusion coefficient, water (proton) resonant frequency,
and the resonant frequency of a methoxy group from a praseodymium-based contrast agent. T1-based
methods may provide better results when field homogeneity is poor or for lower field strength (e.g.,
<0.5 T). Proton resonance frequency (PRF) shift effect still remains the best candidate for reliable
temperature estimation during heat treatments, based on observations that above 60°C most tissues
show thermal coagulation effects which change the basic temperature sensitivities of proton relaxation
times, water diffusion and magnetization transfer but not the water proton resonance frequency.94
Thus the bulk of MR thermometry studies carried out to date have employed the temperature
dependence of the proton resonant frequency (PRF) with the primary imaging tool being single
gradient-echo imaging sequences (Fig. 19-9).
The PRF method is considered the preferred choice for field strengths >1.0 T because of its sensitivity
and linear relation with temperature and near independence with respect to tissue type. Quantitative
measurements of temperature changes in tissue can be made based on the change in the PRF of the
tissue. A pixel-by-pixel measurement of the PRF shift can be made as it changes from baseline. This
provides a calculated temperature for each pixel in an image plane (several planes can be acquired to
cover the tumor). Overall, this technique is estimated to provide accuracy within a few degrees
8 of 26 15-04-2010 21:03
centigrade depending on the field strength of the magnet. PRF methods typically use RF-spoiled
gradient-echo imaging methods to measure the phase change in resonance resulting from the
temperature-dependent change in resonance frequency.
There are two major limitations hindering use of the PRF technique for applications with many lipid-
containing areas such as breast or abdomen: interscan motion and lipid contamination. The single
gradient-echo technique for temperature monitoring relies on a water resonance frequency shift with
temperature of approximately -0.008 ppm/C. This translates to a fairly small frequency shift of
approximately 0.51 Hz/C at field strength of 1.5 T. Even a small frequency shift will, however, result in
a measurably large change in the phase È of the gradient-echo signal when allowed to evolve over an
appropriate echo time (TE) where:
This relation provides the basis for monitoring tissue temperature changes wherein a baseline phase
map is acquired prior to heating followed by baseline subtraction from subsequent phase maps
acquired during heating. Changes in these difference maps will, in principle, reflect locations where
measurable temperature changes have occurred between the two time points due to thermal
treatment. Problems arise, however, when motion occurs between scans or when more than one
resonant frequency is present.
Add to lightbox
9 of 26 15-04-2010 21:03
Add to lightbox
page 527
page 528
Figure 19-9 Spine injections. A, Nerve block: axial CBASS image (TR 8.4/TE 4.2) shows needle at
periradicular space of the exiting S1 nerve. B, Discography: sagittal CBASS image (TR 95/TE 7)
shows needle in the intervertebral disc. (Images courtesy of Roberto Blanco-Sequeiros MD, Oulu
University Hospital, Finland.)
An object introduced into the main magnetic field of an MR scanner invariably causes perturbations in
the magnetic field. This in turn causes the Larmor frequency to become a complicated function of
position. In the case of a single chemical species, e.g., water, the baseline phase map provides a
sensitive, high spatial resolution image of the Larmor frequencies within a particular section. If a later
scan is acquired with the object in a slightly different position, a different set of Larmor frequencies is
measured reflecting a new baseline phase map. This is one of the fundamental problems associated
with temperature monitoring with single gradient-echo sequences. Small interscan motions lead to
displacements that can lead to large, unpredictable local Larmor frequency changes that can easily
exceed small changes due to temperature variations caused by any thermal intervention.
The basic relationship (shown in the equation) between the phase and frequency within a given voxel
assumes the presence of a single chemical species (e.g., water). The presence of fat as well as water
within a voxel, however, renders the assumption of linearity between phase and TE invalid. Since some
tissues (such as breast and abdomen) are generally a mixture of fat and water, the linear relation
between phase and TE will be violated within many voxels. Under these circumstances phase mapping
with a single gradient-echo imaging sequence will not be useful without some form of fat suppression
or other means to deal with the fat. One proposed technical modification is to make separate
measurements of the water and fat with rapid spectroscopic imaging sequences that gather not one
but many gradient-echoes so that the amplitudes and the frequencies of the fat resonance and water
resonance are measured. Not only does this remove the lipid contamination problem altogether but it
will decrease the sensitivity to motion and provides additional temperature-sensitive parameters. These
are the fat and water amplitudes which, with sufficient T1 weighting, also provide temperature change
estimates in tissue95-98 that may prove useful when one resonance, water or fat, predominates within a
given voxel.
To overcome the limitations of current MR temperature monitoring, rapid spectroscopic imaging

methods based on echo-planar spectroscopic imaging (EPSI) readouts allow for a spectroscopic
measurement of water and fat proton frequencies. Signal from both resonances will generally be
available but only the water resonant frequency will display marked temperature dependence. As such,
the frequency difference between the temperature-sensitive water resonance chemical shift and the
temperature-insensitive fat chemical shift will serve as a temperature change measurement free from
the dependence on baseline subtraction methods which are susceptible to interscan motion and local
tissue susceptibility changes. In the voxels where one resonance predominates, such that a frequency
difference measurement is problematic, the T1-weighted amplitude of the resonance will be used for
temperature change estimates. The EPSI approach is a natural extension of the single gradient-echo
method used for current MR temperature monitoring.99 It requires echo-planar gradient hardware for
efficient implementation.
Thermal Ablation Techniques

Radiofrequency
Radiofrequency thermal ablation causes local tissue destruction by inducing ionic agitation that results
in heat deposition within the tissues secondary to increased resistivity of the intervening tissue to the
passage of rapidly alternating current (400-500 kHz). The essential objective is to achieve
temperatures between 50°C and 100°C in order to induce coagulation necrosis (at higher
temperatures tissue vaporizes and carbonizes).
Currently most of the RF ablations are performed without appropriate image guidance. CT and US
guidance cannot depict the 3D extent of critical temperature that causes coagulation. The need for
10 of 26 15-04-2010 21:03
monitoring the thermal treatment is obvious but the temperature-sensitive imaging of RF thermal
ablation is technically more challenging than other thermal ablation techniques because of the
interference between the MR imaging system RF coils and the RF ablation generator. Intraprocedural
monitoring with MR (imaging during treatment) is possible with a recently developed switching
mechanism that briefly interrupts RF deposition during the sampling for imaging pulse sequences.
There is evidence that during RF ablation treatments, MR images of lesions can be used for real-time
monitoring and control. Using an animal-based model, one investigation showed that the region inside
the elliptical hyperintense rim in the MR image closely corresponds to the region of necrosis as
established by histology. The MR region only slightly overestimates the region of necrosis, thus
emphasizing the importance of an adequate ablative margin. Therefore, in MR lesion images, the inner
border of the hyperintense region corresponds to the border of irreversible cell damage.100 MR-guided
RF ablation has been applied predominantly in abdominal applications for liver and kidney neoplasia
(Fig. 19-10).
Cryoablation
Cryoablation (cryotherapy) is the term used to describe all methods of destroying tissue by means of
the application of low-temperature freezing and refers to the in situ freezing of tissues. Cryoablation
destroys tissue mostly by cellular dehydration, cell membrane rupture, and vascular stasis. Cell death
occurs at -30°C to -40°C and is dependent on tissue type, rate, duration and depth of freezing, and
thawing cycles.
page 528
page 529
Add to lightbox
11 of 26 15-04-2010 21:03
Add to lightbox
Figure 19-10 MR thermometry. The proton resonance frequency (PRF) method is based on the
temperature dependence (∆ν/νo = -0.0107 ppm/°C) of the precession (Larmor) frequency of the water
protons used for MR imaging. The PRF shift method of MR thermometry provides an easy and
practical means of quantitatively monitoring in vivo temperatures for MR image-guided thermal
ablation therapy. Temperature-sensitive phase-difference fast spoiled GRE MR images (TR 39.9/TE
19.7) were acquired at peak temperature increase during two sonications of a uterine leiomyoma.
Coronal view (A) was acquired perpendicular to the direction of the ultrasound beam and sagittal view
(B) was acquired parallel to the direction of the beam. An intensity coded temperature scale (lower
right corner of A) is used to estimate the thermal dose (red line) for each sonication.
Compared to hyperthermia ablation tools such as laser or radiofrequency ablation, cryotherapy has
unique MR-specific features. The "iceball" formed at the probed tip (referred to as the cryolesion) is
very conspicuously depicted as a signal void and the profile is a "tear-shaped" or "pear-shaped"
configuration. MR imaging during the procedure without any specific temperature-sensitive pulse
sequence can perform precise monitoring of the cryoablation zone. Conventional gradient-echo,
spin-echo, fast spin-echo, ultra-fast subsecond MR sequences and MR fluoroscopy are all suitable for
monitoring the freezing process. The volume of the damaged tissue, as depicted by MR imaging, does
not increase after completion of the ablation procedure.
Generally, the extent of the iceball corresponds well to the final, histologically confirmed necrosis and
allows a high predictability of the resulting treatment volume. Thus it is felt that the margin of the iceball
defines the thermal "kill" zone. However, this may not be true in highly perfused organs such as the
liver if the iceball is induced during inflow occlusion. Thus some have advocated using intraprocedural
enhanced MRI of the cryolesion to predict tissue damage induced during cryoablation101 but for the
majority of cases, the routine unenhanced MR images should suffice. Thus, one significant advantage
of cryoablation over other thermal tumor treatment methods is that visualizing the ablative zone
intraprocedurally may obviate obtaining immediate follow-up imaging (usually within 24-48 hours) to
document the necrosis zone.
Cryoablation has been applied to treat neoplasms (primary or secondary, benign or malignant) in
several different organ systems including liver,102 kidney,103 uterus,104 prostate, soft-tissues and
105
bone. For liver lesions (Fig. 19-11), patients with tumors less than 3 cm seem to be best suited to
this percutaneous approach.106 MR-guided cryotherapy of the brain is possible and allows a precise
12 of 26 15-04-2010 21:03
prediction of the resulting necrosis and is superior to CT for monitoring of cryoablation, as

demonstrated in an animal model.107
Laser
Interstitial laser therapy (ILT) is a minimally invasive technique for local tumor ablation. Light energy is
delivered directly to tissue through an optical fiber; absorption of the light and subsequent heating
create a volume of thermal damage. The spatial distribution of the laser's energy and the pattern of
heat conduction depend on multiple factors, including the optical and thermal properties of the tissue,
i.e., scattering and absorption, and thermal conductivity and perfusion. Laser energy can be delivered
using different materials. Laser light is converted into heat in the target area, resulting in coagulative
necrosis, secondary degeneration and atrophy, causing tumor shrinkage with minimal damage to
surrounding structures. The feasibility of MRI-guided thermal ablation was first demonstrated using
ILT.108 Applications in the liver and the brain have been reported by several investigators (Fig. 19-12).
page 529
page 530
Add to lightbox
Add to lightbox
13 of 26 15-04-2010 21:03
Add to lightbox
Figure 19-11 Thermal ablation: cryotherapy. MR imaging-guided percutaneous cryotherapy of a liver
tumor (metastatic adenocarcinoma). All are axial intraprocedural T2 FSE images at 0.5 T. A, Two 2.4
mm diameter cryoneedles (arrows) targeting the hyperintense tumor (arrowheads). B, The iceball
(signal void) at maximum size at 15 minutes of freeze covering the tumor. MR imaging allows
monitoring and control and insures an adequate ablative margin while avoiding critical structures. C,
After the majority of the ice thaws, the tissue covered by the iceball is seen to be T2 hyperintense
reflecting necrosis (arrows). (Images courtesy of Stuart G Silverman MD, Kemal Tuncali MD and
Paul Morrison MS, Harvard Medical School, Brigham and Women's Hospital.)
Add to lightbox
Add to lightbox
14 of 26 15-04-2010 21:03
Add to lightbox
Add to lightbox
Add to lightbox
15 of 26 15-04-2010 21:03
Add to lightbox
page 530
page 531
Figure 19-12 Thermal ablation: interstial laser therapy (ILT) of a glioma in the posterior left frontal lobe.
This technique involves the percutaneous introduction of an optical fiber through a needle with light
delivered to the tissue absorbed proximal to the tip and the heat generated creates a localized
coagulative necrosis. A, Axial contrast-enhanced T1-weighted FSE image (TR 400/TE 18, ETL 2)
after the insertion of the guide needle. B, Axial contrast-enhanced T1-weighted FSE image during the
ablation. C, Subtraction image during the ablation shows area of change at the laser tip. D, Mapping
with water proton chemical shift imaging. The intensity level in the black box is equalized for enhanced
visibility. Comparison of pre- and postoperative images. Pre-treatment (E) axial contrast-enhanced
T1-weighted (TR 500/TE 16) CSE image shows left hemispheric abnormality with a 3 cm area of
contrast enhancement. Post-treatment (F) axial contrast-enhanced T1-weighted CSE image shows
rim enhancement with no central contrast enhancement at the center of the lesion reflecting the
coagulation necrosis.
In the liver, some evidence from clinical trials is available. One group of investigators performed 125
MR-guided laser thermal ablations on 35 patients with a variety of lesions: hepatocellular carcinoma,
hepatic metastases, and carcinoid liver tumors. These results showed that MR-guided ILT of primary
and secondary liver tumors is safe, feasible, and significantly decreased the amount of enhancing or
viable tumor. MR-guided ILT produces a better survival in patients with hepatocellular carcinoma than
would be expected in untreated patients and has a mean survival in patients with metastases at least
109
equal to the longest median survival in untreated patients. In patients with liver metastases, local
tumor destruction using minimally invasive, percutaneous ILT results in improved clinical outcomes and
survival rates (7148 laser applications) and can be a potential alternative to surgical resection. 110 In the
treatment of recurrent extrahepatic abdominal tumors, ILT is also a practical, minimally invasive,
well-tolerated technique that can produce large areas of necrosis within recurrent tumors, substantially
111
reducing active tumor volume. ILT has also been applied to a small group of patients to treat
symptomatic uterine leiomyomas with good clinical outcomes followed for one year. 112
Laser ablation of cerebral tumors is an alternative to surgical excision and radiosurgery. However, data
are sparse due to its limited application until now and the value of this approach for tumor control and
survival time remains to be defined.113 A better understanding of the energy dose/tissue response in
human brain tumors is important to optimize this treatment modality. The primary indications of this
treatment are small tumors, location in eloquent regions and deep seated, as well as in older patients
114
or patients in poor functional status. Unfortunately, MR-guided ablation does not solve the problem
of defining a precise target in high-grade tumors of the central nervous system.
Microwave
MR-guided microwave therapy of the liver has recently been described.115 Microwaves induce cell
necrosis by thermocoagulation with the formation of microbubbles. The microbubbles formed by the
ablation do not disturb the visibility of the target by MR images throughout the procedure (unlike US
where there may be obscuration distal to the bubbles). A microwave coagulator at 2.45 GHz did not
seriously disturb the MR image acquisition, even during microwave irradiation. A simple notch filter
prevents electromagnetic interference in the MR images. This is more favorable than radiofrequency
ablation at 500 KHz, where a switching circuit for time-sharing is required to reduce the noise in the
MR images. However, a disadvantage of microwave ablation is the relatively small size of coagulation,
which requires multiple punctures. Overall, the combination of MR images and microwave ablation
appears feasible and synergistic.
MRI-Guided FUS
One of the most promising treatment methods of MRI-guided thermal ablations is noninvasive focused
ultrasound surgery (FUS). Unlike the above-mentioned probe-delivered thermal energy deposition
methods, FUS uses extracorporeal acoustic energy for heating tissue only within the focal volume
where most of the energy is absorbed. The probe-delivered thermal ablation methods deposit energy
at a large tissue volume within which there is a wide temperature gradient, with high temperature at the
16 of 26 15-04-2010 21:03
probe and lower temperature at the periphery. Due to this large gradient the biological effects are also
variable within the treated volume and the boundary within which the tissue is irreversibly destroyed is
ill defined.
In FUS the treatment volume is small but the thermal gradient is very narrow, ensuring a more
complete coagulation of the targeted tissue volume. The treatment of larger tissue volumes is not
based on thermal diffusion and convection (which requires longer deposition times) but the focal
volume is moved from target point to target point using a series of successive short energy
depositions. Correct targeting is achieved by identifying the temperature elevations at the targeted
tissue volume below the level of thermal coagulation (under 56°C). If the alignment is correct, one
repeats the sonication at a higher temperature to insure the desired tissue kill effect. MRI-based
targeting and temperature-sensitive imaging provide a closed loop control of this thermal ablation
method (Fig. 19-13).
116-118
The feasibility of MRI-guided FUS was originally demonstrated in a series of animal experiments
and, shortly after, by treating benign fibroadenoma of the breast.119 Using a commercial system
(Exablate 200, Insightec, Haifa, Israel), clinical trials have been conducted for the treatment of breast
120
cancer and uterine leiomyoma (fibroids). Based on the experience in more than 600 patients, it can
be concluded that this completely noninvasive MRI-guided thermal ablation has the greatest potential
for replacing invasive tumor surgery and radiosurgery (Fig. 19-14). Potential use of this technique has
been demonstrated for prostate and liver cancer treatment.121
The use of MRI for correct tumor definition and the measurement of effective thermal dose makes this
integrated therapy delivery system the optimal choice for noninvasive brain surgery. So far
experimental evidence shows that the sound waves can be focused and the resulting energy is
sufficient when the treatment is performed through the intact skull. MRI-guided FUS can also be used
for functional neurosurgery and anatomically precise selective opening of the blood-brain barrier.
These latter methods can revolutionize clinical neuroscience by providing a safe, noninvasive way to
target drug delivery to the CNS.122 Since FUS can change cell membrane permeability as well as
target drug delivery, gene therapy can also be accomplished with this novel interventional MRI method.
Radiotherapy
Prostate
page 531
page 532
17 of 26 15-04-2010 21:03
Add to lightbox
Add to lightbox
Add to lightbox
18 of 26 15-04-2010 21:03
Figure 19-13 Focused ultrasound surgery (FUS). This method is based on primary acoustic energy
deposition and secondary thermal effects. FUS is noninvasive, requires no probe insertion, and the
appropriately targeted and focused high-energy US beam causes no tissue damage in front of or
beyond the target. A, Diagram illustrating the physical principles of FUS. Focused ultrasound uses an
externally generated high-frequency alternating pressure wave, which propagates through the body,
causing tissue at the focus to vibrate. Where the vibration is strong enough, heat is generated.
Focusing of the pressure wave is used to localize the effect in a precisely defined volume within the
body. After MRI-based localization of the tumor target, a relatively low-power energy deposition is
used for targeting. The small, MRI-detectable temperature elevation within the focal spot causes no
permanent tissue damage but can be localized by MRI and moved on the target within the field of view.
When targeting is accomplished, the power level can be increased to achieve a temperature elevation
of more than 60°C, which results in irreversible tissue damage. B, Diagram illustrating treatment
planning. The MRI-integrated system is semi-automated. The physician outlines the tumor volume and
the computer deposits a sufficient number of overlapping focal volumes within the defined contours of
the lesion to treat the entire tumor. The temperature levels within the focal spots are defined using
water proton chemical shift-based temperature imaging in order to determine the effective area for
tumor killing. C, Photograph of FUS phased array transducer that is housed within the patient table
gantry (ExAblate 2000, InSightec, Haifa, Israel). (Images courtesy of Kullervo H Hynynen PhD and
Nathan McDonald PhD, Focused Ultrasound Laboratory, Harvard Medical School, Brigham and
Women's Hospital.)
An MR-guided prostate brachytherapy program has been developed in response to unresolved

problems of US-guided methods.123 Use of a dedicated interventional MR system allows the placement
of the radioactive seeds into the prostate, transperineally in the lithotomy position under MR guidance
(Fig. 19-15). A redesigned set-up of the interventional MR suite allows for unobstructed access to the
prostate through the perineum. The feasibility of placing needles into specific targets in the prostate
under MR imaging guidance has been shown. The predefined target for brachytherapy is the peripheral
zone that is identified and contoured on all the T2-weighted images sections through the gland. The
needles are placed under MR guidance, using fast gradient recalled sequences and according to the
radiation dose plan, within 5 mm of the planned target. More specifically, manual contouring is
performed, the data are transferred to the planning software program, and the dose distribution and
125
plan are generated. The implant procedure follows this plan and the needles, preloaded with I
seeds, are inserted under real-time MR guidance. An image is acquired with the needle visualized and
the planning program overlays the actual and planned locations. If these locations match, the seeds
are dropped; if they are off by greater than 5 mm, the needle is repositioned. The real-time MR
imaging and planning systems provide unique and rapid feedback at both levels.
By improving target definition and soft-tissue resolution and by simultaneously providing both geometric
and dosimetric feedback in real time, the feasibility of this approach has been established. Using these
124
additional features, excellent dose distributions have been reported. This technique can achieve at
least 90% coverage of the tumor volume while maintaining the prostatic urethra and most of the
anterior rectal wall below tolerance levels, with minimal acute morbidity. 125 A non-randomized
prospective clinical trial has shown that both MR-guided brachytherapy and radical prostatectomy
managed patients have similar 5-year estimates of PSA control. Other health outcome analyses are
126
pending. Currently high-dose seed implantation has also been performed using MRI guidance.
page 532
page 533
19 of 26 15-04-2010 21:03
Add to lightbox
Add to lightbox
20 of 26 15-04-2010 21:03
Add to lightbox
Figure 19-14 Focused ultrasound surgery of uterine leiomyoma. Transverse (patient prone)
T2-weighted fast SE MR images (2500/98) showing the treatment plan with the target area focal spot
(white cross) and the ultrasound path (blue shaded area) position for ultrasound surgery of uterine
fibroid. The transducer was angled during the treatment. The software in the system allows display of
the ultrasound beam overlaid on all tissues through which it would pass. Care must be taken to avoid
any possible contact with bowel loops. If necessary, the beam could be repositioned or even tilted to
optimize the path. After an appropriate volume is selected, the sonication plan is developed in the
computer; equally spaced overlapping focal volumes were placed such that the entire target was
covered. Positioning of the focal sonications was selected such that the induced tissue coagulation
would provide complete coverage in the selected volume. The volume to be treated can be calculated
by measuring the volume of the sonication cylinder on the basis of the size of the focal spot (range,
4.5-6.0 mm) and the spot length (range, 18.0-28.0 mm). Locations of the planned sonications can be
modified interactively during the treatment by the operator to obtain complete MR thermometry derived
thermal dose and complete coverage of the target volume. Contrast-enhanced GRE MR image (TR
195/TE 1.8) acquired pre-treatment (B) and 2 days after treatment (C) show the development of a
central area of non-enhancement reflecting coagulation necrosis as a result of FUS treatment.
(Images courtesy of Minna So MD and Clare Tempany MD, Harvard Medical School, Brigham and
Women's Hospital.)
Vascular Applications
Endovascular interventional MR (or MR-guided vascular intervention) (see Chapter 20) is emerging as
a promising method. Minimally invasive interventional procedures, such as balloon angioplasty, stent
placement or coiling of aneurysms, are playing an increasingly important role in the treatment of
vascular disease. Potential (but not yet fully realized) advantages of using the MR-guided technique for
endovascular interventions include cross-sectional anatomic images and functional information, such as
blood flow velocities, perfusion and diffusion, together with three-dimensionality and tomographic
imaging capacities.
page 533
page 534
21 of 26 15-04-2010 21:03
Add to lightbox
Add to lightbox
22 of 26 15-04-2010 21:03
Add to lightbox
Add to lightbox
page 534
page 535
Figure 19-15 Brachytherapy for prostate carcinoma. Axial T2-weighted FSE (TR 6400/TE 100, ETL
12) image (A) for pre-therapy planning is loaded into the Slicer software. Critical regions of interest
are outlined for dose calculation: prostate gland peripheral zone in green, urethra in blue (signal void
represents Foley catheter) and rectum in red (signal void represents endorectal coil). The region of
carcinoma is represented by the low signal region in the right posterolateral aspect of the normally T2
hyperintense peripheral zone. Axial real-time T1-weighted image (TR 500/TE 10) pre-treatment (B)
shows intermediate signal prostate gland and during treatment (C) shows small rounded areas of
susceptibility artifact reflecting needle guides perpendicular to the section. This allows intraprocedural
verification and alteration of placement prior to seed deployment based on dose calculation. Axial
T1-weighted SPGR (TR 24.5/TE 12.1, flip angle 90°) image (D) shows final location of brachytherapy
23 of 26 15-04-2010 21:03
seed locations. (Images courtesy of Steven Haker PhD and Clare Tempany MD, Surgical Planning
Laboratory, Harvard Medical School, Brigham and Women's Hospital.)
Special enhancements are necessary in order to improve the efficacy of current interventional MR
platforms for vascular applications. Interventional MR-guided angiography with a floating table enables
the field of view to be moved along with the instrument tip to the region of interest and thus enhances
127
the usability and flexibility of the interventional MR set-up. Integrated systems for performing
interventional magnetic resonance angiography (MRA) with actively visualized instruments and
real-time image fusion have been implemented and tested on animal models for peripheral
vasculature.128 Very fast imaging is needed to provide high acquisition speed combined with high SNR
and contrast-to-noise ratio (CNR) for the simultaneous visualization of instruments and arterial
morphology. Such systems enable simultaneous image reconstruction and image post-processing of
multiple receiver channels, with subsequent image fusion display in real time. Optional interleaved
image acquisition in two planes provides additional important information for biplanar instrument
guidance. This represents a powerful platform for performing interventional MRA procedures.
129
Placement of inferior vena cava (IVC) filters has also been described in an animal model.
A safety concern with endovascular techniques is thermal damage. Thus far, adverse thermal effects
of using MR imaging guidewires have not been found (e.g., abnormal changes of coagulation factors,
clinical manifestations of blood coagulation disorders, histopathologic damage to target vessels using a
fast spin-echo pulse sequence with an average specific absorption rate (SAR) of 0.6 W/kg and 70
minutes of imaging time in an animal aorta130). These types of data are useful but the precise safety
zone boundaries are yet to be determined.
Cardiac Applications
Similar to diagnostic imaging, motion presents additional challenges to interventional MR techniques.
Certain technological enhancements are needed such as creating volume renderings from magnetic
resonance imaging data that can be displayed in real time with user interactivity providing continuous
3D feedback to the operator in order to assist in guiding an interventional cardiac procedure. 131
132
Real-time MR-guided coronary artery stent placement has been tested in animals.
Devices that can be used for tracking and delivering therapy have a significant advantage and economy
for MR-guided cardiovascular interventions. A two-wire electrophysiology catheter that simultaneously
records the intracardiac electrogram and receives the MR signal for active catheter tracking has been
133
designed and tested. The catheter acts as a long loop receiver, allowing for visualization of the
entire catheter length while simultaneously behaving as a traditional two-wire electrophysiology
catheter, allowing for intracardiac electrogram recording and ablation. Simultaneously tracking the
catheter and recording the intracardiac electrogram in canine models demonstrated that the entire
catheter was visualized and guided from the jugular vein into the cardiac chambers, where the
intracardiac electrogram was recorded (Fig. 19-16). Treatment efficacy is readily assessed (Fig.
19-17).
24 of 26 15-04-2010 21:03
Add to lightbox
Add to lightbox
page 535
page 536
Figure 19-16 Cardiac ablation: visualization of ablation lesions in a swine left ventricle created by
cryoablation (blue) or RF ablation (red) utilizing myocardium delayed enhancement (A) or heavily
T2-weighted (TE > 100 ms) FSE (B) imaging techniques. (Images courtesy of Vivek Reddy MD,
Harvard Medical School, Massachusetts General Hospital, and Ehud J Schmidt PhD, General
Electric Healthcare Applied Science Laboratory.)
In the pediatric realm, the goal of achieving real-time MR-guided interventions in congenital heart
25 of 26 15-04-2010 21:03
diseases has proven feasible for some applications of diagnosis and therapy. The development and
testing of MR-compatible devices for therapeutic applications such as atrial septal defect, patent
foramen ovale closure and pulmonary artery dilation is being pursued to facilitate the expansion of this
technology.134,135 High-resolution imaging allows accurate determination of defect size before the
136
intervention, and immediate treatment effects (e.g., changes in right cardiac volumes).
One of the most exciting areas for interventional MRI is facilitating local forms of gene therapy.
Intracardiac applications include the potential to guide intramyocardial stem cell injection to specific
targets-the border between infarcted and normal tissue. Precise targeted delivery of potentially
regenerative cellular treatments to recent myocardial infarction borders is feasible with an MR catheter
delivery system. Interventional MR guidance permits visualization of catheter navigation, myocardial
function, infarct borders, and labeled cells after injection in a swine model. 137
© 2010 Elsevier
26 of 26 15-04-2010 21:03
FUTURE DIRECTIONS
Add to lightbox
Add to lightbox
1 of 3 15-04-2010 21:04
Figure 19-17 Cardiac ablation: 3D contrast-enhanced MRA image of a laser balloon in a dog
pulmonary artery, showing the position of the inflated balloon (A) and a real-time T1 difference thermal
map (B) showing the heating effect (white arrow) with the laser turned on. (Images courtesy of Vivek
Reddy MD, Harvard Medical School, Massachusetts General Hospital, and Ehud J Schmidt PhD,
General Electric Healthcare Applied Science Laboratory.)
The successful development of interventional MRI has required innovative approaches, novel
applications, efficient use of computer technologies, advanced therapy devices, and a more
sophisticated and diverse technological infrastructure. This can only be accomplished and extended by
a multifocused, multidisciplinary effort aimed at the task of translational research for developing and
implementing MR-guided interventions. Several areas of imminent development are in interventional
(intrapatient) coils, thermal ablation mapping, and enhancing the navigation task. Some are directly
related to MR-guided interventions while others have a more general applicability for image-guided
therapy, such as better image processing, segmentation, registration, 3D modeling, and enhanced
surgical planning. The integration of intraoperative MRI guidance and computer-assisted surgery will
greatly accelerate the clinical utility of image-guided therapy in general and interventional MRI in
particular.
page 536
page 537
Interventional coils and hand-held probes are an extension of catheter-based work into the realm of
percutaneous intervention and image-guided surgery. To do better than surface coils, a smaller,
invasive probe can be placed closer to the region of interest. This model has worked well for controlled
probe shape and loading conditions (e.g., endorectal and intravascular). However, more general
interventional coils require remote operation and a variable shape, which both introduce practical
problems with tuning and matching. Prior work provides analysis of such issues and circuitry for
automatic tuning of a flexible interventional probe. A coil equipped with an internal spin source
138
increases the signal-to-noise ratio in comparison to a coil system without internal source. The real
benefit of using a small, flexible interventional coil for clinical work is that it retains high local SNR for
arbitrary depths of target anatomy if well matched to the preamplifier. With automatic tuning, SNR can
be maintained while allowing the coil to conform to anatomy. A proliferation of intracavitary coils and
probes is anticipated for a variety of applications including gastrointestinal and sinus endoscopic,
abdominopelvic laparoscopic, joint (intra-articular), spine (intracanalicular), and some viscera
procedures.
Complete 3D intraprocedural mapping of temperatures throughout the tissue has been lacking and
such visualization of the thermal effects could increase the effectiveness of monitoring and controlling
of the thermal ablation process. Currently, MR-monitored thermal therapies are controlled by direct
observation of selected 2D MR image planes intersecting the therapy volume. Yet during therapy, the
highly irregular 3D boundary of the thermally affected region evolves anisotropically, defying control by
direct observation. No visual perspective, single projection or 2D cross-section can properly, rapidly,
and completely display to the operator the 3D thermally damaged tissue in comparison to a 3D
predetermined boundary. Nor can thermal exposure currently be recorded, summed, and visualized in
a 3D rendering. When attempting control by direct observation using 2D MR imaging, therapy may be
terminated prematurely before the entire tumor volume has been treated, leaving residual tumor, or
terminated after extra damage to normal tissue or vessels has occurred with additional potential risks
of hemorrhage and edema within and around the treatment site. Therefore, incomplete visualization
leads to an incomplete understanding of the phenomena and risks of undertreating a tumor (ineffective
therapy), coagulating adjacent non-targeted normal tissue or causing unwanted collateral damage
(unsafe therapy). Complete 3D control will determine the ultimate safety and efficacy of thermal
ablation as a tumor therapy. The ability to produce a rapid isotropic volumetric MR imaging data set
will greatly facilitate achieving this goal and advance MR-based thermal ablations.
New general-purpose and procedure-specific visualization systems based on advanced display

devices, integrated intuitive human/computer interfaces, and appropriate interaction paradigms can
provide physicians with additional timely and useful information during image-guided therapy
procedures, helping to maintain context between different types of data while not unduly distracting
2 of 3 15-04-2010 21:04
from the procedure. Better integration of MR imaging, navigation, and tracking systems with actively
visualized instruments and real-time image fusion will be needed for performing complex surgical,
needle, and catheter-based vascular procedures. Also, being able to predict what anatomic
consequences a specific procedure will induce can be a useful function during surgery or complex
image-guided therapy procedures. What surgeons and interventionalists will need is the ability to
perform advanced biomechanical simulation accessible from procedure suites. This quantitative
information can be fused with preoperative and, more importantly, intraoperative imaging once some
anatomic deformations have occurred for improving intraprocedural decision making and outcomes.
Augmented reality (AR) systems allow image-guided interventions to take place outside the MR
scanner. The conventional closed magnet designs challenge the interventionalists to focus their
attention alternately on the display screen and then the patient. AR refers to computer displays that
add virtual information to a user's sensory perceptions.139 Most AR research focuses on "see-through"
devices, usually worn on the head, that overlay graphics and text on the user's view of the
surroundings. AR systems track the position and orientation of the user's head so that the overlaid
material can be aligned with the user's view of the world. An AR surgical system allows an intervention
to take place outside the imager and incorporates the image data, along with additional real-time
information, directly into the surgical environment via 3D overlays mapped onto the patient and surgical
equipment. However, a significant technical challenge presented by AR surgical systems is
compensating for organ or target motion due to respiration during the intervention. We anticipate the
AR systems will facilitate the use of the current base of high-field strength MR imagers using the
"in/out" paradigm for localization and biopsy-type procedures (similar for CT-guided procedures).
Preliminary results suggest that AR systems can offer improved accuracy over traditional biopsy
guidance methods.140
© 2010 Elsevier
3 of 3 15-04-2010 21:04
CONCLUSION
The original concept of MR-guided procedures has evolved into complex, integrated image-guided and
computer-assisted applications in surgery and interventional radiology. Successful programs combine
interventional and intraoperative magnetic resonance imaging with high-performance computing and
novel therapeutic devices. Interventional MRI has entered into a new stage in which computer-based
techniques play an increasing role in planning, monitoring, and controlling the procedures. The use of
and need for interactive imaging, navigational image guidance techniques, and image-processing
methods have been demonstrated in various applications. MR-guided intervention could be
implemented similar to a CT- or US-guided procedure room but if the full spectrum of services is to be
provided then an operating suite paradigm should be employed. This "interventional MR" suite is the
result of a combination between an operating room, an interventional radiology suite, and a
conventional MR imaging unit. Because MRI guidance may be provided during endoscopic,
laparoscopic or open surgical procedures, this area must be equipped as an operating room. At
several facilities worldwide the landscape of neurosurgery has changed at a fundamental level because
of intraoperative MR imaging.
page 537
page 538
There is certainly evidence of a growing interest in using interventional MRI in the medical community.
There is already a large and rapidly growing body of medical literature devoted to this topic. Several
medical conferences have emerged that either focus on interventional MRI itself or on the broader
umbrella of image-guided therapy (IGT). All major MRI vendors offer an interventional MRI of some
type and configuration. Some vendors have further embraced this concept by designing integrated
procedure rooms/operating rooms. All these activities suggest that there is a continued and growing
interest in IGT using advanced modalities such as MRI.
IGT using MRI may be considered a "disruptive" technology that will likely eventually change the way
medicine is practiced. This has occurred with other modalities that are routinely used in the operating
room, such as fluoroscopy and ultrasound. The deployment of interventional MRI does require much
more infrastructure but many new technologies that eventually become entrenched in medical practice
require early adopters and developers in order to advance the field to a sufficient level to make it
ready for "prime time". Not too long ago, similar statements were made that every hospital would not
have a CT or MRI scanner when these modalities were initially introduced yet these "big ticket" items
have become nearly ubiquitous.
Overall there is a trend for increased use of MR imaging and guidance for planning, targeting,
monitoring, and control of various treatments. Therefore, advancements made in MR-based IGT are
necessary to make this paradigm widely available. Some components are translatable outside the
operating room environment (e.g., interventional MR, diagnostic imaging) and to other modalities as
well (computer-assisted surgery using fluoroscopy). As with any technology, further refinements will
make this system less expensive and more attainable. Based on the rapid advancement of technology,
very high-field strength interventional magnets may become the standard. In clinical practice a
multidisciplinary program provides for a wide range of interventional and surgical procedures. The cost
and technical support required for an intraoperative MRI system presently limit its use to only a few
sites worldwide. As new technology is developed, clinicians must continue to explore and refine and
make it cost-effective and widely applicable. Health services type research is needed to establish
whether MR image-guided therapy definitively improves clinical outcomes and reduces complication
rates. Radiologists are integral to the development and deployment of interventional MR-guided
diagnostic and therapeutic techniques, not only as the operators but also as facilitators for other
specialists who can benefit from adding image-guided procedures to the armamentarium of therapy
options.
The ultimate goal of IGT (with MR or any modality) is a seamless interface between the eye and hand
in the purest sense (i.e., the mind's eye and hand). Ideally, this seamless interface represents
effortless flow between the procedural goal compared with the present situation and the manipulation
of the tools available to accomplish the task, whether it is the scalpel, drill, laser, aspirator, ultrasound,
1 of 7 15-04-2010 21:04
robot or high-dose irradiator. The physical limits of the human being demarcate the confining boundary
of the system but the person/machine interface should augment tool manipulation and provide other
technical adjuncts. The combination of intraoperative visualization and precise surgical navigation will
be the paradigm needed to foster widespread application.
REFERENCES
1. Knauth M, Wirtz CR, Aras N, Sartor K: Low-field interventional MRI in neurosurgery: finding the right dose of contrast
medium. Neuroradiology 43:254-258, 2001. Medline Similar articles
2. Jolesz FA, Shtern F: The operating room of the future. Report of the National Cancer Institute Workshop. Invest Radiol
3. Salomonowitz E: MR imaging-guided biopsy and therapeutic intervention in a closed-configuration magnet: single-center
series of 361 punctures. Am J Roentgenol 177:159-163, 2001.
4. Pulyer Y, Hrovat MI: An open magnet utilizing ferro-refraction current magnification. J Magn Reson 154:298-302, 2002.
5. Lewin JS, Duerk JL, Jain VR, et al: Needle localization in MR-guided biopsy and aspiration: effects of field strength,
sequence design, and magnetic field orientation. Am J Roentgenol 166:1337-1345, 1996.
6. Nitz WR, Oppelt A, Renz W, et al: On the heating of linear conductive structures as guide wires and catheters in
interventional MRI. J Magn Reson Imaging 13:105-114, 2001. Medline Similar articles
7. Ladd ME, Quick HH: Reduction of resonant RF heating in intravascular catheters using coaxial chokes. Magn Reson Med
8. Silverman SG, Collick BD, Figueira MR, et al: Interactive MR-guided biopsy in an open-configuration MR imaging system.
9. Farzaneh F, Riederer SJ, Lee JN, et al: MR fluoroscopy: initial clinical studies. Radiology 171:545-549, 1989. Medline
Similar articles
10. Riederer SJ, Tasciyan T, Farzaneh F, et al: MR fluoroscopy: technical feasibility. Magn Reson Med 8:1-15, 1988.
11. Wright RC, Riederer SJ, Farzaneh F, et al: Real-time MR fluoroscopic data acquisition and image reconstruction. Magn
12. Panych LP, Jakab PD, Jolesz FA: An implementation of wavelet-encoded MRI. J Magn Reson Imaging 3:649-655, 1993.
13. Zientara GP, Panych LP, Jolesz FA: Dynamically adaptive MRI with encoding by singular value decomposition. Magn
14. Panych LP, Zhao L, Jolesz FA, Mulkern RV: Dynamic imaging with multiple resolutions along phase-encode and slice-
select dimensions. Magn Reson Med 45:940-947, 2001. Medline Similar articles
15. Aksit P, Derbyshire JA, Serfaty JM, Atalar E: Multiple field of view MR fluoroscopy. Magn Reson Med 47:53-60, 2002.
16. Zhao L, Madore B, Panych LP: Reduced field-of-view dynamic imaging with 2D spatially-selective excitation and UNFOLD.
Meeting of the International Society for Magnetic Resonance Medicine, Hawaii, 2002.
17. Pauly J, Nishimura D, Macovski A: A k-space analysis of small-tip-angle excitation. J Magn Reson 81:43-56, 1989.
18. Kyriakos WE, Zhao L, Jolesz FA: Parallel imaging of the heart using SPACE RIP. Meeting of the International Society for
Magnetic Resonance Medicine, Hawaii, 2002.
19. Zhao L, Madore B, Panych LP: Reduced field-of-view imaging with two-dimensional RF excitation and UNFOLD. Mag Res
Med 53:1118-1125, 2005.
20. Zhao L, Panych LP: Reduced field-of-view imaging using the combination of a two-dimensional spatial excitation pulse and
a dynamic rFOV reconstructoin. Meeting of the Radiological Society of North America, Chicago, 2001.
21. Sodickson DK. Manning WJ: Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency
22. Klaas P, Pruessmann M, Scheidegger M, Boesiger P: Sense:sensitivity encoding for fast MRI. Magn Reson Med
23. Kyriakos W, Panych L, Kacher D: Sensitivity profiles from an array of coils for encoding and reconstruction in parallel
(SPACE RIP). Magn Reson Med 44:301-308, 2000. Medline Similar articles
24. Kyriakos WE, Panych LP, Kacher DF, et al: Parallel image reconstruction from multiple receiver arrays for fast MRI.
Meeting of the International Society for Magnetic Resonance Medicine, Colorado, 2000.
25. Dumoulin Wright RC, Riederer SJ, Farzaneh F, et al: Real-time MR fluoroscopic data acquisition and image
reconstruction. Magn Reson Med 12:407-415, 1989. Medline Similar articles
26. Gering D, Nabavi A, Kikinis R: An integrated visualization system for surgical planning and guidance using image fusion
and an open MR. J Magn Reson Imaging 13: 967-975, 2001.
2 of 7 15-04-2010 21:04
27. Bakker CJ, Hoogeveen RM, Weber J, et al: Visualization of dedicated catheters using fast scanning techniques with
potential for MR-guided vascular interventions. Magn Reson Med 36:816-820, 1996. Medline Similar articles
28. Dumoulin CL, Souza SP, Darrow RD. Real-time position monitoring of invasive devices using magnetic resonance. Magn
29. Rasche V, Holz D, Kohler J, et al: Catheter tracking using continuous radial MRI. Magn Reson Med 37:963-968, 1997.
30. Elgort DR, Wong EY, Hillenbrand CM, et al: Real-time catheter tracking and adaptive imaging. J Magn Reson Imaging
31. Leung DA, Debatin JF, Wildermuth S, et al: Intravascular MR tracking catheter: preliminary experimental evaluation. Am J
Roentgenol 164:1265-1270, 1995.
32. Araki T, Aoki S, Ishigame K, et al: MR-guided intravascular catheter manipulation: feasibility of both active and passive
tracking in experimental study and initial clinical applications. Radiat Med 20:1-8, 2002. Medline Similar articles
page 538
page 539
33. Leung DA, Debatin JF, Wildermuth S, et al: Real-time biplanar needle tracking for interventional MR imaging procedures.
34. Steiner P, Erhart P, Heske N, et al: Active biplanar MR tracking for biopsies in humans. Am J Roentgenol 169:735-738,
1997.
35. Yeung CJ, Atalar E: RF transmit power limit for the barewire loopless catheter antenna. J Magn Reson Imaging 12:86-91,
36. Zhang Q, Wendt M, Aschoff AJ, et al: A multielement RF coil for MRI guidance of interventional devices. J Magn Reson
37. Roberts TP, Hassenzahl WV, Hetts SW, Arenson RL: Remote control of catheter tip deflection: an opportunity for
interventional MRI. Magn Reson Med 48:1091-1095, 2002. Medline Similar articles
38. Adam G, Neuerburg J, Bucker A, et al: Initial clinical experience with a 1.5-T magnetic resonance system combined with
C-arm fluoroscopy. Invest Radiol 32: 191-197, 1997.
39. Neuerburg J-M, Adam G, Buecker A, et al: MRI-guided biopsy of bone in a hybrid system. J Magn Reson Imaging 8: 85-90,
1998.
40. Hall WA, Martin AJ, Haiying L, et al: Brain biopsy using high-field strength interventional magnetic resonance imaging.
Neurosurgery 44: 807-813, 1999.
41. Vogl TJ, Balzer JO, Mack MG, et al: Hybrid MR interventional imaging system: combined MR and angiography suites with
single interactive table. Feasibility study in vascular liver tumor procedures. Eur Radiol 12:1394-1400, 2002. Medline
Similar articles
42. Fahrig R, Butts K, Rowlands JA, et al: A truly hybrid interventional MR/X-ray system: feasibility demonstration. J Magn
43. Fahrig R, Butts K, Wen Z, et al: Hybrid interventional MR/X-ray system: investigation of in vivo applications. Acad Radiol
44. Fahrig R, Heit G, Wen Z, et al: First use of a truly-hybrid X-ray/MR imaging system for guidance of brain biopsy. Acta
Neurochir 145:995-997, 2003.
45. Sakarya ME, Unal O, Ozbay B, et al: MR fluoroscopy-guided transthoracic fine-needle aspiration biopsy: feasibility.
46. Ferrant M, Nabavi A, Macq B, et al: Registration of 3-D intraoperative MR images of the brain using a finite-element
biomechanical model. IEEE Trans Med Imaging 20:1384-1397, 2001.
47. Hirose M, Bharatha A, Hata N, et al: Quantitative MR imaging assessment of prostate gland deformation before and during
MR imaging-guided brachytherapy. Acad Radiol 9:906-912, 2002. Medline Similar articles
48. Ruiz-Alzola J, Westin CF, Warfield SK, et al: Nonrigid registration of 3D tensor medical data. Med Image Anal 6:143-161,
2002.
49. Ferrant M, Nabavi A, Macq B, et al: Serial registration of intraoperative MR images of the brain. Med Image Anal
50. Jolesz FA: Future perspectives in intraoperative imaging. Acta Neurochir 85 (Suppl):7-13, 2003.
51. Black PM, Moriarty T, Alexander E 3rd, et al: Development and implementation of intraoperative magnetic resonance
imaging and its neurosurgical applications. Neurosurgery 41:831-842; discussion 842-845, 1997.
52. Bradley WG: Achieving gross total resection of brain tumors: intraoperative MR imaging can make a big difference. Am J
53. Schiffbauer H, Ferrari P, Rowley H, et al: Functional activity within brain tumors: a magnetic source imaging study.
Neurosurgery 49:1313-1321, 2001. Medline Similar articles
54. Hall WA, Liu H, Maxwell RE, Truwit CL: Influence of 1.5-T intraoperative MR imaging on surgical decision making. Acta
Neurochir 85 (Suppl):29-37, 2003.
55. Zimmermann M, Seifert V, Trantakis C, Raabe A: Open MRI-guided microsurgery of intracranial tumours in or near
3 of 7 15-04-2010 21:04
eloquent brain areas. Acta Neurochir (Wien) 143:327-337, 2001.

56. Nabavi A, Black PM, Gering DT, et al: Serial intraoperative magnetic resonance imaging of brain shift. Neurosurgery
48:787-797; discussion 797-798, 2001. Medline Similar articles
57. Vitaz TW, Hushek SG, Shields CB, Moriarty TM: Interventional MRI-guided frameless stereotaxy in pediatric patients.
Stereotact Funct Neurosurg 79:182-190, 2002. Medline Similar articles
58. Balmer B, Bernays RL, Kollias SS, Yonekawa Y: Interventional MR-guided neuroendoscopy: a new therapeutic option for
children. J Pediatr Surg 37:668-672, 2002. Medline Similar articles
59. Lam CH, Hall WA, Truwit CL, Liu H: Intra-operative MRI-guided approaches to the pediatric posterior fossa tumors. Pediatr
Neurosurg 34:295-300, 2001. Medline Similar articles
60. Wirtz CR, Knauth M, Staubert A, et al: Clinical evaluation and follow-up results for intraoperative magnetic resonance
imaging in neurosurgery. Neurosurgery 46:1112-1120; discussion 1120-1122, 2000. Medline Similar articles
61. Hall WA, Kowalik K, Liu H, et al: Costs and benefits of intraoperative MR-guided brain tumor resection. Acta Neurochir 85
(Suppl):137-142, 2003.
62. Schwartz RB, Hsu L, Wong TZ, et al: Intraoperative MR imaging guidance for intracranial neurosurgery: experience with the
first 200 cases. Radiology 211:477-488, 1999.
63. Hirose M, Kacher DF, Smith DN, et al: Feasibility of MR imaging-guided breast lumpectomy for malignant tumors in a 0.5-T
open-configuration MR imaging system. Acad Radiol 9:933-941, 2002.
64. Gould SW, Martin S, Agarwal T, et al: Image-guided surgery for anal fistula in a 0.5T interventional MRI unit. J Magn Reson
Imaging 16:267-276, 2002.
65. Hata N, Jinzaki M, Kacher D, et al: MR imaging-guided prostate biopsy with surgical navigation software: device validation
and feasibility. Radiology 220:263-268, 2001. Medline Similar articles
66. Cormack RA, D'Amico AV, Hata N, et al: Feasibility of transperineal prostate biopsy under interventional magnetic
resonance guidance. Urology 56:663-664, 2000. Medline Similar articles
67. Genant JW, Vandevenne JE, Bergman AG, et al: Interventional musculoskeletal procedures performed by using MR
imaging guidance with a vertically open MR unit: assessment of techniques and applicability. Radiology 223:127-136, 2002.
68. Lewin JS, Petersilge CA, Hatem SF, et al: Interactive MR imaging-guided biopsy and aspiration with a modified clinical
C-arm system. Am J Roentgenol 170:1593-1601, 1998.
69. Parkkola RK, Mattila KT, Heikkila JT, et al: Dynamic contrast-enhanced MR imaging and MR-guided bone biopsy on a
0.23 T open imager. Skeletal Radiol 30:620-624, 2001.
70. Adam G, Neuerburg J, Bucker A, et al: Interventional magnetic resonance. Initial clinical experience with a 1.5-T magnetic
resonance system combined with c-arm fluoroscopy. Invest Radiol 32:191-197, 1997.
71. Neuerburg J, Adam G, Bucker A, et al: MRI-guided biopsy of bone in a hybrid system. J Magn Reson Imaging 8: 85-90,
1998.
72. Koenig CW, Duda SH, Truebenbach J, et al: MR-guided biopsy of musculoskeletal lesions in a low-field system. J Magn
73. Ojala R, Sequeiros RB, Klemola R, et al: MR-guided bone biopsy: preliminary report of a new guiding method. J Magn
74. Sequeiros RB, Klemola R, Ojala R, et al: MRI-guided trephine biopsy and fine-needle aspiration in the diagnosis of bone
lesions in low-field (0.23 T) MRI system using optical instrument tracking. Eur Radiol 12:830-835, 2002. Medline
Similar articles
75. Sequeiros RB, Klemola R, Ojala R, et al: Percutaneous MR-guided discography in a low-field system using optical
instrument tracking: a feasibility study. J Magn Reson Imaging 17:214-219, 2003. Medline Similar articles
76. Sequeiros RB, Ojala RO, Klemola R, et al: MRI-guided periradicular nerve root infiltration therapy in low-field (0.23-T) MRI
system using optical instrument tracking. Eur Radiol 12:1331-1337, 2002. Medline Similar articles
77. Sze DY, Mackey SC: MR guidance of sympathetic nerve blockade: measurement of vasomotor response initial experience
in seven patients. Radiology 223:574-580, 2002. Medline Similar articles
78. Ojala R, Klemola R, Karppinen J, et al: Sacro-iliac joint arthrography in low back pain: feasibility of MRI guidance. Eur J
79. Hol PK, Kvarstein G, Viken O, et al: MRI-guided celiac plexus block. J Magn Reson Imaging 12:562-564, 2000.
80. Pfleiderer SO, Reichenbach JR, Azhari T, et al: A manipulator system for 14-gauge large core breast biopsies inside a
high-field whole-body MR scanner. J Magn Reson Imaging 17:493-498, 2003.
81. Viehweg P, Heinig A, Amaya B, et al: MR-guided interventional breast procedures considering vacuum biopsy in particular.
Eur J Radiol 42:32-39, 2002. Medline Similar articles
82. Perlet C, Schneider P, Amaya B, et al: MR-guided vacuum biopsy of 206 contrast-enhancing breast lesions. Rofo Fortschr
Geb Rontgenstr Neuen Bildgeb Verfahr 174:88-95, 2002.
83. Buecker A, Neuerburg JM, Adam GB, et al: MR-guided percutaneous drainage of abdominal fluid collections in
4 of 7 15-04-2010 21:04
combination with X-ray fluoroscopy: initial clinical experience. Eur Radiol 11:670-674, 2001. Medline Similar articles
84. Gohde SC, Pfammatter T, Steiner P, et al: MR-guided cholecystostomy: assessment of biplanar, real-time needle tracking
in three pigs. Cardiovasc Intervent Radiol 20:295-299, 1997. Medline Similar articles
85. Hagspiel KD, Kandarpa K, Silverman SG: Interactive MR-guided percutaneous nephrostomy. J Magn Reson Imaging
86. Jolesz FA, Silverman SG: Interventional magnetic resonance therapy. Semin Interv Radiol 12:20-27, 1995.
87. Anzai Y, Lufkin RB, Castro DJ, et al: MR imaging guided interstitial Nd-YAG laser phototherapy: dosimetry study of acute
tissue damage in an in vivo model. J Magn Reson Imaging 1:553-559, 1991. Medline Similar articles
88. LeBihan D, Delannoy J, Levin RL: Temperature mapping with MR imaging of molecular diffusion: application to
hyperthermia. Radiology 171:853-857, 1989. Medline Similar articles
89. Delannoy J, Chen CN, Turner R, et al: Noninvasive temperature imaging using diffusion MRI. Magn Reson Med
90. Matsumoto R, Mulkern RV, Hushek SG, Jolesz FA: Tissue temperature monitoring for thermal interventional therapy:
comparison of T1-weighted MR sequences. J Magn Reson Imaging 4:65-70, 1994. Medline Similar articles
91. Mulkern RV, Panych LP, McDannold NJ, et al: Tissue temperature monitoring with multiple gradient-echo imaging
sequences. J Magn Reson Imaging 8:493-502, 1998. Medline Similar articles
92. Wlodarczyk W, Hentschel M, Wust P, et al: Comparison of four magnetic resonance methods for mapping small
temperature changes. Phys Med Biol 44:607-624, 1999. Medline Similar articles
93. Wlodarczyk W, Boroschewski R, Hentschel M, et al: Three-dimensional monitoring of small temperature changes for
therapeutic hyperthermia using MR. J Magn Reson Imaging 8:165-174, 1998. Medline Similar articles
94. Graham S, Stanisz G, Kecojevic A, et al: Analysis of changes in MR properties of tissues after heat treatment. Magn Reson
95. Hynynen K, McDannold N, Mulkern R, Jolesz FA: Temperature monitoring in fat with MRI. Magn Reson Med 43:901-904,
96. Bohris C, Jenne J, Rastert R, et al: MR monitoring of focused ultrasound surgery in a breast tissue model in vivo. Magn
Reson Imag 19:167-175, 2001.
page 539
page 540
97. Bohris C, Rastert R, Jenne J, et al: MR monitoring of focused ultrasound surgery (FUS) in a breast tissue model-in vivo
study. Proceedings of the International Society for Magnetic Resonance Medicine, 2000.
98. Bohris C, Schreiber W, Jenne J, et al: Quantitative MR temperature monitoring of high-intensity focused ultrasound therapy.
Magn Reson Imag 17:603-610, 1999.
99. Mulkern RV, Panych LP, McDannold NJ, et al: Tissue temperature monitoring with multiple gradient-echo imaging
sequences. J Magn Reson Imaging 8:493-502, 1998. Medline Similar articles
100. Breen MS, Lancaster TL, Lazebnik RS, et al: Three-dimensional method for comparing in vivo interventional MR images
of thermally ablated tissue with tissue response. J Magn Reson Imaging 18:90-102, 2003. Medline Similar articles
101. Mala T, Frich L, Aurdal L, et al: Intraoperative contrast-enhanced MR-imaging as predictor of tissue damage during
cryoablation of porcine liver. Magn Reson Imaging 21:733-740, 2003. Medline Similar articles
102. Silverman SG, Tuncali K, Adams DF, et al: MR imaging-guided percutaneous cryotherapy of liver tumors: initial
experience. Radiology 217:657-664, 2000. Medline Similar articles
103. Sewell PE, Howard JC, Shingleton WB, Harrison RB: Interventional magnetic resonance image-guided percutaneous
cryoablation of renal tumors. South Med J 96:708-710, 2003. Medline Similar articles
104. Cowan BD, Sewell PE, Howard JC, et al: Interventional magnetic resonance imaging cryotherapy of uterine fibroid
tumors: preliminary observation. Am J Obstet Gynecol 186:1183-1187, 2002. Medline Similar articles
105. Shankar S, Kacher D, Morrison PR, et al: MRI-guided percutaneous cryotherapy of soft tissue and bone metastases.
Radiology 221(Suppl): 1, 2001.
106. Mala T, Edwin B, Samset E, et al: Magnetic-resonance-guided percutaneous cryoablation of hepatic tumours. Eur J Surg
107. Tacke J, Speetzen R, Adam G, et al: Experimental MR imaging-guided interstitial cryotherapy of the brain. Am J
108. Bleier AR, Jolesz FA, Cohen MS, et al: Real-time magnetic resonance imaging of laser heat deposition in tissue. Magn
109. Dick EA, Joarder R, de Jode M, et al: MR-guided laser thermal ablation of primary and secondary liver tumours. Clin
110. Mack MG, Straub R, Eichler K, et al: Percutaneous MR imaging-guided laser-induced thermotherapy of hepatic
metastases. Abdom Imaging 26:369-374, 2001. Medline Similar articles
111. Mack MG, Straub R, Eichler K, et al: MR-guided laser-induced thermotherapy in recurrent extrahepatic abdominal tumors.
Eur Radiol 11:2041-2046, 2001. Medline Similar articles
5 of 7 15-04-2010 21:04
112. Hindley JT, Law PA, Hickey M, et al: Clinical outcomes following percutaneous magnetic resonance image guided laser
ablation of symptomatic uterine fibroids. Hum Reprod 17:2737-2741, 2002. Medline Similar articles
113. Bar NK, Schulz T, Puccini S, et al: [MR-guided laser-induced thermo-ablation of liver tumors: clinical experiences and
therapy control concepts.] Z Med Phys 13:209-213, 2003 (in German).
114. Leonardi MA, Lumenta CB: Stereotactic guided laser-induced interstitial thermotherapy (SLITT) in gliomas with
intraoperative morphologic monitoring in an open MR: clinical expierence. Minim Invasive Neurosurg 45:201-207, 2002.
115. Morikawa S, Inubushi T, Kurumi Y, et al: Advanced computer assistance for magnetic resonance-guided microwave
thermocoagulation of liver tumors. Acad Radiol 10:1442-1449, 2003. Medline Similar articles
116. Cline HE, Hynynen K, Watkins RD, et al: Focused US system for MR imaging-guided tumor ablation. Radiology
117. Daum DR, Smith NB, King R, Hynynen K: In vivo demonstration of noninvasive thermal surgery of the liver and kidney
using an ultrasonic phased array. Ultrasound Med Biol 25:1087-1098, 1999. Medline Similar articles
118. McDannold NJ, Jolesz FA, Hynynen KH: Determination of the optimal delay between sonications during focused
ultrasound surgery in rabbits by using MR imaging to monitor thermal buildup in vivo. Radiology 211:419-426, 1999.
119. Hynynen K, Pomeroy O, Smith DN, et al: MR imaging-guided focused ultrasound surgery of fibroadenomas in the breast:
a feasibility study. Radiology 219:176-185, 2001. Medline Similar articles
120. Tempany CM, Stewart EA, McDannold N, et al: MR imaging-guided focused ultrasound surgery of uterine leiomyomas: a
feasibility study. Radiology 226:897-905, 2003. Medline Similar articles
121. Kennedy J, Wu F, ter Haar G, et al: High-intensity focused ultrasound for the treatment of liver tumours. Ultrasonics
122. McDannold N, Moss M, Killiany R, et al: MRI-guided focused ultrasound surgery in the brain: tests in a primate model.
123. D'Amico A, Cormack R, Tempany C: MRI-guided diagnosis and treatment of prostate cancer. N Engl J Med 344:
776-777, 2001.
124. D'Amico A, Cormack R, Tempany C, et al: Real-time magnetic resonance image-guided interstitial brachytherapy in the
treatment of select patients with clinically localized prostate cancer. Int J Radiat Oncol Biol Phys 42: 507-515, 1998.
125. D'Amico A, Cormack R, Kumar S, Tempany CM: Real-time magnetic resonance imaging-guided brachytherapy in the
treatment of selected patients with clinically localized prostate cancer. J Endourol 14:367-370, 2000. Medline
Similar articles
126. D'Amico AV, Tempany CM, Schultz D, et al: Comparing PSA outcome after radical prostatectomy or magnetic resonance
imaging-guided partial prostatic irradiation in select patients with clinically localized adenocarcinoma of the prostate. Urology
127. Quick HH, Kuehl H, Kaiser G, et al: Interventional MR angiography with a floating table. Radiology 229:598-602, 2003.
128. Quick HH, Kuehl H, Kaiser G, et al: Interventional MRA using actively visualized catheters, TrueFISP, and real-time image
fusion. Magn Reson Med 49:129-137, 2003. Medline Similar articles
129. Bucker A, Neuerburg JM, Adam GB, et al: Real-time MR guidance for inferior vena cava filter placement in an animal
model. J Vasc Interv Radiol 12:753-756, 2001. Medline Similar articles
130. Yang X, Yeung CJ, Ji H, et al: Thermal effect of intravascular MR imaging using an MR imaging-guidewire: an in vivo
laboratory and histopathological evaluation. Med Sci Monit 8:MT113-117, 2002.
131. Guttman MA, Lederman RJ, Sorger JM, McVeigh ER: Real-time volume rendered MRI for interventional guidance. J
Cardiovasc Magn Reson 4:431-442, 2002. Medline Similar articles
132. Spuentrup E, Ruebben A, Schaeffter T, et al: Magnetic resonance-guided coronary artery stent placement in a swine
model. Circulation 105:874-879, 2002. Medline Similar articles
133. Susil RC, Yeung CJ, Halperin HR, et al: Multifunctional interventional devices for MRI: a combined electrophysiology/MRI
catheter. Magn Reson Med 47:594-600, 2002. Medline Similar articles
134. Rickers C, Seethamraju RT, Jerosch-Herold M, Wilke NM: Magnetic resonance imaging guided cardiovascular
interventions in congenital heart diseases. J Interv Cardiol 16:143-147, 2003. Medline Similar articles
135. Buecker A, Spuentrup E, Grabitz R, et al: Magnetic resonance-guided placement of atrial septal closure device in animal
model of patent foramen ovale. Circulation 106:511-515, 2002. Medline Similar articles
136. Rickers C, Jerosch-Herold M, Hu X, et al: Magnetic resonance image-guided transcatheter closure of atrial septal
defects. Circulation 107:132-138, 2003. Medline Similar articles
137. Dick AJ, Guttman MA, Raman VK, et al: Magnetic resonance fluoroscopy allows targeted delivery of mesenchymal stem
cells to infarct borders in Swine. Circulation 108:2899-2904, 2003. Medline Similar articles
138. Erhart P, Ladd ME, Steiner P, et al: Tissue-independent MR tracking of invasive devices with an internal signal source.
6 of 7 15-04-2010 21:04
139. Feiner SK: Augmented reality: a new way of seeing. Sci Am 286:48-55, 2002. Medline Similar articles
140. Rosenthal M, State A, Lee J, et al: Augmented reality guidance for needle biopsies: an initial randomized, controlled trial
in phantoms. Med Image Anal 6:313-320, 2002. Medline Similar articles
© 2010 Elsevier
7 of 7 15-04-2010 21:04
UIDED NDOVASCULAR NTERVENTIONS

Reed A. Omary
INTRODUCTION
Magnetic resonance imaging (MRI)-guided endovascular interventions have been the subject of
progressively increasing research. While mainly performed in animals, this research has also provided
encouraging preliminary results in some human studies.
There are several benefits to using MRI rather than X-rays to guide endovascular procedures. First,
there is no ionizing radiation exposure, which can harm the patient. Additionally, the operating physician
and hospital staff can benefit by avoiding the cumulative damaging effects of a lifetime of radiation
exposure. Second, MRI guidance does not require the use of iodinated contrast medium, with its
attendant risks of renal toxicity and allergic reaction. This advantage is especially important to patients
with poor renal function. Third, MRI is the only imaging modality that permits direct monitoring of
changes in end-organ function at the time of an intervention. For instance, myocardial perfusion could
be assessed directly at the time of a coronary intervention to monitor the effects of therapy. Ultimately,
it is hoped that this capability could lead to changes in the anticipated treatment and may predict the
success of the therapy. Finally, MRI provides outstanding soft-tissue contrast while permitting imaging
in any arbitrary plane. This advantage is especially useful for procedures such as transjugular
intrahepatic portosystemic shunt (TIPS) placement, where the interventionalist needs to know the
relationship of the blood vessels to the organ of interest.
The purpose of this chapter is to provide an overview of the basic principles required for MRI-guided
endovascular interventions. Because a diverse set of techniques are currently available to the
interventionalist, this chapter will discuss the advantages and limitations of competing methods.
Throughout the chapter, illustrative examples will be shown to enhance the text. It is also hoped that
the interested reader will be given potential areas for future research.
© 2010 Elsevier
1 of 1 15-04-2010 21:07
REQUIREMENTS FOR ENDOVASCULAR INTERVENTIONAL MAGNETIC

RESONANCE IMAGING PROCEDURES
page 541
page 542
A typical interventional procedure performed under X-ray guidance uses an X-ray fluoroscopy unit with
digital subtraction angiography (DSA) capabilities. With the patient placed on the procedure table, the
interventionalist has easy access to most parts of the body. Spatial resolution is high (~100 μm), while
temporal resolution is up to 30 frames/s. Endovascular devices, including catheters, guidewires,
balloon catheters, and stents, are designed with materials which are readily visible under X-ray
fluoroscopy. Procedures are depicted in realtime on an in-room monitor, permitting the interventionalist
to alter device position in a fraction of a second. Each of these fundamental tasks used during X-ray
guided procedures presents a significant challenge to those interested in performing MRI-guided
endovascular procedures.
Access to the Patient

Access to the patient is challenging within the MRI environment. Unlike typical X-ray fluoroscopy suites,
the imaging apparatus cannot be moved. Because conventional high-field 1.5 T MRI scanners contain
closed bores ~160-170 cm in length, it can be difficult to maintain access to the catheter insertion site
(groin or neck) while providing a useful field-of-view of the target vascular distribution. One solution is
to use open-bore MRI scanners.1 Horizontal or double donut vertical open magnet configurations
considerably improve access to the patient. However, these open design configurations are hampered
by lower field strengths (typically 0.2-0.7 T) and weaker/slower gradient systems that ultimately
reduce the effectiveness for endovascular interventions from spatiotemporal resolution and signal-
to-noise ratio (SNR) perspectives. Most interventionalists would prefer the trade-off of reduced patient
access with the closed-bore high-field designs for the improved imaging capabilities compared to the
open-bore designs.
Access to X-ray Units

Because MRI guidance still has many technical limitations and patient safety remains unproven, it is
essential that easy access to an X-ray unit is available before these procedures can be accepted
clinically. The easiest, least expensive approach is to site an interventional MRI scanner next door to
an X-ray angiographic unit. While helpful, this approach still leaves unaddressed the issue of how to
rapidly transport patients between both imaging devices. To handle this issue, MRI manufacturers have
begun to offer combined units that offer 1.5 T MRI scanners immediately adjacent to fully functioning
X-ray DSA units.2-4 The patient can easily be transferred between MRI and X-ray using a sliding table,
as shown in Figure 20-1. Another approach is to integrate a digital flat-panel X-ray system into an
interventional magnet, allowing MR and X-ray imaging of the same field-of-view without patient
5
movement. These hybrid units provide the benefit of improved patient safety and reduced procedure
times. The interventionalist can choose to perform part of the endovascular procedure under X-ray
guidance and part of the procedure under MRI guidance. Alternatively, MRI can be used as the sole
guidance modality, with X-ray used as a fallback option in difficult cases. Further experience is
required to discern the relative advantages and disadvantages of competing hybrid units.
Visible and Safe Devices/Instruments
1 of 4 15-04-2010 21:15
Add to lightbox
Figure 20-1 Dual MRI/X-ray angiography system consisting of full-feature C-arm angiography unit (in
the foreground) and an adjoining short-bore 1.5T MR scanner (background). A common floating
tabletop (arrow) allows seamless patient transfer between the MRI and C-arm components. The
component units can be used independently when the leaded, radiofrequency-shielded doors
separating them are closed. (Courtesy of Mark Wilson MD, University of California, San Francisco)
The devices (catheters, guidewires, stents) and instruments (percutaneous access needle, scalpel)
traditionally used for X-ray guided procedures need to be carefully assessed prior to use in
MRI-guided procedures. Ferromagnetic instruments present an extreme safety hazard because of their
potential to travel into the magnet bore. Similarly, many of the traditional devices used in X-ray
procedures, while not ferromagnetic, produce significant susceptibility artifacts that limit their utility, or
they are simply not visible.
Vascular Depiction
Under X-ray fluoroscopy, blood vessels are depicted with the intra-arterial (IA) administration of
iodinated contrast material directly from the catheter. Similarly, in the MRI environment, blood vessels
can be depicted with catheter-based injections of MRI contrast agent, typically gadolinium-based
chelates (Gd). These Gd injections can provide rapid, realtime background vascular roadmaps or can
be used for higher spatial resolution magnetic resonance angiography (MRA). Figure 20-2 portrays a
high-quality MRA obtained via a catheter-based injection of dilute Gd. It is also possible to provide
realtime background vascular depiction without contrast agent injections, using steady-state free
precession (SSFP) imaging.6,7
Realtime Display and Reconstruction

page 542
page 543
2 of 4 15-04-2010 21:15
Add to lightbox
Add to lightbox
3 of 4 15-04-2010 21:15
Add to lightbox
Figure 20-2 Selective catheter-based 3D high-resolution T1-weighted MRA with catheter tip
positioned within the superior mesenteric artery of a pig. The acquisition time for each of the three
data sets was 12 s. Contrast administration was started with a delay of 4 s after the start of data
acquisition and continued for 8 s. The three maximum intensity projection images show the arterial (A),
6
portal-venous (B), and late venous (C) phases. (From Quick HH, Kuehl H, Kaiser G, et al.
Reproduced with permission of Wiley-Liss, Inc., a subsidiary of John Wiley & Sons, Inc.)
Vascular interventions require realtime display of anatomic information on an in-room monitor. Because
of the magnetic field environment, this monitor cannot be a traditional cathode ray tube. While a
complex video projection scheme can be used to depict images in the MRI procedure suite, an easier
approach is to use a liquid crystal display, shown in Figure 20-3. Realtime image processing for
interventional MRI procedures requires substantial computational power. In-room consoles that control
MRI scanner functions are also desirable to improve scanner control.
Team Approach
MRI-guided endovascular procedures require considerable understanding of the technical requirements
to run an MRI scanner, as well as the clinical experience and endovascular skills of an interventionalist.
Analogous to X-ray guided procedures, it is imperative that a specialized team be present during
MRI-guided procedures. At a minimum, this requires a combination of MRI specialists and
interventionalists in order to facilitate procedures in the most safe and effective manner.
© 2010 Elsevier
4 of 4 15-04-2010 21:15
CATHETER-DIRECTED MAGNETIC RESONANCE ANGIOGRAPHY

Whether performed under X-ray or MRI guidance, endovascular procedures require multiple contrast
agent injections to define baseline vascular anatomy, confirm intraluminal position of endovascular
devices, and document vascular anatomic changes following an intervention. Direct catheter-based
injections of Gd under MRI guidance can be used in the same manner as injections of iodinated
contrast material under X-ray guidance: the catheter is placed in the vessel of interest (artery or vein)
and contrast agent is injected.
Add to lightbox
page 543
page 544
Figure 20-3 Realtime images are continuously displayed on the in-room monitor placed adjacent to
the patient couch during MRI-guided coronary catheterization of a dog. (Image courtesy of Ergin
Atalar PhD, Johns Hopkins University)
The major rationale for using catheter-directed MRA rather than conventional intravenous (IV) injections
is to conserve contrast agent while providing rapid vascular depiction. The multiple injections required
during an MRI-guided endovascular intervention would easily exceed the United States Food and Drug
Administration (FDA)-mandated daily dose limit of 0.3 mmol/kg of Gd using IV injections. Because
catheter-directed injections use smaller volumes of dilute contrast agent, the operator should more
easily remain below FDA dose limits. Additional benefits of catheter-based injections include reduction
in background tissue enhancement because less contrast is injected per scan, as well as enhancement
of only the artery of interest. This is especially important when there are other overlapping vascular
beds near the artery of interest, such as with the coronary circulator projection imaging. Compared to
IV injections, there is easier synchronization of the arrival of contrast agent with image acquisition.
Finally, IA injections have reduced contrast agent dispersion.
There are two major approaches to catheter-directed MRA: vascular roadmaps or high-quality
diagnostic MRA. Vascular roadmaps are acquired in realtime without electrocardiographic (ECG)
triggering. They are two-dimensional (2D) acquisitions, often acquired using projection imaging
technique. Projection imaging refers to thick-slice (2-20 cm thick) acquisition, similar to the technique
used in traditional 2D X-ray DSA. The projection method is useful to depict tortuous vessels or those
that extend outside a conventional thin slice. Roadmaps provide speed at the cost of spatial resolution
and MRI signal. Higher quality, diagnostic MRA can be performed using 2D projection techniques or
using three-dimensional (3D) acquisitions. The improved spatial resolution comes at the price of a
decrease in imaging speed. This higher quality MRA is not obtained in realtime. Depending on the
sequence used and anatomic location studied, it may also require ECG triggering. In most animal
studies using catheter-directed MRA, the catheter has been positioned for selective IA injections in the
1 of 9 15-04-2010 21:18
aorta,8-12 carotid arteries,9,11,13,14 renal arteries,9,10,15,16 iliac arteries,11,14,16 and coronary

9,17-20
arteries. Figure 20-4 provides an example of catheter-directed coronary MRA. Catheters have
also been placed in the inferior vena cava (IVC) for direct caval injections.7
Theory of Local Gadolinium Injections

Catheter-directed MRA require injections of dilute contrast agent because Gd induces competitive
effects of T1 and T2* shortening. In a conventional T1-weighted gradient-echo (GRE) sequence, the
T1 shortening increases MRI signal in blood, while the T2* shortening reduces MRI signal. An optimal
Gd concentration ([Gd]) exists where the T1 shortening signal gain is balanced with the competitive
T2* signal loss, and the blood signal is maximal.
For these local injections of Gd, full-strength MRI contrast agent is diluted with saline. The optimal
concentration required for dilution is dependent on the study and pulse sequence used. For
conventional GRE sequences, theoretical expressions, 10,11 static8,9 and dynamic16 phantom studies
9-11,16
and in vivo experiments suggest that optimal MRI signal is obtained using Gd concentrations
ranging from 1% to 6% (0.5 M contrast agent diluted by volume). There is little practical difference in
vessel enhancement or SNR between Gd solutions in this range of concentrations. For SSFP pulse
sequences, the optimal blood concentration of Gd for maximal MRI signal has not yet been elucidated.
However, preliminary studies19,20 of intracoronary injections using SSFP suggest that 8% injected Gd
provides excellent coronary depiction.
Add to lightbox
Add to lightbox
2 of 9 15-04-2010 21:18
Add to lightbox
Add to lightbox
Add to lightbox
Figure 20-4 Realtime catheter-directed projection coronary MRA in a canine artery. Panel of
T1-weighted spoiled GRE images shows coronal views obtained during direct injection of diluted Gd
into left circumflex artery. Discrimination of the wash-in (A-C) and washout (D) arterial phases and
myocardial perfusion phase (E, black star) is evident. (From Serfaty JM, Yang X, Foo TK, et al.63
Reproduced with permission of Wiley-Liss, Inc., a subsidiary of John Wiley & Sons, Inc.)
Intra-arterial Injection Protocols

page 544
page 545
Recognition of the optimal arterial Gd concentration is the first step in performing an IA injection.
However, in most instances the desired arterial Gd concentration differs from the injected Gd
concentration. The difference between injected and arterial Gd concentrations is due to additional
dilution of injected Gd by inflowing blood. Arterial Gd concentration depends upon three factors:
injected Gd concentration, injection rate, and arterial blood flow rate. Frayne et al11 and Bos et al10
11
have described similar relationships between these injection parameters, except that Frayne et al
3 of 9 15-04-2010 21:18
account for the influence of injection rate on overall arterial blood flow rates. The IA injection protocol
originally proposed by Frayne et al11 and subsequently validated16 is:
where [Gd]inj = injected Gd concentration, Qartery = blood flow rate in vessel of interest, Qinj = injection
rate of Gd contrast agent, and [Gd]artery = desired arterial concentration of Gd.
By substituting injection parameters into Equation 20-1, interventionalists can devise injection protocols
for catheter-based MRA. The protocol shows an inverse relationship between injected [Gd] and
injection rate. To obtain a desired arterial [Gd], one can either increase the injection rate and reduce
the injected [Gd] or increase the injected [Gd] and reduce the injection rate. This trade-off occurs
because both approaches deliver the same local Gd mass flux11 to the blood vessel.
Magnetic Resonance Angiography Sequences
Add to lightbox
Figure 20-5 Three-dimensional GRE maximum intensity projection image of bilateral renal artery
stenosis (arrows) in a pig. Catheter-based suprarenal aortic injection used 40 cc of 6% Gd.
MRA sequences should be selected based upon the intended purpose of a catheter-based injection.
For roadmaps with blood vessels located within a defined thin imaging slab, thin-slice 2D time-resolved
imaging is best. For tortuous vessels, 2D projection imaging is more likely to contain the blood vessel
within the imaging slab, at the expense of vessel depiction. If greater diagnostic accuracy or
multiplanar volumetric reconstructions are desired, then 3D sequences should be considered. Figure
20-5 portrays high-resolution catheter-based 3D MRA of renal artery stenosis. For 3D approaches
with the same spatial coverage as 2D projections, temporal resolution is reduced and thus additional
contrast agent dose is required. Electrocardiographic (ECG) gating may be used for some vascular
distributions, such as the heart. However, attention should be paid to the duration of injection. For
instance, intracoronary injections over 4 s in duration can obscure the coronary arteries due to
overlapping myocardial perfusion.20 Myocardial enhancement can be avoided if images are acquired
within a few seconds of injection.
4 of 9 15-04-2010 21:18
T1-weighted spoiled GRE sequences, using short repetition times (TR) and short echo times (TE), are
most commonly used for catheter-based MRA.8-10 These sequences can be similar to those used with
conventional IV Gd-enhanced MRA. However, SSFP sequences may be preferred in at least the
19,20 20
coronary artery distribution. Green et al compared 2D projection GRE versus SSFP for direct
intracoronary injections in swine. They found that SSFP approximately doubled the SNR and contrast-
to-noise ratio (CNR) compared to GRE.20 Sample coronary artery images using the two sequences
are shown in Figure 20-6. The potential benefit of using SSFP over GRE for catheter-based MRA in
other vascular distributions remains to be defined.
Knowledge of the local blood flow rate adjacent to the catheter is necessary to use the injection
protocol relationship described by Equation 20-1. This blood flow rate can be estimated empirically
based on experience obtained from the literature. However, to be more accurate, 2D cine phase
contrast imaging11,12,16,21 can be employed to measure the local blood flow rate.
Catheter-based MRA can be improved by suppression of background tissue. There are several such
22
methods available. Although source imaging data obtained prior to contrast agent can be subtracted,
this method is limited because it requires additional image processing. Motion between data
acquisitions will also cause image artifacts after subtraction. A gradient dephaser can be used in the
slice direction to suppress signals from background tissues,23,24 but the effectiveness of this scheme
depends on the anatomic structure of the imaging slice. Recently, magnetization preparation has been
14,18,20,25
used to suppress background in contrast-enhanced MRA. This method may be extremely
useful for 2D projection MRA.
Strategies to Limit Contrast Agent Dose

26
Several approaches are available to limit injected contrast agent dose. These can be divided into
imaging techniques, injection parameters, and catheter location.
Imaging Techniques
page 545
page 546
Add to lightbox
5 of 9 15-04-2010 21:18
Add to lightbox
Figure 20-6 The right coronary artery of a pig following catheter-directed dilute contrast agent injection
using magnetization prepared (A) GRE and (B) SSFP. Depiction of the proximal (solid arrow) and
middle (dashed arrow) portions of the artery is substantially improved in (B) due to the better SNR and
CNR obtained using the SSFP acquisition scheme. (From Green JD, Omary RA, Tang R, Li D:
Catheter-directed contrast-enhanced coronary MR angiography in swine using magnetization-
prepared True-FISP. Magn Reson Med 2003; 50:1317-1321. Reproduced with permission of
Add to lightbox
6 of 9 15-04-2010 21:18
Add to lightbox
Add to lightbox
Add to lightbox
7 of 9 15-04-2010 21:18
Add to lightbox
Add to lightbox
page 546
page 547
Figure 20-7 Advancement of an actively visualized guidewire and catheter from the iliac artery (A)
through the abdominal aorta (B) into the aortic arch (C) of a pig. SSFP images were acquired in the
coronal plane. Background blood vessels are depicted as bright intravascular signal. The actively
visualized guidewire is displayed in red, while the catheter is displayed in green/yellow. The colored
instrument outlines are overlaid onto the anatomic images that were acquired with the body and the
spine array coils (upper row, A-C). The corresponding images (D-F) in the bottom row show
schematically how the catheter is advanced into the pig while the pig, lying on the floating table, was
moved out of the scanner. Arrowheads indicate the catheter tip. This procedure ensures that the
moving region of interest always stays within the field-of-view that is covered by the body and the spine
array coils of the scanner (black rectangles anterior and posterior to the pig). (From Quick HH, Kuehl
H, Kaiser G, et al,29 with permission)
1. Non-contrast-enhanced SSFP sequences can be used to depict background vascular anatomy

as much as possible. As shown in Figure 20-7, SSFP sequences may be used for both catheter
tracking and background arterial anatomy.6,7 Catheter-based roadmaps and high-quality,
diagnostic MRA can then be employed when truly needed. Alternatively, Wacker et al27 have
applied catheter-based injections of carbon dioxide to reduce arterial signal obtained with
non-contrast-enhanced SSFP bright-blood imaging.
2. Injections should be tailored towards the imaging goal. Because injection duration should cover
at least a substantial portion of the image acquisition period,8 2D catheter-based MRA will use
considerably less contrast agent than 3D methods. Reserve 3D methods for occasions when
improved diagnostic accuracy or multiplanar volumetric reconstructions are desired.
3. Injection duration can be reduced. For 3D imaging, Hwang et al28 showed that injection duration
could be reduced to 50% of the image acquisition time without significant loss of SNR in the
aorta and iliac arteries. For smaller vessels such as the renal arteries, injection duration can be
reduced to 75% of the image acquisition time without significant loss of SNR. In dynamic flow
phantoms, there was no difference in SNR between elliptical centric and conventional sequential
linear encoding schemes for IA injections.28
4. A floating table can follow contrast agent movement by adjusting the field-of-view 29 (see Fig.
20-7).
Injection Parameters
For GRE sequences, arterial [Gd] of 1% can be used. While optimal SNR is obtained with arterial [Gd]
ranging from 1% to 6%, aiming for 1% will reduce contrast agent sixfold over 6%. This reduction is due
to the direct relationship between arterial [Gd] and injected dose (see Equation 20-1).
8 of 9 15-04-2010 21:18
Catheter Location
Catheters can be positioned as selectively as possible. Major reductions in contrast agent dose occur
when catheters tips are placed within smaller vessels. This concept is used routinely to reduce injection
volume with X-ray DSA because smaller vessels have reduced blood flow rates compared to larger
vessels. For example, performing a selective renal artery injection will use substantially less contrast
18
agent than an abdominal aortic injection. One study described how 83 separate selective injections
could be performed into a canine left circumflex artery without exceeding the FDA-mandated daily limit
of 0.3 mmol/kg Gd.
Limitations
There are several important limitations to catheter-directed MRA. First, the FDA has not approved
catheter-based injections of Gd for MRA. These injections represent an off-label use and unapproved
route of administration of MR contrast agent. Second, the safety of IA Gd injections is unproven.
However, there is little incremental risk for catheter-based injections once the catheter has already
been positioned during an endovascular intervention. Interventional radiologists have already adopted
Gd as an alternative contrast agent for use during X-ray DSA in patients with underlying renal
30-32
insufficiency. Third, there is very limited experience of catheter-based MRA in humans. Finally,
there are limited data regarding the accuracy of catheter-based injections. In a swine model of renal
artery stenosis, Omary et al12 showed no significant difference in accuracy between IA- and IV-Gd
enhanced 3D MRA, using X-ray DSA as a gold standard.
© 2010 Elsevier
9 of 9 15-04-2010 21:18
DEVICE TRACKING
Detection of intravascular devices (e.g., catheters, guidewires, stents) in the MRI environment
traditionally employs passive techniques, active techniques or some combination of both approaches.
Passive Tracking
Passive tracking methods employ differences in MRI signal between catheter and background tissues,
without using implanted catheter coils. These methods are analogous to techniques employed in X-ray
DSA. Initial passive methods tracked signal loss caused by the placement of dysprosium oxide
markers on catheters.33,34 The markers created small susceptibility artifacts which could be used to
monitor the position of the catheter. Catheter visualization was limited to the location of the six
dysprosium oxide markers on the catheter. Disadvantages with this approach include relatively poor
temporal resolution, dependence on magnetic field orientation, limited visibility within small or tortuous
vessels, depiction as signal loss rather than positive signal, and inability to detect catheter kinking
between the markers. Intravascular MR contrast agents can potentially enhance the visual conspicuity
of the dysprosium oxide markers in blood vessels.35 However, segmentation of arteries from veins is a
significant issue whenever intravascular contrast agents are used.
page 547
page 548
Add to lightbox
Add to lightbox
Figure 20-8 Comparison of (A) IR-GRE and (B) conventional GRE for catheter tracking in the
abdominal aorta using an 8 Fr inner diameter catheter filled with 4% diluted contrast agent. Images
have a slice thickness of 5 cm and an oblique-sagittal orientation. There are several regions where the
1 of 17 15-04-2010 21:19
catheter is obscured by background tissue in (B) but not in (A) (arrowheads). Bowel (on the left side of
the catheter) is bright, possibly due to short T1 components in the swine's diet. (From Green JD,
36
Omary RA, Finn JP, et al, with permission)
By filling a conventional angiographic catheter with dilute Gd, Unal et al23 obtained bright catheter
signal along the entire length of the catheter using a T1-weighted GRE sequence. The optimal Gd
24
concentration using this approach was shown to be 4% to 6% Gd. A standard angiographic flow
switch or other hemostatic-type valve can be placed on the external end of the catheter to prevent
contrast agent leakage out of the catheter. Significant advantages of the Gd-filled catheter approach
are that conventional, nonbraided angiographic catheters can be used and that the entire catheter can
be visualized. However, providing ample signal for smaller-sized catheters in vivo can be challenging,
especially using thick slab projection imaging, because the overall contribution of background signal
can overwhelm the bright signal within smaller catheters. To improve delineation of these
contrast-agent filled catheters, background tissue should be suppressed using a projection
dephaser23,24 or magnetization preparation.36 Figure 20-8 demonstrates the use of inversion recovery
to suppress background tissue during passive tracking of Gd-filled catheters. The in vivo depiction of
conventional 5 Fr angiographic catheters in realtime using passive Gd-filled catheter methods remains
difficult.
For metallic endovascular devices, intrinsic susceptibility differences between the alloy and background
tissues can help detect the device. While this approach is adequate for large devices, such as
stents37-41 or IVC filters,42,43 smaller diameter devices, such as guidewires, cannot be detected
reproducibly.40 Artifacts vary depending on the composition of the metallic alloy,44,45 with nitinol alloy
offering the potentially best balance between visualization and extensive artifact. While a larger artifact
improves visibility, it also intrinsically distorts local anatomy, hampering precise localization. 46
Visualization of metallic artifacts may be enhanced using an intravascular contrast agent. 38 Unal and
colleagues have also applied a Gd-based coating to guidewires that permits passive detection of the
entire coated length, as shown in Figure 20-9.
Add to lightbox
2 of 17 15-04-2010 21:19
Figure 20-9 Passive tracking of a nitinol guidewire with 60 μm MR-visible Gd-based coating. Coated
guidewire (arrows) in the canine aorta using coronal T1-weighted GRE imaging. (Courtesy of Orhan
Unal PhD, University of Wisconsin-Madison)
While nitinol alloys may be employed in the interventional MRI setting, they are not without risk.
Konings et al47 tested a 0.035-inch diameter nitinol guidewire (Terumo, Tokyo, Japan), commonly used
for both X-ray and interventional MRI settings. Using a worst-case in vitro MRI scanning technique,
they found that the guidewire tip reached temperatures of up to 74º C after 30 s of scanning. Although
nitinol has no ferromagnetic properties, they related the excessive heating to resonating radiofrequency
48
(RF) waves. In an elegant phantom study, Liu et al determined that nitinol guidewire-based heating
was related to location with respect to the RF coil (center versus off-center), deployed length of
guidewire, magnet strength, and TR. In general, convective heat loss from blood flow should reduce in
vivo heating relative to these phantom studies. Quarter-wave length coaxial chokes might also reduce
heating,49 but would tend to alter the mechanical properties of the guidewire.
Active Tracking
Conventional active device tracking methods rely on the presence of one or more RF coils that depict
either the tip of the device or the tissue surrounding the device. 33,50-55 Device location is obtained from
MRI signal detected from the coil. The major advantage of this technique is high temporal resolution.
Depending on the desired spatial resolution, temporal resolutions of over 10 frames/s can be achieved
using common active tracking methods. A significant disadvantage of this type of tracking, however, is
that only the current position of catheter coil can be seen, typically displayed as a colored icon.
Interventionalists need more information than just the location of the catheter tip because the catheter
can buckle without the operator knowing. The build-up of torque may dislodge the catheter out of the
selected vessel, impeding the success of the procedure.
page 548
page 549
Add to lightbox
3 of 17 15-04-2010 21:19
Add to lightbox
Figure 20-10 Active catheter tracking. A, Photograph of a 5 Fr guiding catheter (Bentson 1,
Angiodynamics). B, Drawing of the equivalent MRI guiding catheter showing the conventional guiding
catheter (GC), the shield of the loopless antenna (S), the extended inner conductor (IC), and the
flexible copper wire (CW) attached to the extended inner conductor and wrapped around the distal
part of the guiding catheter with an increasing pitch toward the tip of the guiding catheter. (From
62
Serfaty JM, Yang X, Aksit P, with permission)
Atalar et al56-58 have overcome this limitation by developing a loopless catheter antenna for active
visualization, as shown in Figures 20-10 and 20-11. Because these antennae outline the entire length of
the device, this approach is termed "profiling." Figure 20-12 demonstrates this profiling technique
during a MRI-guided coronary balloon angioplasty in a dog. Although the benefits of profiling the entire
length of catheter are obvious, occasionally imaging the entire catheter in only one projection will miss
key information. Quick et al have used an interleaved technique during active catheter tracking which
6
shows images in two separate orientations. Figure 20-13 demonstrates how their approach can be
helpful during selective catheterization of the superior mesenteric artery.
Add to lightbox
Figure 20-11 Photograph showing the loopless antenna based MRI guidewire (0.014 inch diameter)
inserted inside an inflated coronary balloon angioplasty catheter (upper catheter). The lower catheter
4 of 17 15-04-2010 21:19
is the MRI guiding catheter (7 Fr) built from attaching and coiling a 100 cm-long MRI guidewire (0.032
inch diameter) to a conventional 100 cm-long 5 Fr guiding catheter. (Courtesy of Ergin Atalar PhD,
Johns Hopkins University)
Some other recent advances have benefited active tracking. Elgort et al59 have developed a realtime
continuous feedback system that tracks the position of the active catheter and automatically updates
the scan plane's position, orientation, and field-of-view. Simply slowing the speed of the catheter will
automatically optimize imaging parameters such as field-of-view (Fig. 20-14) and spatial resolution to
59 60
improve visualization of the catheter. Guttman et al have developed a system that produces
continuous 3D feedback using realtime volume renderings. They have implemented several interactive
capabilities to enhance visualization, including complex subtraction, cut planes, and color highlighting. In
subsequent work,61 they implemented a high-performance software system that performs
time-adaptive sensitivity encoding and reconstructions for realtime MRI with interactive, online display.
There are several disadvantages to all active tracking methods that place receiver antennae onto the
catheter surface.51-53,62,63 First, this approach requires specialized catheters that can be expensive,
are difficult to obtain, and are fragile. Second, a completely separate catheter inventory for MRI- and
X-ray-guided endovascular procedures is required. Third, the coils can adversely affect the mechanical
properties of the catheter, increasing the difficulty of navigation into small vessels. Finally, they may
53 64,65
also result in local tissue heating, which remains an area of fertile current research.
Active visualization of nitinol guidewires has been performed using loopless58 and looped RF
antennae.66-68 These active guidewires can be used in combination with active catheters (Fig. 20-15).
These methods traditionally employed GRE sequences to detect the surrounding tissue rather than the
guidewire itself. More recently, SSFP sequences have been applied, as shown in Figure 20-7.
However, the application of SSFP sequences to active guidewires permits direct visualization of the
negative susceptibility artifact from the nitinol guidewire.7 This is helpful to detect guidewire buckling.
SSFP also increases the relative conspicuity of the negative susceptibility artifact because the blood is
bright using T2/T1-weighted SSFP sequences.
page 549
page 550
5 of 17 15-04-2010 21:19
Add to lightbox
Add to lightbox
6 of 17 15-04-2010 21:19
Add to lightbox
Add to lightbox
7 of 17 15-04-2010 21:19
Add to lightbox
page 550
page 551
Figure 20-12 MRI-guided balloon angioplasty of the left circumflex artery in a canine using an active
profile tracking technique. A, Placement of the MRI guiding catheter (arrowhead) in the ascending
aorta using the oblique sagittal view. B, Catheterization with the MRI guiding catheter (arrowhead) of
the left main coronary artery and circumflex artery using the oblique coronal view. C, Realtime
projection angiography of the circumflex artery (arrowhead) on an oblique coronal view after injection
of diluted Gd (31 mM) in the MRI guiding catheter. D, Placement of the MRI guidewire (arrowhead) in
the circumflex artery in the oblique coronal view. The balloon angioplasty catheter can be localized and
advanced on the MRI guidewire by using a black artifact created by a platinum ring localized in the
center of the balloon angioplasty catheter (long arrow). E, Injection of diluted Gd (31 mM) within the
balloon enhances the balloon on the realtime projection angiography images (long arrow). (From
63
Serfaty JM, Yang X, Foo TK, et al. Reproduced with permission of Wiley Liss, Inc., a subsidiary of
John Wiley & Sons, Inc.)
Add to lightbox
8 of 17 15-04-2010 21:19
Add to lightbox
Add to lightbox
Add to lightbox
9 of 17 15-04-2010 21:19
Add to lightbox
Add to lightbox
Add to lightbox
10 of 17 15-04-2010 21:19
Add to lightbox
Add to lightbox
Add to lightbox
Figure 20-13 Biplanar guidance of an active catheter in the abdominal aorta of a pig. Images (A-E)
were acquired in the coronal plane. Advancement of the catheter resulted in visual loss of the proximal
end of the catheter, while the 'false' tip in this plane demonstrated signal accumulation leading to a
changed appearance. Interleaved acquisition of images in the sagittal plane (F-J) revealed entry of the
catheter tip into the superior mesenteric artery. The entire length of the catheter is visible in the sagittal
plane. Further advancement of the catheter resulted in looping of the instrument (images I, J), which
was not obvious from the coronal images. For better instrument visibility, these images were
intentionally windowed and leveled such that the catheter is displayed brighter than the background
6
and the arteries. (From Quick HH, Kuehl H, Kaiser G, et al, with permission)
After initially visualizing active metallic endovascular stents using a connecting coaxial cable, 69 Quick et
70
al developed a method to actively visualize stents without connecting wires (Fig. 20-16). Their stent
11 of 17 15-04-2010 21:19
prototypes were designed to act as active resonant structures that amplified local RF signal, 70 as
shown in Figure 20-17. Their approach, which provides detailed analysis of the stent lumen, might offer
a future means of verifying long-term stent patency noninvasively. Figure 20-18 shows nonocclusive
thrombus within one of these in vivo active stents.
Combined Passive/Active Tracking

Passive and active tracking approaches each have their own relative merits and pitfalls. The ideal
device tracking method would permit fast temporal resolution at high spatial resolutions, employ
devices already available for use with conventional X-ray angiographic equipment, and offer no risk of
tissue heating. This ideal tracking method does not exist.
However, it is possible to combine the relative advantages of passive catheter tracking (depiction of
entire catheter length, use of standard X-ray angiographic catheters) with active tracking (high
temporal resolution, high localized signal detection). Omary et al7 filled conventional angiographic
catheters with 4% Gd and coaxially positioned an active guidewire. Detection of the Gd-filled catheter
was enhanced by the presence of the active guidewire using inversion recovery GRE, while the
guidewire was directly depicted using SSFP. Their approach, depicted in Figure 20-19, allowed
independent MRI-guided tracking of catheters and guidewires using a single loopless antenna located
on the guidewire. It avoided the need for separate loopless antennae to be placed on the guidewire
62,63
and the catheter.
71,72
Another combined approach uses field inhomogeneity catheters to enhance the natural signal void
of the catheter. In this technique, electrical currents are applied to the end of a catheter. As more
current is applied, there is increased susceptibility artifact. The artifact improves catheter depiction at
the expense of localized image distortion.
page 551
page 552
Add to lightbox
12 of 17 15-04-2010 21:19
Add to lightbox
Add to lightbox
13 of 17 15-04-2010 21:19
Add to lightbox
page 552
page 553
Figure 20-14 A temporal series of coronal MR images using SSFP obtained during a pig experiment.
The active catheter has been placed within the abdominal aorta. Using adaptive tracking methods, the
field-of-view was automatically reduced as catheter movement was slowed down. (Courtesy of Daniel
R Elgort MS and Jeffrey L Duerk PhD, Case Western Reserve University)
Add to lightbox
Figure 20-15 Photograph of the active .035-inch diameter guidewire and 6 Fr catheter with integrated
dipole antennas. The Y-connector at the distal end of the catheter includes the lumen for the guidewire
as well as the RF microplug for connecting the catheter to the surface coil port of the MR scanner. The
antenna tuning, matching, and decoupling for the guidewire and the catheter is housed inside
individual RF-shielded boxes at the proximal end of the instruments. The boxes are connected to
separate RF receiver channels of the MR scanner. (From Quick HH, Kuehl H, Kaiser G, et al,6 with
permission)
14 of 17 15-04-2010 21:19
Add to lightbox
Figure 20-16 Schematic of the principle of inductive coupling between two coils and its application to
the visualization of stents. A shows a loop coil that is tuned with a capacitor to resonance. This coil
picks up the MR signal in its immediate vicinity, resulting in a B 1-field vector that can be inductively
coupled to that of a loop surface coil (B). This technique allows the first loop coil to be implanted and
to wirelessly transmit its signal to an outside coil. The RF receiver coil system, consisting of implanted
coil and surface coil, is thus acting as a local signal amplifier and potentially allows high-resolution MR
imaging of deep-sited regions of interest. The implanted resonant circuit does not necessarily require
the shape of a loop coil; various stent-like coil configurations are conceivable as long as a closed-loop
electrical resonant structure is involved (C). (From Quick HH, Kuehl H, Kaiser G, et al,70 with
permission)
Add to lightbox
Add to lightbox
Add to lightbox
15 of 17 15-04-2010 21:19
Add to lightbox
Add to lightbox
page 553
page 554
Figure 20-17 Photograph of a balloon-expandable stent resonator prototype, length 28 mm, inner
diameter before/after expansion 1.8/4 mm. A, Stent in the folded state. B, Folded stent mounted on a
5 Fr balloon catheter (balloon 40 mm × 4 mm). C, Unfolded stent after full inflation of balloon. D, Fully
deployed stent. E, 2D MR GRE image acquired with the stent immersed in saline phantom. (From
70
Quick HH, Kuehl H, Kaiser G, et al, with permission)
Add to lightbox
16 of 17 15-04-2010 21:19
Add to lightbox
Figure 20-18 In vivo MRI using GRE of a stent resonator (length 25 mm, inner diameter 2 mm)
implanted into the right iliac artery of a pig. The outside surface receive coil was placed coaxially
above the position of the stent. The distance from the middle of the stent to the center of the loop coil
was approximately 12 cm. A, The interior of the stent displays high signal. In B, reduction in field-
of-view enables full assessment of the stent lumen over its full length, parallel to the axis of the stent.
The signal void in the middle of the stent lumen (arrow) was identified as thrombus after explantation.
70
(From Quick HH, Kuehl H, Kaiser G, et al, with permission)
© 2010 Elsevier
17 of 17 15-04-2010 21:19
INTERVENTIONS
The range of in vivo MRI-guided endovascular applications is broad. In addition to the many diagnostic
studies of catheter-directed MRA, MRI-guided endovascular interventions in animals have included the
following.
58,73,74 75
1. Percutaneous transluminal balloon angioplasty of the aorta, iliac arteries, renal
15,76 63
arteries, and coronary arteries.
38 37,70 40
2. Stent placement within the aorta, iliac arteries, coronary arteries, and pulmonary
artery/valve.41
42,43
3. IVC filter placement.
4. Atrial septal closure device placement.77,78
4
5. Embolization of the renal arteries using Gd-impregnated particles (Fig. 20-20) and of carotid
artery aneurysms using coils.13
79
6. TIPS.
7. Vascular gene therapy delivery.80
Published clinical studies performed under MRI guidance include iliac stent placement 39 and
81,82
hemodialysis fistula evaluation. Recently in humans, Kee et al performed portal vein punctures
under MRI guidance during TIPS, as shown in Figure 20-21. Limited cardiac chamber catheterization
has been performed under MRI guidance in children and adults with congenital heart disease.83 After
84
initially using an intravascular active guidewire for transvenous imaging of the arterial wall in animals,
Hofmann et al more recently extended this technique to image arterial plaque in humans (Fig. 20-22).
The transvenous approach permits the delineation of arterial pathology, such as dissections (Fig.
20-23), without the inherent risks of arterial catheterization.
© 2010 Elsevier
1 of 1 15-04-2010 21:20
LIMITATIONS
page 554
page 555
Add to lightbox
Add to lightbox
1 of 9 15-04-2010 21:21
Add to lightbox
Add to lightbox
2 of 9 15-04-2010 21:21
Add to lightbox
Add to lightbox
3 of 9 15-04-2010 21:21
Add to lightbox
page 555
page 556
Figure 20-19 Representative sagittal (A-C) and coronal (D-F) oblique images of the aorta obtained
during MRI-guided left coronary artery catheterization. Thin arrows depict device tips. A, SSFP
anatomic reference. B, D and F, SSFP guidewire tracking images with dark guidewire susceptibility
defect (thick arrow) surrounded by bright adjacent blood vessel (arrowheads). C, E and G,
T1-weighted IR-GRE catheter tracking. LCA, left coronary artery. (From Omary RA, Green JD, Schirf
19
BE, et al, with permission)
Add to lightbox
4 of 9 15-04-2010 21:21
Add to lightbox
Add to lightbox
Add to lightbox
5 of 9 15-04-2010 21:21
Add to lightbox
Figure 20-20 Realtime T1-weighted MR images obtained before, during, and after injection of
500-700 μm Gd-impregnated microspheres in a canine. (Courtesy of Mark Wilson MD, University of
California, San Francisco)
Add to lightbox
6 of 9 15-04-2010 21:21
Add to lightbox
page 556
page 557
Figure 20-21 TIPS as performed in a 43-year-old female with hepatitis C-induced cirrhosis using a
truly hybrid combined X-ray/MRI unit. A, Sagittal oblique T1-weighted SPGR image shows puncture
cannula/needle (arrows) during MRI-guided puncture of portal vein (arrowhead, PV) from the hepatic
vein (HV). Following successful portal vein entry on the first needle pass, X-ray guidance was
subsequently used to place a metallic stent across the hepatic parenchymal tract. B, Completion X-ray
splenic portogram showing successful TIPS placement. (Courtesy of Stephen Kee MD, Stanford
University)
Add to lightbox
Figure 20-22 Use of transvenous active intravascular guidewire to detect arterial pathology in a
61-year-old male with lower extremity intermittent claudication. A 0.030-inch diameter intravascular
MR coil/guidewire (IVMRG; Surgi-Vision, Gaithersburg, MD) has been percutaneously placed within
the inferior vena cava (IVC). Axial T1-weighted imaging depicts calcified lipid atherosclerotic plaque
within the adjacent aorta. (Courtesy of LV Hofmann MD, Johns Hopkins University)
7 of 9 15-04-2010 21:21
Add to lightbox
Add to lightbox
8 of 9 15-04-2010 21:21
Add to lightbox
page 557
page 558
Figure 20-23 Use of transvenous active intravascular guidewire to detect arterial pathology in a pig. A
0.030-inch diameter intravascular MR coil/guidewire (IVMRG; Surgi-Vision, Gaithersburg, MD) has
been placed within the inferior vena cava of a pig for imaging of the proximal right renal artery. A, 3D
contrast-enhanced MRA of right renal artery. Bar shows cross-sectional imaging plane obtained with
IVMRG in B. B, T1-weighted Gd-enhanced image of right renal artery dissection. Small arrow points to
false lumen (FL); larger arrow points to true lumen (TL). C, Hematoxylin-eosin stain of pathologic
specimen corresponding to B. (Courtesy of LV Hofmann MD, Johns Hopkins University)
There are several important limitations to MRI-guided endovascular procedures. First, safety has not
been proven. Rigorous safety comparisons with X-ray guided procedures are required before these
procedures can be translated into humans. The FDA has not approved most applications. Local tissue
heating is a major concern of active device tracking but can even occur with passive guidewire devices.
85,86
Yeung et al have proposed a safety index based upon the in vivo temperature change that occurs
with a guidewire in place, normalized to the specific absorption rate of the pulse sequence. Next,
significantly improved endovascular devices need to be developed for use within the MRI environment,
especially guidewires. This issue represents a Catch-22 situation: MRI and endovascular device
manufacturers are each waiting for the other to propel the field forward. From industry's perspective,
considerable financial risk is at stake for this as yet unproven technology. Finally, there remains a need
for improved spatial and temporal resolution for tracking of devices and vascular depiction. However,
as major advances in noninvasive MRI techniques continue, these should provide enhanced benefits to
interventional MRI also.
© 2010 Elsevier
9 of 9 15-04-2010 21:21
CONCLUSIONS AND FUTURE DIRECTIONS

Although the future of MRI-guided endovascular interventions remains promising, these procedures are
currently considered strictly within the research arena. To gain more widespread appeal, a perceived
"killer application" for MRI guidance should be devised, i.e., a procedure where MRI guidance provides
clear benefit over using X-ray guidance. This benefit will be most obvious when MRI can do something
that X-ray cannot. For instance, the research group of Lederman et al have successfully performed
MRI-guided endomyocardial injections, initially using dilute Gd. 87 In subsequent pioneering work, they
88 89
injected iron fluorescent particle-labeled mesenchymal stem cells into myocardial infarct borders
(Fig. 20-24). The ability to determine precise stem cell implantation at infarct borders under MRI
guidance represents one application that cannot currently be performed under X-ray guidance.
Interventionalists might not want to simply repeat applications performed under X-ray guidance;
instead, they should explore applications that are not well performed with X-ray guidance. Potential
novel applications might involve procedures where end-organ function can be assessed, such as
transcatheter embolization of liver tumors or catheter-directed thrombolysis of stroke. The ultimate
goal would be to use these functional changes to gauge the success of a procedure, rather than
relying simply on anatomic changes, as with current X-ray guided procedures. Other applications
include molecular imaging or the monitoring of drug/gene therapy delivery. Figure 20-25 demonstrates
MRI-guided monitoring of vascular gene therapy, as proposed by Yang et al.80 Qui et al90 have
exploited one of the perceived limitations of MRI guidance-active guidewire-induced local tissue
heating-to improve vascular gene transfection.
Add to lightbox
Add to lightbox
1 of 8 15-04-2010 21:22
Add to lightbox
Add to lightbox
Figure 20-24 MR fluoroscopy during endomyocardial injections of tagged mesenchymal stem cells in
a pig. A, Stiletto needle is engaged at apical septal border of anterior myocardial infarction. MRI
signal from needle tip is red and that from guiding catheter green. Arrows indicate previous injections
of iron-labeled stem cells, which show as dark signal voids. B, A 150 μL test injection of Gd is
indicated by arrowhead and shows as white. C, Saturation preparation enhances appearance of test
injectate compared with black myocardium and blood. D, Iron-labeled mesenchymal stem cells (1 ×
6
10 ) are injected into the same spot, extinguishing local signal, and appear dark. (From Dick AJ,
89
Guttman MA, Raman VK, et al, with permission)
Given the considerable patient benefit that has occurred separately over the past two decades from
minimally invasive endovascular procedures and from diagnostic MRI, combining these two methods
should likely improve patient care in the future.
Acknowledgments
The author would like to thank those researchers who contributed images for the figures. He would
also like to acknowledge the contributions of the interventional MRI research group at Northwestern
University: Debiao Li PhD, Jordin Green MS, Brian Schirf MD, and Richard Tang MD.
page 558
page 559
2 of 8 15-04-2010 21:22
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
3 of 8 15-04-2010 21:22
Add to lightbox
Add to lightbox
Add to lightbox
Figure 20-25 High-resolution MR images of Gd/GFP-lentivirus transfer in the iliac artery of a pig. A,
Before Gd/GFP-lentivirus infusion, the balloon is inflated with 3% Gd contrast agent. The open arrow
indicates the artery. V, vein. Scale = 1 mm. B-F, During Gd/GFP-lentivirus infusion from minute 3 to
minute 15 (at 3-minute intervals), the arterial wall is enhanced by the Gd coming from the gene infusion
channels (arrowheads in B) of the gene delivery catheter. At minute 15, the arterial wall is enhanced as
a ring (arrow in F). G and H, Corresponding immunohistochemistry in both control (G) and
GFP-targeted (H) arteries. H, GFP is detected as brown-colored precipitates through all layers of the
intima (arrows) and media as well as the adventitia. Original magnification, 200×. (From Yang XM,
Atalar E, Li D, et al,80 with permission)
REFERENCES
1. Schenck JF, Jolesz FA, Roemer PB, et al: Superconducting open-configuration MR imaging system for image-guided
therapy. Radiology 195:805-814, 1995. Medline Similar articles
2. Adam G, Neuerburg J, Bucker A, et al: Interventional magnetic resonance. Initial clinical experience with a 1.5-tesla
magnetic resonance system combined with c-arm fluoroscopy. Invest Radiol 32:191-197, 1997.
3. Vogl TJ, Balzer JO, Mack MG, et al: Hybrid MR interventional imaging system: combined MR and angiography suites with
single interactive table. Feasibility study in vascular liver tumor procedures. Eur Radiol 12:1394-1400, 2002. Medline
Similar articles
4. Wilson MW, Fidelman N, Weber OM, et al: Experimental renal artery embolization in a combined MR imaging/angiographic
unit. J Vasc Interv Radiol 14:1169-1175, 2003. Medline Similar articles
5. Fahrig R, Butts K, Wen Z, et al: Truly hybrid interventional MR/X-ray system: investigation of in vivo applications. Acad Radiol
6. Quick HH, Kuehl H, Kaiser G, et al: Interventional MRA using actively visualized catheters, TrueFISP, and real-time image
fusion. Magn Reson Med 49:129-137, 2003. Medline Similar articles
7. Omary RA, Green JD, Fang WS, et al: Use of internal coils for independent and direct MR imaging-guided endovascular
device tracking. J Vasc Interv Radiol 14:247-254, 2003. Medline Similar articles
4 of 8 15-04-2010 21:22
8. Omary RA, Frayne R, Unal O, et al: Intraarterial gadolinium-enhanced 2D and 3D MR angiography: a preliminary study. J
Vasc Intervent Radiol 10:1315-1321, 1999.
9. Serfaty JM, Atalar E, Declerck J, et al: Real-time projection MR angiography: feasibility study. Radiology 217:290-295,
10. Bos C, Smits HF, Bakker CJ, Viergever MA: Selective contrast-enhanced MR angiography. Magn Reson Med
11. Frayne R, Omary RA, Unal O, Strother CM: Determination of optimal injection parameters for intraarterial gadolinium-
enhanced MR angiography. J Vasc Intervent Radiol 11:1277-1284, 2000.
12. Omary RA, Henseler KP, Unal O, et al: Comparison of intraarterial and IV gadolinium-enhanced MR angiography with
digital subtraction angiography for the detection of renal artery stenosis in pigs. Am J Roentgenol 178:119-123, 2002.
13. Strother CM, Unal O, Frayne R, et al: Endovascular treatment of experimental canine aneurysms: feasibility with MR
imaging guidance. Radiology 215:516-519, 2000. Medline Similar articles
14. Bos C, Bakker CJ, Viergever MA: Background suppression using magnetization preparation for contrast-enhanced MR
projection angiography. Magn Reson Med 46:78-87, 2001. Medline Similar articles
15. Omary RA, Frayne R, Unal O, et al: MR-guided angioplasty of renal artery stenosis in a pig model: a feasibility study. J
Vasc Intervent Radiol 11:373-381, 2000.
16. Omary RA, Henseler KP, Unal O, et al: Validation of injection parameters for catheter-directed intraarterial gadolinium-
enhanced MR angiography. Acad Radiol 9:172-185, 2002. Medline Similar articles
17. Tsekos NV, Woodard PK, Foster GJ, et al: Dynamic coronary MR angiography and first-pass perfusion with intracoronary
administration of contrast agent. J Magn Reson Imaging 16:311-319, 2002. Medline Similar articles
18. Green JD, Omary RA, Finn JP, et al: Two- and three-dimensional MR coronary angiography with intraarterial injections of
contrast agent in dogs: a feasibility study. Radiology 226:272-277, 2003. Medline Similar articles
19. Omary RA, Green JD, Schirf BE, et al: Real-time magnetic resonance imaging-guided coronary catheterization in swine.
20. Green JD, Omary RA, Schirf BE, et al: Catheter-directed contrast-enhanced coronary MR angiography in swine using
magnetization-prepared True-FISP. Magn Reson Med 50:1317-1321, 2003. Medline Similar articles
21. Schoenberg SO, Knopp MV, Bock M, et al: Renal artery stenosis: grading of hemodynamic changes with cine phase-
contrast MR blood flow measurements. Radiology 203:45-53, 1997. Medline Similar articles
36:345-351, 1996.
23. Unal O, Korosec FR, Frayne R, et al: A rapid 2D time-resolved variable-rate k-space sampling MR technique for passive
catheter tracking during endovascular procedures. Magn Reson Med 40:356-362, 1998.
24. Omary RA, Unal O, Koscielski DS, et al: Real-time MR imaging-guided passive catheter tracking with use of gadolinium-
filled catheters. J Vasc Intervent Radiol 11:1079-1085, 2000.
25. Li D, Carr JC, Shea SM, et al: Coronary arteries: magnetization-prepared contrast-enhanced three-dimensional volume-
targeted breath-hold MR angiography. Radiology 219:270-297, 2001. Medline Similar articles
26. Omary RA, Green J, Finn JP, Li D: Catheter-directed gadolinium-enhanced MR angiography. Radiol Clin North Am
27. Wacker FK, Maes RM, Jesberger JA, et al: MR imaging-guided vascular procedures using CO2 as a contrast agent. Am J
Roentgenol 181:485-489, 2003.
28. Hwang KP, Green JD, Li D, et al: Minimizing contrast agent dose during intraarterial gadolinium-enhanced MR
angiography: in vitro assessment. J Magn Reson Imaging 15:55-61, 2002. Medline Similar articles
29. Quick HH, Kuehl H, Kaiser G, et al: Interventional MR angiography with a floating table. Radiology 229:598-602, 2003.
30. Kinno Y, Odagiri K, Andoh K, et al: Gadopentetate dimeglumine as an alternative contrast material for use in
angiography. Am J Roentgenol 160:1293-1294, 1993.
page 559
page 560
31. Spinosa DJ, Matsumoto AH, Angle JF, Hagspiel KD: Use of gadopentetate dimeglumine as a contrast agent for
percutaneous transluminal renal angioplasty and stent placement. Kidney Int 53:503-507, 1998. Medline Similar articles
32. Spinosa DJ, Kaufmann JA, Hartwell GD: Gadolinium chelates in angiography and interventional radiology: a useful
alternative to iodinated contrast media for angiography. Radiology 223:319-325, 2002; discussion 326-327.
33. Bakker CJ, Hoogeveen RM, Weber J, et al: Visualization of dedicated catheters using fast scanning techniques with
potential for MR-guided vascular interventions. Magn Reson Med 36:816-820, 1996. Medline Similar articles
34. Bakker CJ, Hoogeveen RM, Hurtak WF, et al: MR-guided endovascular interventions: susceptibility-based catheter and
near-real-time imaging technique. Radiology 202:273-276, 1997. Medline Similar articles
35. Bakker CJ, Bos C, Weinmann HJ: Passive tracking of catheters and guidewires by contrast-enhanced MR fluoroscopy.
5 of 8 15-04-2010 21:22
36. Green JD, Omary RA, Finn JP, et al: Passive catheter tracking using MRI: comparison of conventional and magnetization-
prepared FLASH. J Magn Reson Imaging 16:104-109, 2002. Medline Similar articles
37. Buecker A, Neuerburg JM, Adam GB, et al: Real-time MR fluoroscopy for MR-guided iliac artery stent placement. J Magn
38. Dion YM, Ben El Kadi H, Boudoux C, et al: Endovascular procedures under near-real-time magnetic resonance imaging
guidance: an experimental feasibility study. J Vasc Surg 32:1006-1014, 2000. Medline Similar articles
39. Manke C, Nitz WR, Djavidani B, et al: MR imaging-guided stent placement in iliac arterial stenoses: a feasibility study.
41. Kuehne T, Saeed M, Higgins CB, et al: Endovascular stents in pulmonary valve and artery in swine: feasibility study of MR
imaging-guided deployment and postinterventional assessment. Radiology 226:475-481, 2003. Medline
Similar articles
42. Bartels LW, Bos C, van der Weide R, et al: Placement of an inferior vena cava filter in a pig guided by high-resolution MR
fluoroscopy at 1.5 T. J Magn Reson Imaging 12:599-605, 2000.
43. Bucker A, Neuerburg JM, Adam GB, et al: Real-time MR Guidance for inferior vena cava filter placement in an animal
model. J Vasc Intervent Radiol 12:753-756, 2001.
44. Hilfiker PR, Quick HH, Debatin JF: Plain and covered stent-grafts: in vitro evaluation of characteristics at three-dimensional
MR angiography. [comment]. Radiology 211:693-697, 1999. Medline Similar articles
45. Wang Y, Truong TN, Yen C, et al: Quantitative evaluation of susceptibility and shielding effects of nitinol, platinum, cobalt-
alloy, and stainless steel stents. Magn Reson Med 49:972-976, 2003. Medline Similar articles
46. Ladd ME, Quick HH, Debatin JF: Interventional MRA and intravascular imaging. J Magn Reson Imaging 12:534-546, 2000.
47. Konings MK, Bartels LW, Smits HF, Bakker CJ: Heating around intravascular guidewires by resonating RF waves. J Magn
48. Liu CY, Farahani K, Lu DS, et al: Safety of MRI-guided endovascular guidewire applications. J Magn Reson Imaging
49. Ladd ME, Quick HH: Reduction of resonant RF heating in intravascular catheters using coaxial chokes. Magn Reson Med
50. Dumoulin CL, Souza SP, Darrow RD: Real-time position monitoring of invasive devices using magnetic resonance. Magn
51. Leung DA, Debatin JF, Wildermuth S, et al: Intravascular MR tracking catheter: preliminary experimental evaluation. Am J
Roentgenol 164:1265-1270, 1995.
52. Wildermuth S, Debatin JF, Leung DA, et al: MR imaging-guided intravascular procedures: initial demonstration in a pig
model. Radiology 202:578-583, 1997. Medline Similar articles
53. Wildermuth S, Dumoulin CL, Pfammatter T, et al: MR-guided percutaneous angioplasty: assessment of tracking safety,
catheter handling and functionality. Cardiovasc Intervent Radiol 21:404-410, 1998. Medline Similar articles
54. Erhart P, Ladd ME, Steiner P, et al: Tissue-independent MR tracking of invasive devices with an internal signal source.
55. Zimmermann-Paul GG, Ladd ME, Pfammatter T, et al: MR versus fluoroscopic guidance of a catheter/guidewire system: in
vitro comparison of steerability. J Magn Reson Imaging 8:1177-1181, 1998. Medline Similar articles
56. Ocali O, Atalar E: Intravascular magnetic resonance imaging using a loopless catheter antenna. Magn Reson Med
57. Atalar E, Kraitchman DL, Carkhuff B, et al: Catheter-tracking FOV MR fluoroscopy. Magn Reson Med 40:865-872, 1998.
58. Yang X, Bolster BD Jr, Kraitchman DL, Atalar E: Intravascular MR-monitored balloon angioplasty: an in vivo feasibility
study. J Vasc Intervent Radiol 9:953-959, 1998.
59. Elgort DR, Wong EY, Hillenbrand CM, et al: Real-time catheter tracking and adaptive imaging. J Magn Reson Imaging
60. Guttman MA, Lederman RJ, Sorger JM, McVeigh ER: Real-time volume rendered MRI for interventional guidance. J
Cardiovasc Magn Reson 4:431-442, 2002. Medline Similar articles
61. Guttman MA, Kellman P, Dick AJ, et al: Real-time accelerated interactive MRI with adaptive TSENSE and UNFOLD. Magn
62. Serfaty JM, Yang X, Aksit P, et al: Toward MRI-guided coronary catheterization: visualization of guiding catheters,
guidewires, and anatomy in real time. J Magn Reson Imaging 12:590-594, 2000. Medline Similar articles
63. Serfaty JM, Yang X, Foo TK, et al: MRI-guided coronary catheterization and PTCA: a feasibility study on a dog model.
64. Yeung CJ, Atalar E: RF transmit power limit for the barewire loopless catheter antenna. J Magn Reson Imaging 12:86-91,
6 of 8 15-04-2010 21:22

65. Yeung CJ, Atalar E: A Green's function approach to local rf heating in interventional MRI. Med Phys 28:826-832, 2001.
66. Ladd ME, Erhart P, Debatin JF, et al: Guidewire antennas for MR fluoroscopy. Magn Reson Med 37:891-897, 1997.
67. Ladd ME, Zimmermann GG, Quick HH, et al: Active MR visualization of a vascular guidewire in vivo. J Magn Reson
68. Ladd ME, Zimmermann GG, McKinnon GC, et al: Visualization of vascular guidewires using MR tracking. J Magn Reson
69. Quick HH, Ladd ME, Nanz D, et al: Vascular stents as RF antennas for intravascular MR guidance and imaging. Magn
70. Quick HH, Kuehl H, Kaiser G, et al: Inductively coupled stent antennas in MRI. Magn Reson Med 48:781-790, 2002.
71. Glowinski A, Adam G, Bucker A, et al: Catheter visualization using locally induced, actively controlled field
inhomogeneities. Magn Reson Med 38:253-258, 1997. Medline Similar articles
72. Adam G, Glowinski A, Neuerburg J, et al: Visualization of MR-compatible catheters by electrically induced local field
inhomogeneities: evaluation in vivo. J Magn Reson Imaging 8:209-213, 1998. Medline Similar articles
73. Yang X, Atalar E: Intravascular MR imaging-guided balloon angioplasty with an MR imaging guide wire: feasibility study in
rabbits. Radiology 217:501-506, 2000. Medline Similar articles
74. Godart F, Beregi JP, Nicol L, et al: MR-guided balloon angioplasty of stenosed aorta: in vivo evaluation using
near-standard instruments and a passive tracking technique. J Magn Reson Imaging 12:639-644, 2000. Medline
Similar articles
75. Buecker A, Adam GB, Neuerburg JM, et al: Simultaneous real-time visualization of the catheter tip and vascular anatomy
for MR-guided PTA of iliac arteries in an animal model. J Magn Reson Imaging 16:201-208, 2002. Medline
Similar articles
76. Le Blanche AF, Rossert J, Wassef M, et al: MR-guided PTA in experimental bilateral rabbit renal artery stenosis and MR
angiography follow-up versus histomorphometry. Cardiovasc Intervent Radiol 23:368-374, 2000. Medline
Similar articles
77. Buecker A, Spuentrup E, Grabitz R, et al: Magnetic resonance-guided placement of atrial septal closure device in animal
model of patent foramen ovale. Circulation 106:511-515, 2002. Medline Similar articles
78. Schalla S, Saeed M, Higgins CB, et al: Magnetic resonance-guided cardiac catheterization in a swine model of atrial
septal defect. Circulation 108:1865-1870, 2003. Medline Similar articles
79. Kee ST, Rhee JS, Butts K, et al: 1999 Gary J. Becker Young Investigator Award. MR-guided transjugular portosystemic
shunt placement in a swine model. J Vasc Intervent Radiol 10:529-535, 1999. Medline Similar articles
80. Yang X, Atalar E, Li D, et al: Magnetic resonance imaging permits in vivo monitoring of catheter-based vascular gene
delivery. Circulation 104:1588-1590, 2001. Medline Similar articles
81. Bos C, Smits JH, Zijlstra JJ, et al: MRA of hemodialysis access grafts and fistulae using selective contrast injection and
flow interruption. Magn Reson Med 45:557-561, 2001. Medline Similar articles
82. Smits JH, Bos C, Elgersma OE, et al: Hemodialysis access imaging: comparison of flow-interrupted contrast-enhanced
MR angiography and digital subtraction angiography. Radiology 225:829-834, 2002. Medline Similar articles
83. Razavi R, Hill DL, van Vaals JJ, et al: Clinical MR guided cardiac catheterization. Proceedings of the International Society
of Magnetic Resonance in Medicine, Toronto, 2003, p 316.
84. Hofmann LV, Liddell RP, Arepally A, et al: In vivo intravascular MR imaging: transvenous technique for arterial wall imaging.
J Vasc Interv Radiol 14:1317-1327, 2003. Medline Similar articles
85. Yeung CJ, Susil RC, Atalar E: RF safety of wires in interventional MRI: using a safety index. Magn Reson Med 47:187-193,
86. Yeung CJ, Susil RC, Atalar E: RF heating due to conductive wires during MRI depends on the phase distribution of the
transmit field. Magn Reson Med 48:1096-1098, 2002. Medline Similar articles
87. Lederman RJ, Guttman MA, Peters DC, et al: Catheter-based endomyocardial injection with real-time magnetic resonance
imaging. Circulation 105:1282-1284, 2002. Medline Similar articles
88. Hill JM, Dick AJ, Raman VK, et al: Serial cardiac magnetic resonance imaging of injected mesenchymal stem cells.
89. Dick AJ, Guttman MA, Raman VK, et al: Magnetic resonance fluoroscopy allows targeted delivery of mesenchymal stem
cells to infarct borders in swine. Circulation 108:2899-2904, 2003. Medline Similar articles
90. Qiu B, Yeung CJ, Du X, et al: Development of an intravascular heating source using an MR imaging guidewire. J Magn
7 of 8 15-04-2010 21:22
© 2010 Elsevier
8 of 8 15-04-2010 21:22
CREENING AGNETIC ESONANCE MAGING

Susanne C. Ladd
Jörg F. Debatin
INTRODUCTION
History
Experience with preventive radiologic imaging, aiming at the detection of disease prior to its
symptomatic manifestation, is limited as the use of imaging in the radiologic practice is generally
focused on detecting and characterizing suspected or known disease in symptomatic patients. Imaging
tests can, however, play a pivotal role in prevention. Possibly the earliest imaging-based screening
program was started following the introduction in the 1930s of the mobile miniature-film apparatus by
Russell Reynolds and Watsons Ltd for tuberculosis.1,2 Mass radiography was performed for the early
diagnosis of pulmonary tuberculosis, which was important for detecting potentially infectious subjects
within the general population at risk. The program gained in importance when effective drug treatments
for tuberculosis were introduced in the 1950s.
Two decades later screening mammography was first advocated. While the exam has become well
established in the United States as well as several European countries,3,4 a heated debate continues
5
about benefits and risks of breast screening with mammography. Recently, the use of multislice
computed tomography (CT) has been suggested for preventive imaging. Driven by dramatic increases
in scanning speed, early manifestations of cardiovascular disease,6 as well as lung7-9 and colon
10
cancer, are being targeted with this technology. Recently, even elective full-body CT screening
based on contiguous 5 mm sections has become available for the health conscious in the United
States.11
These approaches are all burdened by considerable exposure to ionizing radiation. Associated dangers
have motivated the Federal Drug Administration (FDA) to issue "radiation alerts". 12 The European
Union prohibits the use of imaging techniques using ionizing radiation for screening purposes, with the
13,14
exception of mammography. Recognition thereof has focused attention on an imaging technique
devoid of ionizing radiation or other harmful side-effects: magnetic resonance imaging (MRI).15,16
This chapter discusses the definition of screening and prerequisites for cost-effective screening tests.
Furthermore, an overview of MR screening protocols for various diseases is given. In view of the lack
of hard data regarding the long-term outcome of MR-screened populations, the final recommendations
pertaining to screening MRI were carefully crafted.
page 561
page 562
Add to lightbox
Figure 21-1 Stages of prevention. Prevention is an expression that describes any attempt to lower
morbidity and mortality in the examined population at reasonable costs. Screening is a synonym for
detection of early-stage disease in the not yet symptomatic individual.
1 of 2 15-04-2010 21:23
© 2010 Elsevier
2 of 2 15-04-2010 21:23
SCREENING DEFINITIONS AND PREREQUISITES

A search of medical literature databases for "screening MRI" resulted in the citation of 100,000
studies. Detailed analysis, however, reveals that the term "screening" is used in a rather loose sense
(i.e., search for suspected disease in single individuals). This chapter discusses the definitions of
disease prevention and screening in conjunction with their prerequisites. Furthermore, a survey of
diseases potentially suited for screening is provided.
Types of Prevention
Disease prevention describes attempts to lower morbidity and mortality in an examined population at
reasonable costs (see Fig. 21-1). In this context, primary prevention describes the act of reducing risk
factors in populations with the aim of minimizing the occurrence of disease. The impact of primary
prevention has been known for more than 2000 years; also the ancient scholar Maimonides once said:
"Live sensibly-among a thousand people only one dies a natural death; the rest succumb to irrational
modes of living". Primary prevention thus is defined as the reduction of mortality via reduction in the
incidence of disease. Primary prevention is primarily performed by the communal authorities; examples
are addition of chloride to drinking water and legislation for safety belts in cars. Similarly, vaccinations
or the reduction of risk factors for cardiovascular disease, i.e., avoiding obesity, smoking, arterial
hypertension and hypercholesterolemia, represent forms of primary prevention.
Secondary prevention describes the search for occult disease. Thus, the glucose tolerance test is
available for early detection of diabetes mellitus and conventional mammography permits early
detection of breast carcinoma. The success of secondary prevention is predicated upon the availability
of effective treatment for the targeted disease, if detected at an early stage. The use of MRI for
screening generally falls into this category.
Tertiary prevention is used to avoid worsening of an existing, known disease or to reduce

complications of manifest disease. The use of beta receptor blockers, known to reduce mortality in
patients following myocardial infarction, or routine ophthalmologic exams for retinopathy in patients with
diabetes are good representatives of tertiary prevention.
Screening as referred to in this chapter generally refers to secondary prevention or, as in 'Whole-Body
Tumor Screening' on p 571, to tertiary prevention.
Prevalence and Suitability of Diseases

Disease Prerequisites
For screening to be cost-effective, the targeted disease must be sufficiently prevalent in the examined
group, at the time of the exam. Depending on the cost for a single screening exam, the prevalence of
disease in the screened group should be at least as high as 5-10%. The inherently low prevalence of
most diseases leads to a relatively low positive predictive value for the screening tests, even if the
test's specificity is high (Table 21-1; Fig. 21-2). This means that many subjects who do not suffer from
that disease have to be examined. The prevalence of disease can be enlarged if risk factors can be
defined that make a disease in a subgroup more likely, such as the risk factor "age" for breast cancer
or for colorectal cancer.
Table 21-1. Calculation of Values Necessary to Determine the Accuracy of a

Diagnostic Test*
Disease
Test
result Existing Not existing Total
Positive (a) Tp 27 (b) Fp 35 62 PV+ = a/(a + b) =
27/62 = 44%
1 of 4 15-04-2010 21:24
Negative (c) Fn 10 (d) Tn 77 87 PV- = d/(c + d) = 77/87

= 89%
Total 37 112 149
Se = a/(a + c) = 27/37 Sp = d/(b + d) = 77/112
= 73% = 69%
page 562
page 563
*Values added to allow better visualization.

Tp, true positives;Tn, true negatives; Fn, false negatives; Fp, false positives; Se, sensitivity; Sp,
specificity; PV+, positive predictive value; PV-, negative predictive value; PV+ is the probability that a
candidate with a positive test result has the disease.
+ +
According to the Bayes theorem, PV = (Se*Prev)/((SE*PREV) + (1-SP)*(1-PREV); thus PV
depends on the prevalence of the target disease.
Add to lightbox
Figure 21-2 Theoretical correlation between positive predictive value (PPV) and prevalence for four
tests with different sensitivities (Se) and specificities (Sp) according to the Bayes theorem. Note that
in the lower range of prevalence (as in real disease prevalences) a higher test accuracy does not
increase PPV as much as a higher prevalence would do.
Second, the disease must have high morbidity or mortality if it remains untreated, or if it becomes
treated only at a late stage. Otherwise, early detection of disease would lead to no change in quality
of life. Finally, the disease must have good therapeutic options when treated at an early stage.
Colonic Carcinoma
Colorectal cancer (CRC) is an excellent candidate for screening: high prevalence (approximately 6% of
17
the general population will develop CRC during their lifetime ), lethal if detected late and curable if
diagnosed early. In view of these "ideal" characteristics, CRC has been a focus of many screening
efforts for quite some time. However, despite these efforts, its incidence continues to increase, with
more than 130,000 newly diagnosed patients and 50,000 deaths annually in the United States alone.18
The biology of colorectal cancer, evolving from a precancerous colonic polyp to carcinoma over a
considerable time span,19 has elevated colorectal polyp screening, with subsequent endoscopic
20
polypectomy, to one of the most promising preventive measures in medicine. Poor patient
acceptance due to procedural pain and discomfort in conjunction with the need for bowel cleansing
have limited the impact of colonic screening to date.
2 of 4 15-04-2010 21:24
Cardiovascular Disease
21
Cardiovascular disease is the leading cause of mortality in Western societies. While known risk
factors are readily identifiable by a combination of physical exam, laboratory analysis, and patient
history, MR imaging offers a unique opportunity to assess what damage, if any, has already been
inflicted upon the cardiovascular system. Further reduction of risk factors, as well as minimally invasive
therapies, are treatment options in early disease of peripheral or coronary artery stenosis.
Bronchial Carcinoma
The recent focus on screening for bronchial carcinoma recognizes the far better outcome for smaller
tumors with low T stages. Due to its high spatial resolution, CT represents the first choice for
pulmonary screening. Associated exposure to ionizing radiation has prevented its wider use even in
high-risk populations.
Metastases/Primary Tumors
The use of "screening" in oncology describes the search for metastases in patients with known primary
tumors. The metastases themselves are clinically nonsymptomatic (or suggested only by reduction of
health status in general). In the sense of screening, this subgroup of patients with high tumor stage T
represent those with a high prevalence of metastatic disease if compared to all patients with this kind
of tumor or to the global population. The cost-effectiveness of screening for metastatic disease is not
well known for all tumor types but as this subgroup of patients is relatively small, there is broad clinical
consensus about the need for and effectiveness of screening, especially as it might alter therapy from
surgery to nonsurgical adjuvant or palliative therapy.
Cost and Safety of the Screening Test

The ideal screening test should be widely available and require only minimal resources for completion
and interpretation. Screening costs are determined by the direct cost of the screening test itself as
well as indirect follow-up costs for potential successive tests. Hence, sensitivity, specificity, and
predictive values of the underlying test vastly influence total cost.
While it is ethically justified to accept risks associated with diagnostic tests in patients with specific
complaints or known disease, this is not the case for screening presumably healthy individuals. Lack of
harmful side-effects thus represents a most important requirement for effective screening.
Patient Acceptance
The impact of patient acceptance on the success of screening tests can be illuminated for the case of
cervical cancer. Women with the highest risk for cervical cancer are those most likely not to participate
in screening tests; thus, these women will least likely be diagnosed with early cervical cancer. But
acceptance by clinicians is also a criterion for an effective screening test: on many occasions a
screening test may not be performed, even if it is useful, because the clinician regards it as too
time-consuming or troublesome.
page 563
page 564
Psychological Impact: "Labeling" in Single Patients

Test results might have an important effect on the psyche of patients. A "positive labeling" can be the
result of a negative screening test result: "This means that I can go to work for at least one more
year". The patient's attitude towards work and other daily duties is enhanced. On the other hand, a
positive test result can lead to a "negative labeling": women with false-positive mammography results
will have fear of mammography and carcinophobia for many months or possibly for life.22 Negative
labeling is especially problematic from an ethical point of view, as it can lead to a sensation of threat
instead of better health status.
Screening in the Radiologic Practice
3 of 4 15-04-2010 21:24
Screening attempts to lower morbidity and mortality in a (well-defined) group of subjects rather than
examining and screening for an existing disease in one single individual. Based on this definition,
screening exams remain the exception in clinical radiology today. Mammography screening in the
subgroup of elderly women is one of these exceptions.
Increased availability coupled with vast reductions in data acquisition speed have resulted in a more
liberal use of cross-sectional imaging techniques such as CT and MRI. The total noninvasiveness of the
MR experiment, which also does not rely on exposure to ionizing radiation, has stirred interest in the
use of this imaging technique for the purpose of secondary prevention. Largely, this development is
limited to a few individuals requesting such an exam at a few institutions. This type of "screening MRI"
generally does not do justice to the act of screening in its epidemiologic sense. As we will see,
screening with MRI is only just beginning to enter into radiologic practice and larger screening studies
are still the exception.
© 2010 Elsevier
4 of 4 15-04-2010 21:24
WHY SCREENING WITH MAGNETIC RESONANCE IMAGING?

MRI has emerged as the imaging modality of choice in the evaluation of many organs in the routine
clinical setting. Based on its versatility, MRI is being employed for the assessment of virtually all organ
systems; hence, it can be used for many screening purposes. Compared to the radiation exposure
caused by CT, public health concerns, associated with MRI are minimal. Thus, exposure to magnetic
resonance as a patient has never been associated with any harmful side-effects.23 Side-effects may,
however, be associated with the administration of paramagnetic contrast agents, which must be
considered an integral part of the proposed exam. Although rare, anaphylactoid reactions do occur.
Hence individuals need to be monitored during the examination procedure. On the other hand,
24-28
nephrotoxicity, a worry with iodinated contrast agents, is of no concern. This low profile of
side-effects has led to a rising acceptance in the general population and MRI is increasingly of interest
to healthy persons for screening purposes.
MRI depicts malignant disease as well as vascular disease with high accuracy. Recent developments
in hard- and software put at the radiologist's disposal new MR sequences, which are characterized by
robustness and rapid data acquisition, as well as high temporal and spatial resolution. MRI today no
longer suffers from heterogeneous image quality, as was the case not even 5 years ago. Breath-hold
techniques and navigator-assisted acquisitions, as well as optimized contrast enhancement, have made
MR quality comparable to CT. In addition, MR offers higher tissue contrast, arbitrary scan plane
selection, and faster cardiac triggering. These favorable "imaging" attributes translate into the ability to
depict pathomorphologic changes more comprehensively at earlier stages. Although MR seems well
suited for screening numerous diseases, a review of the literature reveals the utilization of MR for only
a few screening conditions, namely primarily breast cancer and colonic cancer.
MRI appears ideally suited for screening, as it overcomes many limitations inherent to the existing
image-based screening methods. Lack of ionizing radiation, contrast agents void of any nephrotoxicity,
16
and no other harmful side-effects are combined with high diagnostic accuracy based on unsurpassed
soft-tissue contrast, as well as high spatial and temporal resolution. These features inherent to the MR
examination result in high patient acceptance and the ability to perform the exam without special
patient preparation on an outpatient basis. Hence MRI is a natural candidate for preventive imaging. To
date, cost concerns and lengthy data acquisition times have prohibited its use in this regard. However,
recent hard- and software developments have laid the foundation for substantial time and cost savings
in the single MR examination (see Chapter 8), and, as will be discussed on p 570, multiorgan and
multidisease examinations in a single step have become possible. The next two sections of this chapter
discuss the MR protocols, available results, and problems with single- or multiorgan MR screening.
© 2010 Elsevier
1 of 1 15-04-2010 21:26
MAGNETIC RESONANCE SCREENING TODAY: INDICATIONS AND METHODS

To date, mainly organ-specific MR-based screening strategies have been pursued. These are
discussed below.
Magnetic Resonance Mammography

page 564
page 565
Breast carcinoma is the most common malignant tumor in women in Western countries.29 Incidence
and mortality keep increasing and the proportion of younger women rises. Despite all progress in the
development of new therapeutic strategies, the prognosis is first of all determined by the time point of
30
diagnosis, i.e. by the tumor stage. This is the rationale for offering conventional mammography to
women above the age of 50 and, in some countries, even above the age of 40. Unfortunately,
conventional mammography is characterized by relatively poor sensitivity and specificity and also is
burdened by considerable exposure to ionizing radiation.
MR mammography (MRM) has been available since 1983. A number of years ago, MRI was shown to
be more sensitive than conventional mammography for breast cancer detection, particularly in the
31,32
presence of dense breast tissue or breast implants. The diagnosis of malignant breast disease
mainly relies on contrast-enhanced fast dynamic acquisition of 3D spoiled gradient-echo (FLASH)
sequences, which provide information about in- and outflow of intravenously administered contrast
agent (see Chapter 27).
MRM has quickly established itself in the clinical arena.33 Not only does it offer higher sensitivities
compared to conventional mammography; it also seems to be effective in identifying genetically
determined breast cancer, which poses a problem in conventional examinations due to the dense
breast tissue in affected young women. Recent studies have shown MRM to be effective for screening
in women suspected to be carriers of the breast cancer susceptibility gene.34,93,94 Comparative studies
have shown the impact in high-risk patients in comparison to ultrasound and conventional
mammography, with sensitivity values of mammography, ultrasound, and MRM of 33%, 33%, and
100% and specificities of 93%, 80%, and 95%, respectively.34,35
However, despite considerable efforts to optimize the technique, MR mammography has remained
burdened by poor specificity (i.e., benign lesions might be misinterpreted as malignant). Differentiation
between breast cancer and fibroadenomas is frequently not possible, regardless of whether the
distinction is based on quantitative or qualitative criteria. Furthermore, the inability of MRI to detect
microcalcifications hampered its ability to detect ductal cancer in situ.36 The low specificity will most
likely be the main problem with MRM in forthcoming years. The negative impact on the positive
predictive value can be partially overcome if only patients at high risk for breast cancer are examined.
Despite these limitations, MR mammography continues to be proposed and evaluated as a technique
for breast screening but long-term results concerning the socio-economic impact of MRM screening
are not available to date.
Magnetic Resonance Colonography

Insufficient diagnostic accuracy and/or poor patient acceptance characterize most available colorectal
screening modalities, including testing for occult fecal blood, conventional colonoscopy or the double-
contrast barium enema.37,38 Virtual colonography (VC), based on 3D-CT or -MR data sets, has been
found to be highly sensitive for detecting clinically relevant colorectal polyps exceeding 8 mm in
39,40
size. Although CT colonography has some advantages regarding spatial resolution, examination
cost, and scanner availability, the lack of harmful side-effects, including ionizing radiation and high
soft-tissue contrast, renders MRI attractive as a possible alternative imaging modality for colorectal
screening.
MR colonography (MRC) overcomes many of the shortcomings limiting the clinical impact of existing
1 of 11 15-04-2010 21:27
screening techniques, including the gold standard "conventional colonoscopy". Patients undergo a
bowel cleansing procedure the day prior to MRC. Immediately before the examination, a water enema
of 2000-2500 mL of tap water will be applied, rendering the colonic lumen dark in T1-weighted images.
IV scopolamine or glucagon reduces bowel peristalsis, which makes the enema tolerable and at the
same time enhances image quality. A 3D FLASH data set of 96 coronal sections is then acquired
before as well as 60 and 90 seconds following the intravenous administration of paramagnetic
contrast. The data acquisition time for each 3D data set lies below 25 seconds to enable data
collection within a single breath-hold.
The colonic wall as well as colorectal masses41 will be depicted by a contrast agent uptake; remaining
stool is differentiated from polyps by lack of enhancement. The technique has been shown to be both
sensitive and specific regarding the detection of colorectal masses. 42 The diagnostic performance of
MR colonography has been assessed in several studies43,44 using conventional colonoscopy as the
standard of reference. While most mass lesions smaller than 5 mm in size were missed,43 almost all
lesions exceeding 10 mm were correctly identified. In a study by Pappalardo et al, 40 MR colonography
even detected a higher total number of polyps exceeding 10 mm in size than conventional colonoscopy.
MRC identified additional polyps in regions of the colon not reached by colonoscopy.
High patient acceptance of the exam is assured by lack of procedural pain and the prospect of fecal
tagging, which has been shown to successfully eliminate the need for colonic cleansing.45 For fecal
tagging, a barium sulfate-containing contrast agent (Micropaque; Guerbet, Sulzbach, Germany; 1 mg
barium sulfate/mL) is administered at a volume of 200 mL with each of four principal meals, beginning
46
36 hours prior to MR colonography. "Barium-based" fecal tagging renders stool dark and thus makes
it virtually indistinguishable from the water enema administered to distend the colon. Following the
intravenous administration of contrast both the colonic wall and colorectal masses enhance avidly,
while the colonic lumen filled with stool and water remains dark (Fig. 21-3).
Barium-tagged MR colonography detected all polyps larger than 8 mm in a population of 24 patients

45
with known or suspected colorectal tumors. Overall sensitivity of MR colonography amounted to
89.3% for the detection of colorectal masses and specificity was 100%. Although further work is
required to confirm these excellent results, it seems that barium-tagged MR colonography has vast
potential as the examination strategy of choice for the early detection of polyps in asymptomatic
subjects. The technique combines excellent diagnostic accuracy with high patient acceptance based on
a painless exam and no need for colonic cleansing.
page 565
page 566
2 of 11 15-04-2010 21:27
Add to lightbox
Add to lightbox
Figure 21-3 MR colonography without (A) and with (B) fecal tagging with oral barium. Contrast-
enhanced 3D gradient-echo source images of MR colonography after 2000 mL water enema. Native
stool has an inherently high signal in T1 weighting (A); after fecal tagging with barium, stool signal
almost resembles that of the water enema and becomes "transparent" (B).
Direct observational data on growth rates indicate that polyps smaller than 10 mm remain stable over
3 years and are not prone to malignant degeneration.47 Hence, MRC may be considered as reliable as
conventional colonoscopy regarding the assessment of colonic lesions at risk for malignant
44,47
degeneration. Costs for MR colonography are comparable to those of colonoscopy. The major
advantage of MRC with respect to conventional colonoscopy probably lies in the extended view beyond
the interior colonic wall; thus, MRC has the ability to simultaneously detect extraintestinal lesions
affecting the parenchymal abdominal organs, representing a considerable advantage over conventional
colonoscopy.48
3 of 11 15-04-2010 21:27
Magnetic Resonance Angiography

The management of a patient with arterial occlusive disease has to be planned in the context of the
epidemiology of the disease and, in particular, the apparent risk factors or markers predicting
spontaneous deterioration.49 It is obvious that proper management of arterial disease requires a
comprehensive assessment of the underlying vascular morphology.
In recent decades, invasive catheter angiography in digital subtraction technique (DSA) has constituted
the standard of reference for the diagnosis of arterial pathologies. As DSA is invasive, expensive, and
not without risks, alternative noninvasive techniques have entered into clinical routine. Ultrasound
successfully depicts morphology of some vascular territories, particularly the carotid arteries.
Limitations occur in more deeply localized arteries. Also, computed tomography angiography (CTA) is
used for detection of vascular pathologies.50 Although abdominal and pelvic vessels in particular are
imaged with high quality, the need for ionizing radiation makes it less suitable for screening purposes.
Furthermore, CTA relies on the use of potentially nephrotoxic contrast agents.
Compared to catheter-based angiography, MR angiography (MRA) (see Chapter 30) has been shown
51 52
to be almost equivalent in virtually all territories including the carotid, the renal, and the peripheral
arteries.53 Parenchymal enhancement and MR contrast dose limitations had initially curtailed contrast-
enhanced 3D MRA to the display of relatively small arterial territories contained within a single field-
of-view extending over 40-48 cm. The implementation of "bolus chase" techniques extended coverage
to encompass the entire run-off vasculature, including the pelvic, femoral, popliteal, and trifurcation
54-56
arteries. The implementation of faster gradient systems has laid the foundation for a further
extension of the bolus chase technique: whole-body coverage extending from the carotid arteries to the
trifurcation vessels with 3D MRA has become possible in a mere 72 seconds. 57 The whole-body MRA
concept is based on the acquisition of five slightly overlapping 3D data sets obtained in immediate
succession (Fig. 21-4). After administration of a bolus of paramagnetic contrast, coronal T1-weighted
3D FLASH data sets are collected using the lowest achievable TR and TE, permitting the acquisition of
64 sections within 12 seconds. The contrast bolus is chased from the carotid arteries to the ankles,
leading to optimal arterial contrast of the vessel lumen.
page 566
page 567
4 of 11 15-04-2010 21:27
5 of 11 15-04-2010 21:27
Add to lightbox
page 567
page 568
Figure 21-4 Contrast-enhanced whole-body MR angiography; maximum-intensity projections of all five
coronal 3D data sets. Right renal artery stenosis and aortic atherosclerosis. Incidental finding of
varicosity of the left lower leg.
Add to lightbox
Figure 21-5 Medium- to high-grade stenosis of the left proximal carotid artery (arrow). Maximum-
intensity projection of 64 coronal contrast-enhanced 3D gradient-echo sections.
The first data set covers the aortic arch, supra-aortic branch arteries (Fig. 21-5) and the thoracic
aorta, while the second data set covers the abdominal aorta, with its major branches including the
renal arteries (Fig. 21-6). The third data set displays the pelvic arteries and the last two data sets
cover the arteries of the thighs and calves, respectively. Correlation with a limited number of regional
DSA examinations revealed the diagnostic performance of whole-body MRA to be sufficient to warrant
its consideration as a noninvasive alternative to DSA. The performance of whole-body MRA was
further improved with the introduction of AngioSURF (MR-Innovation GmbH, Essen, Germany), which
integrates the torso-surface coil for signal reception. Use of the surface coil results in higher signal-
to-noise and contrast-to-noise values, translating into sensitivity and specificity values of 95.3% and
95.2%, respectively, for the detection of significant stenoses (luminal narrowing >50%) in lower
extremity peripheral vascular disease (PVD).58
59
In a series of 100 consecutive patients with PVD referred for MRA of the peripheral arteries, the
applied whole-body AngioSURF exam revealed additional clinically relevant disease in 25 patients (33
segments): renal artery narrowing (n = 15), carotid arterial stenosis (n = 12), subclavian artery
stenosis (n = 2), and abdominal aortic aneurysm (n = 4). The high degree of concomitant arterial
disease in patients with peripheral vascular disease merely underscores the systemic nature of
6 of 11 15-04-2010 21:27
atherosclerosis.
Cardiac Magnetic Resonance Imaging
Add to lightbox
Figure 21-6 Renal artery stenosis on the left side (arrow). Partial view of a maximum-intensity
projection from a coronal 3D data set of whole-body MRA.
Peripheral vascular disease, due to atherosclerosis, is rarely isolated. Patients with intermittent
claudication are at particularly high risk of atherosclerotic disease affecting other parts of the
circulation. Thus, it is important to recognize the extent of co-existing cardiovascular disease. Studies
on the prevalence of coronary artery disease (CAD) in patients with PVD show that history, clinical
examination, and electrocardiography typically indicate the presence of CAD in only 40% to 60% of
such patients. Furthermore, CAD may often be asymptomatic, as it is masked by exercise restrictions
in these patients (due to arterial insufficiency).60,61 Noninvasive techniques such as ECG suffer from
low sensitivity; invasive techniques such as catheter coronary angiography are not justified if a
subgroup of individuals with a low risk for CAD is examined.
Cardiac MRI (see Chapters 32-38) permits the evaluation of regional and global myocardial
contractibility and valvular function,62 as well as myocardial viability and perfusion.63-65 Recently, many
66
efforts have been focused on the early detection of ischemic heart disease. Thus CINE techniques
permit ready evaluation of ventricular wall motion,67-69 at rest as well as under stress conditions.
Furthermore, myocardial perfusion can be displayed with dynamic MRI protocols. The identification of
"late enhancement" regions provides accurate data about the presence of infarcted myocardium.
Based on delayed enhancement, infarcted myocardium is detected on T1-weighted images with high
70-72
sensitivity and specificity.
page 568
page 569
Only the lumen of the coronary arteries remains unassessed by cardiac MRI. This explains the
relatively low sensitivity of today's MR heart examinations for CAD. While several techniques for the
visualization of the coronary arteries have been proposed, none has gained clinical relevance. Although
the ability to analyze myocardial viability in combination with ventricular function reduces the impact of
this deficit, the incorporation of coronary MRA into a future screening protocol remains highly
desirable. Until that time, it seems crucial that some form of cardiac evaluation under stress conditions
7 of 11 15-04-2010 21:27
is performed.
Magnetic Resonance Imaging of the Right Ventricle

The suspicion of right ventricular dysplasia (RVD) is usually triggered by an episode of ventricular
tachycardia. Since RVD seems to be associated with a genetic component, screening family members
not yet affected by RVD has been suggested.
MR allows the identification of fatty tissue in the right ventricular myocardium with ECG-triggered
non-fat saturated spin-echo sequences. Furthermore, CINE techniques permit a comprehensive
functional assessment of right ventricular function. The ability to characterize the right ventricle has
motivated investigators to assess the ability of MRI to detect and screen for right ventricular
73
dysplasia. Right ventricular abnormalities in asymptomatic subjects at risk showed good correlation
with evoked potentials.74
Fetal Magnetic Resonance Imaging

Less common suggestions for MR screening focus on fetal imaging. Fetal MRI is based on fast
sequences, as motion artifacts pose the largest hindrance to imaging of the small, water-embedded
structures (see Chapter 94). Today HASTE imaging collected in different planes has emerged as the
technique of choice. MR has been shown to detect and characterize cerebrospinal malformations with
high accuracy. Furthermore, fetal brain oxygenation in pregnancies at risk (placentar insufficiency) can
be studied with fetal MRI.75
Magnetic Resonance Imaging of the Prostate

Today, diagnosis of prostatic carcinoma is based on clinical (PSA, digital examination) examinations
and, if necessary, transrectal ultrasound and biopsy. MRI has also been suggested for prostate
screening, as it offers high-resolution images of the pelvic region (see Chapter 89).
MR imaging of the prostate has become successful with the introduction of endorectal coils, which
provide high-quality images of the prostatic morphology. MR protocols make use of the low signal
intensity of prostatic carcinomas, which can best be depicted in non-fat saturated turbo spin-echo
sequences. They also aid diagnosis of tumor invasion into neighboring structures such as the
neurovascular bundle or the seminal vesicles. Recent developments in surface array coils seem likely
to replace the invasive and unattractive transrectal coils, as they offer sufficient signal. But after weak
results with respect to PSA in combination with digital examination, MR of the prostate is today limited
to staging purposes in men with positive tumor markers.76
Magnetic Resonance Imaging of the Lung

Recently, low-dose spiral CT scanning has been demonstrated to detect lung cancer at a preclinical
stage, when surgical resection is still possible.7 While these data are promising, there are no reports
that have demonstrated a definitive reduction in all-cause mortality from any lung cancer screening
program.
For a long time, MR imaging of the lung (see Chapter 76) had been handicapped by susceptibility
effects at the interfaces between the pulmonary interstitium and the air-filled alveoli. These artifacts
can be overcome by using ultra-short echo times. Based on appropriate sequences, the accuracy of
MRI regarding the detection of pulmonary lesions exceeding 10 mm in size has been shown to be quite
high.77-79 In a study involving 30 patients with known pulmonary masses, axial HASTE images
demonstrated 1032 of 1102 lesions seen on computed tomography.80 Reflecting the lack of signal-
providing protons, smaller calcified nodules were missed. Furthermore, lesions with a diameter of less
than 3 mm were also missed. This might explain the low number of pulmonary lesions found in studies
which cover the lungs but could on the other hand increase specificity for malignant pulmonary lesions.
Cerebral Magnetic Resonance Imaging
8 of 11 15-04-2010 21:27
Due to the systemic nature of vascular disease, an MR protocol for the examination of the vascular
system also requires high sensitivity with respect to cerebrovascular disease (CVD); in up to 80% of all
cases, CVD is based on insufficient perfusion. The link between PVD and cerebrovascular disease
(CVD) seems to be weaker than that with CAD. Using duplex sonography, carotid disease has been
found in 26% to 50% of patients with PVD.81,82 Most of these patients will have a history of cerebral
83
events or a carotid bruit and seem to be at increased risk of further events.
The size, number, and distribution of ischemic cerebral regions permit consideration of possible
etiologies: thus microangiopathic changes of the cerebral white matter are highly suggestive of
hypertension,84 while thromboembolic changes are most frequently induced by high-grade carotid
disease. Duplex ultrasound, the screening method of choice, does not cover the whole anatomy;
especially the parts of the carotids near the skull-base cannot be assessed.
Among the organs that are examined with great success with MRI, the central nervous system
features prominently: inflammatory, neoplastic, and vascular disease are reliably detected or
excluded.85,86 Protocols consisting of morphologic imaging (T2-weighted fast spin-echo imaging,
T1-weighted imaging, blood oxygen level-dependent imaging, and diffusion-weighted imaging; Fig.
21-7) as well as vascular imaging (3D time-of-flight (TOF) arterial imaging) have been applied. As
screening also must consider rare but significant cerebral tumors, a contrast-enhanced cerebral
imaging sequence should be part of any screening protocol.
page 569
page 570
Add to lightbox
9 of 11 15-04-2010 21:27
Add to lightbox
Add to lightbox
10 of 11 15-04-2010 21:27
Add to lightbox
Figure 21-7 Microangiopathic changes in the white periventricular matter. T1-weighted SE (A), FLAIR
(B), T2-weighted FSE (C), contrast-enhanced T1-weighted gradient-echo image (D).
© 2010 Elsevier
11 of 11 15-04-2010 21:27
COMPREHENSIVE MAGNETIC RESONANCE SCREENING PROTOCOLS

Cost concerns and lengthy data acquisition times have to date limited the use of MR for preventive
imaging. Vastly improved scanner performance in conjunction with new pulse sequence designs have
led to dramatic reductions in examination times. Hence examination strategies covering more than one
body region and more than one organ system have become possible. This ability opens up new
perspectives for developing cost-effective MR-based screening strategies. Two approaches targeting
different subjects are illustrated below.
Screening the Healthy Subject

page 570
page 571
A comprehensive MR protocol permitting the detection of cardiovascular, peripheral vascular and

cerebrovascular disease as well as lung and colon cancer has been developed. Technical parameters
have been optimized to permit completion of the exam within 1 hour. A pilot study87 documents the
potential findings in a screening environment.
The MR exam can be subdivided into four parts.
1. The cerebrum is assessed by fast T1- and T2-weighted spin-echo sequences, as well as
diffusion-weighted imaging. The intracerebral arterial system is directly visualized by axial 3D
TOF MR angiography. At a later time point, a gradient-echo sequence is performed, making
use of the IV contrast agent, which had been administered for MR angiography; this gives the
important contrast-enhanced images of the brain.
2. To enable whole-body coverage, subjects are examined on a fully MR-compatible rolling table
platform (AngioSURF, MR-Innovation GmbH, Essen, Germany) placed on a 1.5 T system
(Magnetom Sonata®, Siemens Medical Solutions, Erlangen, Germany).58 This device permits
the collection of up to six 3D data sets with a craniocaudal coverage of 380 mm each
(acquisition time 12 seconds each) in immediate succession. Markers permit adjustment to the
desired field-of-view. For whole-body MR angiography, subjects are placed feet first within the
bore of the magnet. A 2 cm overlap at each station's end results in a craniocaudal coverage of
174 cm. Data acquisition is completed in only 72 seconds. Signal reception is accomplished
using posteriorly located spine coils and an anteriorly placed torso phased array coil which rests
in a height-adjustable coil holder. Thus, data for all five or six stations are collected with the
same stationary coil set positioned in the isocenter of the magnet.
3. Cardiac morphology as well as the pulmonary parenchyma are assessed with axial HASTE
images. Subsequent functional assessment of the heart is based on segmented steady-state
free precession cine measurements (TrueFISP) along the long and short axis, as well as along
the left ventricular outflow tract. A 3D segmented inversion recovery turbo gradient-echo
sequence is used to screen for areas of "late enhancement", which denote myocardial
infarction.
4. For MR colonography subjects undergo standard preparation for bowel cleansing on the
previous day: 40 mg of scopolamine is administered intraveneously, to minimize peristaltic
bowel motion. The colon is filled with 1500-2500 mL of warm tap water via a rectal enema tube.
Following the collection of a "precontrast" T1-weighted 3D gradient-echo data set,
paramagnetic contrast is administered (0.1 mmol/kg). After a delay of 60 and 90 s respectively,
the 3D acquisition is repeated with breath-holding over 23 seconds.
Paramagnetic contrast is administered intravenously on two occasions: once for imaging of the
arterial vascular tree (0.2 mmol/kg bodyweight) and a second time for MR colonography (0.1
mmol/kg bodyweight). The total dose amounts to 0.3 mmol/kg bodyweight.
We can today report on 518 mainly healthy subjects who have undergone this combined MR screening
protocol. Their ages ranged from 33 to 77 years, with a mean of 51.4 years; 23.7% showed signs of
arterial disease in the cerebrum, heart or peripheral vascular tree, or colonic polyps. Relevant findings
were not limited to the target organs. Rather, there were a number of 'additional' relevant findings.
Analysis of the parenchymal organs in the abdomen can be based on 3D gradient-echo data sets
1 of 6 15-04-2010 21:28
collected in the arterial, portal venous, and hepatic venous phases. While the first arterial data set is
provided as part of the whole-body MRA exam, the subsequent data sets are collected for MR
colonography. Previous studies have shown this type of dynamic contrast-enhanced 3D imaging of the
abdomen to be very accurate for the identification of pathologies in the parenchymal organs. 88 Similar
42
experiences have been reported based on MR colonography data sets alone. Hence, it was not
surprising that the featured imaging protocol resulted in the identification of multiple "additional findings"
outside the target organs with considerable specificity, i.e., one renal cell carcinoma, disk hernias,
gastric hernias, two cerebral aneurysms, diseased cardiac valves (Fig. 21-8), thyroid nodules (Fig.
21-9), etc.
Whole-Body Tumor Screening

MRI has been shown to be useful in screening for metastatic disease or primary tumors in patients
with carcinomas of unknown primary. Two approaches will be discussed.
Whole-Body MR for Skeletal Metastasis

A couple of studies have demonstrated that STIR imaging in the coronal plane, repeated for four or
five body regions to cover from the head to the knees, results in higher sensitivities and accuracies
than conventional bone scintigraphy.89,90 For this purpose, 5-6 body regions are assessed via a set of
coronal STIR images each. Metastases exhibit a high signal with respect to unaffected dark bones. In
particular, metastases in the vertebral bodies, which are not easy to assess by bone scintigraphy, are
easier to depict by STIR MR.
Whole-Body MR for Detection of Soft-Tissue and Bone Metastases

Staging of malignant disease is crucial for treatment and diagnosis. As metastases can affect any
anatomic region and any kind of tissue, different techniques or multiple examinations of the same
technique have to be performed to cover the whole body. Currently, this is one of the main reasons for
prolonged hospital stay and high costs.
page 571
page 572
Add to lightbox
2 of 6 15-04-2010 21:28
Add to lightbox
Figure 21-8 CINE trueFISP images of the left ventricular outflow tract. Previously unknown combined
aortic valve disease (diastole with retrograde insufficiency jet, systole with stenosis jet over the aortic
valve).
Add to lightbox
3 of 6 15-04-2010 21:28
Add to lightbox
page 572
page 573
Figure 21-9 Thyroid nodules (arrows) as incidental finding in axial HASTE images of the lung (A) and a
coronal section of a 3D MRA data set (B).
Add to lightbox
Figure 21-10 MR protocol for whole-body tumor screening. After a native 3D FLASH data set (64 axial
sections in the abdomen), a single dose of paramagnetic contrast agent is administered intravenously
at 2 mL/s. Repetitive data sets are then acquired as shown to achieve arterial, portovenous, and
venous contrast phases within the upper abdomen. Within the remaining time intervals, brain, thorax,
pelvis, and femura are examined. Between the single acquisitions, which are performed within one
breath-hold, breathing is allowed for 8 seconds.
A more recent approach using a rolling table platform91 allows for the evaluation not only of the bones
(vertebral metastases are depicted with high sensitivity), but also the lung and the abdominal
parenchymal and lymphatic structures. Native (noncontrast enhanced) 3D FLASH (3D VIBE: Volumetric
Interpolated Breathhold Examination) sequences (3D GRE T1w-fat saturation, TR/TE = 3.1/1.2 ms, flip
4 of 6 15-04-2010 21:28
angle 12°, slab thickness 312 mm-104 partitions, slice thickness 3 mm, matrix 240 × 512, axial plane)
build the underlying MR protocol. After a single administration of paramagnetic contrast agent, this 3D
block is repeated eight times by rapidly moving the patient with the table platform from one body
region to the next (Fig. 21-10).
Add to lightbox
Add to lightbox
Figure 21-11 Brain metastasis in a 31-year-old male with seminoma. The metastasis is depicted
equally well with contrast-enhanced 3D VIBE (A) and contrast-enhanced CT (B).
The head, chest, and abdomen are scanned with the same sequence, which provides portal-venous
images of the liver and of the remaining body regions down to the knees. Finally, a venous data set of
the liver is obtained. The results were compared to CT or bone scintigraphy. All brain metastases (Fig.
21-11) seen on CT were correctly diagnosed with MR; furthermore, MR could detect four additional
bone metastases (in the spine and pelvis) if compared to scintigraphy. MR detected 19 of 21 lung
metastases seen by CT. With respect to skeletal scintigraphy, the vertebral and pelvic bones in
particular were better evaluated with respect to metastatic disease. As in-room time amounts to only
11 minutes, this MR approach could be an alternative to skeletal scintigraphy screening for
5 of 6 15-04-2010 21:28
metastases.91
A recent study compared positron emission tomography (PET)-CT with a similar whole-body MRI with
respect to staging accuracy in malignant disease, 92 rendering an accurate TNM stage for PET-CT in
77% and for MRI in 54% of the cases. In detecting distant metastases, both modalities performed
similarly (Fig. 21-12). MRI, as the more widely available modality, therefore could offer a valuable
one-modality alternative for tumor staging.
The results demonstrate that this whole-body protocol could be an all-encompassing alternative to
conventional multimodality tumor staging, permitting dynamic imaging of parenchymal organs in the
abdomen and covering all anatomic regions in rapid succession by using the rolling table platform.
© 2010 Elsevier
6 of 6 15-04-2010 21:28
DECISION STRATEGIES
page 573
page 574
Add to lightbox
Add to lightbox
Figure 21-12 Whole-body tumor imaging. Axial 3D gradient-echo source images after IV contrast
application. Necrotic pulmonary (A) and pelvic (B) metastasis of a thyroid carcinoma (arrows).
MR screening is still in its infancy. Positive and negative predictive values of MRI, however, cannot be
easily extrapolated from clinical data, as the prevalence of disease enters into these values. There
seems to be sufficient potential, however, to study MRI more carefully as a technique for secondary
and tertiary screening. Factors which need to be addressed include diagnostic accuracy of the
underlying MR exam components, the costs of MRI, availability of screening MRI for any target
population, the rate of false positives, the costs for follow-up tests and increased morbidity, the rate of
false negatives and the morbidity from false-negative MR examinations, and many others. Also, the
psychological effects of "whole-body screening" on the screened subjects must be studied, including
avoidance of further check-up examinations and the rate of psychological MR side-effects (Box 21-1).
Cost has to be weighed against the gain measured as a reduction of morbidity and mortality. The
epidemiologically most important unit is the gain in QUALY (quality of life-years), whose determination
depends on repeated detailed questionnaires.
Currently, the effect of mono- or multiorgan screening is not known; therefore, physicians should be
careful when discussing MR screening with interested patients. Patients need to be informed of
potential pitfalls, including follow-up costs and side-effects. Furthermore, information about alternative
screening tests, such as ultrasound or laboratory tests, should be provided.
1 of 2 15-04-2010 21:29
Box 21-1 Criteria for Determining Whether a Medical

Condition or Disease Should Be Included in Routine
Prevention Examinations
What is the impact of the disease/condition on the population with respect

to:
Death?
Disease?
Handicap?
Discomfort/pain?
Unhappiness?
Poverty/misery?
How accurate (sensitive/specific) is the screening test?
How easy is the screening test to perform?
How cost-effective is the screening test?
How safe is the screening test?
What is the level of patient acceptance of the screening test?
How effective is the therapeutic intervention in the primary prevention or
how effective is the therapeutic intervention in the secondary prevention?
What is the level of patient compliance?
How effective is early intervention compared to late intervention?
© 2010 Elsevier
2 of 2 15-04-2010 21:29
MAGNETIC RESONANCE SCREENING IN THE FUTURE

The ongoing development of MR scanner hardware will further facilitate whole-body MR imaging. Thus
the concept of whole-body comprehensive MR screening strategies is likely to remain a focus of active
research. Clearly, one of the big challenges in the screening context will be the depiction of the
coronary arteries. In the context of malignant disease, higher resolution imaging of the pulmonary
parenchyma will raise the diagnostic quality of pulmonary MRI. Colonic imaging with an almost 100%
accuracy for polyps larger than 10 mm seems to be sufficient for colonic carcinoma screening, as the
importance of smaller polyps for mortality is small.
Similar to all other screening tests, MRI will need to be subjected to prospective cohort studies to
provide a detailed investigation of parameters determining cost-effectiveness. Provided such studies
are performed, it seems likely that MRI will play a dominant role in "screening". This potential might
enhance the future of radiology beyond anyone's expectation.
REFERENCES
1. Cordes L, Heine F, Krickau G: Results of mass radiography for tuberculosis in older school children. Offentl Gesundheitswes
2. Lunn JA, Mayho V: Incidence of pulmonary tuberculosis by occupation of hospital employees in the National Health Service
in England and Wales 1980-84. J Soc Occup Med 39:30-32, 1989.
3. Olivotto IA, Kan L, d'Yachkova Y, et al: Ten years of breast screening in the Screening Mammography Program of British
Columbia, 1988-97. J Med Screen 7(3):152-159, 2000.
4. Nystrom L, Andersson I, Bjurstam N, et al: Long-term effects of mammography screening: updated overview of the Swedish
randomised trials. Lancet 359(9310):909-919, 2002.
5. Gotzsche PC, Olsen O: Is screening for breast cancer with mammography justifiable? Lancet 355(9198):129-134, 2000.
6. Kopp AF, Schroeder S, Baumbach A, et al: Non-invasive characterization of coronary lesion morphology and composition
by multislice CT: first results in comparison with intracoronary ultrasound. Eur Radiol 11(9):1607-1611, 2001.
page 574
page 575
7. Henschke CI, McCauley DI, Yankelevitz DF, et al: Early Lung Cancer Action Project: overall design and findings from
baseline screening. Lancet 354:99-105, 1999. Medline Similar articles
8. Diederich S, Wormanns D, Semik M, et al: Screening for early lung cancer with low-dose spiral CT: prevalence in 817
asymptomatic smokers. Radiology 222:773-778, 2002.
9. Ellis JR, Gleeson FV: New concepts in lung cancer screening. Curr Opin Pulm Med 8:270-274, 2002. Medline
Similar articles
10. Johnson CD, Dachman AH: CT colonography: the next colon screening examination? Radiology 216(2):331-341, 2000.
11. Elsberry RB: The invasion of the body scanners. Decis Imag Econ 42:18-21, 2002.
12. FDA: Reducing radiation risk from computed tomography for pediatric and small adult patients. In: Safety alerts, public
health advisories and notices from CDRH. FDA Notice 11/2/2001. Washington: Federal Drug Agency, 2001.
13. Strax P, Venet L, Shapiro S, et al: Mammography and clinical examination in mass screening for cancer of the breast.
Cancer 20:2184-2188, 1967. Medline Similar articles
14. EU Commission: European guidelines for screening mammography. Brussels: EU Commission (EUREF), 1997.
15. Haustein J, Laniado M, Niendorf HP, et al: Triple-dose versus standard-dose gadopentetate dimeglumine: a randomized
study in 199 patients. Radiology 186:855-860, 1993.
16. Haustein J, Niendorf HP, Krestin G, et al: Renal tolerance of gadolinium-DTPA/dimeglumine in patients with chronic renal
failure. Invest Radiol 27:153-156, 1992. Medline Similar articles
17. Neuhaus H: Screening for colorectal cancer in Germany: guidelines and reality. Endoscopy 31(6):468-470, 1999.
18. Landis SH, Murray T, Bodden S, Wingo PA: Cancer statistics, 1998. CA Cancer J Clin 48:6-29, 1998.
19. O'Brien MJ, Winawer SJ, Zauber AG, et al: The National Polyp Study. Patient and polyp characteristics associated with
high-grade dysplasia in colorectal adenomas. Gastroenterology 98:371-379, 1990. Medline Similar articles
20. Liebermann DA, Smith FW: Screening for colon malignancy with colonoscopy. Am J Gastroenterol 86:946-951, 1991.
21. Anderson KM, Wilson PWF, Odell PM, Kannel WB: An updated coronary risk profile: a statement for health professionals.
22. Lerman C, Trock Brimer BK, Boyce A, et al: Psychological and behavioral implications of abnormal mammograms. Ann
Intern Med 114:657-661, 1991. Medline Similar articles
23. Ahmed S, Shellock FG: Magnetic resonance imaging safety: implications for cardiovascular patients. J Cardiovasc Magn
1 of 4 15-04-2010 21:30
Reson 3:171-182, 2001. Medline Similar articles

24. Prince MR, Arnoldus C, Frisoli JK: Nephrotoxicity of high-dose gadolinium compared with iodinated contrast. J Magn
Reson Imag 6:162-166, 1996.
25. Shellock FG, Kanal E: Safety of magnetic resonance imaging contrast agents. J Magn Reson Imag 10:477-484, 1999.
26. Tombach B, Bremer C, Reimer P, et al: Pharmacokinetics of 1M gadobutrol in patients with chronic renal failure. Invest
Radiol 35(1):35-40, 2000.
27. Tombach B, Bremer C, Reimer P, et al: Renal tolerance of a neutral gadolinium chelate (gadobutrol) in patients with
chronic renal failure: results of a randomized study. Radiology 218:651-657, 2001. Medline Similar articles
28. Tombach B, Bremer C, Reimer P, et al: Using highly concentrated gadobutrol as an MR contrast agent in patients also
requiring hemodialysis: safety and dialysability. Am J Roentgenol 178:105-109, 2002.
29. La Veccia C, Luccini F, Negri E, et al: Trends in cancer mortality in europe, 1955-1989: III breast and genital sites. Eur J
Cancer 28:927-998, 1992.
30. Crowe JP, Gordon NH, Shenk RR, et al: Primary tumor size: relevance to breast cancer survival. Arch Surg 127:910-915,
31. Kawashima H, Matsui O, Suzuki M, et al: Breast cancer in dense breast: detection with contrast-enhanced dynamic MR
imaging. J Magn Reson Imag 11:233-243, 2000.
32. Herborn CU, Marincek B, Erfmann D, et al: Breast augmentation and reconstructive surgery: MR imaging of implant rupture
and malignancy. Eur Radiol 12:2198-2206, 2002. Medline Similar articles
33. Heywang SH, Wolf A, Pruss E, et al: MR imaging of the breast with Gd-DTPA: use and limitations. Radiology 171:95-103,
34. Kuhl CK, Schmutzler RK, Leutner CC, et al: Breast MR imaging screening in 192 women proved or suspected to be
carriers of a breast cancer susceptibility gene: preliminary results. Radiology 215:267-279, 2000.
35. Weinreb JC, Newstead G: MR imaging of the breast. Radiology 196:593-610, 1995. Medline Similar articles
36. Boetes C, Strijk SP, Holland R, et al: False-negative MR imaging of malignant breast tumors. Eur J Radiol 7:1231-1234,
1997.
37. Frommer DJ: What's new in colorectal cancer screening? J Gastroenterol Hepatol 13:528-533, 1998. Medline
Similar articles
38. Ahlquist DA, Wieland HS, Moertel CG, et al: Accuracy of fecal occult blood screening for colorectal neoplasia. A
prospective study using Hemoccult and HemoQuant tests. JAMA 269:1262-1267, 1993. Medline Similar articles
39. Fenlon HM, Nunes DP, Schroy PC, et al: A comparison of virtual and conventional colonoscopy for the detection of
colorectal polyps. N Engl J Med 341:1496-1503, 1999. Medline Similar articles
40. Pappalardo G, Polettini E, Frattaroli FM, et al: Magnetic resonance colonography versus conventional colonoscopy for the
detection of colonic endoluminal lesions. Gastroenterology 119:300-304, 2000. Medline Similar articles
41. Lauenstein TC, Debatin JF: Magnetic resonance colonography for colorectal cancer screening. Semin Ultrasound CT MR
42. Luboldt W, Bauerfeind P, Steiner P, et al: Preliminary assessment of three-dimensional magnetic resonance imaging for
various colonic disorder. Lancet 349:1288-1291, 1997. Medline Similar articles
43. Luboldt W, Bauerfeind P, Wildermuth S, et al: Colonic masses: detection with MR colonography. Radiology 216:383-388,
44. Saar B, Heverhagen JT, Obst T, et al: Magnetic resonance colonography and virtual magnetic resonance colonoscopy with
the 1.0-T system: a feasibility study. Invest Radiol 35:521-526, 2000.
45. Lauenstein TC, Goehde SC, Ruehm SG, et al: MR colonography with barium-based fecal tagging: initial clinical
experience. Radiology 223:248-254, 2002. Medline Similar articles
46. Lauenstein TC, Holtmann G, Schoenfelder D, et al: MR colonography without bowel cleansing: a new strategy to improve
patient acceptance. Am J Roentgenol (in press).
47. Villavicencio RT, Rex DX: Colonic adenomas: prevalence and incidence rates, growth rates, and miss rates at
colonoscopy. Semin Gastrointest Dis 11:185-193, 2000. Medline Similar articles
48. Debatin JF, Lubold W, Bauerfeind P: Virtual colonoscopy in 1999: computed tomography or magnetic resonance imaging?
Endoscopy 3:174-179, 1999.
49. Management of peripheral arterial disease (PAD). TransAtlantic Inter-Society Consensus (TASC). J Vasc Surg 31
(2)(suppl):5, 2000.
50. Smith PA, Fishman EK: Clinical integration of three-dimensional helical CT angiography into academic radiology: results of
a focused survey. Am J Roentgenol 173(2):445-447, 1999.
51. Nederkoorn PJ, Mali WP, Eikelboom BC, et al: Preoperative diagnosis of carotid artery stenosis: accuracy of noninvasive
testing. Stroke 33(8):2003-2008, 2002.
52. Korst MB, Joosten FB, Postma CT, et al: Accuracy of normal-dose contrast-enhanced MR angiography in assessing renal
artery stenosis and accessory renal arteries. Am J Roentgenol 174(3):629-634, 2000.
53. Hany TF, Debatin JF, Leung DA, Pfammatter T: Evaluation of the aortoiliac and renal arteries: comparison of breath-hold,
2 of 4 15-04-2010 21:30
contrast-enhanced, three-dimensional MR angiography with conventional catheter angiography. Radiology 204(2):357-362,

1997.
54. Meaney JF, Ridgway JP, Chakraverty S, et al: Stepping-table gadolinium-enhanced digital subtraction MR angiography of
the aorta and lower extremity arteries: preliminary experience. Radiology 211:59-67, 1999. Medline Similar articles
55. Ho KY, Leiner T, de Haan MW, et al: Peripheral vascular tree stenoses: evaluation with moving-bed infusion-tracking MR
angiography. Radiology 206:683-692, 1998. Medline Similar articles
56. Ruehm SG, Hany TF, Pfammatter T, et al: Pelvic and lower extremity arterial imaging: diagnostic performance of three-
dimensional contrast-enhanced MR angiography. Am J Roentgenol 174:1127-1135, 2000.
57. Goyen M, Quick HH, Debatin JF, et al: Whole-body three-dimensional MR angiography with a rolling table platform: initial
clinical experience. Radiology 224(1):270-277, 2002.
58. Goyen M, Quick HH, Debatin JF, et al: Whole body 3D MR angiography using a rolling table platform: initial clinical
59. Goyen M, Herborn CU, Kröger K, et al: Detection of atherosclerosis: systemic imaging for a systemic disease using whole
body 3D MR-angiography-initial experience. Radiology (in press).
60. von Kemp K, van den Brande P, Peterson T, et al: Screening for concomitant diseases in peripheral vascular patients.
Results of a systematic approach. Int Angiol 16:114-122, 1997. Medline Similar articles
61. Hertzer NR, Beven EG, Young JR, et al: Coronary artery disease in peripheral vascular patients. A classification of 1000
coronary angiograms and results of surgical management. Ann Surg 199:223-233, 1984.
62. Friedrich MG, Schulz-Menger J, Poetsch T, et al: Quantification of valvular aortic stenosis by magnetic resonance imaging.
Am Heart J 144:329-334, 2002. Medline Similar articles
63. Pereira RS, Wisenberg G, Prato FS, Yvorchuk K: Clinical assessment of myocardial viability using MRI during a constant
infusion of Gd-DTPA. MAGMA 11(3):104-113, 2000.
64. Hunold P, Brandt-Mainz K, Freudenberg L, et al: Evaluation of myocardial viability with contrast-enhanced magnetic
resonance imaging-comparison of the late enhancement technique with positron emission tomography. Rofo Fortschr Geb
Rontgenstr Neuen Bildgeb Verfahr 174(7):867-873, 2002.
65. Klein C, Nekolla SG, Bengel FM, et al: Assessment of myocardial viability with contrast-enhanced magnetic resonance
imaging: comparison with positron emission tomography. Circulation 105(2):162-167, 2002.
66. van der Wall EE, van Rugge FP, Vliegen HW, et al: Ischemic heart disease: value of MR techniques. Int J Card Imag
13:179-189, 1997.
67. Wilke NM, Jerosch-Herold M, Zenovich A, et al: Magnetic resonance first-pass myocardial perfusion imaging: clinical
validation and future applications. J Magn Reson Imag 10:676-685, 1999.
68. Canet EP, Janier MF, Revel D: Magnetic resonance perfusion imaging in ischemic heart disease. J Magn Reson Imag
10:423-433, 1999.
69. van der Geest RJ, Reiber JHC: Quantification in cardiac MRI. J Magn Reson Imag 10:602-608, 1999.
70. Gerber BL, Lima JA, Garot J, Bluemke DA: Magnetic resonance imaging of myocardial infarct. Top Magn Reson Imag
11:372-382, 2000.
71. Barkhausen J, Ebert W, Debatin JF, Weinmann HJ: Imaging of myocardial infarction: comparison of magnevist and
gadophrin-3 in rabbits. J Am Coll Cardiol 39(8):1392-1398, 2002.
72. Barkhausen J, Ruehm SG, Goyen M, et al: MR evaluation of ventricular function: true fast imaging with steady-state
precession versus fast low-angle shot cine MR imaging: feasibility study. Radiology 219:264-269, 2001. Medline
Similar articles
page 575
page 576
73. Blake LM, Scheinman MM, Higgins CB: MR features of arrhythmogenic right ventricular dysplasia. Am J Roentgenol
162:809-812, 1994.
74. Vignaux O, Lazarus A, Varin J, et al: Right ventricular MR abnormalities in myotonic dystrophy and relationship with
intracardiac electrophysiologic test findings: initial results. Radiology 224:231-235, 2002. Medline Similar articles
75. Wedegartner U, Tchirikov M, Koch M, et al: Functional magnetic resonance imaging (fMRI) for fetal oxygenation during
maternal hypoxia: initial results. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 174:700-703, 2002. Medline
Similar articles
76. Cornud F, Flam T, Chauveinc L, et al: Extraprostatic spread of clinically localized prostate cancer: factors predictive of pT3
tumor and of positive endorectal MR imaging examination results. Radiology 224:203-210, 2002. Medline
Similar articles
77. Li F, Sone S, Maruyama Y, et al: Correlation between high-resolution computed tomographic, magnetic resonance and
pathological findings in cases with non-cancerous but suspicious lung nodules. Eur Radiol 10(11):1782-1791, 2000.
78. Thompson BH, Stanford W: MR imaging of pulmonary and mediastinal malignancies. Magn Reson Imag Clin N Am
8(4):729-739, 2000.
79. Chung MH, Lee HG, Kwon SS, Park SH: MR imaging of solitary pulmonary lesions: emphasis on tuberculomas and
comparison with tumors. J Magn Reson Imag 11(6):629-637, 2000.
3 of 4 15-04-2010 21:30
80. Schroeder T, et al: CXR vs MRI in the detection of primary nodules. Eur Radiol 13(suppl 1):315, 2003. Presented at the
14th European Congress of Radiology, March 2002.
81. Klop RB, Eikelboom BC, Taks AC, et al: Screening of the internal carotid arteries in patients with peripheral vascular
disease by colour-flow duplex scanning. Eur J Vasc Surg 5:41-45, 1991. Medline Similar articles
82. Alexandrova NA, Gibson WC, Norris JW, Maggisano R: Carotid artery stenosis in peripheral vascular disease. J Vasc
Surg 23:645-649, 1996. Medline Similar articles
83. McDaniel MD, Cronenwett JL: Basic data related to the natural history of intermittent claudication. Ann Vasc Surg
84. Kim DE, Bae HJ, Lee SH, et al: Gradient echo magnetic resonance imaging in the prediction of hemorrhagic vs ischemic
stroke: a need for the consideration of the extent of leukoariosis. Arch Neurol 59:425-429, 2002. Medline
Similar articles
85. Hirai T, Korogi Y, Ono K, et al: Prospective evaluation of suspected stenoocclusive disease of the intracranial artery:
combined MR angiography and CT angiography compared with digital subtraction angiography. Am J Neuroradiol 23:93-101,
86. Herskovits EH, Itoh R, Melhem ER: Accuracy for detection of simulated lesions: comparison of fluid-attenuated inversion-
recovery, proton density-weighted, and T2-weighted synthetic brain MR imaging. Am J Roentgenol 176:1313-1318, 2001.
87. Gohde SC, Goyen M, Forsting M, Debatin JF: Prevention without radiation-a strategy for comprehensive early detection
using magnetic resonance tomography. Radiologe 42(8):622-629, 2002.
88. Hawighorst H, Schoenberg SO, Knopp MV, et al: Hepatic lesions: morphologic and functional characterization with
multiphase breath-hold 3D gadolinium-enhanced MR angiography-initial results. Radiology 210:89-96, 1999.
89. Steinborn MM, Heuck AF, Tiling R, et al: Whole-body bone marrow MRI in patients with metastatic disease to the skeletal
system. J Comput Assist Tomogr 23(1):123-129, 1999.
90. Eustace S, Tello R, DeCarvalho V, et al: A comparison of whole-body turboSTIR MR imaging and planar 99mTc-methylene
diphosphonate scintigraphy in the examination of patients with suspected skeletal metastases. Am J Roentgenol
169(6):1655-1661, 1997.
91. Lauenstein TC, Goehde SC, Herborn CU, et al. Three-dimensional volumetric interpolated breath-hold MR imaging for
whole-body tumor staging in less than 15 minutes: a feasibility study. Am J Roentgenol 179(2):445-449, 2002.
92. Antoch G, Vogt FM, Freudenberg LS, et al: Whole-body dual-modality PET/CT and whole-body MRI for tumor staging in
oncology. JAMA 290(24):3199-3206, 2003.
93. Warner E, Plewes DB, Hill KA, et al: Surveillance of BRCA1 and BRCA2 mutation carriers with magnetic resonance
imaging, ultrasound, mammography, and clinical breast examination. JAMA 292(11):1317-1325, 2004.
94. Kriege M, Brekelmans CT, Boetes C, et al: Efficacy of MRI and mammography for breast-cancer screening in women with
a familial or genetic predisposition. M Engl J Med 351(5):427-437, 2004.
© 2010 Elsevier
4 of 4 15-04-2010 21:30
RTIFACTS AND OLUTIONS

Pippa Storey
Artifacts are anomalous features in an image, caused by extraneous factors that affect the acquisition
process. They are important to recognize, and to correct where possible, since they can mask or
mimic pathology. Ferromagnetic objects, for example, such as hairpins and dental braces, alter the
strength and direction of the local magnetic field and can cause large signal voids that obscure the
surrounding tissue.
Whether or not a factor is extraneous, however, depends on the purpose for which the imaging is
performed. Blood flow, for example, can produce unwanted ghost artifacts on anatomic images, due to
the phase changes induced in flowing spins as they move through magnetic field gradients. The
ghosting can be reduced or eliminated by means of flow compensation or ECG gating. Alternatively,
the effect can be exploited for quantitative measurement of flow speed, using phase-contrast imaging
techniques.1 Even the strong signal attenuation produced by ferrimagnetic materials can be used as a
diagnostic tool, allowing tracking of stem cells labeled with ultrasmall superparamagnetic iron oxide
2-8
(USPIO) particles.
This chapter discusses the factors that cause artifacts and possible ways to reduce or eliminate their
effect. It also mentions some of the means by which those same factors can be harnessed in a
controlled manner to glean new physiologic information.
Most artifacts involve an interaction between physiologic variables, acquisition parameters, hardware
limitations, and the choice of pulse sequence. The artifact produced by a ferromagnetic implant, for
example, varies according to the imaging technique used and is generally more severe on a
gradient-echo image than a spin-echo image. In either case it is characterized by both distortion and
intensity alterations. Conversely, a single artifactual feature, such as ghosting, may have a variety of
origins, including blood flow and hardware instabilities. The chapter is organized in terms of the
mechanisms that cause artifacts. However, since most artifacts involve a combination of factors, some
overlap is unavoidable.
Since previous editions of this book were published, many new imaging techniques have become
mainstream. Because each produces its own characteristic artifacts, they are included in a separate
section on "technique-specific" artifacts. The section covers echo-planar imaging (EPI), diffusion-
weighted EPI, steady-state free precession imaging (SSFP), magnetic resonance angiography (MRA),
parallel imaging, high-field imaging (>1.5 T), and non-Cartesian sampling techniques. Due to
technological improvements, certain artifacts that were described in earlier editions of the book are
now very rare and have been omitted. The overall aim of the chapter is to focus on artifacts routinely
encountered in the current clinical setting and to describe potential diagnostic pitfalls and possible
solutions. As far as possible, the example images are taken from recent (2004) clinical studies,
performed on currently available commercial scanners.
page 577
page 578
1 of 2 15-04-2010 21:30
Add to lightbox
Figure 22-1 Aliasing in a 2D localizer, acquired with conventional Cartesian sampling. Tissue that
extends outside the field-of-view in the phase-encoding direction (anterior/posterior) is wrapped
around to the opposite side of the image. Note that aliasing does not occur in the frequency-encoding
direction (superior/inferior), because of frequency filtering.
© 2010 Elsevier
2 of 2 15-04-2010 21:30
ACQUISITION-RELATED ARTIFACTS
Wraparound
One of the most common artifacts is wraparound. An example in a 2D image is shown in Figure 22-1.
The region of tissue extending outside the field-of-view in the phase-encoding direction is correctly
oriented, but "wrapped around" to the opposite side of the image. This is caused by a phenomenon
known in signal processing as "aliasing", which arises because the image data are sampled in a
discrete rather than continuous fashion.
In MRI, signal is acquired from an entire slice (or slab) of tissue simultaneously. In order to reconstruct
an image, the contributions from all the points in the tissue must be correctly identified and mapped
onto the corresponding points in the image. Spatial information identifying the source must therefore be
embedded into the signal. In conventional 2D Cartesian imaging, this is achieved by means of
"frequency encoding" in one direction and "phase-encoding" in the other.
Add to lightbox
Figure 22-2 Each element of raw data collected during an MRI scan corresponds to a single spatial-
frequency component of the object being imaged and is associated with a particular point in k-space,
the Fourier reciprocal of the image domain. In conventional 2D Cartesian imaging, a single line of
k-space data is collected in the frequency-encoding direction during each acquisition period.
Successive lines are collected at discrete intervals δk along the phase-encoding direction, within the
range [-Kmax, Kmax].
Frequency encoding involves the application of a magnetic field gradient during the data acquisition
period. Since the Larmor frequency of the proton spins is proportional to the strength of the magnetic
field, it varies along the direction of the gradient. The location of tissue in this direction can therefore be
identified by the frequency of its signal. In order to resolve individual voxels (volume elements) of
tissue, the data acquisition period must be long enough to induce a phase shift of 2π between adjacent
voxels. The raw data collected over this period are recorded along a line in "k-space", which is the
Fourier reciprocal of the image domain and has dimensions of phase per unit length (Fig. 22-2).
Position in the perpendicular direction is determined by phase-encoding. A second magnetic field

gradient is applied briefly in that direction, immediately prior to the data acquisition period. The
gradient pulse imparts a spatially varying phase shift to the spins, which is imprinted on their signal.
1 of 10 15-04-2010 21:31
The process is repeated many times, with gradient pulses of different amplitudes, and each time a
new line of k-space data is acquired (see Fig. 22-2). As the pulse amplitude is varied, the phase of the
spins changes by an amount that depends on their position along the direction of the gradient; spins
located farther from the center of the gradient coils undergo a proportionally larger phase change. This
phase information, when combined with the frequency information, uniquely identifies the position of the
source within the field-of-view. In practice, the component signals from all the tissue elements are
extracted simultaneously from the raw k-space data, by means of a fast Fourier transform. The
resulting two-dimensional matrix of voxel signals forms the complex image, of which only the magnitude
is usually displayed.
page 578
page 579
Add to lightbox
Figure 22-3 The phase of the signal S emitted by an element of tissue varies with k, at a rate that
depends on its position. The rate of change is used to identify the location of the tissue and forms the
basis of phase-encoding. The curves in the upper and lower graphs show the real and imaginary parts
of the signal that might be detected from two elements of tissue, if k-space were sampled
continuously. The green line illustrates the signal from a tissue element located at a point y within the
field-of-view and the red line corresponds to a point y ± FOV outside the field-of-view. In practice, the
signal is measured only at discrete intervals δk (blue dots). Since the green and red lines intersect at
these points, there is no way to distinguish whether the signal originated from y or y ± FOV. The
reconstruction algorithm assumes that it came from within the field-of-view and maps it to the position
y on the image. This causes wraparound if the tissue is actually located outside the field-of-view.
While the phase-encoding method correctly determines the position of tissue located within the field-
of-view, it cannot identify tissue beyond that region. Since tissue outside the field-of-view lies farther
from the center of the gradient coils, the phase of its signal changes by a larger amount with each
successive increment of the gradient amplitude. A large positive phase change cannot, however, be
distinguished from a smaller negative one, as illustrated in Figure 22-3. The signal is therefore
interpreted by the reconstruction algorithm as coming from a point within the field-of-view, but on its
opposite side, causing wraparound in the image.
2 of 10 15-04-2010 21:31
Add to lightbox
Figure 22-4 Signal from the arm (arrow) is aliased over signal from the abdomen, causing
interference. This image was acquired as part of a 3D MRA using a gradient-echo sequence. Since
the phase of the gradient-echo is very sensitive to magnetic field inhomogeneity, interference involving
tissue far from the isocenter of the scanner produces an oscillatory intensity pattern.
This problem, known as "aliasing", is a classic one in signal processing and arises whenever the
sampling rate is insufficient to monitor adequately the fluctuations in a signal. A more familiar example
occurs in old Western movies, where rapidly spinning wagon wheels appear to move slowly
backwards, due to the finite exposure rate of the film. In MRI, the equivalent of the exposure rate is
the density of phase-encoding lines. Each line is acquired with a phase-encoding gradient pulse of
different amplitude, which is incremented by a fixed amount between consecutive lines. The increment
is chosen so that the pulse produces an additional phase change of 2π across the field-of-view with
each successive line. It is convenient to express this phase change per unit length as:
†
where FOV is the size of the field-of-view . The increment δk can be visualized as a step in k-space
(see Fig. 22-2). At a position y, the phase of the spins changes by δϕ = yδk between consecutive
lines. Spins a distance FOV away undergo a phase change (y ± FOV) δk. It is clear from Equation
22-1, however, that this equals ϕ ± 2π, which is indistinguishable from ϕ. Signal from these spins is
thus aliased to the position y, producing wraparound in the image.
If one part of the anatomy is wrapped over another part, interference occurs between their signals.
For gradient-echo sequences, the interference pattern is characterized by closely spaced intensity
oscillations (Fig. 22-4). The oscillations result from relative phase variations between the signals, which
are due in turn to inhomogeneities in the static magnetic field B0.
†
The expression for δk can alternatively be written without the factor of 2π, depending on the dimensions used to express intervals
in k-space (namely [cycles/length] versus [radians/length]).
page 579
page 580
In 3D acquisitions, phase-encoding is used in the through-plane dimension as well as one of the

in-plane directions. If the slab profile is poorly defined, the images from one end of the slab wrap over
those at the other end (Fig. 22-5). This is common in MRA acquisitions, due to the use of very short
radiofrequency (RF) excitation pulses. The artifact can be avoided by discarding the outermost slices
3 of 10 15-04-2010 21:31
in the slab.
Wraparound does not usually occur in the frequency-encoding direction, since position in that direction
is determined by the frequency of the emitted signal rather than its phase. Signal originating from
outside the field-of-view can therefore be eliminated using an appropriate frequency filter.
Add to lightbox
Figure 22-5 Through-plane wraparound in a 3D acquisition. Since phase-encoding is used to
determine position in the through-plane direction, aliasing can occur if the slab profile is poorly
defined. The result is overlap between tissues from opposite ends of the slab, in this case the head
and shoulders. This image was acquired as part of a 3D MRA.
Wraparound can sometimes be avoided simply by re-centering the image (as in Fig. 22-1). Otherwise
the field-of-view may need to be enlarged. This however involves a loss of spatial resolution, unless the
matrix size is increased accordingly, which lengthens the scan time. Another possibility is to use the "no
phase wrap" feature, which decreases the step size in k-space while maintaining identical spatial
resolution. Depending on the structures to be imaged, alternative solutions may be to swap the phase-
and frequency-encoding directions or for the subject to change position appropriately (for example, by
moving his or her arms outside the excitation volume). Other possibilities include the use of a smaller
surface coil, with a more limited range of sensitivity, or the application of spatial presaturation bands to
the affected regions. A further alternative is to use "inner-volume" techniques9 in which the excitation
volume is limited in both the slice-select and phase-encoding directions.
Gibbs Ringing
A second artifact related to discrete data sampling is Gibbs ringing (Fig. 22-6). This is in some sense
the flip side of the problem of wraparound and occurs when the matrix size of the acquisition is
insufficient to resolve sharp tissue boundaries.
4 of 10 15-04-2010 21:31
Add to lightbox
page 580
page 581
Figure 22-6 An image of a phantom, acquired with a 64 × 256 matrix, exhibits broad Gibbs
oscillations in the phase-encoding direction (left/right). In the frequency-encoding direction (up/down)
the oscillations are barely visible.
The true resolution of the image in a given direction is given by:
where N is the size of the acquisition matrix in that direction. Increasing N while keeping the field-
of-view constant therefore improves the resolution. In k-space this is equivalent to sampling out to
higher k values. The high-order k-space data represent high spatial-frequency components, and are
necessary to resolve fine spatial structures such as abrupt tissue boundaries. Denoting the sampled
range of k-space by [-Kmax, Kmax], the relationship between Kmax and image resolution can be
formalized by substituting Equation 22-1 into Equation 22-2, to give:
where Nδk = 2Kmax.
If the object contains high spatial frequency components that are not sampled during the acquisition,
the result is not only a loss of resolution but also the introduction of Gibbs ringing, which is
characterized by intensity oscillations superimposed on all the fine spatial structures of the object. The
length scale of the oscillations is inversely related to Kmax and identical to the true resolution of the
image, given in Equation 22-3. Increasing Kmax makes the oscillations finer and less noticeable.
However, for the matrix sizes typically used in clinical imaging, the ringing does not usually disappear
altogether, since there are almost always tissue boundaries sharper than the acquisition can resolve.
5 of 10 15-04-2010 21:31
Add to lightbox
Figure 22-7 Gibbs ringing occurs at a tissue interface when the spatial profile of the boundary (red) is
sharper than the acquisition can resolve. The intensity of the pixels in the resulting image (blue dots)
exhibits oscillations, whose width equals the true resolution of the acquisition, namely δx = π/Kmax. If
k-space is fully sampled in a given direction (i.e., the number of acquired data points equals the pixel
number) then the intensity oscillates between adjacent pixels (upper graph). However if fewer k-space
points are sampled, and the missing data are zero-filled, then the pixel number remains the same but
the width of the oscillations increases (lower graph).
While in principle Gibbs ringing can occur in both the phase- and frequency-encoding directions, it is
usually more prominent in the phase-encoding direction because time constraints limit the number of
k-space lines collected. Furthermore, T2* decay introduces some natural filtering in the frequency-
encoding direction (discussed below), which tends to suppress ringing along that axis.
If k-space is fully sampled (i.e., the number of points collected equals the pixel number), then the
intensity oscillations occur on the scale of a single pixel (Fig. 22-7). To minimize scan time, however, it
is common to acquire fewer k-space lines and to fill the remaining lines with zeroes prior to image
reconstruction. This preserves the pixel number but reduces the resolution of the image. The
oscillations are correspondingly broadened, as illustrated in the lower graph of Figure 22-7 and
demonstrated in the earlier phantom image (Fig. 22-6).
While the width of the Gibbs oscillations is determined by the resolution of the acquisition, their
amplitude depends on the signal contrast at the interface. The ringing is therefore more prominent at
boundaries between bright and dark tissues. The artifacts present a potential diagnostic pitfall in
images of the cervical spine, where they can mimic a syrinx (Fig. 22-8). They also frequently affect
contrast-enhanced MR angiograms (discussed later), in which spatial resolution is often compromised
to satisfy the time constraints of a 3D acquisition.
6 of 10 15-04-2010 21:31
Add to lightbox
page 581
page 582
Figure 22-8 Gibbs artifacts in the cervical spine can mimic a syrinx (arrow). As the resolution in the
anterior/posterior direction is increased (from left to right) the Gibbs oscillations become narrower.
The optimal solution for Gibbs ringing is to improve the resolution of the image by increasing the matrix
size or decreasing the field-of-view. This reduces the width of the oscillations, thereby providing better
conspicuity of tissue boundaries. Decreasing the field-of-view, however, may not always be possible
without incurring wraparound, while increasing the matrix size generally requires lengthening the scan
10-12
time. Parallel imaging offers a means to achieve higher resolution without compromising scan time
but involves a loss of signal-to-noise ratio (SNR).
An alternative to increasing the matrix size is to filter the k-space data so that the higher-order spatial
frequencies are progressively attenuated. This avoids the effects of abrupt truncation and reduces the
amplitude of the oscillations. However, the smoothing effect of the filter also decreases the effective
resolution of the image.
Incomplete Fat Suppression

Since fat appears bright both on T1-weighted images and on T2-weighted fast spin-echo (FSE)
13
images, fat saturation is often used to suppress its intensity, thereby improving the conspicuity of
other tissues. This is particularly important in musculoskeletal and breast imaging, and in contrast-
enhanced studies.
Various techniques are used for fat saturation, but most exploit the chemical shift between lipid and
water to null the magnetization of fat protons selectively while leaving water protons fully magnetized.
The difference in Larmor frequency between water and the dominant lipid resonance associated with
methylene (-CH2-) is 3.5 ppm, which at 1.5 T equals 220 Hz. Because fat saturation is frequency
selective, it is extremely sensitive to inhomogeneity in the strength of the static magnetic field B0. This
is particularly problematic at fields below 1.5 T, where the frequency separation between water and
fat is smaller. To achieve complete and uniform fat saturation across the entire area of interest
requires very good shimming and can be particularly difficult when the anatomic structures are
asymmetric, as in the case of an ankle or neck, or far from the isocenter of the magnet, as for a
shoulder. Figure 22-9 shows an example of inhomogeneous fat saturation in a T2-weighted image of
an ankle. The focal enhancement in the bone marrow can mimic the effect of disease processes such
as contusion or edema. Comparison with T1 images can be useful in ruling out such possibilities.
7 of 10 15-04-2010 21:31
The primary remedy against poor fat saturation is to perform a local shim over the area of interest.
This may be more effective at higher field strengths, due to the larger frequency separation between
lipid and water. If no improvement is obtained with a local shim, it may be advisable to check for the
presence of stray metallic objects such as hairclips, which can cause large distortions in the local
magnetic field (see later discussion on magnetic susceptibility). Persistent difficulties in shimming may
alternatively indicate a drift in the center frequency of the scanner, and should be referred to a field
engineer. Poor fat saturation can also result from nonuniformity in the RF excitation field B1, which
causes variations in the effective flip angle across the tissue.
Other methods are available for fat suppression that are relatively insensitive to B0 or B1
inhomogeneities. Dixon techniques14 involve the acquisition of two or more sets of images, in one of
which the signals from fat and water are exactly in phase, and in another of which they are exactly out
of phase. The effects of magnetic field inhomogeneity can be factored out, allowing a set of water-only
images to be obtained from the appropriate combination of in-phase and out-of-phase images. An
alternative approach is to use short TI inversion recovery (STIR) sequences. These eliminate signal
from fat by exploiting its short relaxation time rather than its chemical shift. STIR is not appropriate,
however, for gadolinium-enhanced imaging, due to the suppression of signal from other short-T1
sources.
Analog-to-Digital Converter Overflow

The RF receiver in an MR system detects signal from an entire slice simultaneously. The amplitude of
the net signal varies widely according to the volume and composition of the excited tissue. Since the
signal is ultimately converted to a complex digital number, the gain of the RF receiver must be
appropriately adjusted to achieve adequate dynamic range. The receiver calibration is performed using
signal estimates obtained during the prescan procedure. If the actual signal received during a scan
exceeds the value estimated during calibration, it can cause overflow errors in the analog-to-digital
converter.
8 of 10 15-04-2010 21:31
Add to lightbox
page 582
page 583
Figure 22-9 A T2-weighted fast spin-echo (FSE) image of the ankle exhibits inhomogeneous signal
suppression in the bone marrow (arrow) due to incomplete fat saturation.
Table 22-1. Summary of Acquisition-Related Artifacts

Artifact Causes Solutions
Wraparound Tissue extending outside FOV in phase- Re-center image
encoding direction Increase FOV
"No phase wrap" option
Swap phase- and
frequency-encoding
directions
Use RF coil with smaller
sensitivity range
Spatial saturation bands
Inner volume excitation
Poorly defined slab profile in 3D acquisition Discard outermost slices in
(produces through-plane wraparound) 3D stack
Gibbs ringing Inadequate spatial resolution Increase matrix size
Decrease FOV
Parallel imaging
9 of 10 15-04-2010 21:31
Filter k-space data

Incomplete fat Inhomogeneity in B0 field Local shim
suppression
Inadequate shimming Check for stray metallic
objects
Ferromagnetic materials Dixon techniques
Inhomogeneous RF excitation STIR fat suppression
Scan at higher field
strength
Check for drift in center
frequency
ADC overflow High signal causing overflow in analog- Reduce receiver gain
to-digital converter Fat suppression or spatial
saturation bands
Overflow generally occurs in the low-order components of k-space, since these have the highest
amplitudes. The result is a slowly varying shading pattern that extends across the entire image,
including the background (Fig. 22-10). It is more common in scans that incur high signal from fat and
often affects only certain slices in a multislice acquisition. The affected slices usually lie near the edge
of the stack, since the signal calibration is performed on the central slice and may not be accurate for
more peripheral slices. Overflow errors can also occur in contrast-enhanced images, if the calibration
is performed prior to contrast administration.
Reducing the receiver gain provides an adequate solution in most cases, but may not be appropriate
when subtraction of pre- from post-contrast images is required. In that case, it may be preferable to
repeat the post-contrast acquisition after the agent has dispersed slightly and the signal has
diminished. If the high signal originates from subcutaneous fat, an alternative solution may be to use
fat-selective or spatial presaturation.
A summary of the artifacts discussed in this section is provided in Table 22-1.
© 2010 Elsevier
10 of 10 15-04-2010 21:31
PHYSIOLOGY- AND SUBJECT-RELATED ARTIFACTS

Motion and Flow
Add to lightbox
page 583
page 584
Figure 22-10 An abdominal image exhibits ADC overflow due to the high signal from fat (A). The
adjacent slice (B), however, is unaffected.
1 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-11 An axial image exhibits ghost artifacts (arrows) from cardiac motion. Due to the
periodicity of the motion relative to the sequence repetition time (TR = 175 ms), the ghosts are
manifested as discrete replicas of the heart, which appear along the phase-encoding direction
(anterior/posterior) and overlap the lung, the chest wall, and the heart itself.
Add to lightbox
2 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-12 Axial T1 images of the brain exhibit ghost artifacts due to pulsatile blood flow. The
artifacts propagate along the phase-encoding direction (left/right).
Motion and flow are among the greatest challenges for clinical MRI and have been the primary
motivation behind advances in fast imaging techniques. Cardiac motion, respiration, peristalsis, and
blood flow within the great vessels are determining factors in the design of thoracic and abdominal
imaging protocols. Swallowing and CSF pulsation are important considerations in cervical spine
imaging and head movement is a common problem in brain imaging, particularly in infants and
disoriented patients. Artifacts can arise from motion-induced phase shifts, which cause ghosts,
misregistration, intravoxel dephasing, and banding at tissue interfaces. Intensity anomalies also occur
due to inflow or "time-of-flight" effects.
Ghosting
In standard Cartesian imaging, the motion of spins, either as bulk tissue movement or fluid flow,
produces replicas of the corresponding anatomy in the final image (Figs. 22-11 and 22-12). Only the
moving tissue or fluid is replicated and the "ghosts" propagate primarily in the phase-encoding
direction. The frequency-encoding direction is less affected, since position information in that direction
is acquired during the readout period, which is typically of the order of a few milliseconds and therefore
effectively instantaneous on the time scale of most physiologic motion.
page 584
page 585
3 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-13 Ghosts due to pulsatile flow in the aorta can mimic liver lesions (arrow).
Ghosting arises due to disruption of the phase-encoding process. As described earlier, position in the
phase-encoding direction is imprinted on the signal by means of a spatially dependent phase shift,
which is imparted to the spins via a magnetic field gradient, prior to the acquisition of each line of
k-space. For the method to work, it is essential that the phase-encoding gradient be the only factor
altering the phase of the spins from line to line. Tissue motion causes extraneous phase shifts that
corrupt the spatial encoding and give rise to ghosting.
Depending on the type of motion, the phase shifts can vary in a periodic, random or gradual manner
across k-space. Periodic modulation arises from cardiac motion, blood flow, and CSF pulsation. The
resulting ghosts are discrete and distinct, with a spatial separation that is related to the ratio between
the R-R interval of the cardiac cycle and the sequence repetition time. Where the ghosts overlap other
tissues they can mimic lesions, as demonstrated in Figure 22-13.
If the motion does not exhibit any particular regularity relative to the TR interval, the artifacts are
smeared out along the phase-encoding direction and can obscure pathology (Fig. 22-14). Semi-regular
motion, such as respiration, produces blurred ghosts, which can be difficult to identify as artifacts.
Figure 22-15 shows an example of a motion-induced ghost that mimics an intimal flap within the aorta.
Finally, if the scan time is very short, the motion may be approximately linear over the duration of the
acquisition. The result in this case is not ghosting but rather motional banding at the edges of the
moving structures (discussed later).
4 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-14 An axial image exhibits ghosting due to head motion. Since the movement is random
and displays no particular periodicity relative to the repetition time, the ghosts are smeared out along
the phase-encoding direction (left/right).
The phase shifts that cause ghosts arise from two mechanisms, which may be described as inter-view
and intra-view motion.15 Inter-view motion refers to displacement between successive k-space
acquisitions. It involves bulk movement of tissue rather than flow and is more sensitive to motion in the
phase-encoding direction than in the readout direction. A portion of tissue that is displaced along the
phase-encoding axis will experience a slightly different magnetic field during the application of the
phase-encoding gradient and will accordingly receive a different phase shift. At the nth line, the phase
difference due to a displacement ∆y is:
where δk is the step size in k-space (given by Equation 22-1). If ∆y varies between successive phase-
encoding lines, the resulting phase shifts will cause ghosts in the image. The severity of the ghosting
depends on the size of the displacement relative to the field-of-view. However, even small random
displacements (~5 mm) between k-space lines are sufficient to cause severe artifacts for typical
fields-of-view of around 40 cm or less.
page 585
page 586
5 of 27 15-04-2010 21:41
Add to lightbox
Add to lightbox
Figure 22-15 A source image (A) from a 3D MRA of a clinical patient exhibits blurred ghosts due to
respiration and cardiac motion. A motion artifact in the aortic arch mimics an intimal flap (arrow), which
appears on both the source image and the maximum intensity projection (B). A second acquisition
performed immediately afterwards did not reproduce the intimal flap and a subsequent CT scan ruled
out a diagnosis of aortic dissection.
6 of 27 15-04-2010 21:41
A plethora of techniques exist to reduce motion artifacts and their applicability depends on the type of
motion involved. For imaging the head and extremities, it is often sufficient to use cushioning to
maintain position. Vacuum devices can be particularly useful when immobilization is required over long
16
periods of time, for example in functional MRI (fMRI). They consist of a pump connected to a
specially designed cushion, which molds to the anatomy when lightly inflated and then becomes rigid
when the air is evacuated. In unco-operative patients, image quality can be improved with sedation or
by using PROPELLER techniques (periodically rotated overlapping parallel lines with enhanced
reconstruction).17 In PROPELLER, data are collected in a series of rectangular strips that are rotated
about the origin of k-space (discussed later in the section on non-Cartesian imaging). The central
region of k-space is resampled for each strip, allowing correction of rigid-body in-plane motion and
rejection of inconsistent data resulting from through-plane motion. PROPELLER has proven effective
for motion artifact reduction in pediatric neuroimaging18 and has opened up an avenue for non-EPI
based diffusion imaging.19-21
To avoid artifacts from respiratory motion, the most common approach is to use fast imaging methods
in combination with breath-holding. By incorporating ECG gating, the techniques can also be applied to
imaging the heart. Synchronizing data acquisition to the cardiac cycle reduces artifacts from cardiac
motion and allows reconstruction of cine image series for functional assessment. When breath-holding
is not possible, due to long scan times or patient intolerance, respiratory triggering can be
implemented with the use of navigator techniques, which allow acquisition during free breathing.22 The
diaphragm is monitored by means of a 1D pencil beam acquisition and data are collected only when its
position falls within a certain user-defined range, usually near end-exhalation.
When motion occurs in a region that is not the focus of the study, saturation techniques can be applied
to null its signal. In cervical spine imaging, for example, spatial saturation over the throat is routinely
used to avoid artifacts from swallowing. In thoracic and abdominal imaging, the high-intensity artifacts
resulting from the motion of subcutaneous fat can be suppressed using fat saturation. In certain
circumstances it may be possible to direct ghost artifacts away from the region of interest by
swapping the phase- and frequency-encoding axes.
For applications such as fetal imaging, the best approach is often to use single-shot pulse sequences,
in which all the lines of k-space are acquired in sufficiently rapid succession that the effects of motion
are minimized.23 Single-shot images do not suffer from ghosting but may be subject to band artifacts,
discussed later.
page 586
page 587
7 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-16 In its simplest form (top), the waveform of the frequency-encoding gradient incorporates
a single-lobed preparatory pulse, which ensures that stationary spins are refocused at the desired
echo time. In other words, their phases (blue line) will equal zero at TE regardless of their position.
Moving spins, however, will not be fully refocused and their phases (red line) will be nonzero at the
echo time. Flow compensation (below) involves the use of a dual-lobed preparatory pulse, which
ensures that both stationary spins and those moving with constant velocity are correctly refocused.
Note however that this requires a longer echo time.
Add to lightbox
Figure 22-17 A cervical spine image, acquired with the frequency-encoding direction superior/inferior,
exhibits discrete, uniformly separated ghosts (arrows) due to pulsatile CSF flow (A). The ghosts can be
suppressed using flow compensation in the frequency-encoding direction (B).
Both flow and bulk tissue motion can give rise to artifacts through the mechanism of intra-view motion.
This refers to movement of spins during the time TE between the RF excitation and the center of the
echo that follows. The gradient waveform used to read out a single line of k-space is designed so that
8 of 27 15-04-2010 21:41
stationary spins will be refocused at the desired echo time. In its simplest form, this involves preceding
the readout gradient by a single-lobed preparatory pulse (Fig. 22-16). Spins that move relative to the
gradients will, however, not be fully refocused at the echo time. The residual phase shift is the
cumulative sum of the phases imparted by the magnetic field gradients at all the positions through
which the spin has passed. This is described mathematically by the integral:
where r(t) and G(t) denote the position of the spin and the value of the gradient respectively at the time
t. Only the component of motion in the direction of the gradient contributes to the phase shift. To
determine the effect of the readout waveform, the motion of the spin in this direction can be expressed
as:
so that the integral reduces to the following sum:
The term involving the initial position ro vanishes, since stationary spins are always refocused. Of the
remaining terms, the second usually makes the largest contribution to the phase shift. A rough estimate
of this term can be obtained by neglecting the finite duration of the RF excitation and the ramp times of
the gradients. Under these approximations, the residual phase shift at the center of the echo is:
Note that the phase shift depends on the velocity of the motion in the frequency-encoding direction.
Intra-view motion differs in this respect from inter-view motion (discussed earlier) in which the phase
shifts are determined primarily by motion in the phase-encoding direction.
It is clear from Equation 22-8 that if the velocity alters between successive k-space lines, the phase
shifts will vary also, causing ghosts in the image. This occurs with pulsatile flow, for example blood
flow in the arteries and CSF flow in the cervical spine (Fig. 22-17A). Artifacts due to intra-view motion
can be reduced or eliminated using gradient waveforms with dual-lobed preparatory pulses. The pulses
are designed to refocus spins with constant velocity, by nulling the second term of Equation 22-7. The
technique, illustrated in Figure 22-16, is known as gradient-moment nulling or flow compensation. It
helps in suppressing ghosts from pulsatile flow (Fig. 22-17B) but increases the echo time. Higher-order
terms (involving acceleration, jerk, etc.) can also be nulled by the use of more complicated waveforms,
but at a cost of further lengthening TE.
Pulse sequences that involve multiple spin or gradient-echoes after the initial excitation exhibit some
inherent flow compensation on the even echoes (known as even-echo rephasing). This is due to partial
cancellation of the phase shifts accumulated during successive applications of the readout gradient.
page 587
page 588
Another common technique used to avoid ghost artifacts in the cervical spine is to select phase-
encoding along the superior/inferior direction (in combination with the "no phase wrap" option to
prevent wraparound). This suppresses flow ghosts from CSF pulsatility, since the component of flow in
the frequency-encoding direction (anterior/posterior) is minimized. It also ensures that any motion
artifacts due to swallowing will propagate along the superior/inferior direction and not overlap the
spine. ECG gating is helpful in reducing artifacts due to pulsatile flow but is not routinely used except in
cardiac imaging.
Flow in the through-plane direction can also cause ghosts (see Fig. 22-13) which result from
incomplete refocusing of spins moving relative to the slice-select gradients or from inflow effects
(discussed later). The artifacts can be reduced using flow-compensated slice-select gradients or by
placing spatial saturation bands proximal to the slice of interest.
Since the phase-encoding gradient is not refocused prior to the readout, fluid flow in the phase-
9 of 27 15-04-2010 21:41
encoding direction does not tend to produce ghosts. It can, however, contribute to spatial
misregistration (discussed later) when the flow velocity has components along both the phase- and
frequency-encoding axes.
Fluid flow also produces very distinctive artifacts in steady-state free precession (SSFP) images, in
situations where the moving spins fail to reach a steady state. This is discussed later, in the section on
SSFP imaging.
Motional Band Artifacts

Single-shot techniques and other nonsegmented fast sequences can be useful in minimizing ghost
artifacts due to motion. However, they remain sensitive to movement that occurs during the time taken
to acquire each image. If the motion is approximately uniform during the data acquisition, the result is
neither the random nor periodic phase shifts that give rise to ghosts but rather a smoothly varying
phase modulation of k-space that produces intensity oscillations along the borders of the moving
anatomy.24
Such artifacts may occur in first-pass myocardial perfusion imaging if the data acquisition rate is not
sufficiently rapid with respect to the cardiac motion. To achieve adequate temporal resolution,
perfusion imaging is typically performed using a non-segmented gradient-echo sequence. Motion of the
heart during the time taken to acquire each image can cause artifacts in the myocardium, particularly if
the acquisition coincides with systole. (Note however that artifacts at the endocardial borders may also
occur even in diastole due to Gibbs ringing.) The artifacts appear at the moment of contrast arrival and
may mimic perfusion deficits.
Motional band artifacts can be described analytically for the case in which a single line of k-space is
collected for each RF excitation. The artifacts arise from displacement of tissue between successive
phase-encoding lines, namely "inter-view" motion, and are therefore most severe in tissue moving in
the phase-encoding direction. Assuming that the velocity in this direction is approximately constant over
the course of the acquisition, the displacement will change uniformly with time, ∆y = vt. Assuming
furthermore that k-space is traversed in sequential order, the displacement will have a linear
dependence on the line number n:
The resulting phase shift, which can be determined by inserting Equation 22-9 into 22-4, varies
quadratically with line number:
Writing the k value of the nth line as ky = nδk and using Equation 22-1 gives the following expression
for the phase shift:
Its effect is to multiply the k-space data by the phase modulation:
which is equivalent to convolving the image by the oscillatory function:
The result is to produce intensity oscillations along the borders of structures moving in the phase-
encoding direction. The width of the first lobe of the oscillations is of the order of:
and its amplitude depends on the signal contrast across the interface. The oscillations become
progressively narrower farther from the boundary, as illustrated in Figure 22-18. The calculated
intensity profile closely matches the results of phantom experiments (Fig. 22-19A). Using centric rather
than sequential phase ordering gives an asymmetric intensity pattern on the leading and trailing edges
(Fig. 22-19B), but does not substantially alter the width of the oscillations for a given velocity and
repetition time.
10 of 27 15-04-2010 21:41
In cardiac imaging, the intensity oscillations appear as dark bands near the endocardium and are most
prominent in the phase-encoding direction. In the examples shown in Figure 22-20, the dark bands
occupy almost the entire width of the myocardium, due to the relatively long repetition time used (TR =
5.2 ms). Shorter repetition times produce narrower bands (through the relationship given in Equation
22-14) and may be confined to the subendocardium.
page 588
page 589
Add to lightbox
Figure 22-18 The intensity profile at a tissue boundary moving with constant velocity in the phase-
encoding direction. The simulation assumes a nonsegmented fast gradient-echo acquisition with
sequential phase ordering. The position is plotted in units of , where v is the velocity.
Due to symmetry, the motion could be either to the left or to the right. (From Storey P et al: Band
artifacts due to bulk motion. Magn Reson Med 48:1028-1036, 2002. Copyright © 2002 Wiley-Liss
Inc. Reprinted by permission of Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc.)
In first-pass imaging, band artifacts due to motion or Gibbs ringing may be mistaken for evidence of a
perfusion deficit. Since the amplitude of the oscillations depends on the signal contrast at the tissue
boundary, the bands are generally not visible in the baseline images, which exhibit low signal in both
the myocardium and blood pool. Upon arrival of the contrast material in the ventricle, however, the
intensity of the blood pool enhances dramatically, increasing the signal difference at the endocardial
border. The dark bands then become highly conspicuous, mimicking a perfusion deficit.
To minimize motional artifacts, the acquisition should be timed to coincide with end-diastole, when the
movement of the endocardium is minimal. Traversing k-space as rapidly as possible, by shortening TR
or using parallel imaging techniques, also reduces their severity. Even when the acquisition is fast with
respect to the cardiac motion, however, endocardial rim artifacts can occur due to Gibbs ringing.
Intravoxel Dephasing
In the absence of flow compensation or even-echo rephasing, motion of spins between the RF
excitation and the center of the readout introduces a velocity-dependent phase shift, as discussed
earlier. Within a vessel the flow speed varies, being highest at the center and lowest near the walls.
Since a voxel lying within the vessel generally includes fluid elements with different velocities, it may be
subject to intravoxel dephasing, causing signal loss.
11 of 27 15-04-2010 21:41
Add to lightbox
Add to lightbox
Figure 22-19 A phantom moving in the phase-encoding direction (left/right) and scanned with a
nonsegmented fast gradient-echo technique. The images exhibit band artifacts on the leading and
trailing edges of the phantom (arrows). The artifacts are symmetric if the phase ordering is sequential
(A) and asymmetric if it is centric (B). (From Storey P et al: Band artifacts due to bulk motion. Magn
Reson Med 48:1028-1036, 2002. Copyright © 2002 Wiley-Liss Inc. Reprinted by permission of
Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc.)
Flow Misregistration
When the direction of flow is oblique to the imaging axes, spatial misregistration may occur. For
in-plane flow this is due to the time delay between phase and frequency encoding, which provide
information about the location of spins in the "y" and "x" directions respectively.
12 of 27 15-04-2010 21:41
page 589
page 590
Add to lightbox
Add to lightbox
Figure 22-20 Axial images of a healthy volunteer, acquired with a nonsegmented fast gradient-echo
technique. Dark bands due to cardiac motion appear in the myocardium (arrows) and are most
prominent in the phase-encoding direction, namely anterior/posterior in image A and left/right in image
B. (Note that the phase-encoding direction is evident in each case by the presence of wraparound in
image B and its absence in A.) No exogenous contrast material was used but a slice-selective
inversion pulse (TI = 400 ms) was employed to reduce the signal of the myocardium relative to the
blood pool. The repetition time was TR = 5.2 ms and the phase ordering was centric. (From Storey P
et al: Band artifacts due to bulk motion. Magn Reson Med 48:1028-1036, 2002. Copyright © 2002
Wiley-Liss Inc. Reprinted by permission of Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc.)
13 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-21 Flow misregistration can occur when a vessel is oriented at an oblique angle to the
phase- and frequency-encoding directions, indicated here by the vertical and horizontal axes
respectively. Since the phase-encoding gradient is applied before the frequency-encoding gradient,
the y position of a given fluid element is effectively measured at an earlier time than its x position. If the
fluid element (blue dot) moves from the location (x1, y1) to the location (x2, y2) during the interval
between phase and frequency encoding, its signal will be mapped to the location (x2, y1) (green dot),
which lies outside the vessel.
If an element of the fluid moves in an oblique direction, both its x- and y-positions change with time.
Suppose (x1, y1) denotes the location of the fluid element when the phase-encoding gradient is applied,
and (x2, y2) its location at the echo time. Since the y-position is measured at the earlier time point and
the x-position at the later time point, the signal is mapped to a location (x2, y1) through which the fluid
never passed (Fig. 22-21). The result is that the lumen of the vessel may appear displaced from the
vessel itself (Fig. 22-22). The amount of displacement is proportional to the flow speed and to the time
delay between phase and frequency encoding. Depending on the implementation of the pulse
sequence, this delay may be related to the echo time. The image in Figure 22-22 was acquired with a
relatively long echo time (TE = 15 ms) in order to achieve T2*-weighting, which is useful clinically for
detecting hemorrhage, due to the paramagnetic properties of blood derivatives.
Misregistration can also affect through-plane flow, due to the delays between slice selection and
spatial encoding in the remaining two directions. Possible solutions in both cases include minimizing the
relevant time delays or using a sequence that suppresses signal from flow.
Inflow
Artifacts can result from inflow of fluid into the imaging slice during data acquisition. In gradient-echo
images, the signal of inflowing spins is typically higher than that of stationary spins, provided the
longitudinal relaxation time T1 of the fluid is longer than the repetition time TR of the sequence. This is
due to partial saturation of the stationary spins, whose magnetization does not have time to relax fully
before each subsequent excitation pulse. By contrast, inflowing spins, which have not undergone prior
RF excitations, have relatively higher magnetization. The effect is exploited in time-of-flight MR
angiography, but can cause inhomogeneous intensity in regions of slow or turbulent flow (Fig. 22-23A).
Signal in the blood can be recovered using T1-shortening contrast agents (Fig. 22-23B), which prevent
saturation of slow-flowing spins.
page 590
page 591
14 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-22 An axial T2*-weighted image exhibiting flow misregistration. In vessels oriented at an
oblique angle to the phase- and frequency-encoding axes, the lumen is displaced from the vessel itself
(arrows). This image was acquired with a gradient-echo sequence and an echo time of TE = 15 ms.
The phase-encoding axis is oriented left/right.
Add to lightbox
15 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-23 Axial gradient-echo images of the pelvis show inhomogeneous signal in the left pelvic
vein (arrow) due to slow flow (A). Contrast enhancement with gadolinium provides more homogeneous
signal and consequently better delineation of the veins, due to its shortening effect on the T1 relaxation
time of blood (B). (Courtesy of Robert Edelman MD)
Unlike gradient-echo images, FSE images typically exhibit lower signal from inflowing spins than from
stationary spins. The reason is that inflowing spins may not experience the entire train of RF pulses
(including the excitation pulse and the refocusing pulses). As a consequence, they are not correctly
refocused at the echo time and their signal is suppressed (Fig. 22-24).
Chemical Shift
The net magnetic field experienced by a proton is the sum of the applied field B0 and the much smaller
fields of the surrounding electrons. Protons in dissimilar chemical environments therefore precess at
slightly different Larmor frequencies in a given applied field. The difference, known as chemical shift, is
about 3.5 ppm between water and the dominant lipid peak in fat. The resulting frequency shift is
proportional to field strength and equals about 220 Hz at 1.5 T.
Misregistration
page 591
page 592
16 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-24 A T2-weighted single-shot FSE image exhibits artifacts in the bladder due to fluid flow
into the imaging plane. The inflowing spins exhibit lower signal than the stationary ones since they
have not experienced the entire train of RF pulses (including the excitation and refocusing pulses).
In standard Cartesian imaging, an applied magnetic field gradient is used to induce a spatial variation in
the precession frequencies of the spins, uniquely identifying their position in the readout direction. This
is the basis of frequency encoding. The method relies on the underlying assumption that any
differences in precession frequency are due solely to the effect of the imaging gradients. This
assumption is invalid if the tissue contains protons of different chemical species, for example fat and
water. Because of their chemical shift, fat and water protons at identical positions in the readout
direction precess at slightly different frequencies and the image reconstruction algorithm will interpret
their signals as coming from different locations. This causes spatial misregistration between fat- and
water-containing tissues. The effect occurs most commonly in low bandwidth spin-echo and FSE
images, and produces bright and dark bands on opposite sides of anatomic structures where fat and
water meet (Fig. 22-25). The dark bands correspond to signal voids where fat is displaced away from
water and the bright bands reflect signal summation where the fat and water images overlap.
The magnitude of the displacement equals the distance over which the magnetic field strength changes
by 3.5 ppm. This is determined by the amplitude of the readout gradient, which in turn depends on the
bandwidth of the acquisition. The bandwidth (BW) is simply the frequency range used for image
reconstruction, and equals the spread of precession frequencies across the field-of-view. These
relationships can be used to calculate the chemical shift displacement as follows (Fig. 22-26). Given
that the frequency changes by an amount BW over a distance equal to FOV, it will change by a value
equal to the chemical shift over a distance given by:
This expression shows that the displacement is inversely related to the bandwidth of the readout and
therefore that the artifacts can be reduced by using a larger bandwidth. Since it is also proportional to
field strength, the bandwidth must be further increased at higher field. In certain circumstances it may
be preferable to eliminate the signal from fat entirely, using fat saturation, water-selective excitation or
Dixon techniques.
17 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-25 An axial T2-weighted image exhibits chemical shift misregistration around the thecal
sack where it borders on the epidural fat. The artifacts could potentially be mistaken for hemorrhage or
dural thickening.
Add to lightbox
page 592
page 593
Figure 22-26 Due to their chemical shift, the protons in fat have a slightly lower precession frequency
than those in water at a given magnetic field strength. The signal emitted by fat during the application
of the readout gradient is therefore interpreted by the reconstruction algorithm as coming from an
apparent position that is displaced slightly from its actual location. This causes misregistration in the
image along the frequency-encoding direction. The displacement is proportional to the field-of-view
(FOV) and the strength of the B0 field, and inversely proportional to the bandwidth (BW) of the
acquisition.
Chemical shift misregistration can introduce errors into the estimation of subchondral bone thickness,
25
which is important for the evaluation of morphologic changes in rheumatoid disease. Fat suppression
techniques are not applicable in this context, since they would eliminate signal from the marrow and
prevent delineation of the bone. The magnitude of the error can, however, be calculated using Equation
22-15 and factored into the thickness estimate. In some situations, misregistration artifacts can in fact
provide useful diagnostic information, for example in the identification of fatty lesions such as lipomas,
dermoids, and teratomas.26 Spatial misregistration can also occur in the through-plane direction. The
18 of 27 15-04-2010 21:41
resulting images display fat- and water-containing tissue from slices that are slightly offset from each
other. This occurs because the slice selection process is frequency dependent, relying on a magnetic
field gradient in the through-plane direction to excite only those spins within the chosen slice profile.
The bandwidths used for RF excitation pulses are typically of the order of 1.0-1.5 kHz. At 1.5 T this is
several times larger than the chemical shift (220 Hz) and the resulting offset is therefore only a small
fraction (~15-20%) of the total slice thickness.
Interference
Gradient-echo images are typically acquired with a higher bandwidth than spin-echo and FSE images
and consequently do not suffer from chemical shift misregistration to the same extent. However,
depending on the echo time, they may exhibit dark lines at interfaces between fat- and water-
containing tissues. The lines are another manifestation of chemical shift and result from phase
cancellation, or destructive interference, between fat and water signals in voxels that contain both
tissue types.
Since fat and water protons precess at slightly different frequencies, their magnetization gradually
dephases with time after the initial excitation pulse. At 1.5 T, the frequency of water is approximately
220 Hz higher than that of fat, which means that its magnetization will complete an extra revolution
every 4.5 ms. By choosing the echo time judiciously, images can be obtained in which the
magnetization of water is exactly in phase or exactly out of phase with that of fat. If the magnetization
is out of phase, then destructive interference will occur in voxels containing both tissue types and the
net signal will equal the difference between the fat and water signals. The reduction in intensity on an
out-of-phase image compared to an in-phase image provides information about the composition of the
tissue and is often used diagnostically to detect fatty liver27,28 and to characterize adrenal masses.29,30
Out-of-phase images contain dark borders between fat- and water-containing tissues, due to partial
volume averaging in voxels at the interface (Fig. 22-27). Although the dark outlines are technically
artifacts, they can be useful in delineating organs such as the kidneys, liver, and adrenal glands, which
are surrounded by fat.26
Chemical shift interference does not normally occur in spin-echo and FSE imaging, due to the use of
refocusing pulses, which rephase the fat and water signals at the center of the echo.
Magnetic Susceptibility
Add to lightbox
19 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-27 In-phase (A) and out-of-phase (B) gradient-echo images of the abdomen. The
out-of-phase image exhibits dark lines at the interfaces between fat- and water-containing tissues, due
to chemical shift interference.
Differences in Larmor frequency within the body can disrupt the process of frequency encoding and
slice selection, causing spatial misregistration. On gradient-echo images they can also produce signal
loss through the mechanism of intravoxel dephasing. These effects were discussed above in the
context of chemical shift, but they also occur wherever there are differences in magnetic susceptibility
within the body.
page 593
page 594
Magnetic susceptibility refers to the tendency of a material to become magnetized in the presence of
an applied magnetic field. The susceptibility χ is defined in terms of the amount of magnetization
produced, M, for a given applied field, denoted in electromagnetism by H:
The net B field, which is the quantity that determines the Larmor frequency, depends on the sum of the
applied field and the magnetization:
where μ0 is a constant of proportionality known as the magnetic permeability of free space. B can be
expressed in terms of the susceptibility as:
The value of the susceptibility varies over several orders of magnitude among different materials, and
can be either positive or negative.
The susceptibility of substances with no unpaired electrons is determined by a phenomenon known as

diamagnetism. Just as a magnetic field exerts a force on any moving charged particle, it also exerts a
force on the electrons in an atom, due to their rotational motion about the nucleus. The force modifies
the electrons' orbital motion in such a way as to produce a weak magnetization in the direction
opposite that of the applied field. The result is a small negative susceptibility in the range -10-7 to -10-5.
Although diamagnetism is a fundamental property of all substances, it can be masked by
paramagnetism or ferromagnetism in materials containing unpaired electrons. Most biological
substances are diamagnetic, except for some proteins that contain metal ions, such as
deoxyhemoglobin, methemoglobin, hemosiderin, and ferritin, which are paramagnetic.
Molecules with unpaired electrons have a net magnetic moment. In paramagnetic substances with no
external magnetic field, the moments are randomly oriented as a result of thermal tumbling. When a
magnetic field is applied, the moments exhibit a slight tendency to align themselves in the direction of
the field, producing a weak net magnetization. The resulting susceptibility is small and positive, in the
20 of 27 15-04-2010 21:41
range 10-7 to 10-3. Many of the clinically approved MRI contrast agents contain paramagnetic ions,
3+ 2+
such as Gd and Mn , which are chelated to an organic ligand to reduce their toxicity.
Ferromagnetic materials contain atoms with magnetic moments that are strongly coupled together. The
interaction causes almost perfect alignment among neighboring atoms, creating domains with
macroscopic magnetization. When an external magnetic field is applied, entire domains align with the
field, resulting in a very large susceptibility (up to ~105). The susceptibility depends on field strength in
a nonlinear and often irreversible way, leaving some remnant magnetization even after the applied field
is removed. Whether or not a material is ferromagnetic depends on its composition and crystal
structure. Titanium, tantalum, platinum, and gold are nonferromagnetic and are often used in the
manufacture of medical devices such as stents and aneurysm clips. While iron, nickel, and cobalt are
ferromagnetic, some of their alloys are not. An example is austenitic stainless steel, which contains a
high percentage of nickel. A mild degree of ferromagnetism can, however, be imparted by
cold-working, for example when the material is bent to form surgical clips. Behavior similar to
ferromagnetism is exhibited by materials such as magnetite (Fe3O4) that contain two chemically
distinct species whose magnetic moments are strongly coupled. Such materials are said to be
"ferrimagnetic."
Particles of ferromagnetic or ferrimagnetic materials with length scales in the nanometer range exhibit
a property called "superparamagnetism." Each of the particles contains a single magnetic domain,
which tends to align in the direction of an applied external field, producing a net magnetization. The
interaction among the particles is weak, however, and the domains become randomly oriented due to
thermal fluctuations when the field is removed, leaving no remnant magnetization. This is analogous to
paramagnetism, except that the magnetic moments are not those of single atoms but of entire
5
nanoparticles, each containing ~10 atoms. The resulting susceptibility may be much higher than for
paramagnetic materials, giving rise to the term superparamagnetism. Due to their weak coupling, the
particles do not agglomerate but can instead be suspended as colloids in aqueous media for storage
and administration. Superparamagnetic iron oxide nanoparticles have interesting properties as MRI
contrast agents, because of their high T2 relaxivity and intravascular distribution.
When materials of differing magnetic susceptibility are present in bulk, the magnetization of each atom
affects the magnetic field experienced by the other atoms. The net B field can be calculated in a
self-consistent manner using the Maxwell equations of electromagnetism, in addition to the Equation
22-17 given earlier. It turns out that the susceptibility of a material affects not only the B field within the
material itself, but also in its vicinity. This can be understood intuitively as follows. One of the Maxwell
equations states that the lines of flux of the B field must be continuous. The only way to satisfy both
this requirement and Equation 22-17 is for the field to become distorted in the presence of a material
of different susceptibility. The flux lines are slightly repelled from a diamagnetic substance and weakly
attracted into a paramagnetic one, as illustrated in Figure 22-28. (Note that the susceptibility values
used in the diagram are greatly exaggerated for illustrative purposes.) As shown in the graphs, the
strength of the magnetic field is modified, both within the material and in its immediate neighborhood.
When such materials are present in an MRI scanner, the resulting variations in magnetic field strength
cause shifts in the Larmor frequency, producing distortion and intensity anomalies in the images.
Small inhomogeneities in the magnetic field result from susceptibility differences that occur naturally in
the body.31 Most tissue is diamagnetic, with susceptibility close to that of water (about -0.7 ppm). By
contrast, air, which is present in the lungs, sinuses, bowel, and certain bones of the skull, has
essentially zero susceptibility. Much more severe susceptibility differences are caused by
ferromagnetic implants, such as dental braces, stents, and surgical clips.
page 594
page 595
21 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-28 The field is distorted in the presence of a material of different susceptibility, being slightly
repelled from a diamagnetic substance (left) and weakly attracted into a paramagnetic one (right). Its
strength, determined by the density of the flux lines, is modified not only within the material but also in
its vicinity (lower graphs). The magenta line indicates the strength of the field far from the magnetized
materials. The susceptibility values used in this illustration are greatly exaggerated, being χ = -0.75 for
the diamagnetic material and χ = 3 for the paramagnetic one.
The form and severity of the resulting artifacts depend on the pulse sequence and may involve two
distinct mechanisms, namely intravoxel dephasing and spatial misregistration. Intravoxel dephasing is
most pronounced on gradient-echo images and arises from magnetic field inhomogeneity within
individual voxels. The inhomogeneity causes variations in the precession frequencies of the protons,
resulting in phase dispersion and signal loss. It produces large voids around ferromagnetic implants
and may also cause signal loss on MR angiograms if the concentration of contrast material is very high
(see later discussion). The effect can, however, be useful in identifying hemorrhage32-37 and cavernous
hemangiomas,38-40 which exhibit signal loss on T2*-weighted images due to the presence of
paramagnetic blood derivatives such as deoxyhemoglobin, methemoglobin, and hemosiderin. It can
similarly be exploited to quantify iron overload in the liver,41,42 due to the paramagnetic properties of
ferritin.
Susceptibility artifacts can also arise through the mechanism of spatial misregistration, which causes
both distortion and intensity anomalies in the image. When a material or tissue of different susceptibility
is present within the body, the magnetic field strength in its vicinity is altered. Signals from tissues
experiencing the largest shifts in field strength are displaced farthest and those from regions
experiencing smaller shifts are displaced by a lesser amount. The resulting image is distorted and
exhibits signal enhancement or "pile-up" where it is compressed and signal loss where it is stretched
out. This gives rise to the characteristic dark voids, often with bright rims, that appear around
ferromagnetic objects on MR images. Figures 22-29 and 22-30 show examples of artifacts caused by
43
an aneurysm clip and a spinal fusion device respectively.
22 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-29 A T1-weighted image exhibits susceptibility artifacts due to a ferromagnetic aneurysm
clip (arrow) and a ventricular shunt catheter (arrowhead). The characteristic dark voids with bright rims
are due to nonlinear spatial misregistration and signal pile-up, resulting from alterations in the
magnetic field strength in the vicinity of the implants. The misregistration occurs primarily in the
frequency-encoding direction (anterior/posterior).
Spatial misregistration in the readout direction results from the effect of magnetic field alterations on
the process of frequency encoding. It can also occur in the through-plane direction due to disruption of
slice selection. In Figure 22-31C, for example, the axial image exhibits intensity pile-up from a
ferromagnetic dental expander located outside the imaging slice, although there is no obvious in-plane
distortion. Alterations in the magnetic field due to the presence of ferromagnetic objects can also
cause incomplete fat saturation, as illustrated in Figure 22-32. All such susceptibility-related artifacts
are exaggerated on higher field scanners (discussed in further detail later).
The susceptibility differences that occur naturally in the body are relatively small, the largest being
between air-filled organs (such as the lungs, bowel, and sinuses) and their surrounding tissues.
Artifacts can usually be avoided in these regions with judicious choices of pulse sequence and
bandwidth. Single-shot EPI, for example, performs poorly in the thorax and abdomen and preferred
44
techniques include fast spin-echo, balanced SSFP, and short-TE gradient-echo. For imaging the lung
itself, T2* weighting must be kept to a minimum and FSE sequences (including SSFSE) are the most
common choice.45 Fast gradient-echo techniques can also be used in the lung provided the echo time
46
is in the submillisecond range.
page 595
page 596
23 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-30 A proton density-weighted FSE image exhibits susceptibility artifacts (arrows) from a
plate and fixation device implanted during anterior interbody fusion surgery.
Add to lightbox
24 of 27 15-04-2010 21:41
Add to lightbox
Add to lightbox
Figure 22-31 Sagittal T1-weighted (A), coronal T2-weighted (B), and axial FLAIR (C) images exhibit
susceptibility artifacts from a dental expander. The coronal image is dramatically distorted, while the
axial image shows signal pile-up (arrow) but no obvious in-plane distortion. The signal pile-up results
from through-plane misregistration.
Susceptibility artifacts from ferromagnetic materials can seriously degrade image quality, even with
optimized techniques.47,48 The source may be external (such as clothing), internal (such as orthopedic
hardware) or more subtle, such as metal or carbide particles left by surgical knives, drills, and suction
tools. Artifacts have also been traced to fragments of fractured heart valves that have entered the
circulation.49 Objects such as hairpins, jewelry, buckles, metal pop fasteners, and zippers should be
removed prior to examination. Cosmetics, including mascara, eyeliner and eye shadow, should also be
removed since they may contain iron oxide particles. Even some hair products can cause artifacts50
(Fig. 22-33). Before entering the MRI suite, subjects should be thoroughly screened for embedded or
implanted metal objects, such as shrapnel, IUDs, and surgical clips. Even those objects that are "MR
safe" may not be "MR compatible" and may prevent the acquisition of diagnostic-quality images. The
design of medical devices with minimal susceptibility is the focus of intensive ongoing research.51-55
25 of 27 15-04-2010 21:41
Further discussion of MR-compatible stents is provided in the later section on artifacts in MR

angiography.
The Magic Angle Phenomenon

T2 relaxation or "spin-spin relaxation" describes the irreversible dephasing of spins that arises from
their interaction with each other. One of the most important sources of T2 relaxation is the dipole-
dipole force, which is a function of the distance between two spins and the relative orientation of their
dipoles. Its strength also depends on the angle θ between the external magnetic field and the
displacement of the spins from each other, through the factor (1-3cos2θ).
page 596
page 597
Add to lightbox
Figure 22-32 A coronal T2-weighted image acquired with a preparatory fat-sat pulse exhibits
susceptibility artifacts and incomplete fat saturation (arrow) from a dental prosthesis.
The properties of the dipole-dipole interaction are responsible for the dramatic differences in
transverse relaxation times that exist between liquids and solids. Liquids tend to have long T2 values,
because the rapid tumbling of their molecules causes fluctuations in the strength of the interaction,
which average almost to zero. Solids on the other hand have much shorter T2 values because their
molecules are not so free to move. An example is collagenous tissue, which normally exhibits very
rapid T2 relaxation, because its highly organized structure restricts the motion of water molecules.
26 of 27 15-04-2010 21:41
Add to lightbox
Figure 22-33 A gadolinium-enhanced T1-weighted image exhibits susceptibility artifacts from a
beeswax hair product containing iron oxide. (From McKinstry RC 3rd et al: Magnetic susceptibility
artifacts on MRI: a hairy situation. Am J Roentgenol 182:532, 2004. Reprinted with permission from
the American Journal of Roentgenology.)
The T2 value of collagen is lengthened, however, if the fibers are oriented at a certain "magic angle" to
the direction of the external field. The magic angle is the value of θ at which the strength of the dipole-
dipole interaction equals zero and is the solution of the equation:
namely θ ≈ 55°. It is believed that the water molecules in collagen preferentially orient themselves such
56
that a line joining the two protons is aligned along the collagen fibers. Consequently, when the fibers
are at 55° to the static magnetic field B0, the protons are effectively decoupled from each other and do
not contribute to transverse relaxation. The result is a lengthening of the T2 value. This occurs in
57
tendons and hyaline cartilage at certain orientations, causing signal enhancement on MR images.
Tendons, for example, normally have a very short T2 of around 250 μs, making them dark even on
short-TE images.58 At the magic angle, however, their relaxation time lengthens to about 22 ms,
producing focal enhancement on proton-density and T1-weighted images (Fig. 22-34).
The increased intensity in tendons oriented at 55° to the B0 field can mimic tendinous degeneration,
tendinitis or frank tear. In cartilage, the enhancement can resemble meniscal tear or degeneration. The
artifacts can be eliminated by increasing the echo time, so that the signal from collagen is suppressed
even at the magic angle.58 Alternatively, the effect can be exploited to achieve high intensity in
collagenous tissues, by deliberately positioning the subject so that the fibers are oriented at 55° to the
B0 field, a technique known as magic angle imaging.59-62
© 2010 Elsevier
27 of 27 15-04-2010 21:41
TECHNIQUE-SPECIFIC ARTIFACTS
Echo-Planar Imaging
page 597
page 598
Add to lightbox
Figure 22-34 A T1-weighted image (TR = 500 ms, TE = 10 ms) of an ankle exhibits focal signal
enhancement in the tendon (arrow) due to the magic angle effect.
Table 22-2. Summary of Physiology- and Subject-Related Artifacts

Origin Manifestations Solutions
Motion Discrete ghosts or smearing in Vacuum cushions
phase-encoding direction Sedation
PROPELLER imaging
ECG gating
Breath-holding
Navigator techniques
Fat suppression or spatial
saturation bands
Swap phase- and frequency-
encoding directions
Single-shot imaging
Band artifacts (single- Use gated segmented
shot/nonsegmented sequences) sequences
Acquire cardiac images during
diastole
Parallel imaging
1 of 41 15-04-2010 21:43
Flow Discrete ghosts in phase- Swap phase- and frequency-

encoding direction encoding directions
Flow compensation
ECG gating
Signal loss due to intravoxel Flow compensation or
dephasing even-echo rephasing
Spatial misregistration of Minimize relevant time delays
oblique flow between phase- and frequency-
encoding and/or slice selection
Use sequence that suppresses
signal from flowing spins
Signal loss due to slow flow Contrast enhancement
Chemical shift Displacement in frequency- Increase readout bandwidth
encoding direction
Fat saturation
Water-selective excitation
Dixon techniques
Signal loss at water/fat Use spin-echo or FSE
interfaces (gradient-echo)
For gradient-echo, use
"in-phase" TE value
Magnetic susceptibility Distortion and signal pile-up Screen subjects thoroughly prior
differences and to scanning
ferromagnetic materials
Signal loss due to phase Remove ferromagnetic materials
cancellation where possible(including
cosmetics)
Use optimal sequence (e.g.,
spin-echo or FSE rather than
gradient-echo or EPI)
Increase readout bandwidth
Encourage use of
MRI-compatible implants
Magic angle Anomalous signal enhancement Increase TE
phenomenon in tendons and cartilage Alter subject positioning
oriented at 55° to B0 field
page 598
page 599
2 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-35 In echo-planar imaging, all the lines of k-space are acquired after a single RF excitation
process. The frequency-encoding gradient is rapidly alternated, while a series of small gradient pulses
is applied in the phase-encoding direction, to produce a zigzag trajectory through k-space. During the
positive lobes of the frequency-encoding gradient (drawn in blue) a single line of k-space is acquired
in the direction of increasing k. During the negative lobes (green) the adjacent line is acquired in the
reverse direction. A phase-encoding gradient pulse (red) is applied during the intervening ramp time,
to increment from one line to the next.
Echo-planar imaging (EPI) is currently the technique of choice for functional MRI (fMRI) and diffusion-
weighted imaging (DWI) in the brain. It is chosen for fMRI because of its high speed and T2*
weighting, and for DWI because of its single-shot readout. In EPI all the lines of k-space are acquired
after the same RF excitation process, which may be either a single pulse, in the case of gradient-echo
EPI (GRE-EPI), or a 90-180° combination, in the case of spin-echo EPI (SE-EPI). By repeatedly
switching the sign of the frequency-encoding gradient and simultaneously applying brief gradient pulses
along the phase-encoding axis, k-space is traversed in a zigzag fashion, with alternate lines being
acquired in opposite directions (Fig. 22-35). The technique is extremely sensitive to magnetic
susceptibility differences within the body and to asymmetry between alternate lines. The former
produces image distortion and associated signal pile-up, predominantly in the phase-encoding
direction, and the latter gives rise to so-called "N/2" or Nyquist ghosts.
Susceptibility
3 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-36 An axial echo-planar image exhibits distortion and signal pile-up (arrows) due to
magnetic susceptibility differences between the brain parenchyma and the pneumatized petrous
bones. The artifacts arise from nonlinear spatial misregistration, which occurs predominantly in the
phase-encoding direction (anterior/posterior), due to the low effective bandwidth of the data
acquisition along that axis.
As discussed earlier, the presence of materials or tissues with different susceptibilities alters the
strength of the magnetic field, producing spatial variations in the Larmor frequency of the protons. In
most conventional Cartesian imaging techniques, this causes spatial misregistration in the frequency-
encoding and slice-select directions, but does not affect the phase-encoding direction to the same
extent. In EPI, however, susceptibility differences produce very severe misregistration in the phase-
encoding direction. The reason is that in EPI, unlike other Cartesian imaging techniques, all the lines of
k-space are acquired after the same RF pulse. The variations in Larmor frequency cause phase
offsets among the spins, which accumulate between successive lines and produce phase-encoding
errors. The resulting displacements are very large, because the time required to traverse k-space in
the phase-encoding direction is much longer than in the frequency-encoding direction. In effect, the
phase-encoding axis behaves like a second readout axis but with much lower bandwidth. The
consequence is that EPI is much more vulnerable to susceptibility artifacts than other imaging
techniques. Even the small susceptibility differences that occur naturally between the air-filled cavities
and surrounding tissues in the head are sufficient to cause distortion and intensity anomalies in EPI
images (Fig. 22-36).
The severity of the susceptibility artifacts can be reduced using methods that increase the effective
bandwidth in the phase-encoding direction, such as parallel imaging63-66 and interleaved segmentation.
Segmented GRE-EPI, also known as FGRE with an echo-train readout (FGRE-ET), is commonly used
in the heart since it is faster than conventional FGRE but less vulnerable to susceptibility artifacts than
67
single-shot EPI. This is an important consideration in cardiac imaging because of the proximity to the
air-filled lungs. Segmentation, however, increases the sensitivity of the technique to motion and
consequently makes it less robust for diffusion imaging.
page 599
4 of 41 15-04-2010 21:43
page 600
Efforts have also been made to correct for susceptibility artifacts in post-processing. This can be done
68
by means of field maps or with the aid of additional acquisitions in which k-space is traversed in the
opposite direction.69 A novel remedy for susceptibility effects in the inferior frontal cortex is the use of
an intraoral diamagnetic passive shim.70
For the same reasons that EPI is vulnerable to susceptibility artifacts, it is also extremely sensitive to
chemical shift. For this reason, commercial EPI sequences generally employ spatial-spectral excitation
pulses to suppress signal from fat.
N/2 Ghosts
In EPI, alternate lines of k-space are acquired in opposite directions. Any asymmetry between the
lines therefore causes a periodic modulation in the data, giving rise to ghosts in the image. Since the
modulation repeats every two k-space lines, the ghosts are separated from the parent image by half
the field-of-view in the phase-encoding direction, or N/2 pixels, where N in this case refers to the
number of pixels along the phase-encoding axis.
Asymmetry between alternate lines can result from a variety of factors. Time delays between the
imaging gradients and the RF receiver, for example, can cause misalignment of the even and odd
echoes. The k-space trajectory can also be altered by imperfections in the gradient pulses themselves,
due to the finite inductive rise times of the gradient coils and eddy currents in the hardware.
Furthermore, since the frequency spectrum of the tissue flips between alternate lines, as a function of
the gradient reversals, any asymmetry in the band-pass filter can contribute to ghosts.
The severity of the ghosts can be substantially reduced by means of a reference scan. This is
incorporated into most commercial EPI sequences and is identical to the acquisition scan except that
the phase-encoding gradient is not applied. The resulting data are used to realign the echoes by
means of phase correction in the Fourier domain. However, some residual ghosting often remains (Fig.
22-37). Methods to suppress the artifacts further include the use of phased-array processing to
71
separate the ghost from the parent image. This is similar to the algorithm used in SENSE to unwrap
images in a reduced field-of-view (described later in the section on parallel imaging).
Diffusion-Weighted Echo-Planar Imaging
5 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-37 An axial echo-planar image of an infant's brain exhibits N/2 or Nyquist ghosts (arrows),
which are displaced by half the field-of-view in the phase-encoding direction (anterior/posterior). The
image was acquired using a standard EPI sequence incorporating a reference scan.
Diffusion refers to the random motion of molecules associated with their thermal energy. Since
diffusion of water in intact tissues is constrained by the cell walls, measurements of the apparent
diffusion coefficient in vivo provide information about the integrity and structure of the tissues. Necrosis,
for example, increases the diffusion coefficient, due to breakdown of the cell membranes. In white
matter tracts the diffusivity is anisotropic (direction dependent), reflecting the fact that water can
diffuse more freely along the fibers than across them.
MR pulse sequences can be made sensitive to diffusion by the addition of very strong magnetic field
gradients. The gradients are paired in such a way that stationary spins are not affected, while spins
moving along the diffusion-sensitizing gradients receive a large phase shift. The incoherent component
of the motion, resulting from diffusion, causes intravoxel dephasing, leading to signal loss. The
fractional decrease in signal intensity can be used to calculate the apparent diffusion coefficient in the
direction of the gradients. By acquiring several images with diffusion gradients in different directions,
estimates can be obtained of the average diffusivity and the diffusion anisotropy.
Because diffusion-weighted sequences are extremely sensitive to motion, it is essential to avoid any
random bulk movement between successive lines of k-space, since this would cause very severe ghost
artifacts. Currently the most common approach is to use an echo-planar readout, in which all the lines
are acquired in rapid succession after a single diffusion-weighted excitation process (Fig. 22-38).
Diffusion-weighted EPI suffers from the same artifacts as echo-planar imaging, including susceptibility-
related distortion, signal pile-up, and N/2 ghosts. In addition, the presence of the strong diffusion-
sensitizing gradients causes eddy current-related distortion and increases the sensitivity of the
sequence to bulk motion.
page 600
page 601
Add to lightbox
Figure 22-38 The technique most commonly used for diffusion-weighted imaging consists of a
spin-echo EPI sequence with strong diffusion-sensitizing gradients bracketing the refocusing pulse.
Rapid switching of the diffusion-sensitizing gradients produces eddy currents, which persist during the
EPI readout. The eddy currents create residual magnetic field gradients, which disrupt the phase-
encoding process, causing distortion in the image. The distortion is a combination of displacement,
6 of 41 15-04-2010 21:43
stretching and shearing, according to the components of the diffusion-sensitizing gradients in the
through-slice, phase-encoding and frequency-encoding directions respectively.
Eddy Current-Related Misregistration
Add to lightbox
Figure 22-39 A fractional anisotropy map, calculated pixel-by-pixel from a set of diffusion tensor
images, exhibits artifacts along tissue interfaces. The artifacts arise from misregistration among the
source images, which occurs because of eddy current-induced distortion.
Diffusion-weighted imaging requires the use of very strong magnetic field gradients, which must be
switched on and off as fast as hardware and physiologic constraints allow. The rapid switching of such
strong gradients induces eddy currents in the conducting surfaces of the scanner itself, such as the
cryostat and RF coils. The eddy currents in turn produce small residual magnetic field gradients, which
persist during the acquisition period (see Fig. 22-38). Since the echo-planar readout is extremely
sensitive to variations in magnetic field strength, the residual gradients cause appreciable
misregistration in the phase-encoding direction. The induced gradients are oriented along the same
axis as the diffusion-sensitizing gradients that produced them and affect the image accordingly. If they
lie in the through-slice direction, all the tissue in the imaging plane undergoes an identical shift in
Larmor frequency and the image is displaced uniformly. If the gradients are oriented in the phase-
encoding direction, the resulting displacement varies with distance along the phase-encoding axis and
the image is stretched. If the gradients are in the frequency-encoding direction, the displacement
depends on distance along the frequency-encoding axis and the result is shearing. Finally, if the
gradients have components in multiple directions, the distortion will be a combination of displacement,
stretching, and shearing.
The effect of the distortion is most noticeable when the source images are combined to produce maps
of various tissue parameters, such as the average diffusion coefficient (ADC) and the diffusion
anisotropy. Since the distortion depends on the direction of the diffusion-sensitizing gradients, the
source images will each be distorted in slightly different ways and therefore not perfectly aligned. The
misalignment causes artifacts in the ADC and anisotropy maps, particularly near tissue interfaces (Fig.
7 of 41 15-04-2010 21:43
22-39).
The primary line of defense against eddy current-related distortion is optimization of the hardware.
Most clinical scanners incorporate self-shielded gradient coils, designed to minimize flux penetration
into other conducting elements in the bore, thereby reducing the eddy currents that they will sustain.
The form of the gradient waveforms can also be modified to compensate for residual eddy currents
and this requires accurate on-site calibration by the field engineer (see section on hardware-related
artifacts).
page 601
page 602
The effect of the eddy currents can be further reduced by appropriate sequence design. Since the
degree of image distortion depends on the net amplitude of the eddy currents produced by all the
diffusion gradient pulses, substantial improvements can be achieved with a dual spin-echo
72
arrangement, in which the eddy currents from different pulses largely cancel each other out. Parallel
imaging techniques can also reduce distortion, due to both eddy currents and susceptibility
differences.65
Various image-based correction methods have also been proposed. Maps of the eddy current-induced
magnetic field gradients can be obtained during a calibration scan, for example, and used to
compensate for distortion with the aid of theoretical models.73 Empirical registration schemes can
correct for both image distortion and in-plane shifts in patient position. 74 The robustness of the
methods can be improved by acquiring a second set of images with the polarity of the diffusion
gradients reversed.75 Since the two sets of images have identical signal contrast but opposite
distortion, the degree of distortion can be accurately calculated by pairwise comparison of the images.
A completely different approach is to forego the use of echo-planar acquisitions altogether and to
adopt a multishot FSE-based PROPELLER technique for diffusion imaging. The refocusing pulses
make the technique much less sensitive to field inhomogeneity, including the effect of eddy
19-21
currents.
Bulk Motion
In EPI the acquisition time is sufficiently short that patient motion does not usually affect the quality of
individual images (although it can cause misregistration among different frames). The addition of large
gradient pulses for diffusion imaging, however, increases the sensitivity of the technique to bulk motion
and can cause artifacts or signal loss. The diffusion-sensitizing gradients alter the phase of the spins
by an amount proportional to their velocity. If the motion has a rotational component, the velocity will
vary uniformly with position, producing a linear phase modulation across the tissue. DW-EPI however is
typically performed using partial Fourier techniques, which assume that any phase variation across the
tissue is very slow. The phase modulation due to motion may violate this assumption, causing fine
intensity oscillations across the image (Fig. 22-40). If the phase modulation exceeds the highest
spatial-frequency that is sampled, it will cause signal loss.
Steady-State Free Precession

In fully balanced steady-state free precession (SSFP) imaging (also known by the vendor-specific
acronyms true-FISP, FIESTA, and balanced FFE), the transverse magnetization is preserved between
successive phase-encoding lines, resulting in much higher signal than is possible with spoiled
gradient-echo sequences of similar scan time. The drawback is dramatically increased sensitivity to
off-resonance effects, which can result from inhomogeneity in the applied magnetic field, susceptibility
differences among tissues, and chemical shift.
Static Stripe Artifacts
8 of 41 15-04-2010 21:43
Add to lightbox
Add to lightbox
page 602
page 603
Figure 22-40 Diffusion images of an infant, acquired at adjacent slices. Image (A) exhibits fine
intensity ripples in the phase-encoding direction (anterior/posterior) as a result of head motion. Image
(B) is unaffected.
9 of 41 15-04-2010 21:43
Add to lightbox
Add to lightbox
Figure 22-41 The amplitude and phase of the transverse magnetization as a function of precession
angle ϕ for balanced steady-state free precession cine imaging. The signal from stationary spins
(heavy black line) drops almost to zero at the off-resonance points, where the precession angle is an
odd multiple of π. The solid blue and dashed red lines show representative curves for slow and fast
flow respectively. In this context, the flow can be regarded as slow if the local precession angle
changes by much less than 2π during the relaxation time of the fluid. The calculations were performed
with a flip angle of 45°, a repetition time of TR = 3.3 ms, and relaxation times of T1 = 2.5 s and T2 =
1.9 s (approximating those of water). (From Storey P et al: Flow artifacts in steady-state free
precession cine imaging. Magn Reson Med 51:115-122, 2004. Copyright © 2004 Wiley-Liss Inc.
Reprinted by permission of Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc.)
Any difference between the precession frequency of protons in the tissue and the center frequency of
the scanner causes the phase of the transverse magnetization to change with time. The phase change
accumulated between successive RF excitation pulses is known as the precession angle and is
proportional to the frequency offset ∆f and the repetition time TR of the sequence:
After many TR intervals, the magnetization of the tissue reaches a "quasi-steady state"; although it
undergoes excitation and precession within each cycle, it always returns to the same state at the
beginning of the next cycle. The value of the steady-state magnetization depends on both the flip angle
of the excitation pulses and the precession angle of the spins (Fig. 22-41). In particular, if the
precession angle is an odd multiple of π:
then the transverse magnetization exactly reverses direction
over course of the TR period and the signal falls almost to zero. This "off-resonance condition" is met
at points where the precession frequency of the spins differs from the center frequency of the scanner
by:
or any odd multiple thereof. Because the precession frequency is proportional to the local magnetic
10 of 41 15-04-2010 21:43
field, the off-resonant points occur along lines, which are effectively contour lines of the magnetic field.
They appear as dark stripes in the image, usually near the edge of the field-of-view where the applied
magnetic field is least homogeneous. They can be avoided in the region of interest by performing a
local shim and by minimizing the repetition time TR of the sequence. Even with good shimming,
however, the off-resonance stripe artifacts may still occur in fat (Fig. 22-42), because of the chemical
shift between lipid and water.
Dark Flow Artifacts
Add to lightbox
page 603
page 604
Figure 22-42 An image acquired with a balanced SSFP cine technique exhibits off-resonance stripe
artifacts in subcutaneous fat.
Add to lightbox
11 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-43 An image acquired with a balanced SSFP cine technique exhibits a dark flow artifact in
the ventricle (arrow), due to motion of blood through an off-resonant point in the magnetic field (A). The
field is off-resonance most likely because the scanner tuned erroneously to the lipid peak instead of
the water peak. After manual tuning, the flow artifact disappears (B). Note however that the pericardial
fat is now off-resonance (arrowhead). (From Storey P et al: Flow artifacts in steady-state free
precession cine imaging. Magn Reson Med 51:115-122, 2004. Copyright © 2004 Wiley-Liss Inc.
Reprinted by permission of Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc.)
The establishment of a quasi-steady state depends on the continuity of the RF excitation train, and on
the constancy of the precession angle from cycle to cycle. If the spins are moving, either or both of
these conditions may be violated. If motion occurs perpendicular to the imaging plane, for example, the
76
spins may not experience sufficient RF excitations to reach a steady state. Alternatively, if the spins
move through regions of inhomogeneous magnetic field but remain within the imaging volume, their
precession angle will change with successive TR intervals. This also will prevent the magnetization
from reaching a steady state (see Fig. 22-41). The result is particularly dramatic when the spins move
across a point satisfying the off-resonance condition. The amplitude of their magnetization does not
then recover until much further downstream, leaving a dark flow artifact,77,78 as shown in Figure 22-43.
The artifact is most commonly seen in cardiac images of large patients, where the scanner may
erroneously tune to the lipid peak instead of the water peak. The frequency shift between water and
-1
fat at 1.5 T is about 220 Hz, which is close to the off-resonance condition ½TR for typical repetition
times on state-of-the-art scanners. If the scanner tunes to the fat peak, water protons may then be off
resonance near the center of the field-of-view. Blood that crosses an off-resonance point as it flows
into the ventricles will lose signal. The resulting flow artifact darkens the blood pool downstream and
may obscure the endocardial border. It can be eliminated by performing a local shim or manually tuning
the scanner frequency to the water peak.
Magnetic Resonance Angiography

The techniques used most commonly for magnetic resonance angiography exploit either time-of-flight
(inflow) effects or the T1 relaxivity of intravenously injected contrast agents, to enhance the signal of
blood above that of surrounding tissue.79 In both cases, the resulting 3D image data are usually
reformatted using maximum intensity projections (MIPs) to facilitate visualization of the vessels.
Artifacts can be introduced during the initial data acquisition and in the subsequent image processing.
Time-of-Flight Angiography
Time-of-flight techniques use fast gradient-echo sequences with high flip angles and short repetition
times to saturate the magnetization of stationary spins. Inflowing blood, which is fully magnetized,
produces relatively higher signal than stationary tissue, creating signal contrast between the vessel
12 of 41 15-04-2010 21:43
lumen and its surroundings. Imaging can be performed using a stack of 2D slices or thin 3D slabs,
each approximately perpendicular to the direction of flow. Spatial saturation bands can be placed distal
or proximal to the imaging volume to permit selective visualization of arteries or veins respectively.
Add to lightbox
Add to lightbox
page 604
page 605
Figure 22-44 The MIP of a 2D time-of-flight MRA (A) exhibits signal loss at points where the vessels
lie within the imaging plane (arrows). Comparison with a USPIO-enhanced MRA (B) shows that the
vessels are patent at these points. Note however the uniformly poor vascular signal in the USPIO-
enhanced image, which is due to the high susceptibility of the iron oxide nanoparticles.
Since the technique relies on the rapid passage of blood through the imaging volume, false or
exaggerated stenoses may occur in situations of slow or turbulent flow, in tortuous vessels, and in
those that lie within the imaging plane (Fig. 22-44). To avoid misdiagnosis, it can be useful to acquire a
contrast-enhanced MRA in addition to the time-of-flight study. In 3D time-of-flight imaging, saturation
effects can also cause progressive signal loss along the direction of flow if the slabs are too thick (Fig.
22-45).
Two-dimensional time-of-flight MR angiograms involve the acquisition of many contiguous slices and
typically take longer than a breath-hold to acquire. Respiratory motion during the scan can cause
misregistration among adjacent slices, resulting in apparent tortuosity of the vessels (Fig. 22-46).
When the artifacts affect the vessels of interest, it may be advisable to acquire a breath-hold contrast-
enhanced MRA.
13 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-45 The MIP of a 3D time-of-flight MRA exhibits progressive loss of signal across the
imaging slabs, due to saturation of the moving spins. The effect is particularly noticeable in the slower
flowing veins and can be avoided by using thinner slabs.
Add to lightbox
Figure 22-46 The MIP of a 2D time-of-flight MRA of the neck exhibits artifacts in the vessels where
they pass through the upper thorax (arrow). The artifacts are due to respiratory motion, which causes
misregistration among the source images.
Artifacts can also arise from pulsatile flow, producing ghosts on the source images, which appear as
discontinuous replicas of the vessels in the MIPs (Fig. 22-47). These can be avoided with ECG gating
or contrast enhancement.
The staircase appearance in Figure 22-48 is a discretization artifact and can be minimized by reducing
the slice thickness.
Contrast-Enhanced Angiography
Contrast-enhanced angiography enables large territories of vascular anatomy to be imaged in a short
period of time and is particularly advantageous for scans that require breath-holding. It employs fast
3D gradient-echo sequences with high flip angles and short repetition times, and exploits the
T1-shortening effect of exogenous contrast materials to enhance the signal of blood above that of
surrounding tissue. The background can be further suppressed by subtracting baseline data from the
post-contrast images. The technique is vulnerable, however, to artifacts related to mistiming and
susceptibility differences.80
14 of 41 15-04-2010 21:43
Add to lightbox
Add to lightbox
page 605
page 606
Figure 22-47 Ghost artifacts (arrows) due to pulsatile blood flow appear in a source image (A) of a 2D
time-of-flight MRA. They are manifested as discontinuous replicas of the vessels on the maximum
intensity projection (B).
Add to lightbox
15 of 41 15-04-2010 21:43
Figure 22-48 The MIP of a 2D time-of-flight MRA exhibits a staircase appearance (arrow), due to the
finite slice thickness of the source images.
Accurate timing of the data acquisition with respect to contrast administration is particularly critical
when arterial-venous separation is required. To obtain an arteriogram, the entire 3D volume of interest
must be imaged within the arterial phase of the contrast agent. Imaging too late results in venous
contamination (Fig. 22-49), whereas imaging too early risks missing the contrast agent altogether. If
the acquisition is started slightly prematurely, some lines of k-space may be collected before arrival of
80,81
the contrast material and some afterwards, resulting in a "Maki artifact" (Fig. 22-50). The use of a
test bolus, interactive fluoroscopic trigger or automated bolus detection algorithm (e.g., SmartPrep83
82
79
or CareBolus) is helpful for timing the data collection accurately.
The duration of the arterial phase, however, limits the length of the acquisition (as do the constraints of
breath-holding, in the case of thoracic imaging). This restricts the achievable spatial resolution and can
give rise to Gibbs artifacts (Fig. 22-51). Methods to circumvent the trade-off between spatial and
temporal resolution include the use of parallel imaging, 10-12 TRICKS (time-resolved imaging of contrast
kinetics),84,85 and undersampled radial imaging.86,87 TRICKS achieves greater temporal resolution by
increasing the sampling rate for lower spatial frequencies, interpolating the k-space views, and
zero-filling the higher-order data points in the slice-encoding dimension. The greater temporal
resolution improves the chances of capturing the arterial phase and permits observation of the
passage of contrast material. In undersampled radial imaging, the increased sampling rate for lower
spatial frequencies is an inherent feature of the k-space trajectory pattern (see later discussion on
non-Cartesian imaging).
Add to lightbox
Figure 22-49 The MIP of a 3D contrast-enhanced MRA exhibits venous contamination as a result of
acquiring the data too late.
16 of 41 15-04-2010 21:43
Add to lightbox
page 606
page 607
Figure 22-50 A source image from a 3D contrast-enhanced MRA exhibits a Maki artifact as a result of
starting the data acquisition too early. Since some of the lines of k-space were collected before arrival
of the contrast agent, the arteries display anomalous intensity oscillations.
Add to lightbox
17 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-51 A source image (A) and a MIP (B) from a 3D contrast-enhanced MRA exhibit Gibbs
ringing (arrows) due to inadequate spatial resolution. The artifacts extend parallel to the vessel walls
along their entire length.
18 of 41 15-04-2010 21:43
Add to lightbox
page 607
page 608
Figure 22-52 The lumen of the right renal artery in a contrast-enhanced MRA is obscured due to the
presence of a stent (arrow).
Another common cause of artifacts in MR angiograms is metallic stents (Fig. 22-52). The artifacts
obscure the vascular lumen within the stent itself, which is often where evaluation is most critical, given
the high incidence of in-stent restenosis. Artifacts can arise from both susceptibility effects and RF
shielding. The shielding is a result of eddy currents induced in the electrically conducting wire mesh of
the stent. The eddy currents produce an opposing RF field, which reduces the effective flip angle of the
88
excitation pulses within the stent and also attenuates the emitted signal. The severity of the artifacts
depends on both the material and geometry of the stent51,89,90 as well as its orientation with respect to
91
the static magnetic field. Although stainless steel has favorable mechanical and biological properties
for stent design, it produces extensive signal voids on MR images. Alloys such as NiTinol (a nickel-
titanium alloy) have much lower susceptibility but may still cause RF shielding. Progress in lumen
visualization is being made with the introduction of MR-invisible stents92-94 and the optimization of MRA
95-97
imaging techniques, including the use of higher flip angles to compensate for RF shielding. An
98,99
alternative approach is the design of "active" stents that behave as local RF amplifiers. A stent can
be made into a resonator at the Larmor frequency by the addition of an appropriate capacitor.
Following implantation, it can then be inductively coupled to an external RF coil without the aid of wires.
The coupling amplifies both the excitation pulse within the stent and the detection efficiency of the
emitted signal, thereby improving visualization of the lumen.
Susceptibility artifacts can also arise from the contrast material itself, in regions where it is present in
high concentration.79,80,100 The artifacts occur most commonly in the vicinity of the subclavian and
brachiocephalic veins, ipsilateral to the site of intravenous injection. If the initial bolus has not yet
cleared these veins by the time of data acquisition, the strong paramagnetic properties of the
concentrated material cause loss of signal in neighboring vessels (Fig. 22-53). Diluting the bolus and
19 of 41 15-04-2010 21:43
shortening the echo time help to reduce the incidence of artifacts.101 Comparison with later images is
also useful in distinguishing artifacts from genuine occlusions. Signal attenuation throughout the
vasculature can occur with superparamagnetic iron oxide agents, due to their very high susceptibility
(see Fig. 22-44B). The dose of such agents must be optimized for maximum signal.
Add to lightbox
Figure 22-53 A contrast-enhanced MRA exhibits susceptibility-related signal loss in the subclavian
artery (arrow), due to the high concentration of contrast material in the neighboring subclavian vein.
Post-Processing
The source images acquired in time-of-flight and contrast-enhanced angiography exams are typically
rendered into a series of MIPs of the vascular anatomy from different angles. The post-processing
itself can introduce artifacts and reference should always be made to the source images to confirm
apparent occlusions and stenoses. Pseudo-occlusions can arise when sections of vessels are excluded
from the MIP, either because they were omitted from the original imaging volume or because they
were inadvertently cropped from the projection volume along with extraneous tissues. Stenoses are
often exaggerated on MIPs because of suboptimal intensity thresholds. Narrowed vessels are the
most commonly affected, since their signal may be reduced due to partial volume averaging or spin
dephasing, and may not substantially exceed the background intensity.
Parallel Imaging
In parallel imaging, signal detection is performed using an RF receiver with multiple coil elements,
which are distributed around the anatomy of interest. Since each coil is most sensitive to the region of
tissue closest to itself, the signals obtained with the different elements contain complementary spatial
information and provide a means to acquire data in parallel. This reduces the number of phase-
encoding lines required to achieve the desired resolution, thereby shortening the scan time. 102 The
acceleration factor is limited by the number of coil elements and the degree of overlap in their
sensitivity profiles.
In the original SMASH formulation,103 k-space is sampled sparsely and information about the spatial
harmonic content of the coil sensitivity profiles is used to estimate the values of the unsampled lines. In
the SENSE formulation104 the acquisition is performed with a reduced field-of-view and maps of the
20 of 41 15-04-2010 21:43
coil sensitivity profiles are used to unwrap the resulting aliased images. While the first method is
implemented in k-space and the second in the image domain, the techniques are essentially equivalent,
since reducing the k-space sampling density is identical to reducing the field-of-view.
Parallel imaging is vulnerable to two characteristic types of artifacts, namely inhomogeneous

104,105
noise and residual aliasing. Although they are inherent to any formulation of the method, they are
perhaps easier to explain in the SENSE framework.
Inhomogeneous Noise
page 608
page 609
In the SENSE technique, data acquisition is performed using a reduced field-of-view in the phase-
encoding direction and separate images are reconstructed from each of the coil elements. All the
single-coil images exhibit wraparound, which can be multifold, depending on the acceleration factor.
The relative intensity of the aliased structures differs among the images, however, due to the unique
spatial profiles of the coils. By calculating the appropriate weighted combinations of the pixel intensities
in each of the source images, the aliased structures can be "unwrapped" to obtain a final image with a
full field-of-view. The weightings are determined from the sensitivity profiles of the coil elements and
vary from pixel to pixel.
The way in which noise in the source images propagates to the final image depends on the magnitude
of the weightings used and on the number of overlapping structures at any given point. The weightings
vary from pixel to pixel, as do the number of overlapping structures, since one or more of the aliased
points may fall in the background. As a result, the noise in the final image is spatially inhomogeneous
and can be greatly amplified in regions where multiple structures overlap or where the reconstruction is
ill conditioned (Fig. 22-54). Excessive noise can be avoided by judicious placement of the coil
elements, in combination with an adequate field-of-view and a conservative acceleration factor.
Add to lightbox
21 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-54 An image acquired using the SENSE technique, with a four-element torso phased-array
coil and an acceleration factor of 2, exhibits inhomogeneous noise (A). The noise is amplified in the
central region (dashed ellipse) because the phase-FOV (aspect ratio) of the image is insufficient to
encompass the entire cross-section of the abdomen. The SENSE algorithm therefore has three
overlapping structures to unwrap in the center rather than two, and is less stable there. By increasing
the phase FOV to cover the entire abdomen (B), the SNR in the central region is improved.
Residual Aliasing
To unwrap the source images correctly requires accurate knowledge about the coil sensitivity profiles
within the imaging volume. Any factors that compromise the accuracy of the sensitivity maps can
produce residual aliasing in the final image (Fig. 22-55). A common cause is displacement of the
anatomy between the calibration and diagnostic scans, particularly when breath-holding is required.
Ghost artifacts, wraparound, and analog-to-digital converter overflow in the calibration images may
also be responsible. Even small inaccuracies in the calibration are sufficient to cause substantial
aliasing if the acceleration factor is too large. Remedies therefore include repeating the calibration and
reducing the acceleration factor.
High-Field Magnetic Resonance Imaging

Although 1.5 T remains the standard magnetic field strength for clinical MRI at the present time, higher
field scanners are gaining popularity. The higher field strengths offer improvements in signal-to-noise
ratio but pose new challenges on several fronts, including higher power deposition, longer T1 relaxation
times, and an increase in the incidence and severity of image artifacts.106 In particular, the effects of
chemical shift and magnetic susceptibility are exaggerated at higher field, aggravating existing artifacts
and introducing additional artifacts that would not occur at lower fields. Inhomogeneity in the B1 field
also becomes a significant problem above 1.5 T, due to the shorter RF wavelength.107 The B1
inhomogeneity causes variations in the amplitude of the RF excitations, producing intensity modulations
in the images and compromising techniques such as inversion recovery and magnetization transfer that
require very accurate or extremely large flip angles.
22 of 41 15-04-2010 21:43
Add to lightbox
page 609
page 610
Figure 22-55 An image acquired using the SENSE technique displays residual aliasing (arrow), due to
displacement of the anterior coil element between calibration and image acquisition.
Add to lightbox
23 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-56 A comparison of images acquired at 1.5 T (A) and 3 T (B) demonstrates the increase in
chemical shift displacement with field strength. The effect is particularly conspicuous around the
kidneys (arrows). The images were acquired with identical bandwidth and field-of-view (15.6 kHz and
24 × 18 cm respectively).
Chemical Shift and Magnetic Susceptibility

Chemical shift and magnetic susceptibility differences alter the proton Larmor frequency by amounts
proportional to the magnetic field strength B0. The resulting artifacts, including spatial misregistration,
distortion, and signal pile-up or voids, are therefore aggravated at higher field. The chemical shift
displacement of fat with respect to water, for example, is doubled at 3 T compared to 1.5 T for a given
bandwidth (Fig. 22-56). Although increasing the bandwidth compensates for the greater frequency
shift, it also results in a loss of signal-to-noise ratio, which largely eliminates the advantage of imaging
at higher field.
Add to lightbox
Figure 22-57 Axial T1-weighted FSE images acquired at 3 T exhibit focal hyperintensities (arrows)
due to differences in magnetic susceptibility between air-filled cavities and surrounding tissue. The
four images are taken from contiguous slices (in order A-D from superior to inferior). Note that the
hyperintensity in image A occurs immediately above the sphenoid sinus (arrowhead) in image B.
Similarly, the hyperintensity in image C is located just over the pneumatized petrous bone (arrowhead)
in image D. The hyperintensities represent signal pile-up in the through-plane direction and are
analogous to the more dramatic example shown in Figure 22-31C, which was caused by a dental
expander.
Susceptibility artifacts also appear in situations where they would not have occurred at lower field
strengths. At 3 T, for example, even standard anatomic images may exhibit regions of focal
enhancement due to the small susceptibility differences that exist between the air-filled sinuses and
surrounding tissues of the head (Fig. 22-57). The hyperintensities are due to signal pile-up in the
through-plane direction and may mimic lesions such as ischemic changes. Reference should be made
24 of 41 15-04-2010 21:43
to the adjacent slices to avoid misdiagnosis.
RF Inhomogeneity
The RF coils used for excitation and signal reception must be tuned to the Larmor frequency of the
protons, which is proportional to B0. Since each coil can operate only within a narrow range of
frequencies, different coils must be manufactured for each field strength. For fields above 1.5 T it
becomes increasingly difficult to design coils that provide adequate B1 homogeneity. The reasons are
twofold, namely shorter wavelength and reduced RF penetration.108
page 610
page 611
Add to lightbox
Add to lightbox
Figure 22-58 Inhomogeneity in the B1 field of a standard 3 T birdcage head coil causes signal
modulation across the brain. Image A demonstrates signal fall-off at the base, while image B exhibits
enhancement at the center with respect to the edges. Some compensation can be provided by
surface-coil intensity correction (SCIC) algorithms.
The wavelength of the RF field is inversely related to the Larmor frequency and therefore decreases at
higher field. It is further shortened within the body as compared to air, due to the high dielectric
constant of tissue.109 The net result is that at 3 T and above, the wavelength in tissue becomes
comparable to or smaller than the dimensions of the human body, causing standing waves in the RF
field, with "hot-spots" at the antinodes. In brain imaging, this is typically manifested as a central
brightening artifact, as illustrated in Figure 22-58 (and discussed further in Chapter 18). The field
profile is also affected by RF attenuation, which is caused by eddy currents within the tissue itself. The
attenuation increases with frequency, causing reduced RF penetration at higher field strengths. 109 The
dependence of the RF field on tissue parameters and geometry considerably complicates the design of
coils for high-field applications.
25 of 41 15-04-2010 21:43
Inhomogeneity in the B1 field causes variations in the effective flip angle across the tissue, resulting in
anomalous signal modulation within individual images and among different slices. Partial compensation
can be achieved in post-processing using surface coil intensity correction (SCIC) algorithms. The
algorithms use a priori knowledge of the coil characteristics, in combination with information about the
low frequency intensity modulation of the image itself, to derive a correction map. They are
inadequate, however, for imaging techniques that require very accurate flip angles, such as inversion
recovery, or very high flip angles, such as magnetization transfer (Fig. 22-59). The solution in these
110,111
cases lies in improved coil designs.
Non-Cartesian Sampling
Add to lightbox
Figure 22-59 A magnetization transfer image acquired with a standard 3 T birdcage head coil exhibits
much stronger signal saturation at the center of the brain than towards the edges, due to
inhomogeneity in the B1 field.
Add to lightbox
page 611
26 of 41 15-04-2010 21:43
page 612
Figure 22-60 An illustration of k-space trajectories for the common non-Cartesian sampling
techniques. Color differences are used only as an aid to visualization.
The discussion of artifacts thus far has assumed conventional Cartesian sampling, in which data are
acquired along lines of a rectangular grid in k-space (see Fig. 22-2). Alternative sampling schemes
may, however, offer advantages in certain applications. The most commonly used non-Cartesian
sampling techniques are radial, spiral, and PROPELLER imaging (illustrated in Fig. 22-60). Radial
imaging is also called projection imaging or projection reconstruction. The radial and PROPELLER
techniques are less vulnerable to motion artifacts than standard Cartesian imaging because they
oversample the central region of k-space. PROPELLER imaging17 in particular is designed to
compensate for motion by repeatedly sampling a small region about the center of k-space and using
the relative phase information to correct or reject data. Spiral imaging offers reduced sensitivity to flow
and provides a very efficient way of covering k-space; it can even be run in single-shot mode.
However, it is more sensitive to off-resonance effects due to the correspondingly longer readout times.
In non-Cartesian techniques, image reconstruction is usually performed by interpolating the data onto a
Cartesian grid, with appropriate weightings to correct for the nonuniform sampling density. This
procedure, known as "gridding", is followed by the application of a fast Fourier transform. Filtered
backprojection offers an alternative reconstruction method for radial imaging, but is less commonly
used.
In conventional Cartesian imaging, most artifacts are influenced greatly by the mechanisms of phase
and frequency encoding. Motion, for example, produces ghosting in the phase-encoding direction,
whereas chemical shift causes displacements in the frequency-encoding direction. In non-Cartesian
techniques, however, there are no uniquely defined phase- and frequency-encoding directions and the
manifestation of many artifacts is correspondingly more complex. Because of the inherent symmetry of
the sampling geometry, artifacts in radial imaging often take the form of radiating streaks, while in
spiral imaging they may appear as circular smearing. Blurring is common in both methods. Additional
sources of artifacts also arise in non-Cartesian techniques, including undersampling and gradient timing
errors.
Undersampling
As discussed at the beginning of this chapter, the sampling density in k-space limits the field-of-view
over which an image can be free from aliasing. In many non-Cartesian imaging techniques, however,
the sampling density is not uniform across k-space; typically the center is sampled more densely than
the edges. Undersampling refers to the situation in which the outer regions of k-space are sampled at
a density lower than that used to determine the field-of-view of the reconstructed image.
Undersampling is a common way to reduce scan time, particularly in radial imaging. The scan time
required for a radial acquisition is proportional to the number of projections Np, and therefore imposes
a constraint on the azimuthal sampling density. By convention, however, the field-of-view of the
reconstructed image is determined by the radial sampling density, which in turn depends on the number
of data points acquired per projection Nr. The high spatial frequencies are therefore undersampled if:
Note that this relation assumes that each trajectory traverses the full diameter of k-space. It is also
possible to start the trajectories at the center of k-space and sample along only the radii. While this
allows extremely short echo times, it also results in a twofold increase in scan time.
27 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-61 The magnitude of the point spread function for a radial acquisition with 64 projections,
gridded onto a 256 × 256 matrix. The window intensity is chosen to accentuate the artifacts. The
radius of the artifact-free region is equal to (2Np/πNr)×FOV, where FOV is the field-of-view of the
reconstructed image (determined from the radial sampling density).
In radial imaging, it is not uncommon to reconstruct frames with as few as 64 or even 32 projections,
for high-speed applications such as interventional scanning112 and MR angiography.113 This produces
aliasing in the image, even from objects within the field-of-view. The aliasing is manifested not as
wraparound, however, but rather as streaks and diffuse noise. This can be understood by considering
the point spread function for a radial acquisition (Fig. 22-61), which exhibits an artifact-free region
centered on the signal source and a spoke-like pattern beyond. The artifact-free region is known as
113
the "reduced field-of-view" and has a radius equal to:
where FOV is the size of the reconstructed image. The intensity and angular separation of the spokes
increase as the number of projections is reduced.
Since each point source in the object produces its own point spread function in the image, it is properly
reconstructed only within its own local reduced field-of-view and contributes artifacts outside that
region. The combined artifacts from all the point sources may appear as streaks or diffuse noise in the
image, depending on the signal intensity and distribution of the sources within the object (Fig. 22-62).
As the number of projections is reduced, the streaks become more numerous and intense, and the
amount of diffuse noise increases, producing an apparent background signal in regions of the image
that would otherwise be dark (Fig. 22-63).
page 612
page 613
28 of 41 15-04-2010 21:43
Add to lightbox
Add to lightbox
Figure 22-62 A radial MRA of the head exhibits streak artifacts and diffuse noise due to
undersampling (A). The artifacts can be eliminated by increasing the number of projections (B).
(Courtesy of Walter Block PhD)
Tissue Outside the FOV

In Cartesian imaging, signal from tissue lying outside the field-of-view in the frequency-encoding
direction is eliminated by frequency filtering prior to reconstruction. In non-Cartesian imaging, however,
there is no uniquely defined frequency-encoding direction and tissue lying outside the field-of-view in
any direction can cause artifacts.
In radial imaging, the artifacts occur through two distinct mechanisms. The first is analogous to
29 of 41 15-04-2010 21:43
undersampling; each signal source outside the field-of-view produces a point spread function that is
nonzero inside the field-of-view, contributing streaks and diffuse noise to the image. Extremely bright
streaks may occur when signal is detected from peripheral sources beyond the range of the transverse
gradient coils. Since that signal is not spatially encoded, it collapses to a point on the longitudinal axis
of the scanner. If this point lies outside the field-of-view, it produces a very bright aliasing pattern in the
image (Fig. 22-64). The phenomenon is similar to the so-called "star artifact", discussed later in the
section on gradient nonlinearity. It is most commonly observed when the body coil is employed for
signal reception and can be avoided by using localized RF coils with a more limited range of sensitivity.
Add to lightbox
Figure 22-63 Radial images of the heart, acquired with 256, 128, 64 and 32 projections. Note the
increased prevalence and intensity of the streak artifacts as the number of projections is reduced. The
amount of diffuse noise also increases, producing an apparent background signal in the lung. (From
Peters DC et al: Myocardial wall tagging with undersampled projection reconstruction. Magn Reson
Med 45:562-567, 2001)
The second cause of artifacts is data inconsistencies among projections. If tissue extends outside the
field-of-view in a certain direction, its signal is truncated by the frequency filter in projections that are
oriented at the same angle. Projections at oblique angles will, however, include the signal, causing
inconsistencies among views. The resulting image exhibits a belt of diffuse brightness around the edge
of the field-of-view, near where the object has been cut off (Fig. 22-65). An analogous artifact occurs
114
in CT imaging and has been successfully suppressed in that context with appropriate data filtering.
In MRI it can usually be eliminated by judicious placement of the RF coil, so as to minimize sensitivity
to bright objects outside the field-of-view.
page 613
page 614
30 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-64 A contrast-enhanced radial MRA of the abdomen exhibits high-intensity streak artifacts
that appear to radiate from a point on the longitudinal axis of the scanner, outside the field-of-view. The
signal was emitted by peripheral tissue, located beyond the range of the transverse gradient coils.
(Courtesy of Dana Peters PhD)
In spiral imaging, aliasing from a point source outside the field-of-view is manifested as a ring,
centered on the source and of radius equal to the FOV. Signal from a finite-sized object is therefore
smeared out over large circles, which are visible on the opposite side of the image (Fig. 22-66).
Motion
Radial imaging is much less sensitive to motion than conventional Cartesian imaging, because the
center of k-space is vastly oversampled. Some blurring and streak artifacts do occur, although they
are more tolerable than the ghosting associated with Cartesian techniques. 115 Blurring occurs around
the edges of moving objects and is due to position averaging over the range of motion. The streak
artifacts lie tangential to the moving object and perpendicular to the direction of motion116 (Fig. 22-67).
They can be minimized by choosing a view order that distributes the error azimuthally over k-space,
rather than concentrating it within a narrow range of angles.117 Simple rigid-body translation can also
118
be corrected by exploiting the phase information contained in successive views.
31 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-65 A radial image exhibits a band of diffuse signal around the edge of the image (arrow)
where tissue extends outside the field-of-view. The artifact is due to data inconsistencies among the
projections. (Courtesy of Karl Vigen PhD)
Add to lightbox
Figure 22-66 A spiral image of a resolution phantom, which is offset from the center of the field-
of-view. Signal from points outside the field-of-view is smeared out over large circles (arrow) centered
on the source.
Spiral imaging has reduced sensitivity to motion by virtue of its acquisition speed. Very rapid motion,
however, produces circular smearing artifacts centered on the moving object (Fig. 22-68). Rigid-body
motion may be amenable to correction using orbital navigators.119
page 614
page 615
32 of 41 15-04-2010 21:43
Add to lightbox
Add to lightbox
Figure 22-67 Radial imaging is relatively robust against motion, due to oversampling at the center of
k-space, but nevertheless produces some residual blurring and streak artifacts (A). Image B shows
the same slice without motion. (Courtesy of Ajit Shankaranarayanan PhD)
33 of 41 15-04-2010 21:43
Add to lightbox
Add to lightbox
page 615
page 616
Figure 22-68 Rapid motion during a spiral acquisition produces circular artifacts centered on the
moving object (A). Image B shows the same slice without motion.
34 of 41 15-04-2010 21:43
Add to lightbox
Add to lightbox
Figure 22-69 Images of an oil/water phantom acquired with (A) Cartesian and (B) spiral sampling. In
the Cartesian image the oil is displaced from the water in the frequency-encoding (vertical) direction,
leaving a small signal void between them (arrow). Spiral imaging has no uniquely defined frequency-
encoding direction and the chemical shift causes blurring of the oil signal (arrowhead).
PROPELLER imaging is designed for robustness against motion and uses phase information from the
resampled region near the center of k-space to correct for in-plane rotation and translation, and to
reject inconsistent data resulting from through-plane motion.17 PROPELLER offers marked
improvements in image quality in the context of rigid-body motion, as occurs in the head, but is less
35 of 41 15-04-2010 21:43
efficient in compensating for nonrigid body movement, such as respiratory and cardiac motion.
Off-Resonance Effects
In conventional Cartesian imaging, signal from off-resonant spins is displaced within the image along
the frequency-encoding direction. Since non-Cartesian techniques have no uniquely defined frequency-
encoding direction, the effects of off-resonant spins are considerably more complicated and depend on
the accumulation of phase errors along the k-space trajectories. Spiral imaging is particularly sensitive
to off-resonance effects because of its long readout times. Imperfect shimming, susceptibility
differences, and chemical shift are all possible sources of artifacts, which can be manifested in the
image as blurring or signal loss due to dephasing (Figs. 22-69 and 22-70). Artifacts from fat can be
120,121
minimized using spectral-spatial excitation pulses or fat-saturation prepulses. Dixon techniques
have also been applied to spiral imaging as an alternative means to suppress signal from fat. 122
Several methods have been devised to compensate for B0 inhomogeneity and susceptibility
differences, using direct acquisition of the field map.123-125
Trajectory Errors
Errors in the k-space trajectories can result from both eddy currents and delays in the physical
gradients. In Cartesian imaging, such errors usually have little impact on the image since adjacent lines
are affected equally (a notable exception being EPI, where they cause N/2 ghosts). In non-Cartesian
techniques, however, they constitute a significant source of artifacts. Radial imaging is particularly
sensitive to trajectory errors, since all the trajectories are intended to intersect the origin of k-space.
Errors in either the path or the timing of the trajectories can corrupt the data in the central region of
k-space, where most of the energy of the spatial-frequency spectrum is concentrated. Unless the
errors are accurately accounted for in the gridding procedure, they may cause smearing in the image
(Fig. 22-71).
Gradient timing errors can be corrected with the addition of compensatory areas to the prewinder and
rephasing gradients.126 The effect of eddy currents can be minimized by means of hardware
adjustments and the use of dummy pulses to attain a steady state prior to data acquisition. The steady
state can be maintained during acquisition by avoiding large increments in projection angle between
115
views.
Relaxation-Dependent and Transient Effects

page 616
page 617
36 of 41 15-04-2010 21:43
Add to lightbox
37 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-70 A contrast-enhanced MRA acquired with 3D radial sampling exhibits signal loss in the
aorta near the diaphragm (arrow), due to air/tissue susceptibility differences (A). The signal can be
recovered by correcting for B0 inhomogeneity (B). (Courtesy of Walter Block PhD)
38 of 41 15-04-2010 21:43
Add to lightbox
page 617
page 618
Figure 22-71 A radial image of a phantom exhibits smearing artifacts due to trajectory errors, which
result from delays of 4 μs and 8 μs respectively in the physical x and y gradients. (Courtesy of Dana
Peters PhD)
Add to lightbox
Add to lightbox
Figure 22-72 Radial viability images, acquired with a segmented inversion recovery sequence. Image
39 of 41 15-04-2010 21:43
A exhibits smearing artifacts (arrow) due to variations in the effective inversion time among different
projections, which were collected in order of increasing angle. The artifacts can be reduced by
interleaving projections with different TI values (B). (Courtesy of Dana Peters PhD)
In Cartesian imaging, the view ordering is usually chosen to take advantage of the fact that the low
spatial frequencies have a greater influence over image quality and contrast than the high spatial
frequencies. Transient effects, for example, are typically minimized by sampling the outer lines of
k-space first and the inner lines later, thereby ensuring that the magnetization will have settled into a
steady state by the time the central region of k-space is reached. In a similar fashion, sequence
parameters such as echo time TE and inversion time TI are always calculated to the center of
k-space, since the relaxation dependence of the low spatial frequencies most accurately determines
the contrast of the image as a whole.
In most non-Cartesian techniques, however, the center of k-space is continually resampled throughout
the acquisition. In this situation there is no equivalent of "outer" and "inner" k-space lines; all the views
contain some high and some low spatial frequencies and thus have similar influence on image quality
and contrast. For this reason, non-Cartesian techniques are much more sensitive to changes in the
magnetization over the course of the acquisition, which may result from longitudinal or transverse
relaxation, or transient effects.
Figure 22-72 shows examples of myocardial viability images acquired using a radial inversion recovery
sequence. Figure 22-72A exhibits smearing due to longitudinal relaxation of the magnetization over the
course of data acquisition. The severity of the artifacts, however, depends on view order. Whereas in
Figure 22-72A, the effective TI increased monotonically with projection angle, in Figure 22-72B the
different TI values were interleaved. This distributes the relaxation-related variations azimuthally in
k-space and suppresses the artifacts.
117
A similar result has been obtained in the context of radial FSE. Since the echo time varies among
different projections, the technique is vulnerable to artifacts from T2 relaxation. The effect can,
however, be minimized by interleaving projections with different echo times in a non-periodic fashion.
Timing in CE-MRA
Add to lightbox
40 of 41 15-04-2010 21:43
Add to lightbox
Figure 22-73 Radial MRAs acquired during the first pass of contrast material. Signal is absent from
the left iliac artery in image A, because projections at the corresponding angle were collected prior to
contrast arrival. The vessel becomes visible in the later image B. (From Peters DC et al:
Undersampled projection reconstruction applied to MR angiography. Magn Reson Med 43:91-101,
2000. Copyright © 2000 Wiley-Liss Inc. Reprinted by permission of Wiley-Liss Inc, a subsidiary of
John Wiley & Sons Inc.)
Timing an MRA acquisition correctly with respect to contrast administration is crucial; starting it too late
may cause venous contamination whereas starting it too early may result in the collection of all or
some of the data before the arrival of contrast material. In Cartesian imaging, the latter may give rise
to the so-called Maki artifact81 (see Fig. 22-50).
Radial imaging is gaining popularity for contrast-enhanced MR angiography and exhibits different
artifacts in response to mistiming. If some projections are acquired prior to contrast arrival, any
arteries lying at the corresponding angles may be absent from the image, mimicking a stenosis (Fig.
22-73A). The affected vessels will, however, appear on later images (Fig. 22-73B). Misdiagnoses can
therefore be avoided by reconstructing multiple time frames, using a sliding-window approach in
combination with undersampling.113
© 2010 Elsevier
41 of 41 15-04-2010 21:43
HARDWARE- AND SOFTWARE-RELATED ARTIFACTS

Magnetic Field Inhomogeneity
In all scanners the static magnetic field B0 is strongest and most homogeneous near the isocenter of the magnet
and weaker further away. For this reason it is important to position the subject so that the anatomy of interest is as
close as possible to the center of the scanner. Certain artifacts related to field inhomogeneity are almost inevitable
far from the isocenter, particularly with techniques that are sensitive to off-resonance effects, such as SSFP and
frequency-selective fat saturation.
page 618
page 619
Table 22-3. Summary of Technique-Specific Artifacts

Technique Typical artifacts Solutions
EPI Susceptibility-related distortion and Segmentation of readout (e.g., FGRE-ET)
signal pile-up Parallel imaging
Post-processing (e.g., using B0 maps)
N/2 ghosts in phase-encoding Data correction using reference scan
direction
Phased-array acquisition to separate
ghost from parent image
Diffusion-weighted Eddy current-induced Hardware optimization to minimize eddy
EPI misregistration among raw images currents
Dual-spin-echo
Parallel imaging
Post-processing to correct image
distortion
Use of PROPELLER for diffusion imaging
Ripple artifacts or signal loss due to Sedation and/or restraint
bulk motion
SSFP (aka Static stripe artifacts due to B0 Local shim
true-FISP, FIESTA, inhomogeneity Reduce TR
balanced FFE)
Dark flow artifacts due to Local shim
off-resonance effects
Manually tune center frequency to water
peak
Time-of-flight MRA Signal loss due to slow or turbulent Use 3D contrast-enhanced MRA (with
flow breath-hold where required)
Misregistration between adjacent
2D slices due to respiration
Signal fall-off across slab (3D TOF) Reduce slab thickness
Pulsation artifacts Contrast enhancement
ECG gating
Staircase artifact Reduce slice thickness
Contrast-enhanced Timing-related artifacts (Maki Use of a test bolus
MRA artifact and venous contamination) Interactive fluoroscopic trigger
Automated bolus detection (e.g.,
SmartPrep, CareBolus)
Gibbs artifacts from insufficient Parallel imaging
spatial resolution TRICKS
Undersampled radial imaging
Signal loss near stents MR-invisible stents
Active stents
1 of 17 15-04-2010 21:44
Signal loss from concentrated Dilute bolus prior to injection

contrast material
Shorten echo time
Optimize contrast dose
MIP Pseudo-occlusions due to omission Prescribe imaging volume to include all
of vessels vessels of interest
Take care not to crop vessels from MIP
volume during post-processing
Refer to raw images to avoid
misdiagnosis
Exaggerated stenoses due to low Optimize intensity threshold
lumen signal
Parallel imaging Inhomogeneous noise amplification Position coil elements judiciously
Increase FOV
Reduce acceleration factor
Residual aliasing Ensure identical coil placement, subject
position, and respiratory phase between
calibration and scan
Ensure calibration images are artifact free
(e.g., no wraparound, ghosting or ADC
overflow)
Reduce acceleration factor
High-field MRI Aggravated chemical shift and Increase bandwidth
susceptibility artifacts
Signal modulation due to Post-processing (SCIC)
inhomogeneity of RF field Optimize RF coil design
Radial imaging (aka Undersampling artifacts (streaks Increase number of projections
PR imaging) and increased noise)
Aliasing from tissue outside FOV Use local RF coil
(streaks and increased noise) Fat suppression or spatial saturation
bands
Signal enhancement around edge of Judicious RF coil placement
image where tissue extends outside
FOV
Smearing/streaks due to trajectory Compensate or account for gradient
errors delays
Minimize eddy currents
Use dummy pulses to reach steady state
Avoid large increments in projection angle
between views
Smearing/streaks due to relaxation Optimize view order to distribute
or transient effects during readout variations azimuthally
Pseudo-occlusion in CE-MRA from Acquire multiple time-frames using sliding-
starting acquisition too early window reconstruction in combination with
undersampling
Spiral imaging Aliasing from tissue outside FOV Use local RF coil
(circular smearing) Fat suppression or spatial saturation
bands
Motion artifacts (circular smearing) ECG gating for cardiac acquisitions
Orbital navigators for rigid-body motion
Blurring from off-resonance effects Fat suppression
Post-processing using B0 maps
2 of 17 15-04-2010 21:44
If, however, there is evidence of substantial field variation even at small fields-of-view, a hardware adjustment may
be required. Severe B0 inhomogeneity causes image distortion, due to variations in the Larmor frequency. The
effect is especially dramatic on echo-planar images (Fig. 22-74), due to the low effective bandwidth in the phase-
encoding direction. Large B0 variations also alter the amplitude of the RF excitations, causing artifacts in techniques
such as inversion recovery where the accuracy of the flip angle is critical.
Inhomogeneities in the B0 field are an inevitable result of imperfections in the manufacture and materials of the
primary magnet, and the presence of metal structures in the environment of the scanner. Corrections can be made
via a procedure known as "passive shimming", which involves the placement of metal pieces within the bore to
cancel unwanted field variations. The initial passive shims are glued into the bore at the factory. Some vendors
provide for further passive shimming on site, which is done by loading additional metal pieces onto slide rails that
are then inserted into the bore. Alternatively the scanner is equipped with superconducting or resistive shim coils, to
allow "active shimming". The current through each of the shim coils is adjusted to minimize the quadratic and
higher-order spatial components of the field. Since superconducting coils require a special power supply, their
current cannot be altered dynamically but must be adjusted by the field engineer.
Add to lightbox
page 620
page 621
Figure 22-74 A diffusion image acquired with a DW-EPI sequence exhibits distortion in the phase-encoding
direction (anterior/posterior) due to inhomogeneity in the static magnetic field B0. The higher-order shimming had
been inadvertently omitted during the scanner installation.
When the scanner is in use, the homogeneity of the B0 field within the patient is affected by anatomic geometry and
tissue susceptibility, which vary from subject to subject. Some compensation is provided by the dynamic shim
performed at scan time. On most scanners the effect of the dynamic shim is simply to optimize the DC current
levels through the gradient coils, which affect only the linear components of the field. Persistent inhomogeneities
that cannot be adequately corrected with dynamic shimming may require adjustments to the higher-order shims by
the field engineer.
Gradient Nonlinearity
The magnetic field gradients used for image formation are strongest and most linear near the isocenter of the
scanner. Both their strength and linearity decrease rapidly with distance from the center, particularly in the
transverse direction. Without appropriate correction, the nonlinearity in the imaging gradients produces distortion in
the image, causing tissues near the periphery (where the gradients are weaker) to appear contracted relative to
3 of 17 15-04-2010 21:44
those near the isocenter (where the gradients are stronger). Since the gradient profiles are known, it is relatively
straightforward to compensate for distortion within the imaging plane and this is an inbuilt feature of the image
reconstruction algorithm on commercial scanners.
Add to lightbox
Figure 22-75 Distortion due to nonlinearity in the imaging gradients affects the aliased portion of a conventional 2D
scout image. With the phase-encoding axis in the superior/inferior orientation, the pelvis is wrapped around to the
top of the image. While the image-processing algorithm corrects adequately for the effects of gradient nonlinearity
in tissue lying within the field-of-view, it aggravates the distortion of the aliased tissue, producing a cone-shaped
artifact (arrow).
The correction methods (e.g., GradWarp) do not, however, allow for wraparound and this can result in
cone-shaped artifacts, such as the one shown in Figure 22-75. This image was acquired as part of a three-plane
localizer, with a large field-of-view. Since the phase-encoding direction was chosen parallel to the longitudinal axis
of the scanner, tissue located outside the field-of-view on the inferior side is aliased to a position closer to the
isocenter on the superior side. The image-processing algorithm therefore aggravates the distortion rather than
correcting it. The cone artifact can be avoided using any means that will eliminate the wraparound.
Distortion also occurs in the through-slice direction, with the result that the image does not represent a perfectly flat
slice within the tissue. This so-called "potato chip" effect can be corrected in 3D acquisitions, but is much more
difficult to avoid in 2D imaging.
Very far from the isocenter, the magnetic fields produced by the gradient coils tend to zero, and provide no spatial
encoding at all. This is of no consequence provided all such regions fall outside the sensitivity range of the RF
transmitter or receiver. If signal is detected from such a region, however, the reconstruction algorithm interprets it
as coming from the isocenter of the scanner, where the imaging gradients have zero magnetic field amplitude. The
signal is therefore collapsed to the corresponding point on the image, producing a localized high-intensity "star
artifact" (Fig. 22-76). Note that since the longitudinal gradient generally has a larger range than the transverse
gradients, the signal may be slightly dispersed along the longitudinal axis.
The star artifact typically occurs when the body coil is used for signal reception and is more common in 3D
acquisitions, where a large volume of tissue is excited. It also arises more frequently on scanners equipped with
special-purpose short gradient coils, such as those designed for cardiac imaging. It can generally be avoided by
using local RF coils for signal reception, since they have a more limited range of sensitivity than the body coil.
Eddy Currents
Eddy currents have already been discussed in the context of diffusion-weighted EPI and radial imaging, since both
techniques are particularly sensitive to their effects. Eddy currents are electrical currents generated in conductive
materials in response to magnetic field changes. In MRI they can arise in the metal structures of the scanner, such
as the cryostat and RF coils, as a result of gradient switching. They are a consequence of Faraday's law of
induction, according to which a time-varying magnetic field will induce a current in any conductor that is present.
4 of 17 15-04-2010 21:44
The induced current in turn gives rise to a secondary magnetic field, which opposes the change in the first.
page 621
page 622
Add to lightbox
Add to lightbox
Figure 22-76 A 3D MRA exhibits a star artifact (arrows), due to signal from peripheral tissue outside the range of
the gradient coils. Since the signal has not been position-encoded, it is collapsed to a point in the 3D image
volume that corresponds to the center of the gradient coils. Note that some dispersion occurs along the longitudinal
axis, since the longitudinal gradient coil has a better range than the transverse gradient coils. A raw image is shown
from near the center of the 3D stack (A), together with a MIP (B).
5 of 17 15-04-2010 21:44
Add to lightbox
Figure 22-77 An illustration of the effects of eddy currents on the gradient waveform and the use of pre-emphasis
currents for compensation.
Eddy currents due to rapid switching of the gradient coils in a scanner produce residual secondary magnetic field
gradients, which decay with time constants ranging from a few milliseconds to several hundred milliseconds. The
slower components reduce B0 homogeneity, while the faster ones distort the pulse profiles of the imaging gradients
(as illustrated in Fig. 22-77). The result is an overall degradation of scanner performance. The effects are
particularly severe in diffusion-weighted EPI, where large diffusion gradients are applied in combination with a long
readout time. Radial imaging is also very sensitive to eddy currents, since the residual gradients can cause
miscentering of the trajectories in k-space.
Most scanners use actively shielded gradient coils, which reduce eddy currents by limiting the penetration of
magnetic flux into the metallic structures of the scanner. The shielding consists of a secondary coil counterwound
around the primary one to cancel its external field. The cancellation is not perfect, however, and some residual
eddy currents remain. Their effect can be compensated through the use of pre-emphasis currents in the gradient
amplifier input pulses.127 The pre-emphasis currents build some overshoot into the pulses, to cancel out the effect
of eddy currents in the final gradient waveform (see Fig. 22-77). The amplitude and decay characteristics of the
pre-emphasis currents must be accurately calibrated and this is done by the field engineer. Evidence of substantial
uncorrected eddy currents may warrant a recalibration of the compensation parameters.
Gradient and Radiofrequency Instability

Instabilities in the gradients or RF transmitter can introduce anomalous amplitude and phase modulations into the
k-space data, producing ghosting throughout the image (Fig. 22-78). It can be difficult to determine from the
images alone whether the instability is in the gradient or RF systems. The fault may lie in the coils themselves or in
the amplifiers that drive them. It may alternatively result from a loose connection or a failure in the power supply.
page 622
page 623
6 of 17 15-04-2010 21:44
Add to lightbox
Figure 22-78 Ghost artifacts in the phase-encoding direction (left/right) due to failure of a gradient amplifier.
(Courtesy of Brian Roen PhD)
Spikes
Spikes are data errors at individual points in k-space, which cause intensity oscillations throughout the image,
known as "corduroy artifacts" (Fig. 22-79). Each corrupted data element produces its own pattern of regularly
spaced lines, whose direction and separation depend on the location within k-space of the affected point. Several
such patterns can be superimposed if multiple k-space points are involved. Spikes result from transient electrical
currents, due to arcing or intermittent metal-on-metal contact. They can be caused by loose washers on any of the
mountings or by gradient cables rubbing against each other or against the magnet end flange. Foreign metallic
objects within the bore such as coins and paperclips may also be responsible. Other possible culprits are electrical
discharges from patient blankets or flickering light bulbs.
Radiofrequency Interference
Add to lightbox
7 of 17 15-04-2010 21:44
Figure 22-79 Data errors, or "spikes", at discrete points in k-space produce oscillatory patterns throughout the
field-of-view, often described as "corduroy" artifacts.
In conventional Cartesian imaging, position in the readout direction is determined through the mechanism of
frequency encoding. The readout gradient alters the precession frequency of the spins in a spatially dependent
manner and the position of a given tissue element is identified by the frequency of its emitted signal. Since the
precession frequencies fall in the RF range of the electromagnetic spectrum, any extraneous sources of radio
waves can cause artifacts, which appear as spurious lines or points on the image (Figs. 22-80 and 22-81). The
position and width of the artifacts in the readout direction are determined by the frequency and bandwidth
respectively of the source. In the phase-encoding direction, the signal is usually distributed along a line, with a
spatial modulation that depends on the imaging sequence and temporal correlations in the source. In the special
case of EPI, the artifacts appear as a set of four discrete points, equidistant from the isocenter of the scanner in
the frequency-encoding direction and separated by half the field-of-view in the phase-encoding direction (see Fig.
22-80B). This particular distribution arises because all the lines of k-space are acquired in very rapid succession
and the polarity of the readout gradient is reversed between alternate lines.
RF interference can result from a variety of sources. Most electronic devices containing a CPU chip are RF
emitters, since the oscillator required to drive the CPU operates in the megahertz range. Using such a device within
the magnetically shielded enclosure of the scanner can cause RF noise in the images (as shown in Fig. 22-80).
Artifacts also arise from breaches in the shielding, such as an open door, which exposes the system to RF energy
from external sources such as radio transmissions and other nearby MR scanners (Fig. 22-81).
Stimulated Echoes
page 623
page 624
Add to lightbox
Add to lightbox
Figure 22-80 A scout image (A) and an echo-planar image (B) of a neonate exhibit artifacts due to RF interference
from an infuser inside the magnetically shielded enclosure of the MR suite. Since the radio waves emitted by the
8 of 17 15-04-2010 21:44
infuser have a well-defined frequency, they produce a narrow line on the scout image (arrow), perpendicular to the
frequency-encoding direction (superior/inferior). On the echo-planar image the artifacts appear as a set of four
discrete spots, equidistant from the isocenter of the scanner in the frequency-encoding direction (left/right) and
separated by half the field-of-view in the phase-encoding direction (anterior/posterior). Note that the scout image
(A) also exhibits dark bands in the superior/inferior direction (arrowhead) due to saturation of magnetization in three
sagittal slices that had just been imaged as part of the three-plane localizer.
Add to lightbox
Figure 22-81 RF interference from a nearby scanner through an open door causes a broad band of noise (arrow)
perpendicular to the frequency-encoding direction (superior/inferior).
Whereas spin-echoes are produced by a combination of two RF pulses, stimulated echoes result from a
succession of three RF pulses. The first excites the spins, which then undergo dephasing until the application of the
second. The second pulse rotates some of the transverse magnetization onto the longitudinal axis, where it is
preserved from further dephasing. The third excitation returns the stored magnetization to the transverse plane,
where it is partially refocused, resulting in a stimulated echo after a time delay τ equal to the interval between the
first and second pulses. The process is illustrated in Figure 22-82 for the case of three 90° pulses. Note that in
addition to the stimulated echo, the sequence also produces spin-echoes and free induction decays (not shown).
Stimulated echoes form the basis of the STEAM technique (stimulated echo acquisition mode) used in
spectroscopy to acquire signal from a localized volume of tissue. When they occur inadvertently, however, they can
cause artifacts.
Stimulated echoes are produced most efficiently using 90° pulses, which exchange the largest amount of
magnetization between the transverse plane and the longitudinal axis. However, all RF pulses can contribute to
stimulated echoes, even if the flip angle is nominally 180°. As a consequence, stimulated echoes can arise in any
pulse sequence, unless adequate measures are taken to crush unwanted magnetization or to let it relax fully
between successive excitations. The resulting artifacts include ghosts and "zippers", which are lines in the
frequency-encoding direction that pass through the point of zero phase-encoding. Both the ghost and zipper
artifacts can be reflected through the center of the field-of-view along the phase-encoding axis.
page 624
page 625
9 of 17 15-04-2010 21:44
Add to lightbox
Figure 22-82 A stimulated echo is produced by a succession of three RF excitations, illustrated here with 90°
pulses. The first excites the spins and the second rotates some of the transverse magnetization onto the
longitudinal axis, where it is preserved from further dephasing. The third pulse returns the magnetization to the
transverse plane, where it is refocused after a time τ equal to the interval between the first and second pulses.
Below the spin diagrams are expressions for the magnetization in the product operator formalism. The angle
brackets indicate averages over the frequency shifts ∆ω governing the inhomogeneous T2* dephasing. For
simplicity, T1 and T2 decay are neglected.
Add to lightbox
10 of 17 15-04-2010 21:44
Add to lightbox
page 625
page 626
Figure 22-83 Phantom images exhibiting a reflected ghost (A) and a zipper artifact (B), due to stimulated echoes.
The frequency-encoding direction is vertical. The reflected ghost was obtained by removing the crusher gradients
from around the refocusing pulses in the standard Carr-Purcell-Meiboom-Gill sequence shown in Figure 22-84. The
zipper artifact was created by instead removing the crusher gradient at the end of the TR period.
Add to lightbox
Figure 22-84 A conventional Carr-Purcell-Meiboom-Gill sequence, in which each echo is used to reconstruct a
separate image. To prevent the formation of free induction decays and stimulated echoes, crusher gradients are
applied around each of the refocusing pulses and at the end of the TR period. Note that the crushers on either side
of a given refocusing pulse must be balanced, in order to preserve the spin-echoes. However, their amplitude must
differ from one refocusing pulse to the next, in order to suppress stimulated echoes effectively.
An example of a reflected ghost is shown in Figure 22-83A. Reflection occurs when the phases encoded in the
stimulated echo are reversed with respect to those of the primary echo. The example shown was obtained by
removing the crusher gradients around the refocusing pulses in a conventional Carr-Purcell-Meiboom-Gill sequence
(Fig. 22-84). This allows stimulated echoes to occur after the second and subsequent refocusing pulses, producing
ghosts in all but the first image. The signal detected after the second refocusing pulse, for example, is the sum of a
spin-echo and a stimulated echo with opposite phase. The phases of the echoes differ because the spin-echo has
11 of 17 15-04-2010 21:44
been refocused twice and the stimulated echo only once. The components of the signal that arise from the
stimulated echo are therefore interpreted by the reconstruction algorithm as coming from the opposite side of the
field-of-view, producing a reflected ghost.
If the magnetization forming the stimulated echo is not phase-encoded at all, the result is a zipper artifact (Fig.
22-83B). The zipper lies in the frequency-encoding direction and coincides with the line along which the phase-
encoding gradient has zero magnetic field amplitude. In the axial image of Figure 22-83B, it passes close to the
center of the phantom, since the phantom is located near the center of the scanner. The orientation of the zipper
distinguishes it from RF interference artifacts, which are perpendicular to the frequency-encoding axis. The example
shown in Figure 22-83B was obtained by removing the crusher at the end of the TR period in the Carr-Purcell-
Meiboom-Gill sequence (see Fig. 22-84). The phases imparted by the phase-encoding gradient in one TR interval
are then almost entirely compensated by those of the next and the process repeats every TR period over the
duration of the acquisition, producing a stimulated echo of identical amplitude at each line of k-space. Its signal is
therefore collapsed to the line of zero phase-encoding on the image. The zipper is characterized by alternating
bright and dark pixels, due to the offset of the echo from the center of the readout window.
Crusher gradients suppress stimulated echoes by dephasing their magnetization. Those placed at the end of the TR
period prevent transverse magnetization from being carried over from one TR interval to the next, thereby
eliminating coherence pathways across multiple TR periods. Those bracketing the refocusing pulses prevent the
formation of stimulated echoes within a single TR period. The left and right lobes of each crusher must be balanced
in order to preserve the spin-echo train but must differ from one refocusing pulse to the next to suppress stimulated
echoes effectively, as shown in Figure 22-84.
Acknowledgments
I am deeply indebted to the following people for their assistance and contributions. Radiologists: Robert Edelman
MD, Martin Lazarus MD, Wei Li MD, Joel Meyer MD, Sean Tutton MD and Vahid Yaghmai MD; physicists and
engineers: Walter Block PhD, Andres Carrillo PhD, Andrew Larson PhD, Belinda Li PhD, Charles McKenzie PhD,
Dana Peters PhD, Brian Roen PhD, Ajit Shankaranarayanan PhD and Karl Vigen PhD; technologists: Roland Bejm
RT, Valerie Cecil RT, Nirmal Christian RT and Eugene Dunkle RT.
page 626
page 627
Table 22-4. Summary of Hardware- and Software-Related Artifacts

Origin Manifestations Solutions
Magnetic field Image distortion (particularly in EPI) Adjust high-order shims (passive,
inhomogeneity Errors in effective flip angle superconducting or resistive shims,
depending on scanner)
Persistent off-resonance effects
even with small fields-of-view
Gradient nonlinearity Image distortion Correction algorithm in image
reconstruction (standard feature)
Cone-shaped artifacts due to Avoid wraparound (e.g., by swapping
aggravated distortion of aliased phase- and frequency-encoding directions)
tissue
Star artifact due to signal from Use local RF coil
outside range of gradient coils
Eddy currents Overall deterioration in system Recalibrate eddy current compensation
performance parameters
Misregistration between raw
images in diffusion-weighted EPI
Trajectory errors in radial imaging
RF and gradient Ghosting Check RF/gradient coils and amplifiers
instability
Check power supplies
Check for loose connections
12 of 17 15-04-2010 21:44
Spikes (isolated "Corduroy artifacts" (patterns of Replace burnt-out/flickering light bulbs

k-space errors) regularly spaced lines throughout Check for loose cables/washers and
image) foreign metallic objects in bore
RF interference Lines of noise in phase-encoding Check for breaches in RF shielding (e.g.,
direction open door)
Spurious dots in echo-planar Remove any electronic devices from scan
images room
Stimulated echoes Rotated or displaced ghosts Use crusher gradients to spoil unwanted
magnetization
Zipper artifact in frequency-
encoding direction
REFERENCES
1. Van Goethem JW, van den Hauwe L, Ozsarlak O, et al: Phase-contrast magnetic resonance angiography. JBR-BTR 86:340-44, 2003.
2. Hill JM, Dick AJ, Raman VK, et al: Serial cardiac magnetic resonance imaging of injected mesenchymal stem cells. Circulation
3. Kraitchman DL, Heldman AW, Atalar E, et al: In vivo magnetic resonance imaging of mesenchymal stem cells in myocardial infarction.
4. Hinds KA, Hill JM, Shapiro EM, et al: Highly efficient endosomal labeling of progenitor and stem cells with large magnetic particles allows
magnetic resonance imaging of single cells. Blood 102:867-872, 2003. Medline Similar articles
5. Bulte JW, Ben-Hur T, Miller BR, et al: MR microscopy of magnetically labeled neurospheres transplanted into the Lewis EAE rat brain. Magn
6. Zelivyanskaya ML, Nelson JA, Poluektova L, et al: Tracking superparamagnetic iron oxide labeled monocytes in brain by high-field magnetic
resonance imaging. J Neurosci Res 73:284-295, 2003. Medline Similar articles
7. Jendelova P, Herynek V, DeCroos J, et al: Imaging the fate of implanted bone marrow stromal cells labeled with superparamagnetic
nanoparticles. Magn Reson Med 50:767-776, 2003. Medline Similar articles
8. Walter GA, Cahill KS, Huard J, et al: Noninvasive monitoring of stem cell transfer for muscle disorders. Magn Reson Med 51:273-277, 2004.
9. Hwang KP, Flask C, Lewin JS, et al: Selective missing pulse steady state free precession (MP-SSFP): inner volume and chemical shift
selective imaging in a steady state sequence. J Magn Reson Imaging 19:124-132, 2004. Medline Similar articles
10. Quick HH, Vogt FM, Madewald S, et al: High spatial resolution whole-body MR angiography featuring parallel imaging: initial experience.
Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 176:163-169, 2004. Medline Similar articles
11. Maki JH, Wilson GJ, Eubank WB, et al: Utilizing SENSE to achieve lower station sub-millimeter isotropic resolution and minimal venous
enhancement in peripheral MR angiography. J Magn Reson Imaging 15:484-491, 2002. Medline Similar articles
12. Chen Q, Quijano CV, Mai VM, et al: On improving temporal and spatial resolution of 3D contrast-enhanced body MR angiography with
parallel imaging. Radiology 231:893-899, 2004.
13. Henkelman RM, Hardy PA, Bishop JE, et al: Why fat is bright in RARE and fast spin-echo imaging. J Magn Reson Imaging 2:533-540,
14. Ma J, Singh SK, Kumar AJ, et al: Method for efficient fast spin echo Dixon imaging. Magn Reson Med 48:1021-1027, 2002. Medline
Similar articles
15. Xiang QS, Henkelman RM: K-space description for MR imaging of dynamic objects. Magn Reson Med 29:422-428, 1993. Medline
Similar articles
16. Bale RJ, Lottersberger C, Vogele M, et al: A novel vacuum device for extremity immobilisation during digital angiography: preliminary
clinical experiences. Eur Radiol 12:2890-2894, 2002. Medline Similar articles
17. Pipe JG: Motion correction with PROPELLER MRI: application to head motion and free-breathing cardiac imaging. Magn Reson Med
18. Forbes KP, Pipe JG, Karis JP, et al: Brain imaging in the unsedated pediatric patient: comparison of periodically rotated overlapping
parallel lines with enhanced reconstruction and single-shot fast spin-echo sequences. Am J Neuroradiol 24:794-798, 2003. Medline
Similar articles
19. Cheryauka AB, Lee JN, Samsonov AA, et al: MRI diffusion tensor reconstruction with PROPELLER data acquisition. Magn Reson Imaging
20. Forbes KP, Pipe JG, Karis JP, et al: Improved image quality and detection of acute cerebral infarction with PROPELLER diffusion-
weighted MR imaging. Radiology 225:551-555, 2002. Medline Similar articles
21. Pipe JG, Farthing VG, Forbes KP: Multishot diffusion-weighted FSE using PROPELLER MRI. Magn Reson Med 47:42-52, 2002.
22. Stuber M, Botnar RM, Danias PG, et al: Submillimeter three-dimensional coronary MR angiography with real-time navigator correction:
comparison of navigator locations. Radiology 212:579-587, 1999. Medline Similar articles
23. Coakley FV, Glenn OA, Qayyum A, et al: Fetal MRI: a developing technique for the developing patient. Am J Roentgenol 182:243-252,
2004.
24. Storey P, Chen Q, Li W, et al: Band artifacts due to bulk motion. Magn Reson Med 48:1028-1036, 2002. Medline Similar articles
13 of 17 15-04-2010 21:44
25. McGibbon CA, Bencardino J, Palmer WE: Subchondral bone and cartilage thickness from MRI: effects of chemical-shift artifact. MAGMA
26. Hood MN, Ho VB, Smirniotopoulos JG, et al: Chemical shift: the artifact and clinical tool revisited. Radiographics 19:357-371, 1999.
27. Rinella ME, McCarthy R, Thakrar K, et al: Dual-echo, chemical shift gradient-echo magnetic resonance imaging to quantify hepatic
steatosis: implications for living liver donation. Liver Transpl 9:851-856, 2003. Medline Similar articles
28. Venkataraman S, Braga L, Semelka RC: Imaging the fatty liver. Magn Reson Imaging Clin North Am 10:93-103, 2002.
page 627
page 628
29. Gabriel H, Pizzitola V, McComb EN, et al: Adrenal lesions with heterogeneous suppression on chemical shift imaging: clinical implications.
J Magn Reson Imaging 19:308-316, 2004. Medline Similar articles
30. Fujiyoshi F, Nakajo M, Fukukura Y, et al: Characterization of adrenal tumors by chemical shift fast low-angle shot MR imaging: comparison
of four methods of quantitative evaluation. Am J Roentgenol 180:1649-1657, 2003.
31. Truong TK, Clymer BD, Chakeres DW, et al: Three-dimensional numerical simulations of susceptibility-induced magnetic field
inhomogeneities in the human head. Magn Reson Imaging 20:759-770, 2002. Medline Similar articles
32. Hermier M, Nighoghossian N, Derex L, et al: Hypointense transcerebral veins at T2*-weighted MRI: a marker of hemorrhagic transformation
risk in patients treated with intravenous tissue plasminogen activator. J Cereb Blood Flow Metab 23:1362-1370, 2003. Medline
Similar articles
33. Fiebach JB, Schellinger PD, Geletneky K, et al: MRI in acute subarachnoid haemorrhage; findings with a standardised stroke protocol.
Neuroradiology 46:44-48, 2004. Medline Similar articles
34. Imaizumi T, Chiba M, Honma T, et al: Detection of hemosiderin deposition by T2*-weighted MRI after subarachnoid hemorrhage. Stroke
35. Mitchell P, Wilkinson ID, Hoggard N, et al: Detection of subarachnoid haemorrhage with magnetic resonance imaging. J Neurol Neurosurg
Psychiatry 70:205-211, 2001. Medline Similar articles
36. Perl J 2nd, Tkach JA, Porras-Jimenez M, et al: Hemorrhage detected using MR imaging in the setting of acute stroke: an in vivo model. Am
J Neuroradiol 20:1863-1870, 1999.
37. Anzalone N, Scotti R, Riva R: Neuroradiologic differential diagnosis of cerebral intraparenchymal hemorrhage. Neurol Sci 25 (Suppl
1):S3-5, 2004.
38. Haque TL, Miki Y, Kanagaki M, et al: MR contrast of ferritin and hemosiderin in the brain: comparison among gradient-echo, conventional
spin-echo and fast spin-echo sequences. Eur J Radiol 48:230-236, 2003. Medline Similar articles
39. Denier C, Labauge P, Brunereau L, et al: Clinical features of cerebral cavernous malformations patients with KRIT1 mutations. Ann Neurol
40. Brunereau L, Leveque C, Bertrand P, et al: Familial form of cerebral cavernous malformations: evaluation of gradient-spin-echo (GRASE)
imaging in lesion detection and characterization at 1.5 T. Neuroradiology 43:973-979, 2001.
41. Kreeftenberg HG Jr, Mooyaart EL, Huizenga JR, et al: Quantification of liver iron concentration with magnetic resonance imaging by
combining T1-, T2-weighted spin echo sequences and a gradient echo sequence. Neth J Med 56:133-137, 2000. Medline
Similar articles
42. Ito K, Mitchell DG, Gabata T, et al: Hepatocellular carcinoma: association with increased iron deposition in the cirrhotic liver at MR imaging.
43. Rudisch A, Kremser C, Peer S, et al: Metallic artifacts in magnetic resonance imaging of patients with spinal fusion. A comparison of
implant materials and imaging sequences. Spine 23:692-699, 1998. Medline Similar articles
44. Earls JP, Ho VB, Foo TK, et al: Cardiac MRI: recent progress and continued challenges. J Magn Reson Imaging 16:111-127, 2002.
45. Mai VM, Bankier AA, Prasad PV, et al: MR ventilation-perfusion imaging of human lung using oxygen-enhanced and arterial spin labeling
techniques. J Magn Reson Imaging 14:574-579, 2001. Medline Similar articles
46. Carr JC, Laub G, Zheng J, et al: Time-resolved three-dimensional pulmonary MR angiography and perfusion imaging with ultrashort
repetition time. Acad Radiol 9:1407-1418, 2002. Medline Similar articles
47. Olsen RV, Munk PL, Lee MJ, et al: Metal artifact reduction sequence: early clinical applications. Radiographics 20:699-712, 2000
48. Gonner F, Lovblad KO, Heid O, et al: Magnetic resonance angiography with ultrashort echo times reduces the artefact of aneurysm clips.
Neuroradiology 44:755-758, 2002. Medline Similar articles
49. van Gorp MJ, van der Graaf Y, de Mol BA, et al: Bjork-Shiley convexoconcave valves: susceptibility artifacts at brain MR imaging and
mechanical valve fractures. Radiology 230:709-714, 2004. Medline Similar articles
50. McKinstry RC 3rd, Jarrett DY: Magnetic susceptibility artifacts on MRI: a hairy situation. Am J Roentgenol 182:532, 2004.
51. Wang Y, Truong TN, Yen C, et al: Quantitative evaluation of susceptibility and shielding effects of nitinol, platinum, cobalt-alloy, and stainless
steel stents. Magn Reson Med 49:972-976, 2003. Medline Similar articles
52. Melodelima D, Salomir R, Mougenot C, et al: Intraluminal ultrasound applicator compatible with magnetic resonance imaging "real-time"
temperature mapping for the treatment of oesophageal tumours: an ex vivo study. Med Phys 31:236-244, 2004. Medline Similar articles
53. Steiger HJ, van Loon JJ: Virtues and drawbacks of titanium alloy aneurysm clips. Acta Neurochir (Wien) 72 (Suppl):81-88, 1999.
54. Matsuura H, Inoue T, Konno H, et al: Quantification of susceptibility artifacts produced on high-field magnetic resonance images by various
biomaterials used for neurosurgical implants. Technical note. J Neurosurg 97:1472-1475, 2002. Medline Similar articles
55. Weishaupt D, Quick HH, Nanz D, et al: Ligating clips for three-dimensional MR angiography at 1.5 T: in vitro evaluation. Radiology
214:902-907, 2000.
56. Henkelman RM, Stanisz GJ, Kim JK, et al: Anisotropy of NMR properties of tissues. Magn Reson Med 32:592-601, 1994. Medline
Similar articles
57. Xia Y: Magic-angle effect in magnetic resonance imaging of articular cartilage: a review. Invest Radiol 35:602-621, 2000. Medline
14 of 17 15-04-2010 21:44
Similar articles
58. Li T, Mirowitz SA: Manifestation of magic angle phenomenon: comparative study on effects of varying echo time and tendon orientation
among various MR sequences. Magn Reson Imaging 21:741-744, 2003. Medline Similar articles
59. Oatridge A, Herlihy AH, Thomas RW, et al: Magnetic resonance: magic angle imaging of the Achilles tendon. Lancet 358:1610-1611,
60. Oatridge A, Herlihy A, Thomas RW, et al: Magic angle imaging of the achilles tendon in patients with chronic tendonopathy. Clin Radiol
61. Bydder GM: New approaches to magnetic resonance imaging of intervertebral discs, tendons, ligaments, and menisci. Spine
62. Gray ML, Burstein D, Xia Y: Biochemical (and functional) imaging of articular cartilage. Semin Musculoskelet Radiol 5:329-343, 2001.
63. Klarhofer M, Dilharreguy B, van Gelderen P, et al: A PRESTO-SENSE sequence with alternating partial-Fourier encoding for rapid
susceptibility-weighted 3D MRI time series. Magn Reson Med 50:830-838, 2003.
64. Preibisch C, Pilatus U, Bunke J, et al: Functional MRI using sensitivity-encoded echo planar imaging (SENSE-EPI). Neuroimage
65. Bammer R, Auer M, Keeling SL, et al: Diffusion tensor imaging using single-shot SENSE-EPI. Magn Reson Med 48:128-136, 2002.
66. de Zwart JA, van Gelderen P, Kellman P, et al: Application of sensitivity-encoded echo-planar imaging for blood oxygen level-dependent
functional brain imaging. Magn Reson Med 48:1011-1020, 2002. Medline Similar articles
67. Epstein FH, Arai AE: Optimization of fast cardiac imaging using an echo-train readout. J Magn Reson Imaging 11:75-80, 2000.
68. Roopchansingh V, Cox RW, Jesmanowicz A, et al: Single-shot magnetic field mapping embedded in echo-planar time-course imaging.
69. Andersson JL, Skare S, Ashburner J: How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion
tensor imaging. Neuroimage 20:870-888, 2003. Medline Similar articles
70. Wilson JL, Jezzard P: Utilization of an intra-oral diamagnetic passive shim in functional MRI of the inferior frontal cortex. Magn Reson Med
71. Kellman P, McVeigh ER: Ghost artifact cancellation using phased array processing. Magn Reson Med 46:335-343, 2001. Medline
Similar articles
72. Reese TG, Heid O, Weisskoff RM, et al: Reduction of eddy-current-induced distortion in diffusion MRI using a twice-refocused spin echo.
73. Jezzard P, Barnett AS, Pierpaoli C: Characterization of and correction for eddy current artifacts in echo planar diffusion imaging. Magn
74. Rohde GK, Barnett AS, Basser PJ, et al: Comprehensive approach for correction of motion and distortion in diffusion-weighted MRI. Magn
75. Bodammer N, Kaufmann J, Kanowski M, et al: Eddy current correction in diffusion-weighted imaging using pairs of images acquired with
opposite diffusion gradient polarity. Magn Reson Med 51:188-193, 2004. Medline Similar articles
76. Markl M, Alley M, Elkins C, et al: Flow effects in balanced steady-state free precession imaging. Magn Reson Med 50:892-903, 2003.
77. Storey P, Li W, Chen Q, et al: Flow artifacts in steady-state free precession cine imaging. Magn Reson Med 51:115-122, 2004.
78. Li W, Storey P, Chen Q, et al: Dark flow artifacts with steady-state free precession cine MR technique: causes and implications for cardiac
MR imaging. Radiology 230:569-575, 2004. Medline Similar articles
79. Tatli S, Lipton MJ, Davison BD, et al: From the RSNA refresher courses: MR imaging of aortic and peripheral vascular disease.
Radiographics 23 Spec No: S59-78, 2003.
80. Lee VS, Martin DJ, Krinsky GA, et al: Gadolinium-enhanced MR angiography: artifacts and pitfalls. Am J Roentgenol 175:179-205, 2000.
81. Maki JH, Prince MR, Londy FJ, et al: The effects of time-varying intravascular signal intensity and k-space acquisition order on three-
dimensional MR angiography image quality. J Magn Reson Imaging 6:642-651, 1996. Medline Similar articles
82. Riederer SJ, Bernstein MA, Breen JF, et al: Three-dimensional contrast-enhanced MR angiography with real-time fluoroscopic triggering:
design specifications and technical reliability in 330 patient studies. Radiology 215:584-593, 2000.
83. Ho VB, Foo TK: Optimization of gadolinium-enhanced magnetic resonance angiography using an automated bolus-detection algorithm
(MR SmartPrep). Original investigation. Invest Radiol 33:515-523, 1998. Medline Similar articles
84. Korosec FR, Frayne R, Grist TM, et al: Time-resolved contrast-enhanced 3D MR angiography. Magn Reson Med 36:345-351, 1996.
85. Mazaheri Y, Carroll TJ, Du J, et al: Combined time-resolved and high-spatial-resolution 3D MRA using an extended adaptive acquisition. J
86. Du J, Carroll TJ, Wagner HJ, et al: Time-resolved, undersampled projection reconstruction imaging for high-resolution CE-MRA of the distal
runoff vessels. Magn Reson Med 48:516-522, 2002. Medline Similar articles
87. Barger AV, Block WF, Toropov Y, et al: Time-resolved contrast-enhanced imaging with isotropic resolution and broad coverage using an
undersampled 3D projection trajectory. Magn Reson Med 48:297-305, 2002.
88. Bartels LW, Smits HF, Bakker CJ, et al: MR imaging of vascular stents: effects of susceptibility, flow, and radiofrequency eddy currents. J
Vasc Interv Radiol 12:365-371, 2001. Medline Similar articles
page 628
page 629
89. Meyer JM, Buecker A, Schuermann K, et al: MR evaluation of stent patency: in vitro test of 22 metallic stents and the possibility of
determining their patency by MR angiography. Invest Radiol 35:739-746, 2000.
15 of 17 15-04-2010 21:44
90. Maintz D, Kugel H, Schellhammer F, et al: In vitro evaluation of intravascular stent artifacts in three-dimensional MR angiography. Invest
91. Klemm T, Duda S, Machann J, et al: MR imaging in the presence of vascular stents: a systematic assessment of artifacts for various stent
orientations, sequence types, and field strengths. J Magn Reson Imaging 12:606-615, 2000. Medline Similar articles
92. Buecker A, Spuentrup E, Ruebben A, et al: Artifact-free in-stent lumen visualization by standard magnetic resonance angiography using a
new metallic magnetic resonance imaging stent. Circulation 105:1772-1775, 2002. Medline Similar articles
93. Spuentrup E, Ruebben A, Stuber M, et al: Metallic renal artery MR imaging stent: artifact-free lumen visualization with projection and
standard renal MR angiography. Radiology 227:897-902, 2003. Medline Similar articles
94. Hietala EM, Maasilta P, Stahls A, et al: Magnetic resonance evaluation of luminal patency after polylactide stent implantation: an
experimental study in a rabbit aorta model. Eur Radiol 13:1025-1032, 2003. Medline Similar articles
95. Meyer JM, Buecker A, Spuentrup E, et al: Improved in-stent magnetic resonance angiography with high flip angle excitation. Invest Radiol
96. van Holten J, Wielopolski P, Bruck E, et al: High flip angle imaging of metallic stents: implications for MR angiography and intraluminal
signal interpretation. Magn Reson Med 50:879-883, 2003. Medline Similar articles
97. Bartels LW, Bakker CJ, Viergever MA: Improved lumen visualization in metallic vascular implants by reducing RF artifacts. Magn Reson
98. Kivelitz D, Wagner S, Schnorr J, et al: A vascular stent as an active component for locally enhanced magnetic resonance imaging: initial in
vivo imaging results after catheter-guided placement in rabbits. Invest Radiol 38:147-152, 2003. Medline Similar articles
99. Quick HH, Kuehl H, Kaiser G, et al: Inductively coupled stent antennas in MRI. Magn Reson Med 48:781-790, 2002. Medline
Similar articles
100. Tirkes AT, Rosen MA, Siegelman ES: Gadolinium susceptibility artifact causing false positive stenosis isolated to the proximal common
carotid artery in 3D dynamic contrast medium enhanced MR angiography of the thorax-a brief review of causes and prevention. Int J
Cardiovasc Imaging 19:151-155, 2003.
101. Neimatallah MA, Chenevert TL, Carlos RC, et al: Subclavian MR arteriography: reduction of susceptibility artifact with short echo time and
dilute gadopentetate dimeglumine . Radiology 217:581-586, 2000. Medline Similar articles
102. Heidemann RM, Ozsarlak O, Parizel PM, et al: A brief review of parallel magnetic resonance imaging. Eur Radiol 13:2323-2337, 2003.
103. Sodickson DK, Manning WJ: Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magn
104. Pruessmann KP, Weiger M, Scheidegger MB, et al: SENSE: sensitivity encoding for fast MRI. Magn Reson Med 42:952-962, 1999.
105. Sodickson DK, Griswold MA, Jakob PM, et al: Signal-to-noise ratio and signal-to-noise efficiency in SMASH imaging. Magn Reson Med
106. Frayne R, Goodyear BG, Dickhoff P, et al: Magnetic resonance imaging at 3.0 tesla: challenges and advantages in clinical neurological
imaging. Invest Radiol 38:385-402, 2003.
107. Norris DG: High field human imaging. J Magn Reson Imaging 18:519-529, 2003. Medline Similar articles
108. Alecci M, Collins CM, Smith MB, et al: Radio frequency magnetic field mapping of a 3 Tesla birdcage coil: experimental and theoretical
dependence on sample properties. Magn Reson Med 46:379-385, 2001.
109. Yang QX, Wang J, Zhang X, et al: Analysis of wave behavior in lossy dielectric samples at high field. Magn Reson Med 47:982-989, 2002.
110. Alecci M, Collins CM, Wilson J, et al: Theoretical and experimental evaluation of detached endcaps for 3 T birdcage coils. Magn Reson
Med 49:363-370, 2003.
111. Liu W, Collins CM, Delp PJ, et al: Effects of end-ring/shield configuration on homogeneity and signal-to-noise ratio in a birdcage-type coil
loaded with a human head. Magn Reson Med 51:217-221, 2004. Medline Similar articles
112. Peters DC, Guttman MA, Dick AJ, et al: Reduced field of view and undersampled PR combined for interventional imaging of a fully
dynamic field of view. Magn Reson Med 51:761-767, 2004. Medline Similar articles
113. Peters DC, Korosec FR, Grist TM, et al: Undersampled projection reconstruction applied to MR angiography. Magn Reson Med
114. Ohnesorge B, Flohr T, Schwarz K, et al: Efficient correction for CT image artifacts caused by objects extending outside the scan field of
view. Med Phys. 27:39-46, 2000. Medline Similar articles
115. Schaeffter T, Weiss S, Eggers H, et al: Projection reconstruction balanced fast field echo for interactive real-time cardiac imaging. Magn
116. Glover GH, Pauly JM: Projection reconstruction techniques for reduction of motion effects in MRI. Magn Reson Med 28:275-289, 1992.
117. Theilmann RJ, Gmitro AF, Altbach MI, et al: View-ordering in radial fast spin-echo imaging. Magn Reson Med 51:768-774, 2004.
118. Shankaranarayanan A, Wendt M, Lewin JS, et al: Two-step navigatorless correction algorithm for radial k-space MRI acquisitions. Magn
119. Moriguchi H, Lewin JS, Duerk JL: Novel interleaved spiral imaging motion correction technique using orbital navigators. Magn Reson Med
120. Bornert P, Stuber M, Botnar RM, et al: Comparison of fat suppression strategies in 3D spiral coronary magnetic resonance angiography.
J Magn Reson Imaging 15:462-466, 2002.
121. Nayak KS, Cunningham CH, Santos JM, et al: Real-time cardiac MRI at 3 tesla. Magn Reson Med 51:655-660, 2004.
122. Moriguchi H, Lewin JS, Duerk JL: Dixon techniques in spiral trajectories with off-resonance correction: a new approach for fat signal
suppression without spatial-spectral RF pulses. Magn Reson Med 50:915-924, 2003. Medline Similar articles
16 of 17 15-04-2010 21:44
123. Nayak KS, Tsai CM, Meyer CH, et al: Efficient off-resonance correction for spiral imaging. Magn Reson Med 45:521-524, 2001.
124. Ahunbay E, Pipe JG: Rapid method for deblurring spiral MR images. Magn Reson Med 44:491-494, 2000. Medline
Similar articles
125. Moriguchi H, Dale BM, Lewin JS, et al: Block regional off-resonance correction (BRORC): a fast and effective deblurring method for spiral
126. Peters DC, Derbyshire JA, McVeigh ER: Centering the projection reconstruction trajectory: reducing gradient delay errors. Magn Reson
127. Terpstra M, Andersen PM, Gruetter R: Localized eddy current compensation using quantitative field mapping. J Magn Reson
© 2010 Elsevier
17 of 17 15-04-2010 21:44
MAGE ROCESSING
Vincent J. Argiro
INTRODUCTION
The digital nature of magnetic resonance (MR) image data is such that it lends itself to a variety of
post-processing techniques. Of course, the cross-sectional images produced by MR scanners are the
result of processing frequency-domain (k-space) acquisition data in the first place. However, the
resulting images, while often quite useful on their own, can be made significantly more so through the
application of filtering, segmentation, and three-dimensional (3D) rendering techniques.
As pointed out by Sonka et al,1 image processing does not add information to images, strictly
speaking. Image processing is valuable because it enhances particular information that carries the
greatest significance for human interpretation and decision-making, while suppressing obscuring or
confounding information. Therefore, all the techniques we will discuss here represent processes of
selection; choosing, by the design of the algorithms and when to apply them, the salient diagnostic
message of the image.
We will consider each of these categories of image processing in turn, examining the principles that
underlie them, the techniques and alternatives that make them practical, and some of the characteristic
applications for them. We will then walk through three example case studies, applying a selection of
the techniques discussed in typical clinical workflows. The reader is advised that this chapter should be
considered as an introduction and summary of these techniques; further reading in the cited references
2-5
is highly recommended.
© 2010 Elsevier
1 of 1 15-04-2010 21:46
DYNAMIC RANGE ADJUSTMENT AND FILTERS

Window and Level
The most basic form of post-processing necessary for effective examination of MR images is adjusting
the dynamic range of such images to match both the range of intensities of the image features of most
interest and the capabilities of the display medium. Commonly specified as "window width/window
level" or simply "window/level", this is a linear mapping of a portion of the range of supplied image pixel
intensity values to the range of gray-levels available on the display medium. In the most familiar terms,
this amounts to adjusting the overall brightness and the contrast between dark and bright regions of
the image for viewing.
Adjustment of this mapping is particularly important with MR images, because of their mostly
noncalibrated nature. That is, unlike CT, with its consistent Hounsfield unit (HU) scale of image
intensities, MR images contain varying ranges of intensity values, depending on pulse sequence and
other acquisition parameters, and the normalization scheme chosen by the scanner manufacturer.
Image dynamic range can vary widely from scanner to scanner and even from exam to exam with the
same scanner.
page 630
page 631
In the case of film viewing, this matching can only be done once, before the film is exposed, and is
hence static during interpretation. In the now common practice of viewing MR images electronically on
computer monitors, the interpreting radiologist can make the adjustment dynamically. Window/level
adjustment permits the radiologist to compensate for exam-to-exam variability, ambient light conditions
in the viewing area, and personal interpretation preference.
Filters
Images are filtered to improve their interpretability, both by human observers and by computer-based
segmentation algorithms (considered in the next section). The most common goal of filtering is to
reduce random noise in an image, thereby improving its signal-to-noise (SNR) and contrast-to-noise
(CNR) ratios. We can categorize de-noising filters into simple filters and adaptive filters.
Simple Filters
Simple filters are applied consistently across an image, pixel by pixel. They consist of convolution
algorithms that compute the value of each pixel from a weighted combination of signal values (gray-
levels) of a number of that pixel's neighbors. The two- or three-dimensional array of these weighting
factors is referred to as the "filter kernel." Small filter kernels, which only take into account the eight
immediate neighbors of a 2D pixel or 26 neighbors of a 3D voxel, are fast to apply in computation but
give least control over the filtering operation. Larger kernels can improve the precision of the result at
the cost of greater computational time.
The advantage of such simple filters is that they are easy to implement and fast to execute, so they
can be employed on demand and interactively during image viewing. It is now common for scanner
console software, PACS viewing software, and 3D workstations to contain one or more of these
simple filters as selectable options. The disadvantage of simple kernel-based de-noising filters is that
they tend to blur the edges of intrinsic structures in the image as they smooth out or diffuse noise in the
interior of such structures or regions (Fig. 23-1). This disadvantage has motivated the development of
a second more sophisticated category of de-noising techniques we will call adaptive filters.
Adaptive Filters
Adaptive filters use a variety of methods to restrict the modification of the pixel values of an image to
the interior of approximately homogeneous regions in the image, avoiding the sharper transitions in
pixel values (intensity gradients) at the edges of structures or regions. This results in preservation of
edge detail, even as noise in the region interiors is smoothed.
1 of 3 15-04-2010 21:48
One of the most common techniques is called "anisotropic diffusion." The term refers to a variation or
adaptation in the amount of smoothing (diffusion) according to the strength of gradients or edges in the
image.6,7 The technique has been applied to MR images for improved visual interpretation,8 but also as
a preprocessing step before computation of cerebral perfusion, a process that is highly sensitive to
9
noise in the source images.
Add to lightbox
Add to lightbox
page 631
2 of 3 15-04-2010 21:48
page 632
Figure 23-1 Simple blurring filter applied to a coronal T1-weighted knee exam. A, Unfiltered original
image. B, Blurred image showing decreased tissue speckle, but also edge blurring.
Another approach involves selectively extracting the edge content of an image through the application
of a convolution kernel, as discussed above, and then applying a de-noising filter to the whole image.
10
Finally, the edge content is factored back into the image to restore edge detail. A third approach
consists of the local application of a shape-adaptive template chosen from a defined set of templates.
The template matches shape characteristics of intrinsic structures in the local region of the image and
hence preserves anatomic structure while diffusing noise.11 Still another approach involves operating at
the appropriate range of spatial frequencies or spatial scale to preserve objects in the image while
12
suppressing noise. Prior knowledge of the spatial scale of anatomic features captured in the image is
used to calibrate the adaptive filter.
13
As suggested by Westin et al, the use of adaptive de-noising filters may allow a shorter image
acquisition time or remove the need for contrast agents in some MR angiography studies. In addition to
de-noising, filters have also been devised to remove the overall (low spatial frequency) variation in
signal intensity which may result from bias or RF field inhomogeneities at the time of acquisition.14,15
This type of filtering may be particularly important to improve the performance of automated
16
segmentation algorithms, which we consider in the next section.
© 2010 Elsevier
3 of 3 15-04-2010 21:48
SEGMENTATION
Segmentation is the process of partitioning a 2D image or 3D volume into logically discrete regions of
interest. There are two motivations for this. First, segmentation is used to reduce the visual clutter in
an image, which may interfere with interpretation. We will refer to this as visual segmentation. Second,
segmentation is employed to precisely delineate a region of interest to measure its properties, such as
volume or pixel value statistics.
Visual segmentation is particularly relevant in 3D renderings, where overlying structures may directly
obscure the area of interest given a chosen viewing or projection direction. Often the human interpreter
is only interested in a portion of the acquired image, such as the arteries in a magnetic resonance
angiography exam or the brain in a head exam. Semi-automated or fully automated algorithms can be
applied, for example, to remove extravascular tissue in the MRA case or scalp, eyes, and facial
structures in the neurologic exam.
Manual Segmentation
The simplest methods of visual or qualitative segmentation are manual or semi-automated in their
application. Software tools are provided for sculpting away extraneous or obscuring regions of a 2D
cross-section or a 3D volume, or outlining and selecting regions to be retained for closer inspection.
The most basic tools involve drawing delineating contours over a series of cross-sections and including
or excluding the area inside the contour from further visualization. Multiple contours drawn on several
levels of a single section plane, or contours drawn on orthogonal planes through a volume can be
interpolated to define a boundary surface around a volume of image voxels to be retained or discarded
(Fig. 23-2). When 3D renderings are available as a reference for such sculpting segmentation, the
process can be made more efficient. The software user draws a delineating contour over the 3D
projection image and the algorithm projects this contour through the data volume to produce a
volumetric region to be included or excluded (Fig. 23-3). Rotating the 3D image to another projection
angle allows for repeating the process to refine the included or excluded region.
Add to lightbox
1 of 10 15-04-2010 21:48
Add to lightbox
page 632
page 633
Figure 23-2 Visual editing by serial contouring. A, Contours drawn on axial sections through the brain.
B, Resulting segmentation of the cerebral cortex displayed as volume rendering.
2 of 10 15-04-2010 21:48
Add to lightbox
Add to lightbox
Add to lightbox
3 of 10 15-04-2010 21:48
Add to lightbox
Figure 23-3 Visual editing by projected hand-drawn contour. A, Before segmentation. B, Contour
drawn to isolate heart. C, Heart isolated in front view. D, Heart rotated to lateral view.
Thresholding
Another basic form of segmentation can be accomplished by applying thresholds to the intensity values
present in the source image data, excluding values below or above these designated values from
further consideration. This technique can be of some value in removing low-intensity background from
image volumes, but is of little value for more detailed segmentation, particularly of most MR data,
since voxels of overlapping intensity ranges are usually present throughout the image. That is, a given
range of intensities present throughout a 2D image plane or 3D volume does not often correlate well
with anatomically distinct structures of interest.
page 633
page 634
4 of 10 15-04-2010 21:48
Add to lightbox
Add to lightbox
5 of 10 15-04-2010 21:48
Add to lightbox
Figure 23-4 Segmentation by connected region selection (seed fill) to isolate one of a number of
structures with similar gray-levels. A, Before segmentation. B, Vertebral arteries selected and
highlighted. C, Vertebral arteries extracted; carotid arteries and other tissues hidden.
Thresholds may be chosen manually and interactively, with the visual guidance of immediately
displaying the resulting image, or they may be chosen through an automated algorithm. These
algorithms operate on the principle of histogram analysis-examining the distribution of intensity values
present in the whole 2D or 3D image. An image composed of two distinct regions based on intensity
will exhibit a bimodal histogram. Each region will be characterized by a most frequent pixel value
indicated by a peak in the histogram, and the distinction between the regions will be represented by an
intervening trough or local minimum point in the histogram between these two peaks. Placing a
threshold at or near this minimum point may segregate the regions effectively. The separation is rarely
perfect, since the minimum is often not at zero occupancy. This means that there is some overlap in
the pixel values composing the two regions and a compromise of assignment into two logical groups
must be made in placing the threshold.
Connectedness
Selection of pixels or 3D voxels based on their intensity values can be made more useful by taking
clustering or connectedness of such pixels or voxels into account. That is, a region to be included or
excluded from further viewing or analysis can be identified by beginning at a start or "seed" pixel and
visiting all the neighbors of that pixel and testing them against a chosen intensity threshold or
thresholds (upper or lower or both). If the neighbor pixel is within the threshold criteria, it becomes a
new seed for the neighbor search and the process is repeated. This neighbor testing continues until all
boundaries of the growing connected region have been explored and are at the threshold value or
values.
By restricting the inclusion of pixels or voxels into a region of interest (or disinterest) to those adjacent
to each other, one can select an image structure from other discontinuous structures that possess the
same pixel values. For example, one might select one arterial tree in a 3D MRA exam from other
equally contrast-enhanced vessels in the same plane or volume, as long as the two vessel regions do
6 of 10 15-04-2010 21:48
not come into contact within the bounds of the data set (Fig. 23-4).
A variation of this "seed fill" or "connected components" method involves using gradient thresholds
rather than pixel value thresholds. A gradient value estimates the rate of change of pixel values over a
local area of an image. Gradient values are high at edges between structures or regions and are low in
the interior of such regions. When a gradient threshold is used, slow changes in pixel intensity in the
interior of a region, such as might be artifacts of magnetic field inhomogeneity, will not trigger the
bounding of a region fill, as would be the case with a pixel value or intensity threshold.
page 634
page 635
Add to lightbox
Add to lightbox
7 of 10 15-04-2010 21:48
Figure 23-5 Segmentation by size of connected regions. A, Before segmentation. B, After

segmentation only intracranial vessels are retained. Volume rendered 3D time-of-flight MRA exam.
Another way to use connectedness is to set threshold criteria as described above, exhaustively seed
and fill all regions in image or volume, and then sort the resulting clusters based on size (volume in
3D). Applying a threshold to that size criterion, one can select in or out regions or objects of a
particular size range. This can be particularly useful in removing small extraneous regions from MR
images before or after more targeted segmentation algorithms. For example, applying this cluster size
filter to a 3D MRA study can remove extravascular tissue densities, leaving the arterial tree free for
clear visualization and further analysis (Fig. 23-5). The sorting criterion need not be size or size alone.
Other criteria describing shape of the clustered regions, such as form factor (perimeter/area ratio in
2D or surface/volume ratio in 3D), can be employed to create a more elaborate region filter or
segregation.
Level Sets
Recently, a segmentation methodology has emerged which achieves pixel or voxel clustering, but by a
very different mathematical and physical model. The class of methods is known as "level sets" and
was first proposed and developed by Adalsteinsson and Sethian.17,18 In general terms, the method
models the region aggregation process as the movement of an expanding fluid front over time. An initial
seed curve in 2D or surface in 3D is placed manually or automatically within the region whose
boundaries are to be found and then the curve or surface is expanded into the region iteratively. The
rate of expansion, or size of each iteration step, of each location along the front is determined by a
"speed function." The function relates properties of the image, such as pixel intensity and gradient
magnitude, as well as properties of the expanding surface itself, such as minimum rate of curvature or
stiffness, to the rate of progress of each element of the front. The speed function is chosen in
advance, taking into account a priori knowledge about properties of the imaging method and the shape
of the region to be segmented.
Advantages of the level set method over the simple seed fill method discussed above include the ability
to deal with more complex topology, construct complex speed functions, and providing a way to
accomplish the region aggregation incrementally. Because each iteration of the algorithm produces a
complete boundary, balanced in its extent, control over the termination of the algorithm may be left to
the user, stopping when a desired clustering result has been achieved, even if an objective termination
criterion has not been established or reached. This cannot be done with a simple seed fill, because the
order of filling of the region is based not on any property of the data values themselves, but merely on
the software logic of traversing the voxels.
The method is challenging to implement effectively, since it is much more compute-intensive than the
simple recursive seed fill. However, recent approximations to the full mathematical formulation of level
sets ("fast marching") and optimized software implementations have provided a basis for tools that can
operate on medical image volumes in a matter of seconds.19
In one example, Farag et al20 applied the level sets method to the problem of segmenting the
intracranial arterial tree from volumetric MRA data sets of the head. In this case, a simple thresholding
technique does not work well, due to the overlap in the voxel values of the vessel lumen and of the
subcutaneous and extraocular fat. The study provides evidence that the application of the level set
method permits complete segregation of the intracranial arterial tree, substantially improving 3D
visualization.
page 635
page 636
21
In another study, van Bemmel and colleagues applied a level sets segmentation to the problem of
measuring stenosis in the internal carotid artery of contrast-enhanced MRA exams. Their technique
initializes the level sets propagation from a central axis of the vessel determined from user-placed
seed points. The stenosis measurement results were promising, yielding less interobserver variability
8 of 10 15-04-2010 21:48
than manual measurement methods.
In a third example, Baillard and colleagues applied the level sets method to the task of automated
segmentation of brain regions. The technique used a novel registration scheme utilizing prior data sets
to initialize the level sets boundary evolution.22
Active Shape Models

Another class of segmentation algorithms is also based on successive evolution of a boundary curve or
surface, but not based solely on properties of the voxel values comprising the data volume. Instead,
the algorithm is initialized and guided by an "active shape model" developed directly from a "training
set" of many instances of the type of anatomic region targeted.23 This prototype shape is adapted to
fit the specific boundaries of the newly presented data set through a series of approximating iterations,
until a preset minimum change criterion is reached. It is important that the training set represents a
good sample of the natural variation that the algorithm will need to cope with when presented with new
instances of the shape, so its search space will be large enough to accommodate the variation.
The active shape model concept has spawned a number of research projects in recent years. One
area of application is automated segmentation of subcortical brain regions in MR image sets. For
example, Duta and Sonka24 describe one such approach capable of segmenting a number of deep
brain structures, such as the thalamus, putamen, and ventricular system. The algorithm was trained on
just eight patient studies and tested on another 15. Regions were reliably labeled and border position
errors were approximately one pixel on 256 × 256 pixel MR slice images.
Effective segmentation of MR data sets, or any medical images for that matter, is rarely accomplished
with just one of the techniques described so far. It is usually necessary to combine a number of these
methods into a pipeline of processing steps to achieve an optimal result. 25 Segmentation algorithms
that produce results consistent and precise enough for quantitative analysis almost always require
multiple steps, accomplished in a pre-programmed sequence, or with intermediate results presented to
the user for verification or adjustment of parameters to guide the next step.
Active Appearance Models

An example of that combination is an extension of the active shape model method to the active
26
appearance model. This formulation combines both shape and image gray-level information into the
model built up from a training set of instance images. The image information is introduced as
normalized first derivative profiles; that is, a measure of the gray-level appearance of the edge at each
point on the model boundary. This refinement of the active shape model is more suitable for complex
images, such as MR, where the appearance of the boundary of a structure changes considerably
depending on what its various neighbors are.
For example, one of the most medically useful and challenging segmentation tasks is finding the inner
and outer boundaries of the left ventricle of the heart in cardiac MR exams. Segmenting the ventricle in
time-resolved data sets enables measurements of cardiac function, such as ejection fraction and
27
regional myocardial wall motion. Mitchell et al applied a 3D implementation of the active appearance
model concept to this task. The 3D surface model was trained from a set of images manually traced
by expert readers. Information about the ventricular wall shape as well as its gray-level appearance
was represented in the model. Resulting endocardial and epicardial volume measurements agreed well
with those derived from the manual tracings.
Van Ginneken and colleagues have extended the active appearance model to incorporate "optimal
features" in place of the normalized first-order derivative profiles of the original formulation.28
Landmarks are chosen and the algorithm automatically extracts appearance features from the training
images. These optimized appearance features are then matched between the trained model and the
target data set. They applied and tested the method in two difficult segmentation tasks on MR images
of the brain: delineating the boundaries of the cerebellum and of the corpus callosum. The results were
9 of 10 15-04-2010 21:48
compared with those obtained with an active shape model reference and significant improvements in
the reliability of the boundary following were noted.
Semantic Segmentation
Some of the most advanced segmentation techniques combine one or more of the methods described
so far with the notion of semantic or knowledge-based model of the objects and structures to be
parsed from the image data. Cabello and colleagues presented one of the earliest formulations of this
approach in medical imaging in 1990, in a study focused on anatomic feature extraction from plain-film
X-rays.29 The concept is to supplement the bottom-up approach of extracting features from images
using filters and models with a top-down approach that starts with human understanding and
classification of anatomy. The features extracted from the source images are matched against
preexisting notions of spatial and logical relationships, such as "the chest contains two lungs left and
right of a central mediastinum" or "the thalamus is below and between the cerebral hemispheres" or
"the aorta and vena cava run approximately parallel."
page 636
page 637
In simultaneously segmenting the gray/white and white/CSF boundaries in MR images of the brain,
MacDonald and colleagues applied such spatial relationship constraints to a deformable surface model
segmentation scheme.30 Knowledge of the topology (no self-intersecting surfaces), the relationship
between the two surfaces (CSF boundary inside gray/white boundary), and the expected range of
thicknesses of the gray and white layers were incorporated into the method. The design also allowed
for the two surfaces to influence and guide each other as the algorithm progressively refined their
shape toward an optimum result.
31
Shan et al have developed a scheme to delimit the frontal lobe of the cerebral cortex in MR images
by using knowledge of the position and appearance of key sulci. The sulci are located first, in a
hierarchic sequence, beginning with the longitudinal fissure. The sulci are extracted from the images
using morphologic operators and separated into a set of feature components by connectivity. The
feature components are evaluated against the anatomic model using fuzzy membership functions.
Results were compared against manual segmentation by experts and were favorable.
The semantic model may include information of a physiologic as well as structural nature. In developing
a hierarchic scheme to segment blood vessels from noncontrast MR angiograms of the brain,
Summers et al incorporated information about flow from velocity images, as well as caliber and
extension, into their model describing the appearance of blood vessels in these imaging conditions. 32
Multi-echo Segmentation
One more segmentation approach, which is unique to MR, deserves mention despite less activity in the
area over the last few years. Because MR acquisitions can consist of multiple images (T1, T2, PD,
various spin-echoes) at each location in the tissue volume, each with its own tissue contrast properties,
methods to examine the correlation of those contrast properties have been attempted.33-35 The goal of
these multi-echo techniques is to extract tissue identification and characterization information from the
image data directly, without resort to shape information or models. The general approach is to
examine correlation scatter plots of pixel values from the various images produced of a given region,
identifying discrete and characteristic clusters of values. Using expert knowledge of the anatomy, the
clusters are assigned to one tissue type or another and criteria established to separate adjacent
clusters. The resulting pixel-by-pixel classification may be displayed using arbitrary colors to represent
the various tissue types.
© 2010 Elsevier
10 of 10 15-04-2010 21:48
THREE-DIMENSIONAL REFORMATTING AND RENDERING

Not long ago, 3D rendering of medical images was considered a research obsession, with little
practical value to clinical radiology. That perception has changed dramatically over the last several
years. The shift is probably due to a number of factors acting in concert. One factor is the dramatic
improvement in the speed, usability, and affordability of 3D workstation hardware and software.36
Deployment of 3D workstations is now practical on a broad scale, entertaining routine use in clinical
practice. Another factor that has increased the perceived value of 3D rendering is a shift in applied
software algorithms. Surface rendering has largely been replaced by volume rendering for medical
visualization. As we will see later in the section, volume rendering is a more robust methodology, which
can be adapted appropriately to the varying character of source data. The third factor influencing the
rise of 3D utility and utilization is the massive increase in the size and resolution, both spatial and
temporal, of source cross-sectional data. This has been perhaps most dramatic with multi-row
detector CT, but is also true with MR. Acquisitions of MR data are now more often than previously
done as true 3D acquisitions, where a contiguous volume of data without interslice gaps results.
Particularly in the area of contrast-enhanced MR angiography, the high spatial resolution and now
time-resolved (4D) nature of these studies make them ideal candidates for 3D rendering operations.
Multiplanar Reformatting
Before considering the projection 3D techniques, those that combine all or most of the source cross-
sectional data into a view of the scanned volume, let us consider the techniques that form new cross-
sections from the original data. Magnetic resonance imaging is fundamentally different from CT in this
regard, since the MR scanner can be set up to produce a slice sequence or volume acquisition in any
chosen plane or orientation. In fact, conventional (2D) scanning protocols usually include separate
acquisitions in two of three planes (axial, sagittal, coronal or oblique). Given this flexibility at acquisition
time, multiplanar reformatting (MPR) would seem redundant at first.
However, especially when 3D or volume acquisitions are performed, it may be quite useful to derive
cross-sections in new planes after the fact, at the time of segmentation, viewing, and interpretation.
Most 3D post-processing workstations will derive and include views in the two planes orthogonal to the
acquisition plane automatically when viewing MR slice series. Whether these additional slicing planes
are useful depends on whether the underlying acquisition has sufficiently fine sampling or close spacing
between slices. When this is the case, additional oblique planes can also be produced on demand.
Reformatted cross-sections that follow a curved line projected through the volume can also be quite
useful. Common applications include generating cross-sections that follow the natural curve of the
spinal column (Fig. 23-6) or tracking through the centerline of blood vessels (Fig. 23-7). Recently,
researchers have produced software that creates planar images at various depths parallel to the
overall curved surface of the brain. This can be effective for mapping epileptic lesion position and
37 38
extent in the cerebral cortex and for assisting electrode placement in functional studies.
page 637
page 638
1 of 10 15-04-2010 21:49
Add to lightbox
Add to lightbox
Figure 23-6 Curved section through spine to examine disk spaces. A, Original sagittal section used
for reference in placing control points for reformatted section. B, Resulting reformatted section,
showing disk space measurement.
Maximum Intensity Projection

The first introduced and still most common projection-based 3D rendering technique is the maximum
intensity projection (MIP). The algorithm consists of picking a projection direction or viewpoint,
searching along each projection ray for the volume pixel (voxel) with the highest intensity value and
2 of 10 15-04-2010 21:49
assigning this value to the pixel in the projection image corresponding to this ray. The algorithm is easy
to understand and can be implemented with less computational power than that required for surface or
volume rendering, contributing to its early and continued popularity with both users and system
developers.
Add to lightbox
Add to lightbox
Figure 23-7 Curved section through carotid artery at bifurcation on MRA. A, 3D volume-rendered view
of isolated carotid artery, showing course of centerline for vessel tracking (green overlay). B,
Reformatted curved section through centerline of carotid artery, showing measurement of length of
carotid bulb.
3 of 10 15-04-2010 21:49
The MIP technique lends itself to data sets wherein the structures of greatest interest are also those
with the highest imaged intensities. This is particularly true for contrast-enhanced MR angiography.
Since the extravascular background intensities of MRA data sets are often quite a bit lower than the
range of intensities in the contrast-filled lumen of the blood vessels, MIP projections showing only the
vessels are feasible. This is an advantage that 3D MRA has over CT angiography. In the CT case, the
inevitable presence of bright bone in the line of projection means that some of the vessel course will be
confounded, unless segmentation is used to remove the bone-containing voxels from the data set
before the MIP projection is made.
The MIP method has the advantage that its projection images can be produced with no a priori
adjustment of parameters, except for the window/level transform. This means that it is easy to
generate MIP images in an unattended batch process, with no interactive user intervention. In the early
days of MRA, this meant that series of MIP images from various projection angles could be produced
right on the scanner console computer and included in the images that were printed to film and
provided to the interpreting radiologist. Now with the advent of PACS, much faster post-processing
workstations, and electronic on-screen reading of images, batch generation is less important. MIP
images can be generated, viewed, and manipulated almost instantly on demand. Interactive
window/level, rotation, and zooming on the workstation give complete freedom to the reading
radiologist to find the very best contrast relationship and projection angle to confirm a finding or
illustrate it to the referring physician.
page 638
page 639
Add to lightbox
Add to lightbox
Figure 23-8 MIP thickness in examining renal MRA. A, Standard MIP projection through entire data
volume. B, Limited thickness MIP (40 mm), showing clearer visualization of renal arteries free of
soft-tissue clutter and overlying vessels.
4 of 10 15-04-2010 21:49
A useful variation on this theme, again possible in the current interactive, on-demand electronic reading
environment, consists of limiting the projection to a thick planar subset of the source image volume.
These thick MIP "slabs" allow potentially confounding overlying or underlying structures to be easily
removed from a view, while still extending the "depth of field" beyond that available in a single-
pixel-thick cross-section (Fig. 23-8).
While certainly valuable, the MIP projection does have significant limitations. Probably the greatest is
that it fails to preserve and depict information about the relative depth of overlying structures within an
imaged volume. This arises from the nature of the algorithm itself. It chooses the brightest voxel along
the viewing ray, regardless of its depth in the volume. For example, when two blood vessels cross
each other in a given plane of projection, the viewer cannot tell which is in front of the other because
the two vessel profiles merge at the crossing point (Fig. 23-9). Prior knowledge of anatomy and
examining a rotation sequence of varying projection angles to gain parallax cues may help reduce this
ambiguity, but it cannot be definitively removed, particularly in the face of normal anatomic variation or
pathologic displacement of normal spatial relationships. In addition, some researchers have questioned
the suitability of MIP images for quantitative work, such as vascular stenosis measurement.39
Surface Rendering
Add to lightbox
Add to lightbox
Figure 23-9 Crossing vessel problem with MIP and MRA. A, Frontal projection of abdomen; note that
the superior mesenteric artery appears to disappear into the aorta in the center of the image. B, When
the projection angle is rotated 40° to the left, the full course of the SMA reappears.
5 of 10 15-04-2010 21:49
One response to the limitations of MIP images in communicating the relative depth of structures was
the application to medical imaging of surface rendering. The concept is to extract a 3D geometric
model of the surfaces of relevant anatomic structures from volumetric data and then render those
surfaces, using the techniques originally developed for modeling constructed objects in computer-aided
design. The necessity of extracting a surface model a priori implies that a segmentation algorithm must
be available for the structures to be displayed. If the data contain distinct edges, boundaries can be
defined as isosurfaces. The first successful isosurface extraction algorithm for volumetric medical
image data was developed by Lorenson and Cline40 and dubbed "marching cubes".
In contrast to MIP rendering, the relative depth of overlying structures is taken into account in the
surface rendering process. Underlying occluded structures are properly removed from the view, clearly
communicating a sense of depth, especially in rotating image sequences. To increase further the sense
of depth and surface shape and orientation, a simulated lighting model is applied during rendering, so
surfaces oriented to reflect light from the imaginary source toward the viewer appear brighter than
those aspects angled away from the light source and/or the viewer. Since modern computers are now
almost universally equipped with graphics hardware specifically designed to accelerate the display of
3D-rendered surfaces, viewing surface renderings is generally fast and interactive. However, the initial
segmentation and surface extraction process can be more compute-intensive, especially when
interactive adjustments are necessary, as described below.
page 639
page 640
MR data are particularly difficult to work with in the surface-rendering paradigm, for three reasons.
First, MR data are not calibrated to consistent gray-levels for particular tissue types and features,
meaning that the isosurface thresholds defined in the segmentation and modeling step will probably
need to be adjusted for each data set. Second, MR data are frequently noisy enough to disrupt the
smooth contours that make for smooth-appearing 3D surface models. When isosurfaces are extracted
in the presence of significant noise, those surfaces take on an artifactually rough texture, sometimes
obscuring their true shape. Third, with the exception of MRA exams, MR data, particularly T1-weighted
images, may contain many anatomic edges of common gray-levels, making the extraction of just one
or a few isosurfaces more difficult.
Various improvements have been made or proposed for surface rendering of medical images over the
years,41,42 but all these surface-based methods suffer from the same limitation-the necessity of
abstracting the data into a geometric model prior to any visualization of the source data in a 3D
context. A loss of most of the content of the original volume image is the inevitable result of model
extraction. In situations where the target of the imaging investigation is very well defined, and that
target has consistent properties from exam to exam, making reliable segmentation for model extraction
feasible, this immediate data reduction may be appropriate. Nevertheless, in situations where multiple
unpredictable findings may be contained in the image data set, the presegmentation involved in surface
rendering can rule it out as an effective 3D visualization methodology.
Volume Rendering
In the mid-1980s, George Lucas organized a group of particularly talented scientists and engineers in
the fields of image processing and 3D computer graphics to develop fundamentally new techniques for
special effects in movie production. That group was spun off from Lucas' studio to form the company
Pixar. Before Pixar's metamorphosis into the animation studio of today, the company's scientists
developed key techniques at the junction point of the previously separate disciplines of 3D computer
graphics and image processing. They called this new field "image computing" and one of its most
43
significant results was the technique of volume rendering. Once introduced, the method was
developed and optimized by a growing academic community of visualization researchers. 44,45
Early on, the routine and practical application of volume rendering to disciplines such as radiology with
such sharp time and productivity constraints was limited by the much greater computing requirements
of the method, compared with surface rendering. However, over the past decade or so this
disadvantage has largely been erased, both by the relentless advance in computing speed and
6 of 10 15-04-2010 21:49
capacity and by increasingly optimized rendering software implementations. Today, it is quite practical
to interactively volume render medical data sets of the typical sizes encountered clinically even on
laptop computers.
Volume rendering (also called direct rendering and volume imaging in the technical literature) involves
the building up of a 3D projection image of a volumetric source image data set by compositing together
a weighted combination of all the voxel values along each viewing ray. Instead of just selecting the
brightest single voxel value, as in MIP, each voxel value in the data volume is first assigned a
corresponding opacity value (alpha value in the computer graphics literature) that will determine its
contribution, along with all other voxels in the same projection ray, to the gray-level or color of the
resulting pixel in the final image.
With volume rendering, the correspondence between the MR signal level and the assigned opacity
value can be completely arbitrary, providing a new level of control over what gray-level ranges
contribute opacity and are visible in the rendered image, and which are transparent. Generally
speaking, a transfer function that yields a direct relationship between MR signal intensity and volume
rendering opacity will produce the most intuitive result. That is, areas that are dark in cross-sectional
images are transparent in the rendering and those that are bright are more opaque. However, the
precise shaping of this transfer function can be used to control the relative contribution or prominence
of various ranges of signal intensity.
The volume rendering process integrates the opacity-weighted voxel values along the viewing direction
in an order-dependent fashion. With the appropriate compositing or combining function, this can be
done either back-to-front or front-to-back in relation to the chosen viewpoint. In either case, voxels in
the foreground have a much greater contribution to the final pixel result for that viewing ray than voxels
in the background. In this way the depth cue of occlusion results-overlying structures hide those
behind. The result is that foreground and background are as clearly distinguished as in surface
rendering.
This basic form of volume rendering, with voxels contributing their native gray-level to the final scene
according to an applied opacity transfer function, results in images that appear to be "extruded" from
the source cross-sectional images. Gray-levels are preserved, so one can relate familiar contrast
relationships from the source cross-sections to the rendered 3D projections. For example, in such
basic volume renderings reconstructed from T1-weighted MR images, fat tissue is still bright, aqueous
tissue is gray and air is dark (Fig. 23-10). In T2-weighted renderings, aqueous spaces such as the
brain ventricles will still appear bright.
One of the prime advantages of volume rendering over surface rendering is apparent already with this
basic technique. Because the rendering is an integrative process, adding together multiple voxels to
produce the final gray-level of each pixel in the rendering, random noise in the source data tends to be
averaged out and suppressed. Residual noise is perceived more as a graininess superimposed on
structures than as definitive shape elements of those structures, as can happen with surface rendering.
Hans and colleagues demonstrated this effect elegantly in a comparative visualization study of MR
high-resolution imaging of the cochlea. 46 Neri and colleagues confirmed the suitability of volume
47
rendering for this application.
page 640
page 641
7 of 10 15-04-2010 21:49
Add to lightbox
Figure 23-10 Volume-rendered T1-weighted coronal scan of a knee. View from rear aspect showing
bright subcutaneous fat, gray bone, and nearly transparent dark muscle.
While such basic volume renderings can be quite useful, it is often quite desirable to employ the
artificial lighting method used with surface rendering in order to provide cues about surface shape and
orientation. In volume rendering, since the picture is built up directly from the source voxels and there is
no extracted surface model for the lighting model to interact with, one must create an orientation
attribute for each voxel. This is accomplished by estimating a gradient vector at each voxel by
examining changes in the native voxels' value in the local vicinity of that voxel. The gradient or trend of
change is assessed by a difference calculation in each of the X, Y, and Z directions and then
normalized to produce a unit vector. The direction in which this unit vector points can be considered
normal or perpendicular to the surface this voxel lies on. When a strong consistent gradient is present
indicating an underlying coherent surface, the vectors calculated in that patch of voxels all point in
similar directions. In the interior of more or less homogeneous regions, the gradient vectors will point in
random directions, influenced by the residual random noise in these source data.
When the lighting model is applied, the brightness that each voxel contributes to a viewing ray is no
longer associated with its original gray-level, but rather with the amount of artificial light reflected from
it, based on its orientation in relation to the incident light source and to the viewer, just as in surface
rendering. Therefore, surface orientation cues will be well represented, but at the cost of losing
reference to the original gray-levels in the source data. How can we get the best of both cues? A
solution lies in the application of a pseudo-color scale. If a continuous range of colors, such as a heat
or hot-metal scale (black-to-red-to-orange-to-yellow-to-white), is mapped to the original gray-levels of
the source data and voxels rendered with these colors, information about MR signal level or original
gray-level can be communicated in the rendering, even in the presence of surface lighting and shading
(Fig. 23-11).
8 of 10 15-04-2010 21:49
Add to lightbox
Add to lightbox
Figure 23-11 Use of lighting and color in volume rendering a peripheral MRA. A, Lighting applied
without color. B, With heat scale color applied, greater distinction between contrast-filled vessels and
soft-tissue landmarks is evident.
One might ask-why go to all the trouble of volume rendering an MR data set voxel by voxel if one ends
up with an image that appears similar to a surface rendering? The answer is the summation of the
various points of advantage we have already enumerated: no requirement for a priori modeling and
segmentation, greater flexibility in determining what is visible, higher immunity to noise, and the ability
to retain information about original gray-level differences. There is one more advantage and it is a
subtle one, relating to quality and fidelity of the resulting image. In a surface rendering, the position of
the surface is determined as a binary, all-or-none determination; the result is an artificially sharp
boundary and jagged horizons in the 3D view (Fig. 23-12A). In contrast, volume rendering can depict
surfaces as "soft" boundaries, preserving the finite resolving power of the original scan; the result is a
smoother, albeit somewhat fuzzier edge, an edge we would argue is more faithful to the resolving
power present in the original scan (Fig. 23-12B).
If the reader is new to the use of volume rendering in medical imaging, one might be concerned that so
many degrees of freedom (opacity transfer function, lighting model, color scheme) could make volume
rendering too difficult or time-consuming or too prone to individual user skill for routine clinical use. This
was initially the case with early research systems and some remaining workstation software. However,
the best current workstation software products now use application-specific visualization protocols and
presets to standardize these rendering parameters and deliver reproducible rendered results.
9 of 10 15-04-2010 21:49
© 2010 Elsevier
10 of 10 15-04-2010 21:49
CASE STUDIES
Considering all these methodologic points, we now present a few specific examples of applying
segmentation and 3D rendering to MR exams. In each example, we will illustrate the techniques in
practical and clinically relevant scenarios.
page 641
page 642
Add to lightbox
Add to lightbox
Figure 23-12 Quality of edges and small structures, volume versus surface rendering. A, Detail of
surface rendering of mesenteric branch arteries. B, Same scene in volume rendering showing more
continuous vessel profiles and smoother edges.
1 of 9 15-04-2010 21:50
CASE STUDY 1 BRAIN SURFACE AND VENOGRAPHY

First, let us consider a common contrast-enhanced T1-weighted brain exam, with an eye toward
providing a referring surgeon with information about the relationship of the patient's cortical surface and
overlying veins. Using volume rendering and applying a simple opacity "ramp" transfer function that
directly relates gray-levels with voxel opacity yields an image with the scalp surface prominent, given
the bright (therefore opaque) subcutaneous fat surrounding the vault (Fig. 23-13A). Visualizing the
cortical surface will require segmenting this layer away, either using manual sculpting tools or
preferably a segmentation algorithm tuned to this particular purpose. With this segmentation
accomplished, we have a satisfactory view of the exposed cortical surface and the overlying contrast-
filled veins (Fig. 23-13B). Adding lighting and a "hot metal" pseudo-color scale produces a more
photorealistic image. Note that the color scale helps delineate the veins by rendering the surrounding
aqueous tissue in a deeper shade of orange (Fig. 23-13C).
CASE STUDY 2 RENAL MAGNETIC RESONANCE ANGIOGRAPHY
In a second example, we consider the 3D presentation of a thoracic and abdominal contrast-enhanced
MRA data set. Such exams are performed to evaluate the renal artery configuration in a prospective
living renal donor.48 This is a good opportunity to compare the properties of various 3D rendering
algorithms we have described-MIP, surface rendering, basic volume rendering, and volume rendering
with lighting, surface shading, and color applied. In each of the five images presented in Figure 23-14,
we present identical frontal projections of the 3D data set; only the rendering method differs.
In the first image (Fig. 23-14A), we present the data in a conventional MIP rendering. Clearly, this
image is useful, particularly in its high sensitivity to the fine distal branches of the various abdominal
arteries, as well as in providing some soft-tissue context and landmarks for the vasculature. However,
the image aptly illustrates the limitation of MIP in depicting crossing vessels. The crossing point of the
right renal artery and the superior mesenteric artery is ambiguously presented.
In the second image (Fig. 23-14B), the same data set is surface rendered, displaying the aorta and
major abdominal arteries as an isosurface. Note that the isosurface approach yields a fragmented
appearance, since the distribution of vascular contrast is in fact not uniform. However, we see that
crossing vessels are now clearly distinguishable and their relative position in depth is apparent.
The third image (Fig. 23-14C) is a basic volume rendering; only an opacity transfer function is applied
to increase contributions to the image from the contrasted vasculature and decrease contributions from
the soft-tissue surround. Note that, unlike the surface rendering, the transition from the tissue surround
to the contrasted regions is gradual, so one can still distinguish the renal parenchyma and some of the
finer arterial branches. Here again, in contrast to the MIP rendering, crossing vessels are depicted
unambiguously and a sense of depth is apparent.
page 642
page 643
2 of 9 15-04-2010 21:50
Add to lightbox
Add to lightbox
3 of 9 15-04-2010 21:50
Add to lightbox
Figure 23-13 Brain surface and venography for surgical planning from T1-weighted MR exam. A,
Basic volume rendering with scalp included. B, Basic volume rendering with brain surface and
cerebral veins exposed. C, Same exposed surface view with lighting and color applied.
The fourth image (Fig. 23-14D) introduces artificial lighting to the volume rendering. The light source is
coincident with the viewpoint, like a miner's headlamp. As described earlier, light and dark in the
rendering no longer correspond to the signal levels in the source MR data. Instead, they relate to
orientation of neighborhoods of the source voxels with respect to the light source and viewer. As a
result, this image exhibits greater sensitivity to the darker structures in the image volume such as fine
vessel branches, when compared with the opacity-only volume rendering, and is comparable in this
regard to the MIP image. In comparison to the surface rendering, note that the shapes of structures
are just as well depicted, but the lack of an isosurface constraint produces a more continuous and
natural-looking image.
The fifth and final image in this sequence (Fig. 24-14E) takes the fourth lit, volume-rendered image and
adds the pseudo-color scale. As before, this refinement restores a cue about relative signal level to the
rendering in the presence of the light reflection model. One sees that the vessels and soft-tissue
surround are somewhat better distinguished. The choice between the monochrome image and color
image is probably best left to the individual preference of the reading radiologist or the referring
physician who may be the primary client of such renderings.
Considering the illustrated benefits of the volume renderings (retained sensitivity for small and faint
structures, good sense of depth and spatial relationship, depiction of surface shape and orientation,
and continuous naturalistic image), one can appreciate why increasingly researchers and practicing MR
angiographers alike are beginning to favor volume rendering over MIP.49
page 643
page 644
4 of 9 15-04-2010 21:50
Add to lightbox
Add to lightbox
Add to lightbox
5 of 9 15-04-2010 21:50
Add to lightbox
Add to lightbox
Figure 23-14 Renal MRA with comparison of 3D rendering methods. A, MIP. B, Shaded surface
rendering. C, Basic volume rendering. D, Volume rendering with lighting. E, Volume rendering with
lighting and pseudo-color.
CASE STUDY 3 BRAIN TUMOR VISUALIZATION AND VOLUME MEASUREMENT

Our next example is another brain scan, this time with the aim of characterizing the location and
6 of 9 15-04-2010 21:50
measuring the volume of a large central mass. In this case, we will examine the value of volume
measurement by semi-automated segmentation50 and using transparent volume rendering to illustrate
the overall position of the tumor. The source data are a 119-slice 3D sagittal acquisition. The brain was
also segmented from the rest of the head using a semi-automated algorithm with subsequent manual
editing of the boundary. The tumor was segmented using an automated algorithm that seeks to
distinguish between enhancing and nonenhancing regions of the tumor, corresponding to metabolically
active tissue and the necrotic core, respectively.
page 644
page 645
The first image (Fig. 23-15A) is a representative original sagittal slice from the data set. The next two
(Figs. 23-15B and 23-15C) are coronal and axial sections reformatted from the same volume data.
While the resolution and sampling of this data set are not isotropic, resulting in some stair-step
artifacts in the reformatted sections, it is nonetheless evident that the reformatted sections are
valuable in providing additional context in characterizing the tumor position.
The fourth image (Fig. 23-15D) again shows a sagittal slice but now after the brain and tumor
segmentation steps have been performed. The red outline indicates the boundary of the segmented
brain; the dotted cyan line indicates the search perimeter with which the tumor algorithm was initialized;
the yellow outline is the outer boundary of the segmented enhancing region of the tumor; and the
magenta lines indicate the boundary between the nonenhancing and enhancing regions of the tumor.
The volumes of the enhancing, nonenhancing, and entire tumor are calculated from these last two
boundaries.
Add to lightbox
Add to lightbox
7 of 9 15-04-2010 21:50
Add to lightbox
Add to lightbox
Add to lightbox
8 of 9 15-04-2010 21:50
Add to lightbox
Figure 23-15 Brain tumor visualization and volume measurement from contrast-enhanced MR. A,
Original mid-sagittal slice. B, Reformatted coronal slice through mass. C, Reformatted axial slice
through mass. D, Segmented sagittal slice showing brain boundary (red), tumor segmentation search
region (cyan), outer boundary of enhancing mass (yellow), boundary between enhancing and
nonenhancing regions of mass (magenta). E, Transparent volume rendering with embedded boundary
contours of mass. F, Volume rendering with red-tinted brain and embedded contours of mass.
The final two images (Figs. 23-15E and 23-15F) are two volume renderings illustrating the overall 3D
position of the tumor with respect to important landmarks. The first of these shows the transparent
brain with the yellow outlines of the segmented tumor visible within. The sulci and gyri of the cortex are
visible, as are the cerebral veins. The second volume rendering shows the entire head with the
red-tinted segmented brain within and the yellow outlines of the tumor. The red tinting is accomplished
by applying a color map selectively inside the segmented brain region. Note the appearance of the left
orbit, the sylvian fissure and lateral aspect of the cerebellum. The circular object in front of the left ear
is an adhesive landmark applied to the patient's skin to aid in registration of the data set during
intrasurgical image-guided navigation.
© 2010 Elsevier
9 of 9 15-04-2010 21:50
CONCLUSION
page 645
page 646
We have presented a survey of the basis, development, current research directions, and applications
of image processing applied to MR. The progress in this area has been quite dramatic over the past
several years. This progress is the result of a stronger conceptual foundation for filtering,
segmentation, and rendering techniques but it is also due, as is the case in other imaging modalities, to
the dramatic improvements in the quality of the source images now being produced from MR scanners.
High-quality input images provide a much more fertile ground for successful segmentation,
measurement, and 3D rendering results. In addition, the rapid advance of computing power and
capacity makes more ambitious and complex algorithms practical to use in a busy clinical setting. Many
of the techniques we have illustrated in the case studies, which took only minutes to apply with current
technology, would have been impractical 10 years ago. Given the encouraging trends in both
technology and algorithm research we have described, clinical users of these methods have much to
look forward to in the coming years.
REFERENCES
1. Sonka M, Hlavac V, Boyle R: Image Processing, Analysis and Machine Vision, 2nd ed. Pacific Grove, CA: Brooks/Cole
Publishing, 1999, p 57.
2. Russ JC: The Image Processing Handbook. Boca Raton, FL: CRC Press, 1992.
3. Suri JS, Setarehdan SK, Singh S: Advanced Algorithmic Approaches to Medical Image Segmentation: State of the Art
Applications in Cardiology, Neurology, Mammography and Pathology. New York: Springer-Verlag, 2002.
4. Robb RA: Biomedical Imaging, Visualization, and Analysis. New York: Wiley-Liss, 1999.
5. Gonzalez RC, Woods RE: Digital Image Processing, 2nd ed. Boston, MA: Addison-Wesley, 2002.
6. Perona P, Malik J: Scale-space and edge detection using anisotropic diffusion. IEEE Trans Pattern Anal Machine Intell
12:629-639, 1990.
7. Black MJ, Sapiro G, Marimont DH, et al: Robust anisotropic diffusion. IEEE Trans Image Processing 7:421-432, 1998.
8. Gerig G, Kubler O, Kikinis R, et al: Nonlinear anisotropic filtering of MRI data. IEEE Trans Med Imag 11:221-232, 1992.
9. Murase K, Yamazaki Y, Shinohara M, et al: An anisotropic diffusion method for denoising dynamic susceptibility contrast-
enhanced magnetic resonance images. Phys Med Biol 46:2713-2723, 2001. Medline Similar articles
10. Placidi G, Alecci M, Sotgiu A: Post-processing noise removal algorithm for magnetic resonance imaging based on edge
detection and wavelet analysis. Phys Med Biol 48:1987-1995, 2003. Medline Similar articles
11. Ahn CB, Song YC, Park DJ: Adaptive template filtering for signal-to-noise ratio enhancement in magnetic resonance
imaging. IEEE Trans Med Imaging 18:549-559, 1999. Medline Similar articles
12. Saha PK, Udupa JK: Scale-based diffusive image filtering preserving boundary sharpness and fine structures. IEEE Trans
Med Imaging 20:1140-1155, 2001. Medline Similar articles
13. Westin CF, Wigstrom L, Loock T, et al: Three-dimensional adaptive filtering in magnetic resonance angiography. J Magn
14. Lai SH, Fang M: A dual image approach for bias field correction in magnetic resonance imaging. Magn Reson Imaging
15. Cohen MS, DuBois RM, Zeineh MM: Rapid and effective correction of RF inhomogeneity for high field magnetic resonance
imaging. Hum Brain Mapp 10:204-211, 2000. Medline Similar articles
16. Gispert JD, Reig S, Pascau J, et al: Method for bias field correction of brain T1-weighted magnetic resonance images
minimizing segmentation error. Hum Brain Mapp 22:133-144, 2004. Medline Similar articles
17. Adalsteinsson D, Sethian JA: A fast level set method for propagating interfaces. J Computat Phys 118:269-277, 1995.
18. Sethian J: Level Set Methods and Fast Marching Methods, 2nd ed. Cambridge, UK: Cambridge University Press, 1999.
19. Whitaker RT, Breen DE, Museth K, et al: A Framework for Level Set Segmentation of Volume Datasets. Proceedings of
the International Workshop on Volume Graphics, 2001, pp 159-168.
20. Farag, AA, Hassan H, Falk R, et al: 3D volume segmentation of MRA data sets using level sets: Image processing and
display. Acad Radiol 11:419-435, 2004. Medline Similar articles
21. van Bemmel CM, Viergever MA, Niessen WJ: Semiautomatic segmentation and stenosis quantification of 3D contrast-
enhanced MR angiograms of the internal carotid artery. Magn Reson Med 51:753-760, 2004.
22. Baillard C, Hellier P, Barillot C: Segmentation of brain 3D MR images using level sets and dense registration. Med Image
Anal 5:185-194, 2001.
23. Cootes TF, Taylor CJ, Cooper DH, et al: Active shape models-their training and application. CompVision Image
1 of 3 15-04-2010 21:50
Understanding 61:38-59, 1995.

24. Duta N, Sonka M: Segmentation and interpretation of MR brain images: an improved active shape model. IEEE Trans Med
25. Atkins MS, Mackiewich BT: Fully automatic segmentation of the brain in MRI. IEEE Trans Med Imaging 17:98-107, 1998.
26. Cootes TF, Edwards GJ, Taylor CJ: Active appearance models. In Burkhardt H, Neumann B (eds): 5th European
Conference on Computer Vision, vol 2. Berlin: Springer, 1998, pp 484-498.
27. Mitchell SC, Bosch JG, Lelieveldt BP, et al: 3-D active appearance models: segmentation of cardiac MR and ultrasound
images. IEEE Trans Med Imaging. 21:1167-1178, 2002. Medline Similar articles
28. van Ginneken B, Frangi AF, Staal JJ, et al: Active shape model segmentation with optimal features. IEEE Trans Med
29. Cabello D, Delgado A, Carreira MJ, et al: On knowledge-based medical image understanding. Cybernet Systems 21:
277-289, 1990.
30. MacDonald D, Kabani N, Avis D, et al: Automated 3-D extraction of inner and outer surfaces of cerebral cortex from MRI.
Neuroimage 12: 340-356, 2000.
31. Shan ZY, Liu JZ, Yue GH: Automated human frontal lobe identification in MR images based on fuzzy-logic encoded expert
anatomic knowledge. Magn Reson Imaging 22:607-617, 2004. Medline Similar articles
32. Summers PE, Bhalerao AH, Hawkes DJ: Multiresolution, model-based segmentation of MR angiograms. J Magn Reson
33. Kao YH, Sorenson JA, Winkler SS: MR image segmentation using vector decomposition and probability techniques: a
general model and its application to dual-echo images. Magn Reson Med 35:114-125, 1996. Medline Similar articles
34. Schad LR, Bluml S, Zuna I: MR tissue characterization of intracranial tumors by means of texture analysis. Magn Reson
35. Velthuizen RP, Hall LO, Clarke LP: Feature extraction for MRI segmentation. J Neuroimaging 9:85-90, 1999. Medline
Similar articles
36. ter Haar Romeny BM, Zuiderveld KJ, Van Waes PF, et al: Advances in three-dimensional diagnostic radiology. J Anat
37. Meiners LC, Scheffers JM, De Kort GA, et al: Curved reconstructions versus three-dimensional surface rendering in the
demonstration of cortical lesions in patients with extratemporal epilepsy. Invest Radiol 36:225-233, 2001. Medline
Similar articles
38. Schulze-Bonhage AH, Huppertz HJ, Comeau RM, et al: Visualization of subdural strip and grid electrodes using curvilinear
reformatting of 3D MR imaging data sets. Am J Neuroradiol 23:400-403, 2002.
39. Baskaran V, Pereles FS, Nemcek AA Jr, et al: Gadolinium-enhanced 3D MR angiography of renal artery stenosis: a pilot
comparison of maximum intensity projection, multiplanar reformatting, and 3D volume-rendering postprocessing algorithms.
Acad Radiol 9:50-59, 2002.
40. Lorenson W, Cline H: Marching cubes: a high resolution 3D surface construction algorithm. Computer Graphics
21:163-169, 1982.
41. Cline HE, Lorensen WE, Ludke S, et al: Two algorithms for the three-dimensional reconstruction of tomograms. Med Phys
42. Wood Z, Desbrun M, Schröder P, et al: Semiregular mesh extraction from volumes. Proceedings of Visualization 2000, pp
275-282.
43. Drebin RA, Carpenter, L, Hanrahan P: Volume rendering. Computer Graphics 22:65-74, 1988.
44. Levoy M: Volume rendering. IEEE Comput Graph Appl 10:33-40, 1990.
45. Kaufman A, Cohen D, Yagel R: Volume graphics. Computer 26:51-64, 1993.
46. Hans P, Grant AJ, Laitt RD, et al: Comparison of three-dimensional visualization techniques for depicting the scala vestibuli
and scala tympani of the cochlea by using high-resolution MR imaging. Am J Neuroradiol 20:1197-1206, 1999. Medline
Similar articles
47. Neri E, Caramella D, Cosottini M, et al: High-resolution magnetic resonance and volume rendering of the labyrinth. Eur
48. Carr JC, Nemcek AA Jr, Abecassis M, et al: Preoperative evaluation of the entire hepatic vasculature in living liver donors
with use of contrast-enhanced MR angiography and true fast imaging with steady-state precession. J Vasc Interv Radiol
49. Mallouhi A, Schocke M, Judmaier W, et al: 3D MR angiography of renal arteries: comparison of volume rendering and
maximum intensity projection algorithms. Radiology 223:509-516, 2002. Medline Similar articles
50. Sorensen AG, Patel S, Harmath C, et al: Comparison of diameter and perimeter methods for tumor volume calculation. J
Clin Oncol 19:551-557, 2001. Medline Similar articles
page 646
page 647
*Portions of this chapter were excerpted with permission from Shellock FG, Crues JV: MR procedures: biologic effects, safety, and
2 of 3 15-04-2010 21:50
patient care. Radiology 2004; 232:635-652 and Shellock FG: Reference Manual for Magnetic Resonance Safety, Implants, and
Devices. Los Angeles, CA: Biomedical Research Publishing Group, 2004.
© 2010 Elsevier
3 of 3 15-04-2010 21:50
AGNETIC ESONANCE IOEFFECTS AFETY AND

ATIENT ANAGEMENT
Frank G. Shellock
INTRODUCTION
Magnetic resonance (MR) procedures have been utilized in the clinical setting for approximately 20
years. During this time, the technology has continued to evolve, yielding MR systems with higher static
magnetic fields, faster and higher gradient magnetic fields, and more powerful radiofrequency (RF)
transmission coils. For the more than 150 million MR examinations performed to date, no documented
serious harm has been caused to patients from short-term exposures to the electromagnetic fields
used for MR procedures performed at the levels currently recommended by the United States Food
and Drug Administration (FDA) and according to proper safety guidelines.1-4
Most reported cases of MR-related injuries and the few fatalities that have occurred have been the
result of not following safety guidelines or using inappropriate or outdated information related to the
safety aspects of biomedical implants and devices. 1-7 Notably, the preservation of a safe MR
environment requires constant attention to the management of patients and individuals with metallic
implants and devices because the variety and complexity of these objects constantly change.5-7
Therefore, to guard against accidents in the MR environment, it is necessary to revise bioeffects and
safety information according to changes that have occurred in MR technology and with regard to the
latest guidelines for biomedical implants and devices.1,2,5-16
In consideration of the above, this chapter provides an overview and update with regard to MR
bioeffects, discusses new or controversial MR safety topics and issues, presents evidence-based
guidelines to ensure safety for patients and staff members, and describes MR safety information for
various implants and devices that have recently undergone evaluation.
page 647
page 648
While a comprehensive discussion of MR bioeffects, safety, and patient management is not within the
scope of this chapter, these topics have been addressed in recently published review articles8-12,16 and
textbooks.5-7 In addition, there are two web sites devoted to MR safety that are updated with content
on a frequent, ongoing basis ( http://www.mrisafety.com-MR safety resource that presents information
for over 1300 implants and devices, discussion of over 60 different safety topics, summary of
bioeffects and safety literature, as well as other features; and http://www.IMRSER.org-the web site of
the Institute for Magnetic Resonance Safety, Education, and Research, (IMRSER) that provides safety
guidelines and recommendations developed by the advisory boards of the IMRSER, as well as
recently published MR safety articles posted from the peer-reviewed literature). Therefore, the reader
is directed to these additional sources of information for MR safety.
© 2010 Elsevier
1 of 1 15-04-2010 21:51
BIOEFFECTS OF STATIC MAGNETIC FIELDS

The introduction of MR technology as a clinical imaging modality in the early 1980s has been
responsible for a substantial increase in human exposure to strong static magnetic fields.1,9 Most MR
systems in use today operate with magnetic fields ranging from 0.2 to 3.0 tesla. In the research
setting, the most powerful clinical MR systems in the world are located at Ohio State University (8
tesla) and at the University of Illinois at Chicago (9.4 tesla). According to the latest guidelines from the
US FDA, clinical MR systems using static magnetic fields up to 8.0 tesla are considered a
2
"nonsignificant risk" for patients above the age of 1 month. The exposure of research subjects to
fields above 8.0 tesla requires approval of the research protocol by an Institutional Review Board and
the informed consent of the subjects.1,2
Schenck1,9 recently conducted a comprehensive review of bioeffects associated with exposure to

static magnetic fields. With regard to short-term exposures, the available information for effects of
static magnetic fields on biological tissues is extensive.1,9,17-55 Investigations include studies on
alterations in cell growth and morphology, cell reproduction and teratogenicity, DNA structure and gene
expression, pre- and postnatal reproduction and development, blood-brain barrier permeability, nerve
activity, cognitive function and behavior, cardiovascular dynamics, hematologic indices, temperature
regulation, circadian rhythms, immune responsiveness, and other biological processes.18-55 The
majority of these studies concluded that exposure to static magnetic fields on a short-term basis does
not produce substantial harmful bioeffects. Although there have been reports of potentially injurious
effects of static magnetic fields on isolated cells or organisms, none has been verified or firmly
established as a scientific fact.1,9,17 The relatively few documented injuries that have occurred in
association with MR system magnets were attributed to the inadvertent presence or introduction of
ferromagnetic implants or objects (e.g., oxygen tanks, aneurysm clips, etc.) into the MR
environment.1,5-7,9,17
Regarding the effects of long-term exposures to static magnetic fields, there are several physical
mechanisms of interaction between tissues and static magnetic fields that could theoretically lead to
1,9,17
pathologic changes in human subjects. However, quantitative analysis of these mechanisms
indicates that they are below the threshold of significance with respect to long-term adverse
bioeffects.1,9,17
Presently, the pertinent literature does not contain carefully controlled studies that demonstrate the
absolute safety of chronic exposure to powerful magnetic fields. With the increased clinical use of
interventional MR procedures, there is a critical need for such investigations.
© 2010 Elsevier
1 of 1 15-04-2010 21:52
BIOEFFECTS OF GRADIENT MAGNETIC FIELDS

Under certain conditions during MR procedures, gradient magnetic fields may stimulate nerves or
muscles by inducing electrical fields in patients. This topic has been reviewed by Schaefer et al, 8
55 56
Nyenhuis et al, and Bourland et al. The potential for interactions between gradient magnetic fields
and biological tissue is dependent on a variety of factors including the fundamental field frequency, the
maximum flux density, the average flux density, the presence of harmonic frequencies, the waveform
characteristics of the signal, the polarity of the signal, the current distribution in the body, the electrical
properties, and the sensitivity of the particular cell membrane.8,55-64
Gradient Magnetic Field-Induced Stimulation in Human Subjects

Several investigations have been conducted to characterize MR-related, gradient magnetic field-
induced stimulation in human subjects.57-64 At sufficient exposure levels, peripheral nerve stimulation is
perceptible as "tingling" or "tapping" sensations. At gradient magnetic field exposure levels 50% to
100% above perception thresholds, patients may become uncomfortable or experience pain.8 At
extremely high levels, cardiac stimulation is of concern. However, the induction of cardiac stimulation
requires exceedingly large gradient fields, more than an order of magnitude greater than those MR
systems presently available.8,55,56 Fortunately, the current safety standards for gradient magnetic
fields associated with modern-day MR systems appear to adequately protect patients from potential
risks.2,8,16,55
page 648
page 649
Interestingly, studies performed in human subjects indicated that anatomic sites of peripheral nerve
stimulation vary depending on the activation of a specific gradient (i.e., x-, y- or z-gradient).8
Stimulation sites for x-gradients included the bridge of the nose, left side of thorax, iliac crest, left
thigh, buttocks, and lower back. Stimulation sites for y-gradients included the scapula, upper arms,
shoulder, right side of thorax, iliac crest, hip, hands, and upper back. Stimulation sites for z-gradients
8
included the scapula, thorax, xyphoid, abdomen, iliac crest, and upper and lower back. Typically, sites
of peripheral nerve stimulation were associated with anatomic regions that have bony prominences.
According to Schaefer et al,8 since bone is less conductive than the surrounding tissue, it may increase
current densities in narrow regions of tissue between bone and skin, resulting in lower nerve stimulation
thresholds than expected.
© 2010 Elsevier
1 of 1 15-04-2010 21:52
ACOUSTIC NOISE
Various forms of acoustic noise are produced in association with the operation of an MR system. 65,66
The primary source of acoustic noise is the gradient magnetic field activated during the MR procedure.
This noise occurs during rapid alterations of currents within the gradient coils that, in the presence of
the MR system's powerful static magnetic field, produce substantial (i.e., Lorentz) forces. Acoustic
noise, manifested as loud tapping, knocking or chirping sounds, is generated when these forces cause
motion or vibration of the gradient coils as they impact against their mountings.
Problems associated with acoustic noise for patients and healthcare workers include simple
annoyance, difficulties in verbal communication, heightened anxiety, temporary hearing loss, and
possible permanent hearing impairment.65-81 Acoustic noise may pose a particular hazard to specific
patient groups who are at increased risk. Patients with psychiatric disorders, elderly, and pediatric
patients may be confused or suffer from heightened anxiety.65,66,68 Sedated patients may experience
discomfort due to high noise levels. Certain drugs are known to increase hearing sensitivity.69
Neonates with immature anatomic development may have an increased reaction to acoustic noise, as
has been reported by Philbin et al.70
Characteristics of Magnetic Resonance-Related Acoustic Noise

Variations in MR-related acoustic noise occur with alterations in the gradient output (rise time or
amplitude) associated with different MR parameters.65,66,71-84 Noise is enhanced by decreases in
section thickness, field of view, repetition time, and echo time. The physical features of the MR
system, especially whether or not it has special sound insulation, and the material and construction of
gradient coils and support structures also affect the transmission of acoustic noise and its perception
by the patient.
The presence of a patient and the size of the patient also affect the level of acoustic noise in an MR
system. For example, an increase in acoustic noise has been reported with a patient or volunteer
present in the bore of the scanner, which may be due to pressure doubling (i.e., an increase in sound
pressure) close to an object, as sound waves reflect and undergo an in-phase enhancement.83 Noise
characteristics also have a spatial dependence. For example, noise levels have been found to vary by
83
as much as 10 dB as a function of patient position along the magnet bore.
MR-related acoustic noise levels have been measured during a variety of pulse sequences for MR
systems with static magnetic field strengths ranging from 0.2 to 4.7 tesla.71-73,78-84 Recent studies
performed using MR parameters including "worst-case" pulse sequences showed that, not surprisingly,
fast gradient-echo, fast spin-echo, and echo-planar pulse sequences produced the greatest acoustic
noise levels.72,73,79-81
Magnetic Resonance-Related Acoustic Noise and Permissible Limits

The FDA indicates that MR-related acoustic noise levels must be below the level of concern
2
established by pertinent federal regulatory or other recognized standards-setting organizations. If the
acoustic noise is not below this level, the sponsor (i.e., the manufacturer of the MR system) must
recommend steps to reduce or alleviate the noise perceived by the patient. A single upper limit of 140
dB is applied to peak acoustic noise.2 The instructions for use must advise the MR system operator to
2
provide hearing protection to patients for operation above an acoustic noise level of 99 dB.
In general, acoustic noise levels recorded by various researchers in association with conventional or
routine MR procedures have been found to be below the maximum limit permissible by the
Occupational Safety and Health Administration of the United States. This is particularly the case when
one considers that the duration of exposure is one of the more important physical factors that
85,86
determine the effect of noise on hearing.
1 of 2 15-04-2010 21:53
Prevention of Acoustic Noise Problems

Various techniques have been described to attenuate noise and, thus, prevent problems or hazards
associated with exposure to MR-related acoustic noise.65,66,84 The simplest and least expensive
65,66
means is to use disposable earplugs or commercially available noise abatement headphones.
Earplugs, when properly used, can decrease noise by 10 to 30 dB, which usually affords adequate
protection for MR environments that have relatively loud scanners. Regardless of the technique utilized,
facilities operating with MR systems that generate substantial acoustic noise should require all patients
undergoing examinations to wear protective hearing devices. Exposure of staff members, healthcare
workers, and other individuals (e.g., relatives, visitors, etc.) to "loud" MR systems is also of
concern.65,66,73 As such, these individuals should likewise be required to use an appropriate means of
65,66
hearing protection if they remain in the room during the operation of these scanners.
© 2010 Elsevier
2 of 2 15-04-2010 21:53
BIOEFFECTS OF RADIOFREQUENCY FIELDS

page 649
page 650
The majority of the radiofrequency (RF) power transmitted for MR imaging or spectroscopy (especially
for carbon decoupling) is transformed into heat within the patient's tissue as a result of resistive
losses.11,87-89 Not surprisingly, the primary bioeffects associated with exposure to RF radiation are
related to the thermogenic qualities of this energy.11,87-103 Prior to 1985, there were no published
reports concerning thermal or other physiologic responses of human subjects exposed to RF radiation
during MR procedures. Since then, many investigations have been conducted to characterize the
thermal effects of MR-related heating.88-100,104 This topic was recently reviewed by Schaefer87,102 and
Shellock.11
Magnetic Resonance Procedures and the Specific Absorption Rate of

Radiofrequency Radiation
Thermoregulatory and other physiologic changes that a human subject exhibits in response to exposure
to RF radiation are dependent on the amount of energy that is absorbed. The dosimetric term used to
describe the absorption of RF radiation is the specific absorption rate or SAR. 11,87,102,105 The SAR is
the mass normalized rate at which RF power is coupled to biological tissue and is typically indicated in
units of watts per kilogram (W/kg). The relative amount of RF radiation that an individual encounters
during an MR procedure is usually characterized with respect to the whole-body averaged and peak
SAR levels (i.e., the SAR averaged in one gram of tissue).
Measurements or estimates of SAR are not trivial, particularly in human subjects. There are several
methods of determining this parameter for the purpose of RF energy dosimetry in association with MR
87,102,105,106
procedures. The SAR produced during an MR procedure is a complex function of numerous
variables including the frequency (i.e., determined by the strength of the static magnetic field of the MR
system), the type of RF pulse used (e.g., 90° versus 180° pulse), the repetition time, the type of RF
coil used, the volume of tissue contained within the coil, the configuration of the anatomic region
exposed, the orientation of the body to the field vectors, as well as other factors. 11,87,102,103
Thermophysiologic Responses to Magnetic Resonance-Related Heating

Thermophysiologic responses to MR-related heating depend on multiple physiologic, physical, and
environmental factors.11,87,102,103 These include the duration of exposure, the rate at which energy is
deposited, the status of the patient's thermoregulatory system, the presence of an underlying health
condition, and the ambient conditions within the MR system.
With regard to the thermoregulatory system, when the human body is subjected to a thermal
challenge, it loses heat by means of convection, conduction, radiation, and evaporation. Each
mechanism is responsible to a varying degree for heat dissipation, as the body attempts to maintain
thermal homeostasis.11,87,103,105 If the thermoregulatory effectors are not capable of dissipating the
heat load, then there is an accumulation or storage of heat along with an elevation in local and/or
overall tissue temperature.2,3,25
Various underlying health conditions may affect an individual's ability to tolerate a thermal challenge,
including cardiovascular disease, hypertension, diabetes, fever, old age, and obesity. 107-112 In addition,
medications including diuretics, β-blockers, calcium blockers, amphetamines, muscle relaxants, and
sedatives can greatly alter thermoregulatory responses to a heat load. In fact, certain medications
have a synergistic effect with respect to tissue heating if the heating is specifically caused by exposure
113
to RF radiation.
The environmental conditions that exist in and around the MR system will also affect the tissue
temperature changes associated with RF energy-induced heating. During an MR procedure, the
1 of 2 15-04-2010 21:53
amount of tissue heating that occurs that is tolerable by the patient is dependent upon environmental
factors that include the ambient temperature, relative humidity, and airflow.
MR-Related Heating and Human Subjects

The first study of human thermal responses to RF radiation-induced heating during an MR procedure
was conducted by Schaefer et al.114 Temperature changes and other physiologic parameters were
assessed in volunteer subjects exposed to a relatively high, whole-body averaged SAR (approximately
4.0 W/kg). The findings from this investigation indicated that there were no excessive temperature
14
elevations or other deleterious physiologic consequences related to exposure to RF radiation.
Several subsequent studies were conducted involving volunteer subjects and patients undergoing
clinical MR procedures with the intent of obtaining information that would be applicable to patient
populations typically encountered in the MR setting.88,90-101 These investigations demonstrated that
changes in body temperatures were relatively minor (i.e., less than 0.6°C). While there was a tendency
for statistically significant increases in skin temperatures to occur, they were of no serious physiologic
consequences.
Interestingly, studies have reported that there is a poor correlation between body and skin temperature
changes versus whole-body averaged SAR levels.90,92,97 These findings are not surprising considering
the range of thermophysiologic responses possible to a given SAR that are dependent on the
individual's thermoregulatory system and the presence of one or more underlying condition(s) that can
alter or impair the ability to dissipate heat. Furthermore, there appear to be differences in how a given
MR system's software estimates SAR values.
An extensive investigation was conducted in volunteer subjects exposed to an MR procedure using a

whole-body averaged SAR of 6.0 W/kg,101 which is the highest level of RF energy that human subjects
have ever been exposed to using an MR system. Tympanic membrane temperature, six different skin
101
temperatures, heart rate, blood pressure, oxygen saturation, and skin blood flow were monitored.
The findings indicated that an MR procedure performed at a whole-body averaged SAR of 6.0 W/kg
can be physiologically tolerated by an individual with normal thermoregulatory function.101
page 650
page 651
MR-Related Heating and Very High Field MR Systems

There are over 300 MR systems operating with static magnetic field strengths of 3 T, several
operating at 4 T, one operating at 8 T,100 and one operating at 9.4 T. For a given application, these
very high field MR systems are capable of generating RF power depositions that greatly exceed those
associated with a 1.5 T scanner. Therefore, investigations are needed to characterize thermal
responses in human subjects to determine potential thermogenic hazards associated with the use of
these powerful MR devices. Unfortunately, to date, with the exception of the study conducted at 8 T by
Kangarlu et al,100 there has been little work on this topic.
© 2010 Elsevier
2 of 2 15-04-2010 21:53
MAGNETIC RESONANCE SAFETY AND PATIENT MANAGEMENT

Screening Patients for Magnetic Resonance Procedures and Individuals for the
Magnetic Resonance Environment
The establishment of thorough and effective screening procedures for patients and other individuals is one of the
most critical components of a program that guards the safety of all those preparing to undergo MR procedures or to
5,15,116,117
enter the MR environment. An important aspect of protecting patients and individuals from MR system-
related accidents and injuries involves an understanding of the risks associated with the various implants, devices,
5,6,15,116,117
accessories, and other objects that may cause problems in this setting. This requires obtaining
information and documentation about these objects in order to provide the safest MR setting possible. In addition,
because MR-related incidents have been due to deficiencies in screening methods and/or a lack of properly
controlling access to the MR environment (especially with regard to bringing personal items and other potentially
3,4
problematic objects into the MR system room), it is crucial to set up procedures and guidelines to prevent such
incidents from occurring.
MR Procedure Screening for Patients

Certain aspects of screening patients for MR procedures may take place during the scheduling process. This must
be conducted by a healthcare worker trained in MR safety. That is, the individual should be trained to understand the
potential hazards and issues associated with the MR environment and MR procedures and familiar with the
information contained on screening forms for patients and individuals. During the scheduling process, it may be
ascertained if the patient has an implant that may be contraindicated for the MR procedure (e.g., a ferromagnetic
aneurysm clip, pacemaker, etc.) or if there is any condition that needs careful consideration (e.g., the patient is
pregnant, has a disability, etc.). Preliminary screening helps to prevent scheduling patients who may be
inappropriate candidates for MR examinations.
At the facility, every patient must undergo comprehensive screening in preparation for the MR examination.
Comprehensive patient screening involves the use of a printed form to document the screening procedure, a review
of the information on the screening form, and a verbal interview to verify the information and allow discussion of any
116,117
questions or concerns that the patient may have. An MR safety-trained healthcare worker must conduct this
important aspect of patient screening.
115-117
A screening form for patients developed by Sawyer-Glover and Shellock was recently revised by the IMRSER
in consideration of new information in the peer-reviewed literature. This two-page form, entitled Magnetic
Resonance (MR) Procedure Screening Form for Patients, is shown in Figure 24-1. A downloadable version of this
form is available at http://www.MRIsafety.com.
Page 1 of the screening form requests general patient-related information (name, age, sex, height, weight, etc.) as
well as information regarding the reason for the MR procedure and/or symptoms that may be present. Pertinent
information about the patient is required not only to ensure that the medical records are up to date but also in case
the MR facility needs to contact the referring physician for additional information regarding the examination or to
verify the patient's medical condition. The form requests information regarding prior surgery or operation to help
determine if there may be an implant or device present that could create a problem for the patient. Information is
also requested pertaining to prior diagnostic imaging studies that may be helpful to review for assessment of the
patient's condition. Next, important questions are posed in an effort to determine if there are issues that should be
discussed with the patient prior to permitting entry to the MR environment. For example, information is requested
regarding any problem with a previous MR examination, an injury to the eye involving a metallic object or any injury
from a metallic foreign body. Questions are posed to obtain information about current or recently taken medications.
There are also questions provided to assess past and present medical conditions that may affect the MR procedure
or use of an MRI contrast agent.
page 651
page 652
1 of 11 15-04-2010 21:54
Add to lightbox
page 652
page 653
2 of 11 15-04-2010 21:54
Add to lightbox
page 653
page 654
Figure 24-1 Magnetic Resonance (MR) Procedure Screening Form For Patients. (Reprinted with permission from
the Institute for Magnetic Resonance Safety, Education, and Research, 2004.)
3 of 11 15-04-2010 21:54
At the bottom of page 1, there is a section for female patients that poses questions that may impact MR
procedures. For example, questions regarding the date of the last menstrual period, pregnancy or late menstrual
period are included. A definite or possible pregnancy must be identified prior to permitting the patient into the MR
environment so that the risks versus the benefits of the MR procedure can be considered and discussed with the
patient. Questions pertaining to the date of the last menstrual period, use of oral contraceptives or hormonal
therapy, and fertility medication are necessary for female patients undergoing MR procedures that are performed to
evaluate breast disease or for obstetrics and gynecology applications, as these may alter the tissue appearance on
MR imaging. An inquiry about breastfeeding is included if administration of MRI contrast media is being considered
for nursing mothers.
The second page of the form has an important statement at the top that reads:
WARNING: Certain implants, devices or objects may be hazardous to you and/or may interfere with
the MR procedure (i.e. MRI, MR angiography, functional MRI, MR spectroscopy). Do not enter the
MR system room or MR environment if you have a question or concern regarding an implant, device
or object. Consult the MRI Technologist or Radiologist BEFORE entering the MR system room. The
MR system magnet is ALWAYS on.
Next, there is a section that lists various implants and devices to identify objects that could be hazardous to the
patient in relation to the MR procedure or that may produce a substantial artifact that could interfere with the
interpretation of the examination. Figures of the human body are included as a means of showing the location of any
object inside or on the body. This information is particularly useful so that the patient may indicate the approximate
position of any object that may be hazardous or that could interfere with the interpretation of the MR procedure as a
result of producing an artifact.
Page 2 of the screening form also has an Important Instructions section that states:
Before entering the MR environment or MR system room, you must remove all metallic objects
including hearing aids, dentures, partial plates, keys, beeper, cell phone, eyeglasses, hair pins,
barrettes, jewelry, body piercing jewelry, watch, safety pins, paperclips, money clip, credit cards,
bank cards, magnetic strip cards, coins, pens, pocket knife, nail clipper, tools, clothing with metal
fasteners, & clothing with metallic threads. Please consult the MRI Technologist or Radiologist if you
have a question or concern BEFORE you enter the MR system room.
Finally, there is a statement on the Magnetic Resonance (MR) Procedure Screening Form for Patients that indicates
hearing protection is "advised or required" to prevent possible problems or hazards related to acoustic noise.
With the use of any type of written questionnaire, limitations exist related to incomplete or incorrect answers
116,117
provided by the patient. For example, there may be difficulties associated with patients who are impaired with
respect to their vision, language fluency or level of literacy. Therefore, an appropriate accompanying family member
or other individual (e.g., referring physician) should be involved in the screening process to verify information that
may impact patient safety. Versions of this form should also be available in other languages, as needed (i.e.,
specific to the demographics of the MR facility). If the patient is comatose or unable to communicate, the form
should be completed by the most qualified individual (e.g., physician, family member, etc.) who has knowledge about
the patient's medical history and present condition. If the screening information is inadequate, it is advisable to look
for surgical scars on the patient and obtain plain films of the skull and/or chest to search for implants that may be
particularly hazardous in the MR environment (e.g., aneurysm clips, cardiac pacemakers, etc.). Following completion
of the Magnetic Resonance (MR) Procedure Screening Form for Patients, an MR safety-trained healthcare worker
must review the form's content. Next, a verbal interview should be conducted by the MR safety-trained healthcare
worker to verify the information and to allow discussion of any concern that the patient may have. This allows a
mechanism for clarification or confirmation of the answers to the questions posed to the patient so that there is no
miscommunication regarding important MR safety issues.
It should be noted that undergoing a previous MR procedure without incident does not guarantee a safe subsequent
MR examination. Various factors (e.g., the static magnetic field strength of the MR system, the orientation of the
patient, the orientation of a metallic implant or object, etc.) can substantially change the scenario. Therefore, a
comprehensive screening procedure must be conducted each time a patient prepares to undergo an MR procedure.
This is not an inconsequential matter because a surgical intervention or accident involving a metallic foreign body
may have occurred that could impact the safety of entering the MR environment.
MR Environment Screening for Individuals

Similar to the procedure conducted for screening patients, all other individuals (e.g., MRI technologists, patient's
family members, visitors, allied health professionals, maintenance workers, custodial workers, fire fighters, security
officers, etc.) must undergo screening using appropriate guidelines before being allowed into the MR environment.
4 of 11 15-04-2010 21:54
This involves the use of a printed form to document the screening procedure, a review of the information on the
form, and a verbal interview to verify the information and allow discussion of any question or concern that the
individual may have before permitting entry to the MR environment.
In general, MR screening forms were developed with patients in mind and therefore pose many questions that are
inappropriate or confusing to other individuals who may need to enter the MR environment. Therefore, the IMRSER
created a screening form specifically for such individuals. This form, entitled Magnetic Resonance (MR) Environment
Screening Form for Individuals, is shown in Figure 24-2. A downloadable version is available at
http://www.MRIsafety.com.
page 654
page 655
5 of 11 15-04-2010 21:54
Add to lightbox
page 655
page 656
Figure 24-2 Magnetic Resonance (MR) Environment Screening Form for Individuals. (Reprinted with permission
from the Institute for Magnetic Resonance Safety, Education, and Research, 2004.)
6 of 11 15-04-2010 21:54
At the top of this form, the following statement is displayed:
The MR system has a very strong magnetic field that may be hazardous to individuals entering the
MR environment or MR system room if they have certain metallic, electronic, magnetic or
mechanical implants, devices or objects. Therefore, all individuals are required to fill out this form
BEFORE entering the MR environment or MR system room. Be advised, the MR system magnet is
ALWAYS on.
The form requests general information (name, age, address, etc.) and poses important questions to determine if
there are issues that should be discussed with the individual prior to permitting entry to the MR environment. A
warning statement is also provided on the form, as follows:
WARNING: Certain implants, devices or objects may be hazardous to you in the MR environment or
MR system room. Do not enter the MR environment or MR system room if you have any question or
concern regarding an implant, device or object.
In addition, there is a section that lists various implants, devices, and objects to identify the presence of anything
that could be hazardous to an individual in the MR environment (e.g., an aneurysm clip, cardiac pacemaker,
implantable cardioverter defibrillator (ICD), electronic or magnetically activated device, metallic foreign body, etc).
Finally, there is an "Important Instructions" section that indicates that the individual must remove all metallic objects
and other devices that may be problematic in the MR environment.
Metallic Orbital Foreign Bodies and Screening

118
The single case report in 1986 by Kelly et al regarding a patient who sustained an ocular injury from a retained
metallic foreign body has led to controversy over the procedure required to screen patients and individuals prior to
119-121
allowing entry to the MR environment. Importantly, this incident is the only serious eye-related injury that has
occurred in association with the MR setting (i.e., based on a recent review of the peer-reviewed literature and
review of data files from the US Food and Drug Administration, Center for Devices and Radiological Health,
Manufacturer and User Facility Device Experience Database, MAUDE, http://www.fda.gov/cdrh/maude.html, and US
Food and Drug Administration, Center for Devices and Radiological Health, Medical Device Report,
http://www.fda.gov/CDRH/mdrfile.html).
In the past, any individual or patient with a suspected orbital foreign body typically underwent screening using plain
film radiographs of the orbits. Thus, screening plain films of the orbits were performed routinely on individuals not
only with a history of injury from a foreign body but also those who simply had a history of exposure to metallic
objects, such as welders, grinders, metal workers, sculptors, and others. Obviously, plain films of the orbits were
unnecessarily obtained in many individuals and patients based on this policy.
121
Recently, Seidenwurm et al presented research and new guidelines for radiographic screening of patients and
individuals with suspected metallic foreign bodies. This investigation addressed the cost-effectiveness of using a
121
clinical versus radiographic technique to screen individuals for orbital foreign bodies before MR procedures. The
costs of screening were determined on the basis of published reports, disability rating guides, and a practice survey.
A sensitivity analysis was performed for each variable. For this analysis, the benefits of screening were avoidance of
immediate, permanent, nonameliorable or unilateral blindness. Seidenwurm et al121 implemented the following policy:
"If a patient reports injury from an ocular foreign body that was subsequently removed by a doctor or that resulted in
negative findings on any examination, we perform MR imaging…Those persons with a history of injury and no
subsequent negative eye examination are screened radiographically." The findings of this study indicated that using
clinical screening before radiography increased the cost-effectiveness of foreign body screening by an order of
magnitude (i.e., assuming base case ocular foreign body removal rates). It is worth noting that Seidenwurm et al121
have performed approximately 100,000 MRI procedures under this protocol without incident.
Thus, an occupational history of exposure to metallic fragments, by itself, is not sufficient to mandate radiographic
120,121
orbital screening. Therefore, current practice guidelines for foreign body screening should be altered in
consideration of this new information and because radiographic screening before MR procedures on the basis of
120,121
occupational exposure alone is not cost effective, nor is it deemed clinically necessary.
Updated Guidelines for Orbital Foreign Body Screening

The procedure to follow when managing a patient with a suspected orbital foreign body involves a clinical screening
protocol that entails asking the patient if he or she has had an ocular injury.121 If an ocular injury from a metallic
object was sustained, the patient is asked if a medical examination was conducted at the time of the injury and
121
whether they were informed by the doctor that all of the object had been removed. If there was no injury, the
7 of 11 15-04-2010 21:54
ophthalmologic examination was normal, and/or the foreign body was completely removed at the time of the injury,
then the patient may proceed to MR imaging. Based on the results of the clinical screening protocol, the patient
should be screened using plain film radiographs if an ocular injury related to a metallic object was sustained and the
121
patient was not informed that the postinjury eye examination was normal. In this case, the MR examination is
postponed and the patient is scheduled for screening radiography.
Excessive Heating and Burns Associated with Magnetic Resonance Procedures

The use of radiofrequency coils, physiologic monitors, electronically activated devices, and external accessories or
objects made from conductive materials has caused excessive heating, resulting in burn injuries to patients
3-6,122-134
undergoing MR procedures. Heating of implants and similar devices may also occur in association with MR
procedures, but this tends to be problematic primarily for objects made from conductive materials that have an
elongated shape such as electrodes, leads, guidewires, and certain types of catheters (e.g., catheters with
135-143
thermistors or other conducting components).
page 656
page 657
Notably, more than 30 incidents of excessive heating have been reported in patients undergoing MR procedures in
the United States that were unrelated to equipment problems or the presence of conductive external or internal
3,4,144
implants or materials. These incidents included first-, second-, and third-degree burns experienced by patients.
In many of these cases, the reports indicated that the limbs or other body parts of the patients were in direct
contact with body RF coils or other RF transmit coils of the MR systems or there were skin-to-skin contact points
3,4,144
suspected to be responsible for these injuries.
In consideration of the above, the IMRSER recently developed guidelines to prevent excessive heating and burns
related to MR procedures (Box 24-1). The adoption of these guidelines will help to ensure that patient safety is
maintained, especially as more conductive materials and electronically activated devices are used in association with
MR procedures.
Box 24-1 Guidelines to Prevent Excessive Heating and Burns in

Association with MR Procedures (reprinted with permission from the
Institute for Magnetic Resonance Safety, Education, and Research)
1. Prepare the patient for the MR procedure by ensuring that there are no unnecessary
metallic objects contacting the patient's skin (e.g. metallic drug delivery patches,
jewelry, necklaces, bracelets, key chains, etc.).
2. Prepare the patient for the MR procedure by using insulation material (i.e.
appropriate padding) to prevent skin-to-skin contact points and the formation of
"closed loops" from touching body parts.
3. Insulating material (minimum recommended thickness, 1 cm) should be placed
between the patient's skin and transmit RF coil (alternatively, the RF coil itself should
be padded). For example, position the patient so that there is no direct contact
between the patient's skin and the body RF coil of the MR system. This may be
accomplished by having the patient place his arms over his head or by using elbow
pads or foam padding between the patient's tissue and the body RF coil of the MR
system. This is especially important for those MR examinations that use the body coil
or other large RF coils for transmission of RF energy.
4. Use only electrically conductive devices, equipment, accessories (e.g. ECG leads,
electrodes, etc.) and materials that have been thoroughly tested and determined to
be safe and compatible for MR procedures.
5. Carefully follow specific MR safety criteria and recommendations for implants made
from electrically conductive materials (e.g. bone fusion stimulators, neurostimulation
systems, etc.).
6. Before using electrical equipment, check the integrity of the insulation and/or housing
of all components including surface RF coils, monitoring leads, cables, and wires.
Preventive maintenance should be practiced routinely for such equipment.
7. Remove all non-essential electrically conductive materials from the MR system (i.e.
unused surface RF coils, ECG leads, cables, wires, etc.).
8. Keep electrically conductive materials that must remain in the MR system from
directly contacting the patient by placing thermal and/or electrical insulation between
the conductive material and the patient.
9. Keep electrically conductive materials that must remain within the body RF coil or
other transmit RF coil of the MR system from forming conductive loops. Note: The
8 of 11 15-04-2010 21:54
patient's tissue is conductive and, therefore, may be involved in the formation of a

conductive loop, which can be circular, U-shaped or S-shaped.
10. Position electrically conductive materials to prevent "crosspoints", for example, where
a cable crosses another cable, where a cable loops across itself or where a cable
touches either the patient or sides of the transmit RF coil more than once. Note that
even the close proximity of conductive materials with each other should be avoided
because some cables and RF coils can capacitively couple (without any contact or
crossover) when placed close together.
11. Position electrically conductive materials to exit down the center of the MR system
(i.e. not along the side of the MR system or close to the body RF coil or other
transmit RF coil).
12. Do not position electrically conductive materials across an external metallic
prosthesis (e.g. external fixation device, cervical fixation device, etc.) or similar
device that is in direct contact with the patient.
13. Allow only properly trained individuals to operate devices (e.g. monitoring equipment)
in the MR environment.
14. Follow all manufacturer instructions for the proper operation and maintenance of
physiologic monitoring or other similar electronic equipment intended for use during
MR procedures.
15. Electrical devices that do not appear to be operating properly during the MR
procedure should be removed from the patient immediately.
16. Closely monitor the patient during the MR procedure. If the patient reports
sensations of heating or other unusual sensation, discontinue the MR procedure
immediately and perform a thorough assessment of the situation.
17. RF surface coil decoupling failures can cause localized RF power deposition levels to
become excessive. The MR system operator will recognize such a failure as a set of
concentric semicircles in the tissue on the associated MR image or as an unusual
amount of image non-uniformity related to the position of the RF coil.
Tattoos and Permanent Cosmetics

Traditional (i.e., decorative) and cosmetic tattoo procedures have been performed for thousands of years. Cosmetic
tattooing or "permanent cosmetics" are used to reshape, recolor, recreate or modify eye shadow, eyeliner,
eyebrows, lips, beauty marks, and cheek blush. Additionally, permanent cosmetics are used to hide scars and for
145,146
other esthetic applications.
page 657
page 658
147-156
There is considerable controversy regarding the MR safety aspects of tattoos and permanent cosmetics.
Problems are associated with the use of iron oxide or other metal-based pigments. Because a small number of
patients with permanent cosmetics who underwent MR procedures (fewer than 10 documented cases) experienced
transient skin irritation, cutaneous swelling or heating sensations, 3,4 many radiologists have refused to perform MR
procedures on individuals with permanent cosmetics (unpublished observations, Shellock, 2004). Obviously, this
undue concern for possible adverse events prevents patients with permanent cosmetics from accessing an
150
extremely important diagnostic imaging modality.
150
A recent study conducted by Tope and Shellock determined the frequency and severity of adverse events
associated with MR imaging in a population of subjects with permanent cosmetics. A questionnaire was distributed
to clients of cosmetic tattoo technicians. One hundred and thirty-five (13.1%) study subjects underwent MR imaging
after having permanent cosmetics applied. Of these, only two individuals (1.5%) experienced problems associated
with MR imaging: one subject reported a sensation of "slight tingling" and the other subject reported a sensation of
150 3,4
"burning", both transient in nature. Based on these findings as well as other available information, it is apparent
that MR procedures may be performed in patients with permanent cosmetics without any serious soft-tissue
reactions or adverse events. Therefore, the presence of permanent cosmetics should not prevent patients from
undergoing MR procedures.
Interestingly, decorative tattoos tend to cause worse problems (including first- and second-degree burns) for
patients undergoing MR procedures compared to those that have been reported for cosmetic tattoos. For example,
Kreidstein et al154 reported that a patient experienced a sudden burning pain at the site of a decorative tattoo during
MR imaging of the lumbar spine at 1.5 T. Surprisingly, in order to permit completion of the MR examination, an
154
excision of the tattooed skin was performed. The authors of this report stated: "Theoretically, the application of a
154
pressure dressing of the tattoo may prevent any tissue distortion due to ferromagnetic pull". However, this simple,
9 of 11 15-04-2010 21:54
relatively benign procedure was not attempted for the patient. Kreidstein et al154 also indicated that: "In some cases,
removal of the tattoo may be the most practical means of allowing MRI". Kanal and Shellock155 commented on this
report in a letter to the editor, suggesting that the response to this situation was "rather aggressive". Clearly the
trauma, expense, and morbidity associated with excising a tattoo far exceed those that may be associated with
MR-related tattoo interactions. Because of the remote possibility of an incident occurring in a patient with a
permanent cosmetic or tattoo and due to the relatively minor, short-term complications or adverse events that may
develop (i.e., transient cutaneous redness and swelling), the patient should be permitted to undergo a procedure
without reservation. Any problem in performing an MR procedure in a patient with a permanent cosmetic or tattoo
should not prevent the examination, since the important diagnostic information that is provided by this modality is
typically critical to the care and management of the patient.
Additional information on this topic has been provided for patients by the US Food and Drug Administration, Center
151
for Food Safety and Applied Nutrition, Office of Cosmetics and Colors Fact Sheet, as follows :
… the risks of avoiding an MRI when your doctor has recommended one are likely to be much
greater than the risks of complications from an interaction between the MRI and tattoo or permanent
makeup. Instead of avoiding an MRI, individuals who have tattoos or permanent makeup should
inform the radiologist or technician of this fact in order to take appropriate precautions, avoid
complications, and assure the best results.
Pregnant Patients and Magnetic Resonance Procedures

MR procedures have been used to evaluate obstetrical, placental, and fetal abnormalities in pregnant patients for
157-166
more than 19 years. Initially, there were substantial technical problems with the use of MR imaging primarily
due to image degradation caused by fetal motion. However, several technological improvements, including the
development of high-performance gradient systems and rapid pulse sequences, provided advances that were
especially useful for imaging pregnant patients. Thus, high-quality MR examinations for obstetrical and fetal
166
applications may now be accomplished routinely in the clinical setting.
Diagnostic imaging is often required during pregnancy.157 Safety issues exist related to possible adverse bioeffects
associated with exposure to the static, gradient, and RF electromagnetic fields used for MR procedures.5,13,157
Accordingly, laboratory and clinical research investigations have been conducted to determine the effects of using
167-177
MR procedures during pregnancy. The overall findings from these studies indicate that there is no substantial
evidence of injury or harm to the fetus; however, additional research on this topic is warranted.
Guidelines for the Use of MR Procedures in Pregnant Patients

In 1991, the Safety Committee of the Society for Magnetic Resonance Imaging issued the document entitled
"Policies, Guidelines, and Recommendations for MR Imaging Safety and Patient Management" which stated13:
MR imaging may be used in pregnant women if other non-ionizing forms of diagnostic imaging are
inadequate or if the examination provides important information that would otherwise require
exposure to ionizing radiation (e.g. fluoroscopy, CT, etc.). Pregnant patients should be informed that,
to date, there has been no indication that the use of clinical MR imaging during pregnancy has
produced deleterious effects.
These guidelines have been subsequently adopted by the American College of Radiology and considered to be the
"standard of care" with respect to the use of MR procedures in pregnant patients.
page 658
page 659
In cases where the referring physician and attending radiologist can uphold that the findings of the MR procedure
have the potential to impact the care or management of the mother or fetus (e.g., to address important clinical
problems, to identify potential complications, anomalies or complex fetal disorders, etc.), the MR procedure may be
performed with verbal and written informed consent, regardless of the trimester. 13,157
Pregnant Technologists and Healthcare Workers

Due to the concern with regard to pregnant technologists and healthcare workers in the MRI environment, a survey
of reproductive health among female MR system operators was conducted in 1990 by Kanal et al.271
Questionnaires were sent to all female MR technologists and nurses at the majority of clinical MR facilities in the
United States. The questionnaire addressed menstrual and reproductive experiences as well as work activities. This
10 of 11 15-04-2010 21:54
study attempted to account for known potential confounders (e.g., age, smoking, alcohol use) for this type of data.
Of the 1915 completed questionnaires analyzed, there were 1421 pregnancies: 280 occurred while working as an
MR employee (technologist or nurse), 894 while employed at another job, 54 as a student, and 193 as a
homemaker.
Five categories were analyzed: spontaneous abortion rate, preterm delivery (less than 39 weeks), low birth weight
(less than 5.5 pounds), infertility (taking more than 11 months to conceive), and gender of the offspring. The data
indicated that there were no statistically significant alterations in the five areas studied for MR healthcare workers
relative to the same group studied when they were employed elsewhere, prior to becoming MR healthcare
employees. Additionally, adjustment for maternal age, smoking, and alcohol use failed to markedly change any of
the associations.
Menstrual regularity, cyclicity, and related topics were also examined in this study. These included inquiries
regarding the number of days of menstrual bleeding, the heaviness of the bleeding, and the time between menstrual
cycles.
Admittedly, this is a very difficult area to examine objectively, because it depends upon both subjective memory and
the memory of the respondent for a topic, where subjective memory is notoriously inadequate. Nevertheless, the
data suggested that there was no clear correlation between MR workers and any specific modifications of the
menstrual cycle.
The data from this extensive epidemiologic investigation were reassuring insofar as there did not appear to be any
deleterious effects from exposure to the static magnetic field component of the MR system. Therefore, a policy is
recommended that permits pregnant technologists and healthcare workers to perform MR procedures, as well as to
enter the MR system room, and to attend to the patient during pregnancy, regardless of the trimester.
Importantly, the technologists or healthcare workers should not remain within the MR system room or magnet bore
during the actual operation of the device. This recommendation is especially important for those MR users involved
in interventional MR-guided examinations and procedures to adhere to since it may be necessary for them to be
directly exposed to the MR system's electromagnetic fields at levels similar to those used for patients.
Notably, these recommendations are not based on indications of adverse effects but rather, from a conservative
point of view and the feeling that there are insufficient data pertaining to the effects of the other electromagnetic
fields of the MR system to support or allow unnecessary exposures.
© 2010 Elsevier
11 of 11 15-04-2010 21:54
MAGNETIC RESONANCE PROCEDURES AND IMPLANTS AND DEVICES

The MR environment may be unsafe for patients or individuals with certain biomedical implants or devices primarily
3-7,137,178-195
due to movement or dislodgment of objects made from ferromagnetic materials. While excessive
heating and the induction of electrical currents may also present risks to patients with implants or devices, these
MR safety problems are typically associated with implants that are made from conducting materials and have
elongated configurations and/or that are electronically activated (e.g., neurostimulation systems, cardiac
122,123,134-137
pacemakers, etc.).
To date, more than 1300 objects have been tested for MR safety, with over 300 evaluated at 3.0 T or
5-7,137,178-195
higher. This information is available to MR healthcare professionals and others as published reports,
compiled lists, and in its entirety in an online format at http://www.MRIsafety.com. The topic of MR safety for
5,137
implants and devices was recently reviewed by Shellock. As such, the material presented in this chapter
provides information for implants and devices for which there may be controversy or confusion, with an update on
objects tested at 3.0 T or higher.
Magnetic Resonance Safety and Compatibility

The terms "MR safe" and "MR compatible" are typically used to designate specific aspects of metallic implants and
devices.5-7 Therefore, it is important to appreciate the differences between these terms, as they should not be
used interchangeably. For those in the MR community unfamiliar with these terms, they are defined as follows.
MR safe
The device, when used in the MR environment, has been demonstrated to present no additional risk to the patient
or other individual, but may affect the quality of the diagnostic information. The MR conditions in which the device
was tested should be specified in conjunction with the term "MR safe" since a device that is safe under one set of
conditions may not be so under more extreme MR conditions.
page 659
page 660
MR compatible
A device is considered "MR compatible" if it is MR safe and, when used in the MR environment, has been
demonstrated to neither significantly affect the quality of the diagnostic information nor have its operations affected
by the MR device. The MR conditions in which the device was tested should be specified in conjunction with the
term "MR compatible" since a device which is compatible under one set of conditions may not be so under more
extreme MR conditions.
In general, MR safety testing of an implant or device involves assessment of magnetic field interactions, heating,
and induced electrical currents while MR compatibility testing requires all of these as well as characterization of
artifacts.9,76 In addition, the functional or operation aspects of the implant or device should be evaluated.
Evaluation of Implants and Devices for Safety in the Magnetic Resonance

Environment
The evaluation of an implant or device with regard to the MR environment is not a trivial matter. The proper
assessment of an object typically entails characterization of magnetic field interactions (translational attraction and
torque), MR procedure-related heating, induced electrical currents, and artifacts. A thorough evaluation of the
impact of the MR environment on the functional or operation aspects of certain implants and devices may also be
necessary.
Notably, an object demonstrated to be safe under one set of MR conditions may be unsafe under more "extreme"
conditions (e.g., higher static magnetic field, greater level of RF power deposition, faster gradient fields, different
RF transmission coil, etc.). Accordingly, the specific test conditions for a given implant or device must be known
before making a decision regarding whether a particular object is safe for a patient or individual in the MR
environment.
Magnetic Field-Related Issues

Magnetic field-related translational attraction and torque are known to present hazards to patients and individuals
with certain implants or devices.5-7 Currently, MR systems used in clinical and research settings operate with static
magnetic fields that range from 0.2 to 9.4 tesla. Most previous ex vivo tests performed to assess objects for MR
5-7,137
safety used scanners with static magnetic fields of 1.5 tesla or lower. Accordingly, this could present
problems as it is possible that an object that displayed "weakly" ferromagnetic qualities in association with a 1.5
1 of 10 15-04-2010 21:55
tesla MR system may exhibit substantial magnetic field interactions with a system operating at a higher static
179-182
magnetic field strength. Therefore, investigations have been conducted and are ongoing using 3.0 and 8 tesla
MR systems to determine MR safety for implants and devices relative to these powerful scanners. 179-182
Long-Bore versus Short-Bore MR Systems

Different magnet configurations exist for commercially available 1.5 and 3.0 tesla MR systems. These include
conventional "long-bore" and "short-bore" scanners used for whole-body (1.5 and 3.0 T MR systems) and
head-only (3.0 T MR systems) clinical applications. Studies have indicated that there are significant differences in
the position and magnitude of the highest spatial magnetic gradient for long-bore vs short-bore MR systems,
especially at 3.0 tesla.180,181 This may impact the MR safety aspects of a given implant or device and is an
additional factor that must be taken into consideration when evaluating metallic objects in the MR environment.
Aneurysm Clips
The presence of an intracranial aneurysm clip in a patient referred for an MR procedure or an individual who needs
to enter the MR environment represents a situation that requires careful consideration because of the associated
5-7,137,196-214
risks (Fig. 24-3). Aneurysm clips made from ferromagnetic materials are contraindicated for MR
procedures because excessive, magnet-induced forces may displace these clips, causing serious injury or death.
By comparison, aneurysm clips classified as "non-ferromagnetic" or "weakly ferromagnetic" (e.g., made from
Phynox, Elgiloy, titanium alloy or commercially pure titanium) have been shown to be safe for patients undergoing
208
MR procedures at 1.5 tesla or less. In 1998, Shellock and Kanal provided guidelines for the management of a
patient with an aneurysm clip based on the relevant peer-reviewed literature (Box 24-2).
Add to lightbox
page 660
page 661
Figure 24-3 Examples of aneurysm clips showing a variety of shapes and sizes. Aneurysm clips may be made from
various materials including ferromagnetic and non-magnetic metals.
Box 24-2 Guidelines for the Management of a Patient with an

Aneurysm Clip Referred for an MR Procedure (adapted from
reference 208)
1. Specific
Various studies information (i.e. manufacturer,
have been performed to supporttype or model,
scanning material,
patients with lot and serial
nonferromagnetic aneurysm clips (Fig.
numbers) about 206
theetaneurysm clip must be known so that patients
only patients with
24-4). For example, Pride al reported findings from several with nonferromagnetic aneurysm clips
imagednon-ferromagnetic
at 1.5 tesla. Thereorwas weakly ferromagnetic
no objective adverseclips are allowed
outcome intopatients,
for these the MR confirming that MR procedures
environment. This information is provided in the labeling of the aneurysm197 clip by the
can be performed safely in patients with nonferromagnetic clips. Brothers et al also demonstrated MR safety at
manufacturer. The implanting surgeon is responsible for properly communicating this
1.5 tesla for patients with nonferromagnetic aneurysm clips. This report was particularly important because MR
information in the patient's records.
imaging was found to be better than CT for postoperative assessment of aneurysm patients, especially with regard
2. An aneurysm clip that is in its194 original package and made from Phynox, Elgiloy,
to showing smalltitanium
MP35N, zones of ischemia.
alloy, commercially pure titanium or other material known to be
non-ferromagnetic or weakly ferromagnetic does not need to be evaluated for 204
Notably,ferromagnetism
only one ferromagnetic aneurysm
and, as such, theseclip-related fatality
are considered to has beenfor
be safe reported in the peer-reviewed literature.
MR procedures
This incident was the
performed result
at 1.5 of erroneous
tesla or less. information pertaining to the type of aneurysm clip that was present in the
patient-the
3. The clip was believed
radiologist to be a nonferromagnetic
and implanting Yasargilfor
surgeon are responsible aneurysm clipthe
evaluating (Aesculap Inc., South San Francisco,
information
204
CA) butpertaining
turned outtotothe
beaneurysm
a ferromagnetic Vari-Angle
clip, verifying clip (Codman
its accuracy, & Shurtleff,
obtaining written Randolf, MA).
documentation and deciding to perform the MR procedure after considering the risk
AneurysmversusClips
benefit Tested at a3.0
aspects for and
given 8.0 T
patient.
2 of 10 15-04-2010 21:55
Various aneurysm clips have been tested for magnetic field interactions in association with 3.0 and 8.0 tesla MR
179,180,182
systems. Findings indicated that the clips either exhibited a lack of magnetic field interactions or relatively
"weak" magnetic field-related translational attraction and torque at 3.0 tesla. Accordingly, some aneurysm clips are
considered to be entirely safe for patients undergoing procedures using MR systems operating at 3.0 tesla, while
179,180
others require further characterization of magnetic field-induced torque.
The first investigation to determine magnetic field interactions for medical implants at 8.0 tesla involved an
182
assessment of aneurysm clips. Aneurysm clips representative of those made from nonferromagnetic or weakly
ferromagnetic materials used for temporary or permanent treatment of aneurysms or arteriovenous malformations
were selected for this study.182 Test results showed that MR safety at 8.0 tesla was dependent on not just the
material but also the dimensions, model, shape, size, and blade length of a given clip.
Add to lightbox
Figure 24-4 MR images of the brain (1.5 T) performed in patients with nonferromagnetic aneurysm clips. Note the
presence of relatively small artifacts (signal voids) associated with these implants.
Heart Valve Prostheses and Annuloplasty Rings

Numerous heart valve prostheses and annuloplasty rings have undergone testing for MR safety.5-7,179,215-223 Of
these, the majority showed measurable yet relatively minor translational attraction and/or torque in association with
exposure to the MR systems used for testing. Since the magnetic field-related forces exerted on heart valves and
annuloplasty rings are deemed minimal compared to the force exerted by the beating heart (i.e., approximately 7.2
215,216
N), an MR procedure is considered to be safe for a patient with the heart valve prostheses or annuloplasty
5-7,179,215-223
rings that have undergone testing to date. This includes the Starr-Edwards Model Pre-6000 heart valve
prosthesis previously suggested to be potentially hazardous for a patient in the MR environment.
Heart Valve Prostheses and Annuloplasty Rings Tested at 3.0 T

page 661
page 662
Many heart valve prostheses and annuloplasty rings have now been evaluated for MR safety using 3.0 tesla
179
scanners. Findings indicated that one annuloplasty ring (Carpentier-Edwards Physio Annuloplasty Ring, Mitral
Model 4450, Edwards Lifesciences, Irvine, CA) showed relatively minor magnetic field interactions. Therefore,
similar to heart valves prostheses and annuloplasty rings tested at 1.5 tesla, because the actual attractive forces
3 of 10 15-04-2010 21:55
exerted on these implants are deemed minimal compared to the force exerted by the beating heart, MR procedures
5,179
at 3.0 tesla are not considered to be hazardous for patients or individuals who have these implants.
Additional heart valves and annuloplasty rings from the Medtronic Heart Valve Division have undergone MR safety
testing at 3.0 tesla (Medtronic Inc., Minneapolis, MN; work conducted by E Kanal). These implants were tested for
magnetic field interactions and artifacts using a shielded, 3.0 tesla MR system. According to information provided
by Medtronic (personnel communication, 2002, Kathryn M Bayer, Senior Technical Consultant, Medtronic Inc.),
these specific implants are safe for patients undergoing MR procedures using scanners operating up to 3.0 tesla.
Coils, Filters, and Stents

There are many different types of coils, filters, and stents that are used for a variety of applications (Fig. 24-5).
These implants are usually made from metals that include platinum, titanium, stainless steel, Phynox, Elgiloy, and
5,192,224-240
Nitinol, which are mostly nonmagnetic or "weakly" ferromagnetic at 1.5 tesla or less. Heating and
induced currents have been evaluated for a wide variety of shapes and sizes of these implants and there do not
appear to be any safety issues for these devices. For those coils, filters, and stents found to have no magnetic field
interactions, an MR procedure may be performed immediately after implantation.5-7,224-228,230,236 However, for
implants made from weakly ferromagnetic materials, it is typically recommended to wait 6 to 8 weeks to allow for
5-7,224-228
tissue ingrowth or other mechanism to help retain the implant in place. If there is any possibility that a coil,
filter or stent is not positioned properly or firmly in place, the patient should not be allowed into the MR
environment.
Unfortunately, some implant manufacturers, in their product documentation, may not differentiate between their
nonferromagnetic devices and those that are weakly ferromagnetic (i.e., indicating a waiting period of 6 to 8 weeks
for all implants regardless of the material used to make the device), which results in confusion for the MR safety
aspects of these implants.
It should be noted that because coils, filters, and stents are developed on an ongoing basis, general MR safety
guidelines cannot be provided for these implants. Therefore, it is necessary to obtain documentation that clearly
identifies the device, material, and manufacturer in order to avoid hazardous situations in the MR environment.
A study by Taal et al239 supports the fact that not all stents are safe for patients undergoing MR procedures. This
investigation reported that "an appreciable attraction force and torque" was found for two different types of
239
Gianturco stents. In consideration of these results, Taal et al advised, "…specific information on the type of stent
is necessary before a magnetic resonance imaging examination is planned".
MR Safety at 3.0 T, Coils and Stents

179,229
Several different coils and stents have been evaluated at 3.0 tesla. For the implants tested, two displayed
229
magnetic field interactions exceeding levels that may present risks to patients. However, similar to other coils
and stents, tissue ingrowth may be sufficient to prevent these implants from posing a substantial risk to a patient or
individual in the 3.0 tesla MR environment. Thus, this MR safety issue warrants further study.
Essure Device
The Essure Device (Conceptus, San Carlos, CA) is a new implant developed for permanent female
240
contraception. It is composed of 316L stainless steel, platinum, iridium, nickel-titanium alloy, silver solder, and
polyethylene tetraphthalate (PET) fibers. The Essure Device is a dynamically expanding microcoil that is placed in
the proximal section of the fallopian tube using a non-incisional technique. It then elicits a benign tissue response,
resulting in tissue ingrowth that anchors it and occludes the fallopian tube, resulting in permanent contraception.
An MR safety assessment of this implant involved testing for magnetic field interactions at 1.5 tesla, heating,
induced electrical currents, and artifacts.240 The findings indicated it is safe for a patient with the Essure Device to
undergo an MR procedure using a system operating at 1.5 tesla or less.
Essure Device and Testing at 3.0 T

The Essure Device was recently evaluated for MR safety at 3.0 tesla and found to be safe for patients undergoing
179
MR procedures operating at this field strength.
Implantable Spinal (Bone) Fusion Stimulator

The implantable spinal fusion stimulator (EBI, LLC, Parsippany, NJ) is used to enhance and facilitate the rate of
bone healing. It consists of a generator (which includes a battery and electronics in a titanium shell) and electrodes
4 of 10 15-04-2010 21:55
implanted near the area of treatment of the spine. Two wire leads are connected from the generator to the fusion
sites where they are embedded in pieces of bone grafts. The device remains in place for approximately 24 to 26
weeks.
This device received approval from the FDA designating it to be "MR safe" based on comprehensive investigations,
as long as specific guidelines are followed as provided by the manufacturer in the product insert labeling. These
guidelines are indicated in Box 24-3.
TheraSeed Radioactive Seed Implant

page 662
page 663
Add to lightbox
5 of 10 15-04-2010 21:55
Add to lightbox
page 663
page 664
Figure 24-5 Examples of an intravascular filter (A) and various stents (B) that have undergone MR safety testing.
Box 24-3 Guidelines Recommended for Conducting an MR

Procedure in a Patient with the Implantable Spinal (Bone) Fusion
Stimulator (EBI, LLC, Parsippany, NJ)
The 1. During implantation,

TheraSeed radioactive the seed implantable spinal fusionCorporation,
implant (Theragenics stimulator should
Buford,beGA)placed as far
is used to deliver low-level
as possible
palladium-103 fromto
radiation thethespinal canaltoand
prostate treatbone graft This
cancer. sincerelatively
this will small
decrease the is composed of a titanium tube
implant
with twolikelihood
graphitethat artifacts
pellets and awillleadaffect the inside.
marker area ofTreatments
interest on mayMR images.
involve placement of from 80 to 120
2. The cathodes
TheraSeeds. MR testing of the
for implantable
magnetic field spinal (bone) fusion
interactions, stimulator
heating, inducedshould be positioned
currents, and artifacts revealed that the
a minimum
TheraSeed implantofis 1safecm from nerve roots
for patients to reduce
undergoing MR the possibilityatof1.5
procedures nerve
teslaexcitation.
or less.
3. Plain film radiographs should be obtained prior to the MR procedure to verify that
Cardiacthere
Pacemakers
are no broken leads present for the implantable spinal fusion stimulator. If this
cannot be reliably determined, then the potential risks and benefits to the patient
Cardiac pacemakers are the most common electronically activated implants found in patients referred for MR
requiring the MR procedure must be carefully assessed due to the possible
procedures. Unfortunately, the presence of a pacemaker is considered to be a strict contraindication for the MR
development
5-7,241-259 of excessive heating in the leads.
environment.
4. MR examinations Potential
must only adverse interactions
be performed between
using systems pacemakers
operating at and1.5MR procedures
tesla or include movement
of the pulse generator or leads, electrode heating, induction of ventricular
less and only with conventional imaging techniques such as spin-echo, turbo or fast fibrillation, rapid pacing, reed switch
malfunction, asynchronous
spin-echo pacing, pulse
or gradient-echo inhibition of pacingPulse
sequences. output, alterationorofconditions
sequences programming that with possible damage to
5-7,241-263
the pacemaker
produce circuitry,
exposures and to other problems.
high levels of RF energy (i.e. Some of these a
exceeding issues are theoretical while others have been
whole-body
studied averaged
in vitro, inspecific
laboratory animals or in human subjects.
absorption rate of 1.0 W/kg) or exposure to gradient fields that
exceed 20 T per second (e.g. echo-planar imaging) or any other unconventional MR
More than 13 deaths
technique shouldhavebebeen attributed to MR procedures performed in patients with cardiac
avoided.
3,4,261-263
pacemakers. These fatalities
5. The patient should be continuously were poorly characterized,
observed during the MRas there wasand
procedure no instructed
electrocardiographic monitoring
during the examinations.
to report any unusualIn each case, the
sensations "mode any
including of death" (i.e.,
feelings the mechanism
of warming, burning responsible
or for the adverse
cardiacneuromuscular
pacemaker/MRexcitation procedure or interaction)
stimulation. was not reported and it was unknown whether these patients were
6 of 10 15-04-2010 21:55
3,4,261-263
pacemaker dependent. Importantly, there have been no deaths associated with physician-supervised
scanning.261-263
263
In a letter addressing the controversy over scanning patients with cardiac pacemakers, Gimbel pointed out that
pacemaker-related deaths occurred in patients "'inadvertently' placed in the MRI environment without the attending
physician conducting the MRI knowing that the patient being scanned had a pacemaker. Thus, none of the easily
implemented techniques that might have allowed a harmless scan to proceed were implemented".
To date, more than 230 patients with cardiac pacemakers have undergone MR procedures safely, either
inadvertently or during purposeful, monitored attempts to perform much needed
249,250,254,257,258,260-264
examinations. Thus, there is growing evidence that MR examinations are not as detrimental
as once thought for certain patients and under certain, highly specific MR conditions. Accordingly, restrictions for
conducting MR procedures in patients with cardiac pacemakers may be modified in the near future.
Investigations of human subjects with cardiac pacemakers have suggested various strategies for safe MR
procedures. These strategies included only scanning nonpacemaker-dependent patients, programming the
pacemaker device to an "off" or asynchronous mode, programming to a bipolar lead configuration, limiting the
radiofrequency energy, and only doing MR examinations if the pulse generator was positioned outside the bore of
the MR system.249,250,254,255,257,258,260,263
264
A recent study by Martin et al performed at 1.5 T suggested that these strategies may not be necessary for
nonpacemaker-dependent patients at 1.5 T. In this investigation, in order to examine risk in the broadest possible
population, no restrictions were placed on the anatomy scanned, the type of pulse sequence and imaging
264
parameters used for MR imaging, nor on the type of pacemaker present in the patient. Pacemaker-dependent
patients were excluded to eliminate problems if pacing was inhibited during scanning. Absolute requirements for
performing MR procedures in these nonpacemaker-dependent patients included having resuscitation equipment in
close proximity to the MR system room, an electrophysiology-trained physician present to monitor to the case, and
264
advanced cardiac life support (ACLS)-certified personnel present to respond to any untoward consequence.
Findings from this study indicated that MR procedures at 1.5 tesla did not cause substantial problems or difficulties.
Furthermore, this investigation emphasized that it was not necessary to inhibit the pacing pulse, reprogram the
pulse generator or change MR scan parameters to achieve safety, as was done in prior studies of patients with
cardiac pacemakers.
page 664
page 665
Given the infinite possibilities of pacing systems, cardiac and lead geometry, as well as variable RF and gradient
magnetic fields, absolute safety with regard to pacemaker and MR interactions cannot be assured under all
operational conditions. However, based on information in the peer-reviewed literature, it appears that, given
appropriate patient selection as well as continuous monitoring and preparedness for resuscitation efforts,
performance of MR procedures in nonpacemaker-dependent patients with implanted cardiac pacemakers may be
achieved with reasonable safety, even at static magnetic field strengths of 1.5 T.
In the past, the presence of all electronically activated implants was considered a strict contraindication for a
patient or individual in the MR environment. However, over the years, various studies have been performed to
138,141-143
define safety criteria for electronic devices. As such, if highly specific guidelines are followed, MR
procedures may be conducted safely in patients with various electronically activated implants including
neurostimulation systems, cochlear implants, and programmable drug infusion pumps. 5-7,138,141-143 In fact, some of
these electronically activated devices have received "MR-safe" approval labeling from the US FDA. Accordingly,
given the findings on conducting safe MR procedures that have been published in the peer-reviewed literature, it is
hoped that cardiac pacemaker manufacturers will be encouraged to proactively support and/or conduct
investigations directed towards identifying safety criteria for their respective devices. This will ultimately have a
substantial impact on patient management and the overall healthcare of patients with pacemakers who may require
MR procedures.
Neurostimulation System for Deep Brain Stimulation

Because of the increased interest in the use of deep brain stimulation (DBS) of the thalamus, globus pallidus, and
subthalamic nucleus for treatment of medically refractory movement disorders and other types of neurologic
conditions, the number of patients receiving implantable pulse generators (IPGs) and DBS electrodes is rapidly
141,142,265-267
growing. The use of MR imaging in patients with neurostimulation systems is frequently desired for
surgical planning, as well as for the ongoing management of underlying conditions.141,142,265-267 Additionally, MR
imaging may be needed for various clinical scenarios including verification of lead position, evaluation of patients
with poor or worsening outcomes, and examining patients with other pathologic abnormalities unrelated to DBS
7 of 10 15-04-2010 21:55
142
neurostimulation such as stroke, tumor or hemorrhage.
As with all electronically activated devices in the MR environment, it is generally recommended that patients with
neurostimulation systems should not undergo MR imaging examinations because of the potential for serious risks,
including movement or dislodgment of the leads or implantable pulse generators, excessive MR imaging-related
5-7
heating, induced electrical currents, and functional disruption of the operational aspects of the device. Thus,
before performing MR procedures in patients with DBS neurostimulation systems, it is essential to collect in vitro
141,142
experimental data to define MR conditions that may permit imaging to be performed safely. From an MR
safety perspective, the greatest concern for electronically activated or electrically conductive implants in the brain is
excessive MR imaging-related heating that can cause irreversible tissue damage. 141,142 Studies conducted to date
revealed that there is a realistic potential for injury due to excessive MR imaging-related heating of neurostimulation
141,142
systems used for DBS.
Recent investigations evaluated MRI-related heating for the only neurostimulation system approved by the FDA for
141,142
use in chronic deep brain stimulation (Activa® Tremor Control System, Medtronic, Minneapolis, MN). This
system is a fully implantable, multiprogrammable device designed to deliver electrical stimulation to the thalamus or
other brain structures. The basic implantable system is composed of the neurostimulator (or implantable pulse
generator, IPG), DBS lead, and an extension that connects the lead to the IPG. This system delivers
high-frequency electrical stimulation to a multiple contact electrode placed in the ventral intermediate nucleus of the
thalamus or other anatomic site.
Studies conducted by Rezai et al141 and Finelli et al142 on neurostimulation systems indicated that MR safety is
highly dependent on a number of critical factors. To simulate a "worst-case" clinical application of DBS, these
investigations evaluated bilateral DBS applications such that two neurostimulators, two extensions, and two leads
were assessed during in vitro experiments (Fig. 24-6). Different configurations were evaluated for the bilateral
141,142
neurostimulation systems to characterize worst-case and clinically relevant positioning scenarios. MR imaging
procedures were performed on a gel-filled phantom designed to approximate the head and upper torso of a human
subject. Temperature changes were studied in association with MR examinations conducted at 1.5 T/64 MHz at
various levels of RF energy using the transmit/receive RF body and transmit/receive head RF coil. The findings from
these studies indicated that substantial heating occurs under certain conditions while others produced relatively
minor, physiologically inconsequential temperature increases. Furthermore, factors that strongly influenced local
temperature increases at the electrode tip included the positioning of the neurostimulation system (especially the
electrode), the type of RF coil used, and the specific absorption rate (SAR) used for the MR procedure.
page 665
page 666
8 of 10 15-04-2010 21:55
Add to lightbox
Figure 24-6 Schematic showing bilateral neurostimulation systems (Activa® Tremor Control System, Medtronic,
Minneapolis, MN) with implantable pulse generators (IPGs) in the subclavian pockets (Soletra® Model 7426
neurostimulator, Medtronic), attached to extensions (Model 7495 quadripolar extension, Medtronic) wound around
the IPGs, connected to quadripolar leads (Model 3389 DBS™ lead, Medtronic). Each quadripolar lead is
positioned in the thalamus. Note: There is no small coil placed at the top of the burr hole for this positioning
scheme. As such, it is inadvisable for a patient with bilateral neurostimulation systems placed in this manner to
undergo an MR procedure due to the possibility of excessive heating at the tips of the electrodes.
According to the study by Rezai et al,141 MRI-related heating does not appear to present a major safety concern
for patients with the bilateral neurostimulation systems that underwent testing, as long as highly specific guidelines
pertaining to the positioning of these neurostimulation devices and parameters used for MR imaging are carefully
adhered to. Finelli et al142 reported that MR imaging sequences commonly used for clinical procedures can be
performed safely in patients with bilateral DBS neurostimulation systems at 1.5 tesla with the utilization of a
transmit/receive RF head coil. It should be noted that most present-day high field strength MR systems use the
body coil to transmit RF with a receive-only head coil. As such, additional studies are required to characterize the
impact of the use of this transmit/receive RF coil combination with regard to MR imaging-related heating of
neurostimulation systems used for DBS.
Gastric Electric Stimulation

Gastric electrical stimulation (GES) performed using a specialized neurostimulation device (The Enterra Therapy,
Gastric Electrical Stimulation (GES) System, Medtronic Inc., Minneapolis, MN) is indicated for treatment of patients
with chronic, intractable nausea and vomiting secondary to gastroparesis of diabetic or idiopathic etiology. GES
uses mild electrical pulses to stimulate the stomach to help control symptoms associated with gastroparesis.
The GES device is composed of a neurostimulator, an implantable intramuscular lead, and an external programming
system. Currently, the use of MR procedures in patients with this device is contraindicated due to possible hazards
related to dislodgment or heating of the neurostimulator and/or the leads used for gastric electrical stimulation.
Additionally, the voltage induced through the lead and neurostimulator may cause uncomfortable "jolting" or
5
"shocking" levels of stimulation.
Postoperative Patients and Magnetic Resonance Procedures

Because confusion exists regarding the issue of performing an MR procedure during the postoperative period in a
9 of 10 15-04-2010 21:55
patient with a metallic implant or device, the IMRSER recently developed guidelines pertaining to this MR safety
topic ( http://www.IMRSER.org). Studies have supported that if a metallic object is a "passive implant" (i.e., there
is no electronically or magnetically activated component associated with the operation of the object) and it is made
from a nonferromagnetic material (e.g., titanium, titanium alloy, Nitinol, etc.), the patient may undergo an MR
procedure immediately after implantation using a system operating at 1.5 tesla or
less.5-7,178,186,190,193,206,207,211,224,226-228 In fact, several reports describe placement of vascular stents and other
234,237,268-271
implants using MR-guided procedures that include the use of high field strength (1.5 T) MR systems.
Additionally, a patient or individual with a nonferromagnetic, passive implant would be allowed to enter an MR
environment associated with a 1.5 T or less MR system immediately after implantation of such an object. Currently,
there are few data to provide guidelines for MR environments using scanners operating at 3.0 tesla or higher.
For an implant or device that exhibits "weakly magnetic" qualities, it is typically necessary to wait a period of 6 to 8
weeks after implantation before performing an MR procedure or allowing the individual or patient to enter the MR
5-7,224-228
environment. For example, certain intravascular and intracavitary coils, stents, filters, and cardiac
occluders designated as being weakly ferromagnetic become firmly incorporated into tissue 6 to 8 weeks following
placement. In these cases, retentive or counterforces provided by tissue ingrowth, scarring, or granulation
essentially prevent these objects from presenting hazards to patients or individuals in the MR environment. Patients
with those implants or devices that are weakly magnetic but rigidly fixed in the body, such as a bone screw, may be
186
scanned immediately in the postoperative period.
If there is any concern regarding the ability of the tissue to retain the implant or object in place or the implant cannot
be properly identified, the patient or individual should not be exposed to the MR environment. Specific information
pertaining to the recommended postoperative waiting period may be found in the labeling or product insert for a
"weakly magnetic" implant or device.
© 2010 Elsevier
10 of 10 15-04-2010 21:55
CONCLUSION
With the continued advancements in MR technology and development of more sophisticated implants
and devices, there is an increased potential for hazardous situations to occur in the MR environment.
Therefore, to prevent incidents and accidents, it is necessary to be aware of the latest information
pertaining to MR bioeffects, to use current evidence-based guidelines to ensure safety for patients and
staff members, and to follow proper recommendations pertaining to biomedical implants and devices.
REFERENCES
1. Schenck JF: Health effects and safety of static magnetic fields. In Shellock FG (ed): Magnetic resonance procedures: health
effects and safety. Boca Raton, FL: CRC Press, 2001, pp. 1-30.
2. Criteria for Significant Risk Investigations of Magnetic Resonance Diagnostic Devices, Document issued on July 14 2003
[this document supersedes "Guidance for Magnetic Resonance Diagnostic Devices-Criteria for Significant Risk
Investigations" issued on September 29 1997]. US Department of Health and Human Services, Food and Drug Administration,
Center for Devices and Radiological Health, Radiological Devices Branch, Division of Reproductive, Abdominal, and
Radiological Devices, Office of Device Evaluation, 2003.
3. US Food and Drug Administration, Center for Devices and Radiological Health (CDRH), Medical Device Report (MDR) (
http://www.fda.gov/CDRH/mdrfile.html). The files contain information from CDRH's device experience reports on devices which
may have malfunctioned or caused a death or serious injury. The files contain reports received under both the mandatory
Medical Device Reporting Program (MDR) from 1984 to 1996, and the voluntary reports up to June 1993. The database
currently contains over 600,000 reports.
4. US Food and Drug Administration, Center for Devices and Radiological Health (CDRH), Manufacturer and User Facility
Device Experience Database, MAUDE ( http://www.fda.gov/cdrh/maude.html). MAUDE data represent reports of adverse
events involving medical devices. The data consist of all voluntary reports since June 1993, user facility reports since 1991,
distributor reports since 1993, and manufacturer reports since August 1996.
5. Shellock FG: Reference manual for magnetic resonance safety, implants, and devices: 2005 edition Los Angeles, CA:
Biomedical Research Publishing Group, 2005.
6. Shellock FG: Pocket guide to MR procedures and metallic objects: update 2001. Philadelphia, PA: Lippincott, Williams and
Wilkins, 2001.
7. Shellock FG: Magnetic resonance procedure: health effects and safety. Boca Raton, FL: CRC Press, 2001.
8. Schaefer DJ, Bourland JD, Nyenhuis JA: Review of patient safety in time-varying gradient fields. J Magn Reson Imaging
9. Schenck JF: Safety of strong, static magnetic fields. J Magn Reson Imaging 12:2-19, 2000. Medline Similar articles
10. King KF, Schaefer DJ: Spiral scan peripheral nerve stimulation. J Magn Reson Imaging 12:164-170, 2000. Medline
Similar articles
11. Shellock FG: Radiofrequency energy-induced heating during MR procedures: a review. J Magn Reson Imaging 12:30-36,
12. McJury M, Shellock FG: Auditory noise associated with MR procedures: a review. J Magn Reson Imaging 12:37-45, 2000.
13. Shellock FG, Kanal E: Policies, guidelines, and recommendations for MR imaging safety and patient management. J Magn
14. Kanal E, Shellock FG: Policies, guidelines, and recommendations for MR imaging safety and patient management. Patient
monitoring during MR examinations. J Magn Reson Imaging 2:247-248, 1992. Medline Similar articles
15. Shellock FG, Kanal E: SMRI Report. Policies, guidelines and recommendations for MR imaging safety and patient
management. Questionnaire for screening patients before MR procedures. J Magn Reson Imaging 4:749-751, 1994.
16. International Electrotechnical Commission (IEC): Medical electrical equipment. Particular requirements for the safety of
magnetic resonance equipment for medical diagnosis. International Standard IEC 60601-2-33, 2002.
17. Budinger TF: Nuclear magnetic resonance (NMR) in vitro studies: known thresholds for health effects. J Comput Assist
18. Schenck JF, Dumoulin CL, Redington RW, et al: Human exposure to 4.0 tesla magnetic fields in a whole-body scanner.
Med Phys 19:1089-1098, 1992.
19. Kangarlu A, Burgess RE, Zhu H, et al: Cognitive, cardiac, and physiological safety studies in ultra high field magnetic
resonance imaging. Magn Reson Imaging 17:1407-1416, 1999. Medline Similar articles
20. Hong CZ, Shellock FG: Short-term exposure to a 1.5 Tesla static magnetic field does not affect somato-sensory evoked
potentials in man. Magn Reson Imaging 8:65-69, 1989.
21. Innis NK, Ossenkopp KP, Prato FS, et al: Behavioral effects of exposure to nuclear magnetic resonance imaging: II. Spatial
memory tests. Magn Reson Imaging 4:281-284, 1986. Medline Similar articles
1 of 10 15-04-2010 21:55
22. Shellock FG, Schaefer DJ, Gordon CJ: Effect of a 1.5 Tesla static magnetic field on body temperature of man. Magn
Reson Med 3:644-647, 1986.
23. Shellock FG, Schaefer DJ, Crues JV: Exposure to a 1.5 Tesla static magnetic field does not alter body and skin
temperatures in man. Magn Reson Med 1:371-375, 1989.
24. Brody A, Sorette M, Gooding C, et al: Induced alignment of flowing sickle erythrocytes in a magnetic field: a preliminary
report. Invest Radiol 20:560-566, 1985. Medline Similar articles
25. Brody A, Embury S, Mentzer W, et al: Preservation of sickle cell blood flow pattern(s) during MR imaging: an in vivo study.
Am J Roentgenol 151:139-141, 1988.
26. Klitzing LV: Do static magnetic fields of NMR influence biological signals? Clin Phys Physiol Meas 7:157-160, 1986.
27. Papatheofanis F: A review on the interaction of biological systems with magnetic fields. Physiol Chem Phys Med NMR
28. Tenforde T, Budinger T: Biologicalal effects and physical safety aspects of NMR imaging and in vivo spectroscopy. In
Thomas S, Dixon R (eds): NMR in Medicine: Instrumentation and Clinical Applications. New York: American Association of
Physicists in Medicine, 1986, pp. 493-548.
29. Vogl T, Pasulus W, Fuchs A, et al: Influence of magnetic resonance imaging on evoked potentials and nerve conduction
velocities in humans. Invest Radiol 26:432-437, 1991. Medline Similar articles
30. Von Klitzing L: Do static magnetic fields of NMR influence biological signals? Clin Phys Physiol Meas 7:157-160, 1986.
31. Weiss M, Herrick R, Tabor K, et al: Bioeffects of high magnetic fields: a study using a simple animal model. Magn Reson
32. Yuh WTC, Ehrhardt JC, Fisher DJ, et al: Phantom limb pain induced in amputee by strong magnetic fields. J Magn Reson
33. Ngo F, Blue J, Roberts W: The effect of a static magnetic field on DNA synthesis and survival of mammalian cells irradiated
with fast neurons. Magn Reson Med 5:307-317, 1987. Medline Similar articles
34. Kay H, Herfkens R, Kay B: Effect on magnetic resonance imaging on Xenopus Laevis embryogenesis. Magn Reson
35. Prasad N, Wright D, Ford J, Thornby J: Safety of 4-T MR imaging: study of effects on developing frog embryos. Radiology
174:251-253, 1990.
36. Prasad N, Bushong S, Thronby J, et al: Effect of nuclear magnetic resonance on chromosomes of mouse bone marrow
cells. Magn Reson Imaging 2:37-39, 1990.
37. Geard C, Osmak R, Hall E, et al: Magnetic resonance and ionizing radiation: a comparative evaluation in vitro of oncogenic
and genotoxic potential. Radiology 152:199-202, 1984. Medline Similar articles
38. Cooke P, Morris P: The effects of NMR exposure of living organism. II. A genetic study of human lymphocytes. Br J Radiol
39. Wolff S, Crooks L, Brown P, et al: Test for DNA and chromosomal damage induced by nuclear magnetic resonance
40. Wolff S, James T, Young G, et al: Magnetic resonance imaging: absence of in vitro cytogenetic damage. Radiology
41. Schwartz J, Crooks L: NMR imaging produces no observable mutations or cytotoxicity in mammalian cells. Am J
Roentgenol 139:583-585, 1982.
42. Tyndall D, Sulik K: Effects of magnetic resonance imaging on eye development in the C57BL/6J mouse. Teratology
43. Heinrichs W, Fong P, Flannery M, et al: Midgestational exposure of pregnant balb/c mice to magnetic resonance imaging.
44. McRobbie D, Foster M: Pulsed magnetic field exposure during pregnancy and implications for NMR foetal imaging: a study
with mice. Magn Reson Imaging 3:231-234, 1985. Medline Similar articles
45. Shivers R, Kavaliers M, Teskey G, et al: Magnetic resonance imaging temporarily alters blood-brain barrier permeability in
the rat. Neurosci Lett 76:25-31, 1987. Medline Similar articles
46. Short W, Goodwill L, Taylor C, et al: Alteration of human tumor cell adhesion by high-strength static magnetic fields. Invest
Radiol 27:836-840, 1991.
47. Prato F, Ossenkeopp KP, Kavaliers M, et al: Attenuation of morphine-induced analgesia in mice by exposure to MRI:
separate effects of the static, radiofrequency and time-varying magnetic fields. Magn Reson Imaging 5:9-14, 1987. Medline
Similar articles
48. Ossenkopp KP, Kavaliers M, Prato F, et al: Exposure to nuclear magnetic resonance imaging procedure attenuates
morphine-induced analgesia in mice. Life Sci 37:1507-1514, 1985. Medline Similar articles
49. Kavaliers M, Ossenkopp KP: Tolerance to morphine-induced analgesia in mice: magnetic fields function as environmental
specific cues and reduce tolerance development. Life Scie 37:1125-1135, 1985.
2 of 10 15-04-2010 21:55
50. Muller S, Hotz M: Human brainstem auditory evoked potentials (BAEP) before and after MR examinations. Magn Reson
51. Brockway J, Bream P: Does memory loss occur after MR imaging? J Magn Reson Imaging 2:721-728, 1992. Medline
Similar articles
52. Besson J, Foreman E, Eastwood L, et al: Cognitive evaluation following NMR imaging of the brain. J Neurol Neurosurg
Psychiat 47:314-316, 1984. Medline Similar articles
page 667
page 668
53. Ossenkopp KP, Innis N, Prato F, Sestini E: Behavioral effects of exposure to nuclear magnetic resonance imaging: I.
Open-field avoidance behavior and passive avoidance learning in rats. Magn Reson Imaging 4:275-280, 1986. Medline
Similar articles
54. Doherty J, Whitman G, Robinson M, et al: Changes in cardiac excitability and vulnerability in NMR fields. Invest Radiol
55. Nyenhuis JA, Bourland J, Kildishev AV, Schaefer DJ: Health effects and safety of intense gradient fields. In Shellock FG
(ed): Magnetic Resonance Procedures: Health Effects and Safety. Boca Raton, FL: CRC Press, 2001, pp. 31-54.
56. Bourland JD, Nyenhuis JA, Schaefer DJ: Physiologic effects of intense MRI gradient fields. Neuroimag Clin North Am
9:363-377, 1999.
57. Cohen MS, Weisskoff R, Kantor H: Sensory stimulation by time varying magnetic fields. Magn Reson 14:409-414, 1990.
58. Budinger TF, Fischer H, Hentshel D, et al: Physiological effects of fast oscillating magnetic field gradients. J Comput Assist
Tomogr 15:609-614, 1991.
59. Mansfield P, Harvey PR: Limits to neural stimulation in echo-planar imaging. Magn Reson Med 29:746-758, 1993.
60. Irnich W, Schmitt F: Magnetostimulation in MRI. Magn Reson Med 33:619-623, 1995. Medline Similar articles
61. Ehrhardt JC, Lin CS, Magnotta VA, et al: Peripheral nerve stimulation in a whole-body echo-planar imaging system. J
62. Abart J, Eberhardt K, Fischer H, et al: Peripheral nerve stimulation by time-varying magnetic fields. J Comput Assis
Tomogr 21:532-538, 1997.
63. Ham CLG, Engels JML, van de Weil GT, Machielsen A: Peripheral nerve stimulation during MRI: effects of high gradient
amplitudes and switching rates. J Magn Reson Imaging 7:933-937, 1997. Medline Similar articles
64. Nyenhuis JA, Bourland JD, Schaefer DJ: Analysis from a stimulation perspective of magnetic field patterns of MR gradient
coils. J Appl Phys 81:4314-4316, 1997.
65. McJury M: Acoustic noise and magnetic resonance procedures. In Shellock FG (ed): Magnetic Resonance Procedures:
Health Effects and Safety. Boca Raton, FL: CRC Press, pp. 115-138.
66. McJury M, Shellock FG: Acoustic noise and MR procedures: a review. J Magn Reson Imaging 12:37-45, 2000. Medline
Similar articles
67. Brummett RE, Talbot JM, Charuhas P: Potential hearing loss resulting from MR imaging. Radiology 169:539-540, 1988.
68. Quirk ME, Letendre AJ, Ciottone RA, Lingley JF: Anxiety in patients undergoing MR imaging. Radiology 170:464-466,
1989.
69. Laurell G: The combined effect of noise and cisplatin . Ann Otol Rhinol Laryngol 1001:969-976, 1992.
70. Philbin MK, Taber KH, Hayman LA: Preliminary report: changes in vital signs of term newborns during MR. Am J
71. Counter SA, Olofsson A, Borg E, et al: Analysis of magnetic resonance imaging acoustic noise generated by a 4.7 T
experimental system. Acta Otolaryngol 120:739-743, 2000.
72. Price DL, De Wilde JP, Papadaki AM, et al: Investigation of acoustic noise on 15 MRI scanners from 0.2 T to 3 T. Magn
Reson Imaging 13:288-293, 2001.
73. Shellock FG, Ziarati M, Atkinson D, Chen D-Y: Determination of gradient magnetic field-induced acoustic noise associated
with the use of echo planar and three-dimensional fast spin echo techniques. J Magn Reson Imaging 8:1154-1157, 1998.
74. Strainer JC, Ulmer JL, Yetkin FZ, et al: Functional MR of the primary auditory cortex: an analysis of pure tone activation and
tone discrimination. Am J Roentgenol 18: 601-610, 1997.
75. Ulmer JL, Bharat BB, Leighton PM, et al: Acoustic echo planar scanner noise and pure tone hearing thresholds: the effects
of sequence repetition times and acoustic noise rates. J Comput Assist Tomogr 22:480-486, 1998. Medline
Similar articles
76. Bandettini PA, Jesmanowicz A, Van Kylen J, et al: Functional MRI of brain activation induced by scanner acoustic noise.
77. Goldman AM, Gossman WE, Friedlander PC: Reduction of sound levels with antinoise in MR imaging. Radiology
3 of 10 15-04-2010 21:55
78. Hurwitz R, Lane SR, Bell RA, Brant-Zawadzki MN.: Acoustic analysis of gradient-coil noise in MR imaging. Radiology
79. Shellock FG, Morisoli SM, Ziarati M: Measurement of acoustic noise during MR imaging: evaluation of six "worst-case"
pulse sequences. Radiology 191:91-93, 1994. Medline Similar articles
80. McJury M, Blug A, Joerger C, et al: Acoustic noise levels during magnetic resonance imaging scanning at 1.5 T. Br J
Radiol 67:413-415, 1994.
81. McJury MJ: Acoustic noise levels generated during high field MR imaging. Clin Radiol 50:331-334, 1995. Medline
Similar articles
82. Counter SA, Olofsson A, Grahn, Borg E: MRI acoustic noise: sound pressure and frequency analysis. J Magn Reson
83. Hedeen RA, Edelstein WA: Characteristics and prediction of gradient acoustic noise in MR imagers. Magn Reson Med
84. Cho ZH, Park SH, Kim JH, et al: Analysis of acoustic noise in MRI. Magn Reson Imaging 15:815-822, 1997. Medline
Similar articles
85. Miller LE, Keller AM: Regulation of occupational noise. In Harris CM (ed): Handbook of Noise Control. New York:
McGraw-Hill, 1979, pp. 1-16.
86. Melnick W: Hearing loss from noise exposure. In Harris CM (ed): Handbook of Noise Control. New York: McGraw-Hill,
1979, p. 2.
87. Schaefer DJ: Safety aspects of radio-frequency power deposition in magnetic resonance. MRI Clin North Am 6:775-789,
1998.
88. Morvan, D, Leroy-Willig A, Jehenson P, et al: Temperature changes induced in human muscle by radiofrequency H-1
coupling: measurement with an MR imaging diffusion technique. Radiology 185:871-874, 1992. Medline
Similar articles
89. Bottomley PA, Edelstein WA: Power deposition in whole body NMR imaging. Med Phys 8: 510-512, 1981.
90. Shellock FG, Crues JV: Temperature, heart rate, and blood pressure changes associated with clinical MR imaging at
1.5-T. Radiology 163:259-262, 1987.
91. Shellock FG, Schaefer DJ, Grundfest W, et al: Thermal effects of high-field (1.5 Tesla) magnetic resonance imaging of the
spine: clinical experience above a specific absorption rate of 0.4 W/kg. Acta Radiol Suppl 369:514-516, 1986.
92. Shellock FG, Gordon CJ, Schaefer DJ: Thermoregulatory response to clinical magnetic resonance imaging of the head at
1.5 Tesla: lack of evidence for direct effects on the hypothalamus. Acta Radiol Suppl 369:512-513, 1986.
93. Shellock FG, Crues JV: Corneal temperature changes associated with high-field MR imaging using a head coil. Radiology
94. Shellock FG, Crues JV: Temperature changes caused by clinical MR imaging of the brain at 1.5 Tesla using a head coil.
Am J Neuroradiol 9:287-291, 1988.
95. Shellock FG, Rothman B, Sarti D: Heating of the scrotum by high-field-strength MR imaging. Am J Roentgenol
154:1229-1232, 1990.
96. Shellock FG, Schatz CJ: Increases in corneal temperature caused by MR imaging of the eye with a dedicated local coil.
97. Shellock FG, Schaefer DJ, Crues JV: Alterations in body and skin temperatures caused by MR imaging: is the
recommended exposure for radiofrequency radiation too conservative? Br J Radiol 62:904-909, 1989. Medline
Similar articles
98. Vogl T, Krimmel K, Fuchs A, et al: Influence of magnetic resonance imaging of human body core and intravascular
temperature. Med Phys 15:562-566, 1988. Medline Similar articles
99. Kido DK, Morris TW, Erickson JL, et al: Physiologic changes during high-field strength MR imaging. Am J Neuroradiol
8:263-266, 1987.
100. Kangarlu A, Shellock FG, Chakeres D: 8.0 tesla MR system: temperature changes associated with radiofrequency-
induced heating of a head phantom. J Magn Reson Imaging 17:220-226, 2003.
101. Shellock FG, Schaefer DJ, Kanal E: Physiologic responses to MR imaging performed at an SAR level of 6.0 W/kg.
Radiology 192:865-868, 1994.
102. Schaefer DJ: Health effects and safety of radiofrequency power deposition associated with magnetic resonance
procedures. In Shellock FG (ed): Magnetic Resonance Procedures: Health Effects and Safety. Boca Raton, FL: CRC Press,
2001, pp. 55-74.
103. Shellock FG, Schaefer DJ: Radiofrequency energy-induced heating during magnetic resonance procedures: laboratory
and clinical experiences. In Shellock FG (ed): Magnetic Resonance Procedures: Health Effects and Safety. Boca Raton, FL:
CRC Press, 2001, pp. 75-96.
104. Shuman WP, Haynor DR, Guy AW, et al: Superficial and deep-tissue increases in anesthetized dogs during exposure to
high specific absorption rates in a 1.5-T MR imager. Radiology 167:551-554, 1994.
105. Gordon CJ: Normalizing the thermal effects of radiofrequency radiation: body mass versus total body surface area.
Bioelectromagnetics 8:111-118, 1987. Medline Similar articles
4 of 10 15-04-2010 21:55
106. National Council on Radiation Protection and Measurements: Biological Effects and Exposure Criteria for
Radiofrequency Electromagnetic Fields. Report No. 86. Bethesda, MD: National Council on Radiation Protection and
Measurements, 1986.
107. Rowell LB: Cardiovascular aspects of human thermoregulation. Circ Res 52:367-379, 1983. Medline
Similar articles
108. Drinkwater BL, Horvath SM: Heat tolerance and aging. Med Sci Sport Exer 11:49-55, 1979.
109. Fennel WH, Moore RE: Responses of aged men to passive heating. Am J Physiol 67:118-119, 1969.
110. Kenny WL: Physiological correlates of heat intolerance. Sports Med 2:279-286, 1985. Medline Similar articles
111. Barany FR: Abnormal vascular reaction in diabetes mellitus. Acta Med Scand Suppl 304:556-624, 1955.
112. Buskirk EF, Lundergren H, Magnussen L: Heat acclimation patterns in obese and lean individuals. Ann NY Acad Sci
113. Jauchem JR: Effects of drugs on thermal responses to microwaves. Gen Pharmacol 16:307-310, 1985. Medline
Similar articles
114. Schaefer DJ, Barber BJ, Gordon CJ: Thermal effects of magnetic resonance imaging. In Book of Abstracts. Berkeley,
CA: Society for Magnetic Resonance in Medicine, 1985, p. 925.
115. Sawyer-Glover A, Shellock FG: Pre-magnetic resonance procedure screening. In Shellock FG (ed): Magnetic Resonance
Procedures: Health Effects and Safety. Boca Raton, FL: CRC Press, 2001, pp. 271-326.
116. Sawyer-Glover A, Shellock FG: Pre-MRI procedure screening: recommendations and safety considerations for
biomedical implants and devices. J Magn Reson Imaging 12: 92-106, 2000.
117. Shellock FG: MR safety update 2002: implants and devices. J Magn Reson Imaging 16:485-496, 2002.
118. Kelly WM, Paglen PG, Pearson JA, et al: Ferromagnetism of intraocular foreign body causes unilateral blindness after MR
study. Am J Neuroradiol 7:243-245, 1986. Medline Similar articles
119. Otto PM, Otto RA, Virapongse C, et al: Screening test for detection of metallic foreign objects in the orbit before magnetic
resonance imaging. Invest Radiol 27:308-311, 1992. Medline Similar articles
page 668
page 669
120. Jarvik JG, Ramsey JG: Radiographic screening for orbital foreign bodies prior to MR imaging: is it worth it? Am J
121. Seidenwurm DJ, McDonnell CH 3rd, Raghavan N, Breslau J: Cost utility analysis of radiographic screening for an orbital
foreign body before MR imaging. Am J Neuroradiol 21:426-433, 2000.
122. Dempsey MF, Condon B: Thermal injuries associated with MRI. Clin Radiol 56:457-465, 2001. Medline
Similar articles
123. Dempsey MF, Condon B, Hadley DM: Investigation of the factors responsible for burns during MRI. J Magn Reson
124. ECRI: Health devices alert. A new MRI complication? Health Devices Alert May 27, p. 1, 1988.
125. ECRI: Thermal injuries and patient monitoring during MRI studies. Health Devices Alert 20:362-363, 1991.
126. Hall SC, Stevenson GW, Suresh S: Burn associated with temperature monitoring during magnetic resonance imaging.
Anesthesiology 76:152, 1992. Medline Similar articles
127. Jones S, Jaffe W, Alvi R: Burns associated with electrocardiographic monitoring during magnetic resonance imaging.
Burns 22:420-421, 1996. Medline Similar articles
128. Kanal E, Shellock FG: Burns associated with clinical MR examinations. Radiology 175:585, 1990. Medline
Similar articles
129. Keens SJ, Laurence AS: Burns caused by ECG monitoring during MRI imaging. Anaesthesia 51:1188-1189, 1996.
130. Kugel H, Bremer C, Puschel M, et al: Hazardous situation in the MR bore: induction in ECG leads causes fire. Eur Radiol
131. Nakamura T, Fukuda K, Hayakawa K, et al: Mechanism of burn injury during magnetic resonance imaging (MRI)-simple
loops can induce heat injury. Front Med Biol Eng 11:117-129, 2001. Medline Similar articles
132. Bashein G, Syrory G: Burns associated with pulse oximetry during magnetic resonance imaging. Anesthesiology
133. Brown TR, Goldstein B, Little J: Severe burns resulting from magnetic resonance imaging with cardiopulmonary
monitoring. Risks and relevant safety precautions. Am J Phys Med Rehabil 72:166-167, 1993. Medline Similar articles
134. Shellock FG, Slimp G: Severe burn of the finger caused by using a pulse oximeter during MRI. Am J Roentgenol
153:1105, 1989.
135. Smith CD, Kildishev AV, Nyenhuis JA, et al: Interactions of MRI magnetic fields with elongated medical implants. J Appl
Phys 87:6188-6190, 2000.
136. Smith CD, Nyenhuis JA, Kildishev AV: Health effects of induced electrical currents: implications for implants. In Shellock
FG (ed): Magnetic Resonance Procedures: Health Effects and Safety. Boca Raton, FL: CRC Press, 2001, pp. 393-413.
5 of 10 15-04-2010 21:55
137. Shellock FG: MR safety update 2002: implants and devices. J Magn Reson Imaging 16:485-496, 2002.
138. Chou C-K, McDougall JA, Chan KW: Absence of radiofrequency heating from auditory implants during magnetic
resonance imaging. Bioelectromagnetics 44:367-372, 1997.
139. Davis PL, Crooks L, Arakawa M, et al: Potential hazards in NMR imaging: heating effects of changing magnetic fields and
RF fields on small metallic implants. Am J Roentgenol 137:857-860, 1981.
140. Nyenhuis JA, Kildishev AV, Foster KS, et al: Heating near implanted medical devices by the MRI RF-magnetic field. IEEE
Trans Magn 35:4133-4135, 1999.
141. Rezai AR, Finelli D, Nyenhuis JA, et al: Neurostimulator for deep brain stimulation: ex vivo evaluation of MRI-related
heating at 1.5 tesla. J Magn Reson Imaging 15:241-250, 2002.
142. Finelli DA, Rezai AR, Ruggieri PM, et al: MR imaging-related heating of deep brain stimulation electrodes: in vitro study.
143. Shellock FG, Hatfield M, Simon BJ, et al: Implantable spinal fusion stimulator: assessment of MRI safety. J Magn Reson
144. Knopp MV, Essig M, Debus J, et al: Unusual burns of the lower extremities caused by a closed conducting loop in a
patient at MR imaging. Radiology 200:572-575, 1996. Medline Similar articles
145. Becker H: The use of intradermal tattoo to enhance the final result of nipple-areola reconstruction. Plast Reconstr Surg
146. Halder RM, Pham HN, Hreadon JY, Johnson HA: Micropigmentation for the treatment of vitiligo. J Dermatol Surg Oncol
147. Carr JJ: Danger in performing MR imaging on women who have tattooed eyeliner or similar types of permanent cosmetic
injections. Am J Roentgenol 165:1546-1547, 1995.
148. Jackson JG, Acker H: Permanent eyeliner and MR imaging. (letter). Am J Roentgenol 149:1080, 1987.
149. Lund A, Nelson ID, Wirtschafter ID, Williams A: Tattooing of eyelids: magnetic resonance imaging artifacts. Ophthalm
Surg 17:550-553, 1986.
150. Tope WD, Shellock FG: Magnetic resonance imaging and permanent cosmetics (tattoos): survey of complications and
adverse events. J Magn Reson Imaging 15:180-184, 2002. Medline Similar articles
151. US Food and Drug Administration, Center for Food Safety and Applied Nutrition: Office of Cosmetics and Colors Fact
Sheet-Tattoos and Permanent Makeup. Rockville, MD, 2000.
152. Vahlensieck M: Tattoo-related cutaneous inflammation (burn grade I) in a mid-field MR scanner (letter). Eur Radiol 10:97,
2000.
153. Wagle WA, Smith M: Tattoo-induced skin burn during MR imaging (letter). Am J Roentgenol 174:1795, 2000.
154. Kreidstein ML, Giguere D, Freiberg A: MRI interaction with tattoo pigments: case report, pathophysiology, and
management. Plast Reconstr Surg 99:1717-1720, 1997. Medline Similar articles
155. Kanal E, Shellock FG: MRI interaction with tattoo pigments (letter). Plast Reconstr Surg 101:1150-1151, 1998.
156. Weiss RA, Saint-Louis LA, Haik BG, et al: Mascara and eyelining tattoos: MRI artifacts. Ann Ophthalmol 21:129-131,
157. Colletti PM: Magnetic resonance procedures and pregnancy. In Shellock FG (ed): Magnetic Resonance Procedures:
Health Effects and Safety. Boca Raton, FL: CRC Press, 2001, pp. 149-182.
158. Stark DD, McCarthy SM, Filley RA, et al: Pelvimetry by magnetic resonance imaging. Am J Roentgenol 144:947-950,
1985.
159. McCarthy SM, Filly RA, Stark DD, et al: Obstetrical magnetic resonance imaging: fetal anatomy. Radiology 154:427-432,
160. McCarthy SM, Stark DD, Filly RA, et al: Obstetrical magnetic resonance imaging: maternal anatomy. Radiology
15:421-425, 1985.
161. Smith FW, Kent C, Abramovich DR, et al: Nuclear magnetic resonance imaging-a new look at the fetus. Br J Gynaecol
92:1024-1033, 1985.
162. Smith FW, Sutherland HW: Magnetic resonance imaging: the use of the inversion recovery sequence to display fetal
morphology. Br J Radiol 61:338-341, 1988. Medline Similar articles
163. Kier R, McCarthy SM, Scoutt LM, et al: Pelvic masses in pregnancy: MR imaging. Radiology 176:709-713, 1990.
164. Kay HH, Spritzer CE: Preliminary experience with magnetic resonance imaging in patients with third-trimester bleeding.
Obstet Gyenecol 78:424-429, 1991.
165. Garden AS, Weindling AM, Griffiths RD, et al: Fast-scan magnetic resonance imaging of fetal abnormalities. Br J Obstet
Gynaecol 98:1217-1222, 1991. Medline Similar articles
166. Tempany CMC, Fielding JR: Female pelvis. In Edelman RR, Hesselink JR, Zlatkin MB (eds): Clinical Magnetic
Resonance Imaging, 2nd edn. Philadelphia, PA: Saunders, 1996, pp. 1432-1466.
167. Heinrichs WL, Fong P, Flannery M, et al: Midgestational exposure of pregnant balb/c mice to magnetic resonance
6 of 10 15-04-2010 21:55
imaging. Mag Res Imaging 6:305-313, 1988.

168. McRobbie D, Foster MA: Pulsed magnetic field exposure during pregnancy and implications for NMR foetal imaging: a
study with mice. Magn Reson Imaging 3:231-234, 1985. Medline Similar articles
169. Prasad N, Wright DA, Ford JJ, Thornby JI: Effect of nuclear magnetic resonance on early stages of amphibian
development. Magn Reson Imaging 1:35-38, 1982. Medline Similar articles
170. Prasad N, Wright DA, Ford JJ, Thornby JI: Safety of 4-T MR imaging: a study of effects of developing frog embryos.
Radiology 174:251-253, 1990.
171. Tyndall DA, Sulik KK: Effects of magnetic resonance imaging on eye development in the C57BL/6J mouse. Teratology
172. Tyndall DA: MRI effects on the tertogenicity of X-irradiation in the C57BL/6J mouse. Magn Reson Imaging 8:423-433,
173. Tyndall DA: MRI effects on the teratogenicity of X-irradiation in the C57BL/6J mouse. Magn Reson Imaging 8:423-433,
174. Wolff S, Crooks LE, Brown P, et al: Tests for DNA and chromosomal damage induced by nuclear magnetic resonance
175. Kay HH, Herfkens RJ, Kay BK: Effect of magnetic resonance imaging on Xenopus Laevis embryogenesis. Magn Reson
176. Yip YP, Capriotti C, Norbash SG, et al: Effects of MR exposure on cell proliferation and migration of chick motoneurons. J
177. Yip YP, Capriotti C, Yip JW: Effects of MR exposure on axonal outgrowth in the sympathetic nervous system of the chick. J
178. Shellock FG, Crues JV: High-field strength MR imaging and metallic biomedical implants: an ex vivo evaluation of
deflection forces. Am J Roentgenol 151:389-392, 1988.
179. Shellock FG: Biomedical implants and devices: assessment of magnetic field interactions with a 3.0 tesla MR system. J
180. Shellock FG, Tkach JA, Ruggieri PM, et al: Aneurysm clips: evaluation of magnetic field interactions using "long-bore"
and "short-bore" 3.0 tesla MR systems. Am J Neuroradiol 24:463-471, 2003.
181. Shellock FG, Tkach JA, Ruggieri PM, Masaryk TJ: Cardiac pacemakers, ICDs, and loop recorder: evaluation of
translational attraction using conventional ("long-bore") and "short-bore" 1.5- and 3.0 tesla MR systems. J Cardiovasc Magn
Reson 5:387-397, 2003.
182. Kangarlu A, Shellock FG: Aneurysm clips: evaluation of magnetic field interactions with an 8.0 T MR system. J Magn
Reson Imaging 12:107-111, 2000.
183. Shellock FG, Crues JV: High-field MR imaging of metallic biomedical implants: an ex vivo evaluation of deflection forces.
Am J Roentgenol 151:389-392, 1988.
184. Shellock FG, Schatz C: High field strength MRI and otologic implants. Am J Neuroradiol 12:279-281, 1991. Medline
Similar articles
185. Yuh WTC, Hanigan MT, Nerad JA, et al: Extrusion of a magnetic eye implant after MR examination: a potential hazard to
the enucleated eye. J Magn Reson Imaging 1:711-713, 1991. Medline Similar articles
186. Shellock FG, Mink JH, Curtin S, Friedman MJ: MRI and orthopedic implants used for anterior cruciate ligament
reconstruction: assessment of ferromagnetism and artifacts. J Magn Reson Imaging 2:225-228, 1992. Medline
Similar articles
page 669
page 670
187. Shellock FG, Morisoli SM: Ex vivo evaluation of ferromagnetism and artifacts for cardiac occluders exposed to a 1.5
Tesla MR system. J Magn Reson Imaging 4:213-215, 1994.
188. Nogueira M, Shellock FG: Otologic bioimplants: ex vivo assessment of ferromagnetism and artifacts at 1.5 Tesla. Am J
Roentgenol 163:1472-1473, 1995.
189. Fagan LL, Shellock FG, Brenner RJ, Rothman B: Ex vivo evaluation of ferromagnetism, heating, and artifacts of breast
tissue expanders exposed to a 1.5 T MR system. J Magn Reson Imaging 5:614-616, 1995.
190. Shellock FG, Shellock VJ: Vascular access ports and catheters tested for ferromagnetism, heating, and artifacts
associated with MR imaging. Magn Reson Imaging 14:443-447, 1996. Medline Similar articles
191. Shellock FG: MR imaging and cervical fixation devices: evaluation of ferromagnetism, heating, and artifacts at 1.5 Tesla.
192. Shellock FG, Detrick MS, Brant-Zawadski M: MR-compatibility of Guglielmi detachable coils. Radiology 203:568-570,
193. Shellock FG, Shellock VJ: Evaluation of cranial flap fixation clamps for compatibility with MR imaging. Radiology
194. Shellock FG, Shellock VJ: Cardiovascular catheters and accessories: ex vivo testing of ferromagnetism, heating, and
artifacts associated with MRI. J Magn Reson Imaging 8:1338-1342, 1998. Medline Similar articles
7 of 10 15-04-2010 21:55
195. Shellock FG, Shellock VJ: Metallic marking clips used after stereotactic breast biopsy: ex vivo testing of ferromagnetism,
heating, and artifacts associated with MRI. Am J Roentgenol 172:1417-1419, 1999.
196. Becker RL, Norfray JF, Teitelbaum GP, et al: MR imaging in patients with intracranial aneurysm clips. Am J Roentgenol
9:885-889, 1988.
197. Brothers MF, Fox AJ, Lee DH, et al: MR imaging after surgery for vertebrobasilar aneurysm. Am J Neuroradiol
198. Brown MA, Carden JA, Coleman RE, et al: Magnetic field effects on surgical ligation clips. Magn Reson Imaging
199. Burtscher IM, Owman T, Romner B, et al: Aneurysm clip MR artifacts. titanium versus stainless steel and influence of
imaging parameters. Acta Radiol 39:70-76, 1998. Medline Similar articles
200. Dujovny M, Kossovsky N, Kossowsky R, et al: Aneurysm clip motion during magnetic resonance imaging: in vivo
experimental study with metallurgical factor analysis. Neurosurgery 17:543-548, 1985. Medline Similar articles
201. Kanal E, Shellock FG: MR imaging of patients with intracranial aneurysm clips. Radiology 187:612-614, 1993.
202. Kanal E, Shellock FG: Aneurysm clips: effects of long-term and multiple exposures to a 1.5 tesla MR system. Radiology
210:563-565, 1999.
203. Kanal E, Shellock FG, Lewin JS: Aneurysm clip testing for ferromagnetic properties: clip variability issues. Radiology
204. Klucznik RP, Carrier DA, Pyka R, Haid RW: Placement of a ferromagnetic intracerebral aneurysm clip in a magnetic field
with a fatal outcome. Radiology 187:855-856, 1993. Medline Similar articles
205. New PFJ, Rosen BR, Brady TJ, et al: Potential hazards and artifacts of ferromagnetic and non-ferromagnetic surgical and
dental materials and devices in nuclear magnetic resonance imaging. Radiology 147:139-148, 1983. Medline
Similar articles
206. Pride GL, Kowal J, Mendelsohn DB, et al: Safety of MR scanning in patients with non-ferromagnetic aneurysm clips. J
207. Shellock FG, Crues JV: Aneurysm clips: assessment of magnetic field interaction associated with a 0.2-T extremity MR
system. Radiology 208:407-409, 1998.
208. Shellock FG, Kanal E: Yasargil aneurysm clips: evaluation of interactions with a 1.5 tesla MR system. Radiology
207:587-591, 1998.
209. Romner B, Olsson M, Ljunggren B, et al: Magnetic resonance imaging and aneurysm clips. J Neurosurg 70:426-431,
210. Kato Y, Sano H, Katada K, et al: Effects of new titanium cerebral aneurysm clips on MRI and CT images. Minim Invasive
Neurosurg 39:82-85, 1996. Medline Similar articles
211. Lawton MT, Heiserman JE, Prendergast VC, et al: Titanium aneurysm clips: Part III. Clinical application in 16 patients with
subarachnoid hemorrhage. Neurosurgery 38:1170-1175, 1996.
212. Piepgras A, Guckel F, Weik T, Schmiedek P: Titanium aneurysm clips and their advantages in diagnostic imaging.
Radiologe 35:830-833, 1995. Medline Similar articles
213. Wichmann W, Von Ammon K, Fink U, et al: Aneurysm clips made of titanium: characteristics and artifacts in MR. Am J
214. Shellock FG, Shellock VJ: MR-compatibility evaluation of the Spetzler titanium aneurysm clip. Radiology 206:838-841,
215. Soulen RL: Magnetic resonance imaging of prosthetic heart valves (letter). Radiology 158:279, 1986. Medline
Similar articles
216. Soulen RL, Budinger TF, Higgins CB: Magnetic resonance imaging of prosthetic heart valves. Radiology 154:705-707,
217. Randall PA, Kohman LJ, Scalzetti EM, et al: Magnetic resonance imaging of prosthetic cardiac valves in vitro and in vivo.
Am J Cardiol 62:973-976, 1988. Medline Similar articles
218. Edwards M-B, Taylor KM, Shellock FG: Prosthetic heart valves: evaluation of magnetic field interactions, heating, and
artifacts at 1.5 tesla. J Magn Reson Imaging 12:363-369, 2000.
219. Frank H, Buxbaum P, Huber L, et al: In vitro behavior of mechanical heart valves in 1.5-T superconducting magnet. Eur J
Radiol 2:555-558, 1992.
220. Hassler M, Le Bas JF, Wolf JE, et al: Effects of magnetic fields used in MRI on 15 prosthetic heart valves. J Radiol
67:661-666, 1986.
221. Pruefer D, Kalden P, Schreiber W, et al: In vitro investigation of prosthetic heart valves in magnetic resonance imaging:
evaluation of potential hazards. J Heart Valve Dis 10:410-414, 2001. Medline Similar articles
222. Shellock FG: Prosthetic heart valves and annuloplasty rings: assessment of magnetic field interactions, heating, and
artifacts at 1.5 tesla. J Cardiovasc Magn Reson 3:159-169, 2001.
223. Shellock FG, Morisoli SM: Ex vivo evaluation of ferromagnetism, heating, and artifacts for heart valve prostheses exposed
8 of 10 15-04-2010 21:55
to a 1.5 tesla MR system. J Magn Reson Imaging 4:756-758, 1994.

224. Teitelbaum GP, Bradley WG, Klein BD: MR imaging artifacts, ferromagnetism, and magnetic torque of intravascular
filters, stents, and coils. Radiology 166:657-664, 1988. Medline Similar articles
225. Shellock FG, Shellock VJ: Stents: evaluation of MRI safety. Am J Roentgenol 173:543-546, 1999.
226. Teitelbaum GP, Ortega HV, Vinitski S, et al: Low artifact intravascular devices: MR imaging evaluation. Radiology
227. Teitelbaum GP, Raney M, Carvlin MJ, et al: Evaluation of ferromagnetism and magnetic resonance imaging artifacts of
the Strecker tantalum vascular stent. Cardiovasc Intervent Radiol 12:125-127, 1989. Medline Similar articles
228. Watanabe AT, Teitelbaum GP, Gomes AS, et al: MR imaging of the bird's nest filter. Radiology 177:578-579, 1990.
229. Hennemeyer CT, Wicklow K, Feinberg DA, Derdeyn CP: In vitro evaluation of platinum Guglielmi detachable coils at 3-T
with a porcine model: safety issues and artifacts. Radiology 219:732-737, 2001.
230. Hug J, Nagel E, Bornstedt A, et al: Coronary arterial stents: safety and artifacts during MR imaging. Radiology
231. Girard MJ, Hahn P, Saini S, et al: Wallstent metallic biliary endoprosthesis: MR imaging characteristics. Radiology
232. Kiproff PM, Deeb DL, Contractor FM, Khoury MB: Magnetic resonance characteristics of the LGM vena cava filter:
technical note. Cardiovasc Intervent Radiol 14:254-255, 1991. Medline Similar articles
233. Leibman CE, Messersmith RN, Levin DN, et al: MR imaging of inferior vena caval filter: safety and artifacts. Am J
Roentgenol 150:1174-1176, 1988.
235. Marshall MW, Teitelbaum GP, Kim HS, et al: Ferromagnetism and magnetic resonance artifacts of platinum embolization
microcoils. Cardiovasc Intervent Radiol 14:163-166, 1991. Medline Similar articles
236. Rutledge JM, Vick GW, Mullins CE, Grifka RG: Safety of magnetic resonance immediately following Palmaz stent implant:
a report of three cases. Catheter Cardiovasc Interv 53:519-523, 2001. Medline Similar articles
237. Buecker A, Neuerburg JM, Adam GB, et al: Real-time MR fluoroscopy for MR-guided iliac artery stent placement. J Magn
239. Taal BG, Muller SH, Boot H, Koop W: Potential risks and artifacts of magnetic resonance imaging of self-expandable
esophageal stents. Gastrointest Endosc 46:424-429, 1997. Medline Similar articles
240. Shellock FG: New metallic implant used for permanent female contraception: evaluation of MR safety. Am J Roentgenol
178:1513-1516, 2002.
241. Erlebacher JA, Cahill PT, Pannizzo F, Knowles RJR: Effect of magnetic resonance imaging on DDD pacemakers. Am J
Cardiol 57:437-440, 1986. Medline Similar articles
242. Hayes DL, Holmes DR, Gray JE: Effect of 1.5 Tesla nuclear magnetic resonance imaging scanner on implanted
permanent pacemakers. J Am Coll Cardiol 10:782-786, 1987.
243. Holmes DJ, Hayes DL, Gray JE, Merideth J: The effects of magnetic resonance imaging on implantable pulse generators.
Pacing Clin Electrophysio 9:360-370, 1986.
244. Zimmermann BH, Faul DD: Artifacts and hazards in NMR imaging due to metal implants and cardiac pacemakers. Diagn
Imaging Clin Med 53:53-56, 1984. Medline Similar articles
245. Achenbach S, Moshage W, Diem B, et al: Effects of magnetic resonance imaging on cardiac pacemakers and
electrodes. Am Heart J 134:467-473, 1997. Medline Similar articles
246. Peden CJ, Collins AG, Butson PC, et al: Induction of microcurrents in critically ill patients in magnetic resonance systems.
Crit Care Med 21:1923-1928, 1993. Medline Similar articles
247. Fontaine JM, Mohamed FB, Gottlieb C, et al: Rapid ventricular pacing in a pacemaker patient undergoing magnetic
resonance imaging. Pacing Clin Electrophysiol 21:1336-1339, 1998. Medline Similar articles
248. Fetter J, Aram G, Holmes DR, et al: The effects of nuclear magnetic resonance imagers on external and implantable pulse
generators. Pacing Clin Electrophysiol 7:720-727, 1984. Medline Similar articles
249. Vahlhaus C, Sommer T, Lewalter T, et al: Interference with cardiac pacemakers by magnetic resonance imaging: are
there irreversible changes at 0.5 Tesla? Pacing Clin Electrophysiol 24(Pt. I): 489-495, 2001.
250. Gimbel JR, Johnson D, Levine PA, Wilkoff BL: Safe performance of magnetic resonance imaging on five patients with
permanent cardiac pacemakers. Pacing Clin Electrophysiol 19:913-919, 1996. Medline Similar articles
251. Duru F, Luechinger R, Scheidegger MB, et al: Pacing in magnetic resonance imaging environment: clinical and technical
considerations on compatibility. Eur Heart J 22:113-124, 2001. Medline Similar articles
page 670
page 671
9 of 10 15-04-2010 21:55
252. Pavlicek W, Geisinger M, Castle L, et al: The effects of nuclear magnetic resonance on patients with cardiac
pacemakers. Radiology 147:149-153, 1983. Medline Similar articles
253. Luechinger R, Duru F, Scheidegger MB, et al: Force and torque effects of a 1.5 Tesla MRI scanner on cardiac
pacemakers and ICDs. Pacing Clin Electrophysiol 24:199-205, 2001.
254. Alagona P, Toole JC, Maniscalco BS, et al: Nuclear magnetic resonance imaging in a patient with a DDD pacemaker
(letter). Pacing Clin Electrophysiol 12:619, 1989. Medline Similar articles
255. Shellock FG, O'Neil M, Ivans V, et al: Cardiac pacemakers and implantable cardioverter defibrillators are unaffected by
operation of extremity MR imaging system. Am J Roentgenol 172:165-170, 1999.
256. Inbar S, Larson J, Burt T, et al: Case report: nuclear magnetic resonance imaging in a patient with a pacemaker. Am J
Med Sci 3:174-175, 1993.
257. Iberer F, Justich E, Stenzl W, et al: Nuclear magnetic resonance imaging of a patient with implanted transvenous
pacemaker. Herz 7:196-199, 1987.
258. Garcia-Bolao I, Albaladejo V, Benito A, et al: Magnetic resonance imaging in a patient with a dual chamber pacemaker.
Acta Cardiol 53:33-35, 1998. Medline Similar articles
259. Bhachu DS, Kanal E: Implantable pulse generators (pacemakers) and electrodes: safety in the magnetic resonance
imaging scanner environment. J Magn Reson Imaging 12:201-204, 2000. Medline Similar articles
260. Sommer T, Vahlhaus C, Lauck G, et al: MR imaging and cardiac pacemakers: in vitro evaluation and in vivo studies in 51
patients at 0.5 T. Radiology 215:869-879, 2000.
261. Gimbel JR: Implantable pacemaker and defibrillator safety in the MR environment: new thoughts for the new millennium. In:
2001 Syllabus, Special Cross-Specialty Categorical Course in Diagnostic Radiology: Practical MR Safety Considerations for
Physicians, Physicists, and Technologists. Oak Brook, IL: Radiological Society of North America, 2001, pp. 69-76.
262. Juratli N, Sparker J, Gimbel JR, Wilkoff B: Strategies for the safe performance of magnetic resonance imaging in
selected pacemaker patients (abstract). Circulation 104(17):638-639, 2001.
263. Gimbel JR: Letter to the editor. Pacing Clin Electrophysiol 26 (Part I):1, 2003.
264. Martin ET, Coman A, Willis O, et al: Magnetic resonance imaging and cardiac pacemaker safety at 1.5 tesla. J Am Coll
Cardiol 2004; 43:in press.
265. Limousin P, Krack P, Pollak P, et al: Electrical stimulation of the subthalamic nucleus in advanced Parkinson's disease. N
Engl J Med 339:1105-1111, 1998. Medline Similar articles
266. Rezai AR, Lozano AM, Crawley AP, et al: Thalamic stimulation and functional magnetic resonance imaging: localization of
cortical and subcortical activation with implanted electrodes. J Neurosurg 90:583-590, 1999. Medline Similar articles
267. diPierro CG, Francel PC, Jackson TR, et al: Optimizing accuracy in magnetic resonance imaging-guided stereotaxis: a
technique with validation based on the anterior commissure-posterior commissure line. J Neurosurg 90:94-100, 1999.
269. Rutledge JM, Vick GW, Mullins CE, Grifka RG: Safety of magnetic resonance immediately following Palmaz stent implant:
a report of three cases. Catheter Cardiovasc Interv 53:519-523, 2001. Medline Similar articles
271. Kanal E, Gillen J, Evans J, et al: Survey of reproductive health among female MR workers. Radiology 187:395-399, 1993.
© 2010 Elsevier
10 of 10 15-04-2010 21:55
HE AGNETIC ESONANCE MAGING ENTER

Marc Rothenberg
Advances in hardware and software technologies have brought magnetic resonance imaging from
experimental development to clinical use in hospitals to mainstream use in free-standing MRI centers.
The usefulness and accessibility of MRI in modern American medicine are now amply demonstrated
through a growing list of uses supported by appropriateness criteria, current procedural terminology
(CPT) codes and associated International Classification of Diseases, Ninth Revision (ICD-9) codes.
In the United States, procedure volumes have grown tremendously over the past decade, with an
average annual growth rate of 16.5% per year for the past 4 years,1 far surpassing population growth
2
trends, averaging approximately 1.3% per year, and shifts in demography that might otherwise
account for increased utilization per capita (Fig. 25-1). In 2002, there were an estimated 21.9 million
MRI procedures in the United States alone, of which 84% were performed on an outpatient basis.3 In
absolute terms, MRI procedure volumes, in the United States, have increased an average of 2.5 million
3
procedures per year during the past 4 years.
Add to lightbox
Figure 25-1 Total MRI procedures performed in the United States. (Reprinted with permission from
reference 1.)
Data stratified by procedure type demonstrate that MRI procedure volumes for brain, spine, and
extremities account for the large majority of the caseload, though abdominal, pelvic, vascular, head
and neck MRI procedures are represented at lesser volumes (Table 25-1). The growth in volumes for
procedures such as cardiac MR, MR spectroscopy, functional MR, and other advanced MR
procedures may blossom with continued technological refinement, scientific proof of the clinical
effectiveness of these procedures, and the development of appropriateness criteria and payment
policies from Medicare and other third-party payers.
The caseload for MRI procedures performed on an installed base of 6015 scanners distributed
throughout the United States in hospital and non-hospital settings is as shown in Table 25-2.3 Sixty-five
3
percent of existing fixed scanners have been installed since 1998 and 59% have a magnetic field
strength of 1.5 Tesla.3
Table 25-1. MRI Procedures Performed in 2002

Procedure type Procedures in 2002 (M) Percentage of total
Brain 5.9 27
Spine 5.8 26
Extremities 4.1 19
Head and neck 1.9 9
Vascular 1.9 9
1 of 11 15-04-2010 21:57
Abdominal and pelvic 1.4 6

Chest 0.6 3
Cardiac 0.1 0
Interventional 0.1 0
Miscellaneous 0.1 0
Total 21.9 100
Source: reference 1
Table 25-2. Caseload for MRI Procedures Performed on an Installed Base of

6015 Scanners
Site Number
Hospital 2370
Non-hospital 2245
Mobile 1400
Total 6015
page 672
page 673
Source: reference 1
State-by-state population variances, demography, payer policies, and Certificate of Need (CON)
regulations, among others factors, all impact MRI utilization and the proliferation of MR facilities.
However, the market for MRI procedures is growing. MRI utilization among Medicare beneficiaries has
shown a growth rate of 15% to 20% per year from 1999 to 20024 and overall physician services
4
utilization among these patients is increasing.
As the demand for MRI continues to increase, competition for technical component revenues will
emanate from many sources. Radiologists, non-radiologist physicians, hospitals, and non-physician
investors all will seek to exploit the potentially lucrative opportunities offered through the ownership of
MRI centers. For some, barriers to entry may diminish the opportunity to establish an MRI center, so
each barrier must be adequately addressed during the project planning. Two significant challenges are
the CON laws extant in approximately half of the United States, which regulate and sometimes limit the
development of MRI centers, and the financing necessary to capitalize or secure the acquisition of a
magnet and MRI center. But these barriers may also serve to protect the MRI center from prospective
competition.
Nonetheless, with a large capital expenditure and high operating leverage due to fixed expenses, the
financial feasibility of the MRI center is highly dependent on the revenue side, relying on procedure
volumes, payer mix, and payment levels. While the outlook for increased utilization appears favorable,
increasingly aggressive competition can erode market share, and trends in payer mix and payment
5
levels are less positive.
Therefore, in the face of existing or future, and often well-resourced, competition, radiologists seeking
6
entry to or expansion of ownership of an MRI center cannot rely on a Field of Dreams approach to
planning as a formula for success. On the contrary, whether by acquisition or construction, the financial
magnitude of such an undertaking requires that the core concept of a business plan be utilized to
organize a critical analysis of the MRI market and demand for services, strategic positioning of the
MRI center, operational assumptions, and financial analysis.
Marketing is a cornerstone of a comprehensive business plan, yet is sometimes misunderstood in its
2 of 11 15-04-2010 21:57
overall critical importance to the success of the MRI center. Marketing begins with research in order to
identify opportunities for the MRI center in a community and among patient populations. Detailed
market research delineates unmet need for MRI services through a comparison of the demand for MRI
procedures versus the supply of MRI services by existing providers within the geographic service area.
Detailed research can further stratify the data pertaining to MRI services into more specific categories,
known as market segments, based on certain characteristics such as magnet type and field strength,
site type and location, procedure type, population demographics, etc. Data sources may include
population demographics and utilization estimates available through the state or proprietary sources,
statistical data collected by state regulatory agencies and equipment vendors, curbside consults with
referring physicians, and information provided by existing MRI providers in their newsletters, public
statements, and websites.
From the market research, a radiologist may better distinguish the potentially more effective strategies
and tactics for the competitive positioning of the MRI center in the market, and segue into the
operational requirements and financial consequences of the strategy and tactics undertaken. These
business strategies can be divided into three general categories: high quality; low cost, and service
differentiation.7
The essence of a strategy promoting high quality as the key factor for an MRI center is that the quality
of care and service is superior to that of the competitors. This message requires an understanding of
how the referring physicians and patients determine quality, as compared to the radiologist. Referring
physicians and patients may focus more on service issues, such as scheduling, friendliness and report
turnaround, and less on the technical factors affecting image quality. That is not to say that technical
quality is not important, for it certainly is, but that appreciating the differences in technical quality may
not be sufficient for market positioning. For example, while a hospital-based 3 T scanner with better
signal-to-noise ratio produces superior images, it may be less marketable compared to an MRI center
with a 1.5 T platform and keen attention to customer service.
Nonetheless, an MRI center that wants to convey a market position of high quality as its approach to
attracting patient referrals may consequently expect to incur higher costs in order to achieve and
sustain those levels of quality and service, whether in capital equipment for state-of-the-art magnets,
coils, gradients, and software applications or for operating expenses such as an abundance of clerical
staff, MRI-certified technologists, and fellowship-trained radiologists.
One alternative to the strategy of high quality is low cost. In healthcare, however, much of the demand
for MRI services is less elastic relative to price, since patients frequently seek services at the location
requested by their physician, and third party payers fund much of the patient care, except for
co-payments and deductibles. A low-cost strategy has potential to yield some success in markets with
significant managed care penetration, where competitively negotiated fee contracts may translate into
patient volumes if the managed care entity employs steerage mechanisms favorable to the MRI center.
Service differentiation denotes a strategy whereby a center's services are unlike those of the
competition. A center's differentiation can be based on technology, such as open architecture in an
otherwise high-field market, a niche such as a vertical "standing" magnet, an extremity magnet for
orthopedic and rheumatology patients, or evening and weekend operating hours in an otherwise
nine-to-five world. Whatever the service differentiation, the MRI center must ensure either that the
projected demand for these services will yield a financially viable entity or that the center retains
enough flexibility in its equipment and services to alter its strategy and market to a different or more
general patient population.
page 673
page 674
The above strategies are not necessarily mutually exclusive. It is possible to employ more than one
strategy simultaneously, such as a high-quality, differentiated service or a low-cost, differentiated
service. However, the resources involved to ensure the success of one selected approach may
preclude or simply overshadow the realization or effectiveness of a second approach. Moreover, the
use of two approaches simultaneously may actually reduce the otherwise optimal performance of each
3 of 11 15-04-2010 21:57
of the strategies independently. Whichever strategy is undertaken, it must be emphasized that success
can be undermined by inattention to the expectations of the customers-referring physicians, their office
staffs, patients, and even third party payers.
Once a strategy has been selected, MRI center planning can be further delineated based on the "four
Ps" of marketing: product, place, price, and promotion.8 Contemplation of each of these elements will
help to synthesize the operational and financial assumptions used in the development of the financial
projections and analyses for the MRI center.
The product of an MRI center, more accurately a service, is guided by the market research to identify
and understand the reasons resulting in an unmet need for MRI procedures. Clearly, the product of an
MRI center is imaging based on a magnet type and field strength, but the features and manner in
which that imaging is offered and delivered can further characterize the product so as to accommodate
the unmet need. The MRI center should continuously re-examine the market in order to refine or
re-develop its services to accommodate the changing environment and changing need over time, as
strategy and marketing are necessarily evolving in order to maintain a competitive position in the
community.
The place of service connotes the channels for the referral of MRI patients, specifically the referring
physician offices, where awareness of the MRI center and the need for MRI services originate, and
managed care entities, where participating provider directories help to channel patients to the MRI
center.
Place also encompasses the geographic location of the MRI center relative to the channels of referral.
The MRI center is best served by a location with easy access from major thoroughfares, sufficient
parking, and adequate lighting. The center should be near the referring physicians and patients, such
as in a medical office park or medical office building, so as to maximize the convenience of referring
patients to the center. As legendary bank robber Willie Sutton might have said, location is important
"because it's where the patients are".9
Within a building, an MRI center is best situated on the ground floor, for higher visibility for all entering
and exiting the building, as well as ease of patient access and equipment upgrade. Moreover, siting a
magnet on the ground floor minimizes construction and rigging costs due to the size and weight of the
magnet, and allows for floating floors or any other construction design modifications due to ambient
vibrations or other interference, as may be necessary with some magnets.
Price, to the extent that it connotes a participating provider agreement and negotiated fee schedule
between the MRI center and managed care organizations, influences the target market to seek
services at the MRI center. More and more, patients and referring physicians are preferentially
constrained to using in-network MRI centers to achieve the payer's cost containment goals, and by
using in-network providers, the payer retains responsibility for the MRI center fees. Thus, participating
provider agreements will yield patient volume. Some even offer enhanced steerage of patients based
on more deeply discounted negotiated fee schedules. Nonetheless, some patients may seek MRI
service at the center despite its status as a non-participating provider. For these and also for self-pay
patients, an appropriately set price for MRI services may be a factor in their decision.
Promotion refers to means by which the MRI center communicates its presence and product to the
target market. It may encompass a wide range of activities, including print, radio and television
advertising, patient brochures situated at referring physician offices, participation in and sponsorship of
community activities, embossed mugs, magnets and other trinkets, and the important face-to-face
interaction of personal selling.
Personal selling is intrinsic to every interaction between the patient, MRI center, and referring physician
office. MRI center staff must appreciate that each interaction is an opportunity for them to convey to
each constituency a trust and confidence through clear, courteous, and constructive communication
about the center's potential to meet patients' imaging needs. Moreover, radiologists and designated
4 of 11 15-04-2010 21:57
staff should regularly visit referring physician offices to convey tailored information about MRI center
services and capabilities and receive immediate feedback about those physicians' experiences with the
MRI center's service to their patients. Such personal selling also creates a positive atmosphere in
which to express appreciation for patient referrals to the MRI center.
Whether with referring physician offices or influential leaders in the community, personal selling can be
an effective channel for directing goodwill toward the MRI center. What may appear as idle chats
about family, birthdays and vacations, later recorded in a notebook and later remembered, can result
in a personal connection. This level of personal selling, done well, creates a bond between the
referring physician office and the MRI center, which is then more likely to provide a stream of patient
referrals, as well as information about approaching changes in the market from the perspective of the
referring physician.
page 674
page 675
While gifts to referring physician offices are common, it is imperative to abide by the constraints of
federal and any applicable state laws pertaining to financial interest and inducements to refer. The
federal Ethics in Patient Referrals Act, more commonly known as the Stark legislation, prohibits gifts
10
totaling more than $300 per year. Moreover, the Health Insurance Portability and Accountability Act
prohibits giving gifts to Medicare and Medicaid beneficiaries of more than $10 per gift or $50
annually.11 Also, while the Stark laws do permit "per-click" and "time-leasing" financial arrangements12
between referring physicians and MRI centers, the terms of these financial arrangements, as well as
the charge and claims filing procedures, may be limited by federal and state laws, as well as third
party payer policies. The MRI center should be wary of such arrangements, since even when only
commercially insured patients are referred via one of these financial arrangements, their existence may
be viewed as an inducement to refer patients with federally funded health insurance, thereby potentially
13
invoking the federal Anti-Kickback statute.
In any case, the personal connections established through marketing can serve as an information
channel to garner useful information from these numerous sources. It allows the MRI center to be
proactive, not just reactive, in its market orientation. To the extent that the MRI center's resources
allow, strategies can be re-evaluated periodically as the characteristics of the community change over
time. A continuous multi-pronged marketing effort can assist the MRI center in monitoring those
characteristics as they develop, not only within the changing mix of the center's referring physicians,
patients, and services but in the external market of physicians who are not referring to the MRI center,
patients whose needs are not being served, potential patients who could be served by the center, and
changes in competition and technology.
The infrastructure of the MRI center can be critical to its overall success and must be carefully
evaluated and continually re-evaluated. Proper staffing is necessary for the projected clinical caseload
and customer service expectations of the market, but cross-training of duties can alleviate some
excess capacity. The number and mix of support staff depend on the number of assignments, time
commitment for each task, and capacity of each position to fulfill its duties within its scheduled day. A
capacity analysis can assist in determining the most appropriate staffing levels within the framework of
assumptions used.
Two technologists may be deemed adequate to support a newly developed MRI center, image
acquisition and processing, and patient handling activities, while providing enough coverage for
vacation and other leaves of absence. Other staff provide administrative support for patient scheduling,
registration, and screening. Billing and transcription may be performed on site or outsourced as a
variable cost to the MRI center. Cost-benefit analyses may be applied to decisions, from the number
and mix of staffing, outsourcing versus insourcing of services such as transcription billing,
housekeeping, the use of film or PACS, to the purchase of each signal-focusing coil.
For example, the number and type of support staff can comprise a large portion of the annual
operating budget of the MRI center. However, to counterbalance such concerns, it is important to
recognize the incremental revenue for the additional patient throughput that can be achieved through
5 of 11 15-04-2010 21:57
increased efficiencies, enhanced productivity, and expanded target markets. An MRI center may
employ mid-level providers in addition to its complement of technologists. Although not trained to
operate the magnet, the mid-level providers can assist in all aspects of patient handling but, most
importantly, reducing anxiety and administering and monitoring sedation of certain patients. By having a
mid-level provider on staff, the MRI center may be less likely to lose patients to claustrophobia and, in
fact, may even increase its referrals of claustrophobic and pediatric patients who require special care.
Similarly, the MRI center can exploit technology to add value and must evaluate the financial costs
versus reward for each of its technological decisions. For example, while a picture archiving and
communication system (PACS) may represent a significant capital expenditure, the benefits are
widespread, including cost containment, efficiency, and marketing. PACS eliminates the internal use of
film and its associated expenses, though a dry-laser and film may still be necessary for some patients
and referring physicians, who may want copies from time to time. In addition, PACS eliminates the
annual salary and benefits costs related to film hanging, storage and records retrieval, multi-panel
viewer expenditures, repair and maintenance, and physical space for film storage. In terms of
efficiency, the radiologist can retrieve soft-copy comparison films for display on the PACS monitors
within seconds. Moreover, with an external node using secure servers and connections, radiologists
can interface with the PACS from remote locations to review images and provide services as though
they were on site. Alternatively, they can use such a configuration node to transmit images to
colleagues for subspecialty opinions, or expanded as a marketing tool to bind referring physicians to
the MRI center by allowing them in-office access to their patients' images.
A breakeven analysis can be performed on staffing, technology such as a coil or software package,
the impact operational decisions, and even the MRI center as a whole, in order to quantify financial
feasibility. The analysis calculates the procedure volume necessary for the return of the investment as
described by the incremental revenues achieved due to the investment versus the incremental fixed
costs and variable operating costs of the investment.
To illustrate, a radiologist may pose the question of whether or not the MRI center should purchase a
breast coil. Assume that a MRI center with excess capacity contemplates the purchase of a breast coil
that costs $26,000; average payment per procedure, based on service mix and payer mix, is $800;
and the variable cost of supplies is $25 per procedure. With those assumptions, the breakeven is
calculated as follows:
Thirty-four breast MRI procedures are necessary for the MRI center to recoup its investment in a
breast coil. From a marketing perspective, the center can now evaluate whether or not the demand
exists to justify the purchase of a coil.
page 675
page 676
Table 25-3. Stratification of the Projected Payer Mix

Pay class Percentage of patients Revenue per scan (dollars)
Medicare 30 500
Medicaid 10 450
Commercial 10 1000
Blue Cross Blue Shield 10 750
Managed care 30 500
Self pay 10 1000
Weighted-average revenue 100 620
6 of 11 15-04-2010 21:57
The development of the market research, equipment selection, and operational assumptions
culminates into the composition of a pro forma income statement for the MRI center, typically
projected for 3-5 years into the future. The pro forma income statement serves as a guide and budget
for the anticipated financial performance of the MRI center. It compels a radiologist to commit to a
series of assumptions and then to test the reasonableness of those assumptions and the impact on
profitability. Moreover, the statement serves as a basis for further financial analysis as to the
investment value of the MRI center as compared to alternative investments.
The process may begin with projections of the expected payer mix of the patients to be served by the
MRI center. Payer mix projections may be based on a variety of sources including, but not limited to,
historical data from the center or affiliated imaging services or data from other providers such as
referring physicians or a nearby hospital. The payer mix may be stratified into a few major payer
classifications, and the revenue from each payer classification can be ascertained or estimated based
on existing or intended charges and negotiated fee schedules. A weighted-average revenue per scan
can then be calculated for use in the pro forma income statement (Table 25-3).
Reasonable assumptions must also be made about revenues and expenses, and the effect of changes
in those assumptions on the bottom line over the short term. The use of conservative, yet realistic,
data is generally advisable, relying on lower volumes and revenue, and higher expenses. This method
helps to temper overly optimistic forecasts about MRI center financial performance.
Table 25-4. Assumptions

Average patients per day 8
Days per year 250
Growth rate per year 5%
Revenue per patient 620
Revenue change per year -1%
Bad debt 3%
Equipment cost 1,750,000
Depreciation method Accelerated, mid-Q1 convention
Depreciation period in years 7
Maintenance agreement 8%
Supplies expense per patient 10
Employees 4
Average salary per employee 33,750
Salary increase per year 4%
Benefits rate 20%
Assumptions may be fixed or vary year by year for each year of the projection. No matter which
method is used, assumptions should correspond to the local prevailing market related to each
assumption. Moreover, as the financial outcomes are highly sensitive to the assumptions, it may be
advisable to perform a sensitivity analysis, varying the assumptions from less conservative to more
conservative projections about future volumes, revenues and expenses, in order to discern the range of
possible financial outcomes (Table 25-4).
The construction of the pro forma income statement relies on a synthesis of the various data,
operational and financial assumptions collated during the market research and planning. Table 25-5
reflects a line-item analysis, with certain of the key revenue assumptions and statistics readily
7 of 11 15-04-2010 21:57
viewable.
Table 25-5. Pro Forma Income Statement: Tax Basis of Accounting

Year 1 Year 2 Year 3 Year 4 Year 5
Revenues
Average patients per day 8.0 8.4 8.8 9.3 9.7
Days per year 250 250 250 250 250
Total patients 2000 2100 2205 2315 2431
Average fee per patient 620 614 608 602 596
Total patient revenue 1,240,000 1,288,980 1,339,895 1,392,821 1,447,837
Bad debt 37,200 38,669 40,197 41,785 43,435
Net patient revenue 1,202,800 1,250,311 1,299,698 1,351,036 1,404,402
Expenses
Radiologist salary 300,000 300,000 300,000 300,000 300,000
Radiologist benefits 75,000 75,000 75,000 75,000 75,000
Support staff salaries 135,000 140,400 146,016 151,857 157,931
Support staff benefits 27,000 28,080 29,203 30,371 31,586
Depreciation 437,500 375,025 267,925 191,275 153,125
Supplies 20,000 21,000 22,050 23,153 24,310
Maintenance agreement 0 140,000 140,000 140,000 140,000
Rent 60,000 60,000 60,000 60,000 60,000
Utilities 12,000 12,000 12,000 12,000 12,000
Total expenses 1,066,500 1,151,705 1,052,616 984,323 954,892
Net income/(loss) 136,300 110,513 271,956 405,682 503,780
page 676
page 677
Based on the projected revenues and expenses delineated in the financial pro forma income statement,
several methodologies can be used to quantitatively determine the anticipated financial performance of
the MRI center as more than a source of employment, but as an investment. These methodologies can
help radiologists to develop reasonable expectations as owners for the timeframe and magnitude of
the return on investment in the MRI center, and can be used in concert to optimize decision making as
to investment expectations.
Parenthetically, it is important to note that the pro forma income statement above assumes a purchase
of the magnet. Depreciation is a non-cash expense that represents the gradual devaluation of that
medical equipment, using one of several methodologies, during a prescribed timeframe. Yet a
purchase is only one of several financing alternatives, with leasing being another popular financing
vehicle. The decision to buy versus lease is contingent on several factors, including the availability of
monies to purchase, interest rates, other uses for the available monies, lender-mandated debt ratios,
and the projected useful life and obsolescence of the magnet.14
As demonstrated previously, a breakeven analysis can be performed to determine the volume of

procedures at which the overall investment in the MRI center is recovered, and beyond which a
contribution to the overheads of a larger organization or profits are generated.
The payback period is another tool to estimate the amount of time required for the MRI center to
8 of 11 15-04-2010 21:57
recover its initial investment in capital equipment through cashflows, that being the cumulative profit or
loss, with non-cash expenses such as depreciation or amortization added back. The payback period
can be calculated using the formula:
However, this formula works best only under circumstances of stable cash flow. In growing or less
stable circumstances, the payback period is better reflected in a table format (Table 25-6).
Table 25-6. Payback Period Analysis-Uneven Cashflows

Year Net income Accumulated cashflow
0 -1,750,000 -1,750,000
1 136,300 -1,176,200
2 110,513 -690,662
3 271,956 -150,781
4 405,682 446,176
5 503,780 1,103,082
This payback period analysis demonstrates that the MRI center will recover the cost of its initial
investment in equipment some time during the fourth year of operation. Nonetheless, while the payback
period analysis can provide information as to the timeframe for the recovery of the initial investment, it
is limited in several other respects. The payback period does not account for the time value of money.
In order words, it does not reflect the net present value of cashflows received at a future time when
comparing those cashflows to money used at the present time. For example, the present value of
$100,000 next year, with a 7% interest rate, can be calculated as:
Moreover, the payback period does not offer benchmark data to compare the investment performance
of the MRI center to other alternative uses of money.
A net present value (NPV) analysis provides a firmer basis for determining the comparative economic
potential of a project. The NPV hones the investment decision by reflecting the financial potential in
dollars comparable to each other in time, and is calculated as the annual cashflows (profit with
non-cash expenses added back) reduced to the present by the time value of money, plus the negative
cashflow of the initial investment. An NPV greater than zero suggests that the value of the investment
is greater than the initial cash outlay in the context of a given rate of return.
The NPV of the MRI center is calculated by the following formula:
where "I" is the interest rate.
The NPV for the MRI center is given in Table 25-7. Therefore, the NPV equals:
Since the NPV is greater than zero, the MRI center would be among the candidates for investment for
a radiologist seeking an investment return of greater than 7%.
9 of 11 15-04-2010 21:57
Table 25-7. NPV of the MRI Center

Year Cashflow Interest rate Present value
1 573,800 7% 536,262
2 485,538 7% 424,088
3 539,881 7% 440,703
4 596,957 7% 455,416
5 656,905 7% 468,364
Total 2,324,834
page 677
page 678
Table 25-8. Calculation of the IRR

Year Cashflow IRR Present value
1 573,800 18% 485,448
2 485,538 18% 347,527
3 539,881 18% 326,923
4 596,957 18% 305,825
5 656,905 18% 284,718
Total 1,750,442
Lastly, in order to more precisely quantify the rate of return provided by an investment in the MRI
center, an internal rate of return (IRR) analysis can be used more definitively. IRR is the rate of return
that equates the present value of the cash inflow to the present value of the cash outflow. The IRR
analysis measures the actual economic return earned by the MRI center, subject to the assumptions
used in the pro forma income statement. The IRR calculation uses the same variable as the NPV
calculation but derives the interest rate by setting the NPV at zero in order to derive the actual IRR
yielded by the MRI center investment (Table 25-8):
Subtract the initial investment from both sides of the equation:
At an IRR of 18.2%, the present value of future cashflows equals the initial investment. With this
information, the radiologist can now compare an investment in the MRI center to other investment
opportunities.
The application of business concepts to the practice of radiology can provide valuable tools in the
planning, implementation, analysis, and reorientation of the MRI center toward success. As the ancient
Chinese warrior Sun Tzu said: "The general who wins the battle makes many calculations in his temple
before the battle is fought…Thus do many calculations lead to victory and few calculations to
defeat".15
A wealth of knowledge exists within the archives of radiology and management books, journals and
newsletters, among experienced radiologists and management executives, and in the seminars and
10 of 11 15-04-2010 21:57
conferences offered to the radiology community. This information and infrastructure, available to assist
radiologists in contemplating investment opportunities in MRI centers, must be utilized in today's
market in order to retain control over the practice of radiology and ultimately to achieve success.
REFERENCES
1. MRI Benchmark Report 2002-3. Des Plaines, IL: IMV Medical Information Division Inc, 2003, p 3.
2. Hobbs F, Stoops N: Demographic Trends in the 20th Century. Census 2000 Special Reports. Washington, DC: US Census
Bureau, US Department of Commerce, 2002.
3. MRI Benchmark Report 2002-3. Des Plaines, IL: IMV Medical Information Division Inc, 2003, pp 2-4.
4. Variation and Innovation in Medicine. Washington, DC: Medicare Payment Advisory Commission, 2003, pp 62-63.
5. Drop in doctor payments foreseen without action. Mod Healthcare 34(23): 35, 2004.
6. "If you build it, he will come." Field of Dreams, Universal Studios, 1989.
7. Porter ME: Competitive Strategy. New York: Macmillan Publishing. 1980, pp 35-40.
8. McCarthy EJ, Perreault WD: Basic Marketing: A Managerial Approach, 10th ed. Homewood, IL: Richard Irwin Inc, 1990, pp
36-39.
9. Sutton W: Where the Money Was. New York: Broadway Books, 2004, pp 159-161.
10. HHS News: HHS Issues Final Rule Addressing Physician Self-Referrals. Washington, DC: US Department of Health and
Human Services, 2001.
11. Wieland JB, Kass JE: Inducements to beneficiaries: when good marketing might violate the law. RBMA Bulletin May/June:
29-30, 2003.
12. Special Report: Analysis of Stark II Final Rules. Health Law Briefs. Chicago: Jenner & Block LLC, 2001.
13. Haule AM: Per-click and time-share lease arrangements-too good to be true? Diagnostic Imaging Intelligence Reports
3(2): 5-7, 2004.
14. Koche HS, Romano RL: To own or not to own? RBMA Bulletin April: 10, 21, 2001.
15. Clavell J (ed): The Art of War by Sun Tzu. New York: Delacorte Press, 1983, p 11.
© 2010 Elsevier
11 of 11 15-04-2010 21:57
EASURING THE APACITY RODUCTIVITY AND OSTS

OF ERVICE OF AN ENTER
Abraham Seidmann
Frank J. Lexa
Tushar Mehta
In this chapter, we look at how a manager can better understand the costs of the services provided at
a magnetic resonance imaging (MRI) center, paying particular attention to the dilemma of unused
capacity and the implications for productivity improvements that result from management decisions and
initiatives. We will move beyond simple ledger accounting by extending the principles of activity-based
costing (ABC) to address the unique challenges of radiology-specifically, MRI. In doing so, we develop
what we call the service activity costing system, or SAC. With it, the imaging manager can intelligently
allocate all the costs incurred by the center to the services provided and improve management
decision-making. SAC also provides deeper insights into the exact cost to a center to not provide MR
imaging services for which it has the capacity, i.e., the cost of unused capacity. This accounting
concept is frequently misunderstood and often misapplied, particularly in the medical field.
Radiology directors should use financial information to make better decisions about current and future
processes, resources, and contracts, not merely to reflect on the past. SAC, especially when used
proactively, overcomes some of the limitations of more traditional financial systems. This cost
information can also serve as a basis for pricing and contract negotiation. Once we understand the
relationships between capacity, flexible resources, and productivity improvement initiatives, it will be
easier to see how a radiology director can leverage such initiatives to create new business
opportunities, since the costs of many resources are "fixed" only if the manager cannot, or will not,
exploit the unused capacity he or she helped to create.
page 679
page 680
In the next section, we introduce the concepts of cost accounting in healthcare. In a nutshell, we
present the need for a new methodology that connects the services provided by a radiology center to
the costs of the resources required to produce those services. Next, we see how it leads to an SAC
system and how it helps the center to price the capacity available to it. The center then can also attach
a price to unused capacity. Accurate costing data allow managers to avoid some of the common
pitfalls in cost accounting, notably the "death spiral," which is discussed in the Appendix to this chapter.
Finally, we examine how information about unused capacity can help radiology managers focus their
efforts on those productivity improvements that will be most effective.
© 2010 Elsevier
1 of 1 15-04-2010 21:58
COST ACCOUNTING IN HEALTHCARE

Costing systems, when properly specified, can reduce treatment cost distortion, increase
cost-effectiveness, and improve decision-making in healthcare. With the wrong assumptions, of course,
per treatment costs and profitability information become distorted and can lead to suboptimal
operating decisions. Decision-makers may incorrectly reduce or eliminate apparently unprofitable
activities and procedures in the name of cost containment. Moreover, as providers increase their
commitments to managed care, managers will expect their costing systems to highlight opportunities
for cost containment. Financial survival will require that executives cost more accurately under
managed care. If they bid too high, the contract is lost. If they bid too low, the contract is unprofitable.
When properly specified, our newly developed SAC methodology enables better and more profitable
decisions.
Capettini et al1 refer to several articles that have recently advocated the use of cost accounting
systems by service organizations in general and healthcare organizations in particular. They investigate
how the non-volume-based costs are distributed over the various categories and demonstrate how
activity-based allocations can yield less biased (more accurate) cost estimates than traditional
volume-based allocations. Their findings from a questionnaire distributed to department managers at
three large metropolitan hospitals strongly suggest that there are limitations to the hospital costing
systems currently used and there is a need to seriously consider installing better systems in hospitals.
Ruhl and Hartman2 have a thorough discussion of hospital cost accounting systems and review
examples of hospitals that have successfully implemented ABC systems. Holt3 provides an overview of
the approach taken in the assessment and implementation of activity-based management for the Army
Medical Department (AMEDD).
Most healthcare costing systems are reimbursement driven. The most common healthcare costing
systems in use today are ratio of cost to charges (RCC) and relative value units (RVUs). Although
RCC is used by a majority of provider organizations, a case study by West et al4 that applied these
5
three methods in a renal dialysis clinic found that ABC provided the most accurate cost data. Baker
gives the comparison shown in Table 26-1.
© 2010 Elsevier
1 of 1 15-04-2010 22:01
COMMONLY USED COSTING SYSTEMS IN HEALTHCARE

In a series of articles and books, Cooper and Kaplan6,7 have developed a new perspective on
accounting and budgeting, referred to as activity-based costing and activity-based budgeting. This
work has generated a revolution in the business world in terms of the way cost information is collected
and used.
Table 26-1. Comparison of Healthcare Costing Systems

Advantages Disadvantages
Ratio of cost to Easy method Calculation tied to revenue, forcing the
charges (RCC) Configuration is the same assumption that revenue proportions
as Medicare cost reporting accurately reflect resource consumption
ratios Aggressive reimbursement maximizing (such
Familiarity over time as "grossing up" techniques) increases
(especially for financial revenue amounts, skewing the ratio
managers who have cost No cost containment emphasis
reimbursement experience)
Relative value Recognizes resources Assumes that every RVU consumes exactly
units (RVUs) consumed in delivery of a the same set of resources, in a proportion
service that always remains exactly proportionate
Service-level cost is (this major weakness is not recognized by
determined from a clinical many managers who rely on RVUs for
base instead of a costing purposes)
reimbursement base
Presents a methodology for
the cost of acquiring
resources
Activity-based Resources consumed at The newest of the three methods and
costing (ABC) the treatment level are therefore not yet as well known
more precisely defined and Some members of management may not
reflected want more precise costs to become known
Resources consumed by
the particular cost object
(or cost objective) are more
directly tracked and
identified
page 680
page 681
The authors' approach was motivated by a belief that traditional "general ledger" accounting
information is all but useless to managers who are interested in evaluating the effectiveness of
resource allocation decisions in their companies. This traditional information is geared instead toward
satisfying auditors or other outsiders who are interested in some evidence of financial accountability.
According to Cooper and Kaplan,6 one of the most serious problems lies in the traditional allocation of
overhead costs. Over time, as production processes have become more and more complex, a greater
proportion of total production costs are described as "overhead" and are arbitrarily allocated to output.
The authors suggest that many of these "overhead costs" (e.g., costs of logistics, production,
marketing, sales, distribution, service, technology, financial administration, information resources, and
general administration) can, in fact, be traced to individual products or product groups. Certain
activities and processes consume a disproportionate amount of these activities. Cooper and Kaplan
argue that the misallocation of overhead costs can generate tremendous distortions in production cost
estimates. Specifically, traditional costing strategies tend to attribute too much overhead to less
1 of 2 15-04-2010 22:01
complex products and products produced in high volume. Conversely, they seriously underestimate
low-volume, complex products and services. Because this cost information is often used to evaluate
the profitability of different production strategies, the misallocation of costs can lead managers to
make poor decisions. Cokins, Stratton, and Helbling8 provide a useful implementation-focused overview
of ABC and discuss these trade-offs. For example, organizations might want to focus on particularly
expensive resources, on resources whose consumption varies by product, or on resources whose
demand patterns are not correlated with the traditional allocation measures.
Because activity-based accounting systems are more complex and costly than traditional systems, not
all companies use them. A 2003 empirical study by Kiani, Raj and M Sangeladji9 of The Fortune 500
Largest Industrial Corporations in the USA shows that 51% of the respondents do use an
activity-based costing and management system, but only a few have used them for more than 5 years.
This calls for further empirical and research studies to evaluate the degree of their usefulness. Still,
more and more organizations in both manufacturing and nonmanufacturing industries are adopting
activity-based systems for a variety of reasons, as documented by Horngren, Sundem, and Stratton10:
Fierce competitive pressure has resulted in shrinking margins. Companies may know their
overall margin, but they often do not believe in the accuracy of the margins for individual
products or services.
Business complexity has increased, which results in greater diversity in the types of products
and services as well as customer classes. Therefore, the consumption of a company's shared
resources also varies substantially across products and customers.
New production techniques have increased the proportion of indirect costs-that is, indirect costs
are far more important in today's world-class manufacturing environment. In many industries
direct labor is being replaced by automated equipment. Indirect costs are sometimes over 50%
of the total cost.
The rapid pace of technological change has shortened product life-cycles. Hence, companies do
not have time to make price or cost adjustments once errors are discovered.
Computer technology has reduced the costs of developing and operating cost systems that
track many activities.
© 2010 Elsevier
2 of 2 15-04-2010 22:01
THE CONCEPT OF ACTIVITY-BASED COSTING

Activity-based costing has two major elements-cost measures and performance measures. It is a
methodology that measures the cost and performance of activities, resources, and cost objects. It
recognizes the causal relationships of cost drivers to activities. The basic concept of ABC is that
activities consume resources to produce an output. Expenses should be separated and matched to the
level of activity that consumes the resources. This separation should be independent of how many units
are produced and sold. The ABC approach differs from the traditional approach because of its
fundamental concentration on activities: it uses both financial and nonfinancial variables as bases for
cost allocation, a greater number of cost drivers as cost allocation bases, and more indirect cost
11
pools. As Player has shown, ABC uses a multistage process to convert the traditional view of cost
information (dollars by resource required) into an activity-based view (dollars by activity performed).
Terms applicable to the ABC process include the following:
Activities-the work done in the organization, such as performing X-rays, administering

medication, reviewing test results, and taking patient information.
Resources-the financial and operational inputs required to perform activities. Resources
coincide with traditional cost pools, such as salaries, medical supplies, and depreciation, and
portions of these various resources are consumed by each activity.
Resource drivers-measures of the quantity of resources consumed in performing each activity.
Cost object-anything that requires a separate cost measurement, such as a customer, product,
or service line. If, for example, the goal of the ABC analysis is to evaluate service profitability,
cost objects would be services or service lines, whereas if the goal is to evaluate payer
profitability, the cost objects would be specific payers.
Activity drivers-measures of the frequency and intensity of the demand placed on an activity
by a cost object.
© 2010 Elsevier
1 of 1 15-04-2010 22:01
ACTIVITY-BASED COSTING IN HEALTHCARE

page 681
page 682
Add to lightbox
Figure 26-1 Basics of a radiology center: a cost driver overview.
In the early 1980s, ABC received a warm welcome from industrial companies in the United States. The
manufacture of products was a natural application for ABC. In the early 1990s, implementation by
service organizations began to gather momentum. By the mid-1990s, a trend toward the adoption of
ABC by healthcare organizations had become well established. In one form or another ABC is now
being used in numerous health organizations including approximately 20% of hospitals in the USA and
Canada[CE1].12 In her book on activity-based costing in healthcare, Baker5 argues that there is a need
for improved costing systems in healthcare because competition is a driving force, while productivity
and efficiency remain serious concerns. ABC can respond to the pressures of managed care by
delivering the information needed to maximize resources and to relate costs to performance and
outcome measures without negatively affecting the quality of service.
Two particular circumstances propel the present need for resource consumption and service cost
information: 1. diversity of service delivery, and 2. transition in the payer mix. Managed care and
capitation push the healthcare institutions of today to discover resource consumption and cost of
services. ABC is gaining ascendancy in the healthcare field because of its flexibility in these two areas.
It can be applied across all care levels, and its methodology is particularly suited to the complexities of
healthcare service delivery. Managed care contracts usually include some requirement to measure
outcomes. Outcome measures are, of course, a type of performance measure and can thus be
integrated into an ABC system.
© 2010 Elsevier
1 of 1 15-04-2010 22:02
THE PRINCIPLES OF ACTIVITY-BASED COSTING IN MRI

Let us begin by briefly exploring the principles of ABC.* An MR imaging center provides an array of
types of MR imaging studies in which different pulse sequences and coils image different body parts
producing both anatomic as well as functional sequences. The end result is a collection of imaging data
that a radiologist interprets and then reports to the referring clinician. To provide these services, the
center employs many resources, including real estate, medical equipment, labor (front-office staff,
management, technologists, radiologists, and billing and collection staff), capital, and consumables
(Fig. 26-1).
Some of these resources can be directly tied to a specific study, such as the contrast dose used for a
MRI scan. Yet many of the more significant expenses are independent of the number of studies. For
example, the lease on the MRI magnet is a fixed cost, as is the initial cost of locating and building the
facility. In addition, the costs of front-office staff, back-office employees, technologists, and the like
are relatively fixed, as they are shared across multiple MR services. Tracking these elements provides
a mechanism by which the radiology manager determines how the different resources support each
type of service provided by the center; these decisions then serve as "drivers" for allocating the right
portion of the cost of each resource to each service. For example, the lease or depreciation costs of
the MR systems may be allocated to the different MR services based on scan duration. On the other
hand, patient volume may be the appropriate cost driver to allocate the costs of support and
administrative staff. With ABC, the imaging manager can intelligently allocate all the other costs
incurred by the center to the services provided by the center and improve management decision-
making and financial success.
*For the seminal work in the field, see Cooper and Kaplan6 and Cooper.12
© 2010 Elsevier
1 of 1 15-04-2010 22:02
APPLICATIONS OF ACTIVITY-BASED COSTING IN RADIOLOGY

page 682
page 683
In radiology, report turnaround, technologist productivity, equipment utilization, film library

effectiveness, and financial unit costs are frequently benchmarked to identify best practices and
opportunities for reducing expenses. Levine13 argues that when using unit costs to benchmark best
practices at the modality level, it is important to separate indirect radiology expenses (transcription,
film library) from direct modality expenses such as technologists and supplies. If unit costs at the
modality level have support services included, it is difficult, if not impossible, to isolate the best practice
underlying the metric. This is because very few radiology departments perform true activity-based
costing to accurately allocate these expenses from support services to modality cost centers.
Depreciation and lease expenses should also be unbundled from modality unit costs because the cost
accounting practices for capital equipment are often very different across institutions.
There are many applications of activity-based costing in the radiology literature:
Canby14 uses ABC principles and techniques to determine costs associated with the X-ray
15
process in a mid-sized outpatient clinic; Lievens et al compute radiotherapy costs for the
University Hospitals Leuven (Belgium).
Enzmann et al16 assess the financial status of mammography services at seven university-based
programs by using an extensive financial survey encompassing revenue, direct and indirect
costs, and volume data for 1997 and 1998. At one of the institutions, an activity-based costing
analysis was performed by procedure type: screening mammography, diagnostic
mammography, breast ultrasonography, interventional procedures, and review of outside
mammograms. The authors conclude that the reimbursement rate for mammography
procedures, especially diagnostic mammography, needs to be increased to reflect the current
reality of the resources necessary to maintain the accessibility and accuracy of this evolving mix
of clinical services.
17
Nisenbaum et al examined the costs of computed tomography (CT) procedures in a large
academic radiology department, including both professional (PC) and technical (TC)
components, by analyzing actual resource consumption using an ABC method and comparing
them with Medicare payments. They found that in the setting and time period studied Medicare
under-reimbursed professional costs while technical costs were over-reimbursed.
Laurila et al18 design a study to get an informative and detailed picture of the resource utilization
in a radiology department in order to support its pricing and management. They find that the
allocation of overhead costs was greatly reduced by the introduction of ABC. The overhead cost
as a percentage of total costs dropped from 57% to 16%. The change in unit costs of radiologic
procedures varied from -42% to +82%.
© 2010 Elsevier
1 of 1 15-04-2010 22:02
THE LIMITATIONS OF USING ACTIVITY-BASED COSTING SYSTEMS BASED

ON HISTORICAL DATA
Historical cost systems use data about past expenditures to develop an estimate of the cost
associated with a service provided by the facility. The first step, as for any ABC system, is establishing
activity cost drivers, in this case the number of studies per annum. Next, management uses historical
data to sum up the total fixed costs of the resources used by the center. These would include labor,
equipment, capital, etc. The total historical costs are then divided by the total number of cost driver
units required by each service.
As a simple case, consider the clinical activity MR examination, so that the appropriate activity cost
driver for all resources for this activity is the number of studies. A review of historical data establishes
total fixed costs, which are then divided by the total number of studies. This establishes the rate used
to calculate the cost and therefore the profitability of individual studies.
For example, after management decides that the appropriate cost driver for MRI studies is the number
of studies, it establishes the total cost and the total number of studies for the past year. In doing so, it
finds that the 3600 MRI studies done last year cost a total of $1,200,000. Consequently, the per study
cost (i.e., the per driver unit cost) is $333.33 ($1,200,000 divided by 3600).
Clearly, this historical method is objective, and it is relatively simple to implement. To estimate activity
cost driver rates with the historical method, we:
trace the resource expenses to activities

obtain the quantitative data on the activity cost driver for each activity, and
obtain the quantity of each activity cost driver used for each imaging study during the historical
period.
We know that such a calculation, while much more accurate and detailed than traditional costing
systems, is not quite as useful or correct as it should be. Recent research in managerial accounting
shows that historical cost driver rates have two major limitations.
First, the actual cost driver rate is calculated ex post, i.e., at the end of the period. You can readily
imagine the following statement from your partner:
"Last year's data won't help us; our costs next year are going up by at least a quarter of
a million dollars at this site. Give me accurate information for negotiating and
renegotiating our contracts in the coming year."
In other words, when using traditional cost accounting models for decision support we are forced to
wait until the end of the period to obtain cost driver rates required to calculate medical service costs
and profitability. While shortening the period can partly overcome this limitation, it may increase the
accounting workload, and such cost accounting still remains a backward-looking tool.
The second limitation has to do with the accuracy of the cost driver rate when the capacity of the
available resources is not fully utilized. This is the more subtle and also more treacherous limitation of
the analysis. By ignoring the time when the equipment and the staff were idle and by allocating the
total costs to the studies actually performed, the per study cost is artificially, and incorrectly, inflated.
Let us look at an example of this. If the MRI center really had the capacity to handle 4500 studies a
year, not just the 3600 actually performed, then the correct per procedure cost is considerably less
than the $333.33 calculated from historical data. That rate includes both the cost of resources used to
handle each study and the cost of resources supplied but not used during the period. The actual rate
would be 1,200,000/4500 studies per year or only $266.66 per case. The difference between the two
values reflects the cost of the unused (additional potential) capacity to scan. This will be discussed in
greater depth below.
1 of 2 15-04-2010 22:03
page 683
page 684
Radiology directors should use their financial information to make better decisions about current and
future processes, resources, and contracts. An SAC system provides the tools to overcome some of
the limitations of the more traditional accounting systems discussed above.
© 2010 Elsevier
2 of 2 15-04-2010 22:03
USING BUDGETED CAPACITY TO ESTIMATE ACTIVITY COSTS AND

ACTIVITY COST DRIVER RATES
The standard model must now be adapted to forecast the budgeted expenses for resources in the
upcoming period. In this way, activity cost driver rates will be a function of anticipated expenses rather
than historical costs and therefore have improved accuracy. This enables cost driver rates to be
calculated at the beginning of a period so that radiology managers can use this information, almost in
real time, when making decisions about pricing services and contracting for customers.
Continuing with the MRI Study example, let us assume that your partner is right and for the next year
the budgeted expenses of resources required for the MRI Study will go up to $1,500,000 from
$1,200,000. Let us keep it simple by also expecting that the center will have the same capacity for
4500 MRI studies in the coming year. The ABC model, using budgeted data, gives a proactive
calculation of a cost driver rate of $333.33 per study ($1,500,000/4500). This per study cost will be
charged to every imaging study conducted, reflecting the increase in costs in the coming year.
This cost information can also serve as a basis for pricing and contract negotiation for the next year.
Clearly, management does not have to wait until the end of the year to learn how much each study
cost. A second, and equally important, benefit is that management can more accurately track the cost
of unused capacity.
© 2010 Elsevier
1 of 1 15-04-2010 22:03
ABC + CAPACITY COST = SERVICE ACTIVITY COSTING (MEASURING THE

TRUE COST OF MRI CAPACITY)
The final point in the previous section cannot be overemphasized. Suppose a proposed contract with a
large source of patients falls short and the center performs fewer cases than expected. Since the fixed
resources used for each case remain the same, the per study costs remain unchanged. How can that
be, since the number of studies changes? The paradox is solved by realizing that the center has
capacity that it is paying for but not using. By tracking these missing dollars, the manager avoids
several serious accounting pitfalls.
Let us return to our site that can perform 4500 scans at a fixed cost of $1.2 million. We predict that for
the coming year the cost will be $1,200,000/4500, or $266.67 per study. Now, what if a contract that
fell short of expectations resulted in the center performing only 4000 cases instead of the planned
4500? Your junior partner does a back-of-the-napkin analysis and concludes that the cost per study is
now $1.2 million/4000 or $300 dollars per study. She says this revised number should be used in
negotiating a new contract with another MCO. Is that correct? Did the costs just rise by $33.34 per
study because of the shortfall? The answer, of course, is no. The cost per study is still the same.
Capacity did not change because of the shortfall, even though the center's bank account may not look
very good at this point. The missing information lies elsewhere. There is an unused capacity of 500
cases. At the real cost driver rate ($300), this unused capacity would be valued at $150,000 ($300 ×
500). That is where the shortfall is, and that factor should be used for negotiating and for analyzing the
results of other management initiatives. The extreme case of failing to understand this issue can result
in the "death spiral" discussed in the Appendix.
Failure to take this into account can lead to mental or perhaps even real-world traps. With simplistic
costing methods, when imaging activity levels decline-perhaps because of a change in local referral
patterns or the loss of a major managed care contract-the imaging activity cost driver rate will appear
to increase because expenses, the numerator of the calculation, remain the same, while the cost driver
quantity, the denominator, declines. This can artificially increase costs if one does not account for the
cost of unused capacity. Thinking of this capacity as unused inventory may help make this clearer.
© 2010 Elsevier
1 of 1 15-04-2010 22:04
LINKING CLINICAL SERVICE LEVELS AND CAPACITY IN A SERVICE

ACTIVITY COSTING SYSTEM
Figure 26-2 demonstrates how a typical MR facility would determine its optimal configuration and
capacity.
The long-run decisions involve creating a marketing forecast and establishing desired service levels.
These service levels could use criteria such as how long a patient must wait for an appointment, how
long the patient must wait after checking in at the front desk, or how long the referring doctor must
wait for a report. They could also involve back-office criteria, such as how long the center takes to bill
for its services. The use of Management Science models such as queuing networks helps the center
create a profile of the capacity it must support.
Management must calculate the resources needed to support this capacity, taking into account the
issue of practical capacity. The resources include real-estate space, number of magnets, front-office
and back-office employees, technologists, nurses, radiologists, etc. Management can use this
information to calculate the cost associated with the resources and, by implication, the cost of
providing the particular service level. If management deems the costs excessive, it should revisit its
earlier decisions. One can go back and forth between decisions that affect the demands for resources
and decisions to increase or decrease the supply of resources.
page 684
page 685
Add to lightbox
Figure 26-2 Algorithm for determining optimal configuration and capacity of an MR facility.
Once management is satisfied with its estimates, and the center is in operation, the center should
1 of 2 15-04-2010 22:04
respond to how the observed performance compares with the predicted performance. Depending on
the observed demand, the center can calculate the cost of unused capacity. It can also use both
pricing and marketing initiatives to revise the forecast and the observed demand. The iterative nature
of the design phase is equally applicable to the operational phase, and center management should
regularly adjust its resource allocation and demand forecast based on observed demand and actual
utilization.
© 2010 Elsevier
2 of 2 15-04-2010 22:04
MEASURING THE PRACTICAL COSTS AT THE LIMIT

This analysis can be taken to the limit of efficient use of the center. Even working with budgeted data,
the forecasted activity volume of imaging studies may be well below the quantity that could be handled
by the center's resources. In our example, the $300 activity cost driver rate that your partner came up
with includes expenses of both used and unused resources, at a forecasted activity level of 4000
studies per year. This important point is explained next.
This common application of forecasted activity levels for calculating the MRI Study cost is conceptually
incorrect. If, as we have assumed, the resources supplied to perform an activity such as an MRI Study
are essentially fixed in the short run, we need to obtain an additional and very important new piece of
information: how many MRI studies could be handled during the period by the current resources
supplied at the center? This new information represents the practical capacity of the MRI center's
resources for this activity, the largest number of MRI studies that could be handled at the existing site
without creating unusual delays, forcing overtime, or requiring additional staffing resources to be
supplied.
Suppose, for purposes of illustration, that the practical limit of capacity for this activity is 5000 studies
per year. In this case, the correct cost driver rate is $240 per study, not the $300 per study previously
calculated. Why is $240 "more correct" than the $300?
page 685
page 686
Management authorized an annual supply of resources expected to cost $1,200,000 with the intent
that the resources would provide sufficient capacity to handle 5000 studies per year. What have they
received from this authorization? Assuming that each study requires approximately the same resources
to handle (if not, our ABC model should use a duration or acuity driver, or a weighted index of task
complexity), then approximately $240 of resources are used each time a patient is handled. This
number represents the basic efficiency of the MRI Study processes, $240 of resources for each study.
If, in a particular year, only 4000 studies are performed, the efficiency of the activity should remain
about the same. The staff, the magnet, and all other resources required to perform this task do not
suddenly become less efficient (raising the cost to $300 per order) just because fewer studies are
performed in a particular time period.
The lower number of studies received means that not all the resources supplied during the period are
expected to be used. Because of the contracts and commitments (explicit and implicit) made to the
resources performing this activity, the supply of resources can not be lowered in the short run in
response to the expected lower activity level (it is a "fixed" cost). Alternatively, radiology managers
may want to retain the current level of resources in order to handle higher expected patient volumes in
the future. An example would be a center that has signed a competitive contract with an important
practice that grows more slowly than expected or is in a city where demand is strongly seasonal. The
imaging facility cannot easily decrease its capacity to save money, and if it does then it might be
unable to meet the demand and could lose the entire contract, not just the cases it refused. In either
case, the cost driver rate should reflect the underlying efficiency of the imaging process-the cost of
conducting each study-and this efficiency is measured better by recognizing the capacity of the
resources being supplied. The numerator in an activity cost driver rate calculation represents the costs
of supplying resource capacity to do work. The denominator should match the numerator by
representing the quantity of work the resources can perform.
In our numerical example, the center expects to perform 1000 fewer studies than it could handle with
the resources supplied. When the practical capacity is used to calculate activity cost driver rates, the
MRI center has an additional line item in its periodic financial reports: the budgeted cost of unused
capacity, which equals $240,000 (1000 unused studies at a cost driver rate of $240/study).
The basic principle represented by this calculation is simple yet profound. It is captured by the
following equation:
1 of 2 15-04-2010 22:05
As discussed above, most financial systems-whether general ledger systems that measure expenses
actually being incurred, or budgeting systems that measure expenses expected to be incurred-were
designed historically to measure the left-hand side of this equation. They measure the amount of
organizational expenses incurred to make resources available for productive use. This is an important
measurement and one that needs to continue to be made for any current or future system. It
represents the heart of systems for financial reporting and for operational control by measuring (or
forecasting) the actual spending by the organization. But such a measurement, by itself, is inadequate
for measuring the costs of resources required to actually perform work.
The distinction between the cost of resources supplied and the cost of resources used is critical for
reconciling some confusion about SAC systems. SAC systems measure the first term on the right-hand
side of the equation. They measure the cost of resources used (or, alternatively, the resource costs of
activities performed) for individual activities. The difference between the resources supplied and the
resources actually used during a period represents the unused capacity of resources for the period.
Those managers who interpret SAC systems as predicting that taking on an additional patient to scan
would cause organizational spending to increase by $240 misunderstand and misapply a fundamental
concept underlying activity-based costing. In effect, the money has already been spent. A good
analogy is that you have already purchased that slot at that price. Even if you do not use it, you still
have purchased it. The only additional costs to an additional scan are the smaller true variable costs.
© 2010 Elsevier
2 of 2 15-04-2010 22:05
EXCEEDING THE LIMIT: WHEN DEMAND EXCEEDS THE PRACTICAL

CAPACITY
When the capacity of existing imaging resources is exceeded, the consequence is obvious: delays or
poor service levels. Such shortages can occur on magnets, office space, or even parking, the usual
case that comes to mind when you think about fixed costs and capacity. The SAC approach makes it
clear that such shortages can also occur for resources performing support activities, such as patient
scheduling, maintenance, or billing and collection. Radiology managers, facing such shortages, move to
the second step of making committed costs variable: they spend more to increase the supply of the
appropriate resources and relieve the bottleneck.
© 2010 Elsevier
1 of 1 15-04-2010 22:05
APPLYING SERVICE ACTIVITY COSTING TO MEASURE THE SUCCESS OF

MANAGEMENT INITIATIVES
In some centers, management initiatives are designed to improve efficiency and decrease costs.
These include productivity improvements (also known as total quality management [TQM], continuous
quality improvement [CQI], or reengineering initiatives) to reduce or eliminate inefficiencies both in
imaging activities and in administrative support processes (such as scheduling, billing, and collection).
We look at initiatives that address very specific changes, each of which should be familiar to managers
of radiology centers:
reducing operating costs while maintaining capacity constant

for a center operating at capacity, increasing capacity with constant costs
creating new capacity at an additional cost to meet a planned increase in demand
for a center with existing excess capacity, increasing capacity with no additional increase in
costs.
page 686
page 687
Reducing Operating Costs While Maintaining Capacity Constant

Past or current operations may represent quite inefficient activities with a substantial potential for
improvement. If one can determine the quantity and cost of the inefficiencies, then these costs can be
excluded from the estimated expenses. With this approach, estimates of resource expenses assigned
to certain MRI center activities will represent the (standard) costs of more efficient operations. The
benefits of management initiatives can be a reduction in costs, an increase in capacity, or some
combination of the two.* If costs are reduced, management should use the lower estimated expenses
in planning for the future.
Suppose management estimates that 15% of the expenses in the MR imaging activity can realistically
be eliminated by process improvements without any reduction in capacity. Now the standard cost of
this activity will be estimated at $1,020,000 ($1,200,000 × 85%), with a corresponding reduction in the
activity cost driver rate for this activity. Assuming that the process improvements leave the capacity
unchanged at 4500 studies per year, the cost driver goes from $266.67/study ($1.2 million/4500) to
$226.67 ($1,080,000/4500). This insight can translate into a variety of management initiatives,
including more accurate contract negotiations.
For a Center Operating at Capacity, Increasing Capacity with Constant

Costs
The focus of a productivity improvement project may be to increase available capacity. When the
center is operating at, or near, full capacity, any productivity improvement will directly create additional
capacity that the center can use in a variety of ways. The radiology managers can bid for additional
imaging business, price services more intelligently, and perhaps offer new types of studies based on
these anticipated improvements in productivity.
Let us return to the center of the previous example, with its $1,200,000 annual expenses for 4500
procedures at a per procedure cost of $267. Suppose management estimates that it can increase
capacity to 5000 procedures without any associated increase in cost. Now the standard cost of this
activity $240 ($1,200,000/5000). Of course, the total cost is still unchanged, but the center now has an
excess capacity of 500 procedures. To fill this excess capacity, management may schedule more
patients, offer additional scanning procedures, or take other measures it deems appropriate.
Creating New Capacity at an Additional Cost to Meet a Planned Increase

in Demand
Above, we focused our attention on cases in which either the costs changed or the capacity changed,
but not both simultaneously. However, in most practical cases, both will change simultaneously. The
1 of 4 15-04-2010 22:06
methodology does not change drastically. What is important to remember is that in a beneficial
productivity improvement project the percentage change in the costs is always better than the
percentage change in capacity. The following example illustrates what happens when an intentional
increase in capacity requires an increase in costs.
Consider the center with costs of $1.2 million to maintain a capacity of 5000 studies. The manager
sees an opportunity to capture more of the young professional market by staying open later in the
evening. The anticipated additional cost for labor and incidentals will be approximately $100,000 to
stay open later three nights a week for a year. This would open up as many as 20 new study slots per
week, for an annual total of approximately 1000 new studies. The new cost is $1.3 million; the new
capacity is 6000. Note that in the short run the marginal cost of the new slots is only $100 a piece
($100,000/1000). If these evening slots are to become a permanent fixture of the center's operating
environment, it should reassess its overall cost structure. The center's long-term cost basis becomes
$217 dollars per study ($1,300,000/6000). Be careful about mixing marginal and average cost
analyses. Here the marginal analysis is more relevant to the manager's decision.
For a Center with Existing Excess Capacity, Increasing Capacity with

No Additional Increase in Costs
Finally, consider a management initiative to increase capacity even when the center is not operating at
full capacity.* The net result of the productivity improvement project is twofold. First, as expected, it
will decrease the per procedure cost. Yet it will not decrease the total operating cost. Instead, it will
increase the cost of unused capacity. It might appear that the project is of no benefit, yet that is not
necessarily true. The excess capacity could play a pivotal role in a planned expansion of the center's
services, or it could be a stepping stone in the process of converting committed resources into flexible
resources, a process discussed below.
*Strictly speaking, it is more likely that a reduction in costs will lead to a reduction in capacity (or that an increase in capacity will be
accompanied by an increase in costs). The key is that the percentage reduction in operating costs must be lower than the
percentage reduction in capacity, with the ideal reduction in capacity being zero. Similarly, a process change that leads to an
increase in capacity should be such that the percentage increase in capacity should be greater than any percentage increase in
costs.
*A variant of this scenario is the managerial anticipation of a decrease in demand. In this case, the total operating cost and the per
procedure cost will remain unchanged, but the cost of excess capacity will increase.
page 687
page 688
Even though the cost of excess capacity has increased, there is little that management can do to effect
immediate spending reductions. Remember that most of the resources consumed by a radiology
center are fixed in nature-equipment, computers, real estate, and labor. The reduced demand for
these committed resources will lower the cost of resources used (by imaging services or customer
support), but this decrease will be offset by an equivalent increase in the cost of unused capacity as
defined above.
Management can effect a decrease in the cost of unused capacity in one of two ways: 1. by increasing
the demand for activities for the resources to perform (through additional marketing, for example); or
2. by reducing spending by scaling back the supply of resources.
For committed costs to become variable in the downward direction, management must actively shift
the unused capacity of these resources out of the system. At that point, and only at that point, the
costs of resources will start to decrease. Thus what makes a resource cost "variable" is not inherent in
the nature of the resource; it is a function of management decisions-first, to acknowledge a reduction
in the demands for the resource, and, second, to lower the out-of-pocket spending on the resource.
Unused capacity is at the heart of any effort to correctly implement activity-based costing by the
radiology management. In fact, one can view the entire SAC approach as giving radiology managers
insights about the existence, creation, and deployment of capacity, both used and unused. Even
actions taken to improve the efficiency of activities (such as CQI, business process reengineering, or
operational activity-based management [ABM]) require an understanding of how unused capacity is
2 of 4 15-04-2010 22:06
created and then managed. Radiology managers do not seem to have any problem recognizing how
costs are "variable" in an upward direction. Examination of past history will usually reveal how spending
has increased to cope with increased volume, the variety and complexity of the imaging operations,
and the needed clinical and administrative tasks. The mechanism by which costs head downward,
however, seems to have eluded most managers and academics for a long time.
Consider a center with costs of $1.2 million, a capacity of 5000 slots, and actual usage of 4000. The
annual expenses associated with the MRI studies will be $960,000 (4000 [commat] $240), and the
excess and unused capacity will cost $240,000 (1000 [commat] $240). The center's management now
launches a productivity improvement initiative to reduce waste and delays due to non-value-added
tasks in the MRI Study activity. These improvements may even include some modest technology
investments: a new voice recognition system and an Internet-based scheduling system. Such initiatives
typically result in increased employee productivity and effectively increase the center's capacity. Let us
say that the practical capacity will rise to 6000 MRI studies per year. Yet fixed expenses will remain
the same; they may even increase slightly if any technology investments are made. Given the high
percentage of fixed costs in the facility, one may question whether the cost of MRI studies has really
been reduced by the TQM or reengineering initiative. The answer is yes. Remember that before the
changes an MRI study required $240 of resources and the center had $240,000 in unused capacity;
now, a single study requires only $200 of resources ($1,200,000/6000), or perhaps a little more to
cover any new technology investment. Suppose that the number of studies actually performed remains
unchanged at 4,000. Consequently, the expenses assigned to the actual studies will decrease to
$800,000 (4000 [commat] $200). This will signal that the MRI study cost has been reduced because of
the TQM/reengineering initiative. On the other hand, the cost of unused capacity will rise to $400,000
(2000 [commat] $200), since all the resources previously supplied are still being paid for. This example
clearly shows that applying productivity improvement efforts to resources that are already in excess of
the supply will have the main effect of producing even more excess capacity.
Management Science tools like queuing models and bottleneck analysis can signal those resources
that are currently at, or expected soon to reach, capacity constraints. Improvement initiatives can then
be focused on the activities performed by these constraining resources. Radiology management
initiatives designed to improve the efficiency of activities or to yield improvements in spending must
have a plan to eliminate or redeploy the resources that become available as a result of the initiatives.
Alternatively, the SAC system will signal where unused capacity already exists in the radiology center
or will be created after some operational improvements. Such a signal directs managers' attention to
eliminating the unused capacity in ways described above.
While excess capacity, per se, is not necessarily bad in the long run, refusal to act on it is. We have
visited MRI centers that kept existing resources in place, even though they had increased capacity
through productivity improvement projects or the demand for the activities performed by those
resources had diminished substantially. By failing to find new activities that could be handled by the
imaging and administrative resources already in place, the centers received no benefit from the
changes in their operating environment. It is important to note that the failure to capture benefits from
these changes was not due to imaging costs being intrinsically "fixed." Rather, the failure occurred
when radiology managers were unwilling or unable to exploit the unused capacity they created. The
costs of resources are "fixed" only if the manager cannot, or will not, exploit the new business
opportunities from the unused capacity he or she helped to create.
A center can create new business opportunities in a variety of ways. We discussed one above
whereby it found a niche catering to young professionals. It can also determine where the unused
capacity is (late afternoons, for example) and do targeted marketing to direct the fill rate. If the issue
is no-shows, then using incentives, such as a van service, may improve the fill rate. Alternatively,
management could consider mapping which sources of patients are the most fruitful and focus
marketing and incentive initiatives on only those sources that have particularly low no-show/reschedule
rates.
page 688
page 689
3 of 4 15-04-2010 22:06
SAC is an innovative system for identifying, measuring, creating, and managing capacity.
Understanding the subtle interplay between the actions taken with ABC information and capacity
management is central to the approach.
In each of the three examples above, the analysis was based on anticipated changes. Basing
managerial analysis on budgeted rather than historical expenses allows the use of the SAC model to
forecast the future, not just explain the past.
© 2010 Elsevier
4 of 4 15-04-2010 22:06
FROM A SERVICE ACTIVITY COSTING RESOURCE USAGE MODEL TO

DECISIONS ABOUT RESOURCE SUPPLY
SAC can identify resource usage and quantify the relationship between revenues generated and
resources consumed. On the other hand, ABC cannot automatically recognize the relationship between
changes in resources used and changes in resources supplied.
It should be apparent by now that making business decisions based solely upon resource usage (the
SAC system) is problematic, as there is no guarantee that the spending to supply resources will be
well aligned with the new levels of resources demanded in the near future. For example, if the result of
an operational improvement decision is to eliminate film filing on site, no economic benefit will be
achieved unless the resources supplied to filing films on site, which are no longer needed, are
eliminated or redeployed to higher revenue uses. Consequently, before making business decisions
based on an SAC model, you should analyze the resource supply implications of these decisions.
Management should understand the resource supply implications of any contemplated decision. For
example, consider a decision whether to maintain the existing patient mix, or to eliminate all pediatric
studies. The SAC model has shown that resource consumption by pediatric studies is higher than adult
studies and in fact often higher than the revenues they generate (that is, they are loss services). Past
attempts to improve the pediatric imaging processes or to increase reimbursements have failed.
Careful analysis of the referral base indicates that nearly all existing customers can use substitute
pediatric centers without any inconvenience. These pediatric services are candidates for further review
within the center to determine how much of the current expenses will actually be reduced by eliminating
some or all of the pediatric services. Moreover, since the specialized pediatric centers do not image
adults, the potential loss of related adult services from your site may be quite low. (On the other hand,
parents may prefer an imaging center that caters to both young and old, so if your practice includes
many family groups, the impact could be substantial.)
The SAC resource usage model will not show how the center should take advantage of the resources
freed up by the elimination of pediatric studies. It will, however, identify the decrease in resource
usage that will occur if the unprofitable pediatric services are dropped. Management must determine,
on a case-by-case basis, when and if such newly available resources can be redeployed so that the
resulting out-of-pocket savings exceed the revenue losses. At that time it may be appropriate to drop
certain services, since the elimination or redeployment of the associated resources will result in real
savings.
We have seen in this chapter how our newly developed SAC methodology quantifies the dynamics of
potential changes in imaging study mix, activity management, process performance, or the imaging unit
design characteristics. It shows how these changes will affect the future demands for certain
resources in the center. SAC can also provide signals to radiology managers about services and
customers that are generating revenues in excess of resource costs, and those that require resources
that cost more than the revenues generated by them. These enable you to anticipate where new
bottlenecks for resources will develop in the future and to identify resources whose current and future
supply will likely exceed the future demands for the capabilities they provide. One can iterate back and
forth between decisions that affect the demands for resources and decisions to increase or decrease
the supply of resources. The SAC approach also enables radiology managers to decrease operating
expenses for resources in excess supply, without taking the risk of reducing the supply of a resource
below the level where it becomes a constraint on current or future imaging activity.
© 2010 Elsevier
1 of 1 15-04-2010 22:06
APPENDIX What is a Death Spiral?

What is a "death spiral"? The ill-advised allocation of costs can cause a perfectly viable (and
profitable) venture to go out of operation, as illustrated in the following example (Fig. 26-3):
Suppose a radiology center provides just one service: MRI scans. The simplified economics of the
facility are as in Table 26-2. At the beginning of the year, management expects to perform 4500
procedures through the coming year. To accomplish that goal it plans to incur overhead costs (front-
office staff, service maintenance contracts, contracts with billing and collection agencies, radiologist
fees, etc.) of $3,000,000. Based on the budgeted overheads, the center will earn a per procedure net
profit of $33.33. For the whole year, it expects to earn $150,000.
page 689
page 690
Add to lightbox
Figure 26-3 An ill-advised allocation of costs leading to collapse of a profitable venture.
A few months down the road, and while the practice is in the midst of negotiating a new contract with a
large HMO, one of the major sources of referrals-a group practice-is purchased by a local hospital.
The center's management expects that this will lead to a loss of about 500 patients each year.
Using traditional costing methods, the center reassesses its financial health. Now, it expects to do only
4000 procedures annually. Yet the overhead costs will remain unchanged. The results are shown in
Table 26-3. Allocating the $3,000,000 over the expected 4000 procedures leads to a per procedure
cost of $750. Adopting a hard line in negotiations with the HMO, the center insists that the HMO
increase its average fee for the MRI scan from $700 to $775.
Table 26-2. Simplified Economics of a Radiology Center

Budgeted
Overhead cost of center $3,000,000
Planned annual procedures 4500
1 of 4 15-04-2010 22:07
Average net reimbursement per procedure (reimbursement-direct costs) $700

Overhead cost per planned procedure $666.67
Profit per procedure $33.33
Annual net revenue $3,150,000
Annual overhead cost $3,000,000
Annual expected profit $150,000
Not surprisingly, the HMO takes its patient referrals elsewhere, leading to a further decline in patient
volume. Now, the center expects the annual volume to be 2000 patients. Allocating the $3,000,000
overhead to 2000 scans leads to a per procedure cost of $1500. The center approaches its other
large referral source and indicates that the fee of $700 is far from adequate. As a result, it loses that
source of referrals and now can expect an annual volume of just 100 patients. At this point the
management decides to close shop, effectively closing an otherwise fairly successful radiology center!
This death spiral from a viable center doing 4500 procedures annually to none resulted from the
inappropriate use of cost allocation techniques. If the center had used activity-based costing, with its
inherent accounting for unused capacity, the analysis would have looked as in Table 26-4. Center
management would have realized that the per procedure overhead remains unchanged at $666.67. It
would also have realized that the 500 scans of unused capacity cost the center $333,333 each year.
The solution is not to demand higher reimbursement from existing referral sources but to intensify its
marketing efforts to fill the 500 vacant slots or to renegotiate terms with its vendors-the magnet
manufacturer, the landlord, etc.
Table 26-3. Economics of a Radiology Center Using Traditional Costing

Methods
Actual-Traditional
Actual annual procedures 4000
Average net reimbursement per procedure $700
Profit per procedure ($50.00)
Annual total overhead (must equal Overhead cost of center) $3,000,000
page 690
page 691
Table 26-4. Economics of a Radiology Center Using Cost Allocation Techniques

Actual-ABC & Capacity
Actual annual procedures 4000
Average net reimbursement per procedure $700
Profit per procedure $33.33
Unused capacity 500
Cost of unused capacity $333,333
2 of 4 15-04-2010 22:07
Annual overhead cost of actual procedures $2,666,667

Annual cost of unused capacity $333,333
Annual total overhead (must equal Overhead cost of center) $3,000,000
REFERENCES
1. Capettini R, Chee WC, McNamee AH: On the need and opportunities for improving costing and cost management in
healthcare organizations. Managerial Finance 24:46-59, 1998.
2. Ruhl JM, Hartman BP: Activity-based costing in the service sector. Advances in Management Accounting 6:147-161, 1998.
3. Holt T: Developing activity-based management system for the Army Medical Department. J Health Care Finance 27:41-49,
4. West TD, Balas A, West DA: Contrasting RCC, RVU, and ABC for managed care decisions. Healthcare Financial
Management 50:54-61, 1996. Medline Similar articles
5. Baker JJ: Activity-based Costing and Activity-Based Management for Healthcare. New York, NY: Aspen Publishers, 1998.
6. Cooper R, Kaplan RS: Measure costs right: Make the right decision. Harvard Business Review, Sept-Oct 1988, 96-103.
7. Cooper R, Kaplan RS: Profit priorities from activity-based costing. Harvard Business Review, May-Jun 1991, 130-135.
8. Cokins G, Stratton A, Helbling J: An ABC Manager's Primer. New York: McGraw-Hill, 1992.
9. Kiani R, Sangeladji M: An empirical study about the use of the ABC/ABM models by some of the Fortune 500 Largest
Industrial Corporations in the USA. Journal of American Academy of Business 3:174-182, 2003.
10. Horngren CT, Sundem GL, Stratton WO: Introduction to Management Accounting, 12th ed. Chapter 5: Cost allocation and
activity-based costing systems. Upper Saddle River, NJ: Prentice Hall, 2002.
11. Player S: Activity-based analyses lead to better decision making. Healthcare Financial Management 52:66-70, 1998.
12. Cooper R: The rise of activity-based costing-Part One: What is an activity-based cost system? Journal of Cost
Management 2:45-54, 1988.
13. Levine L: Unit-cost financial benchmarking identifies improvement opportunities for imaging centers.
http://www.auntminnie.com, April 2001.
14. Canby JB IV: Applying activity-based costing to healthcare setting. Healthcare Financial Management 49:50-56, 1995.
15. Lievens Y, van den Bogaert W, Kesteloot K: Activity-based costing: A practical model for cost calculation in radiotherapy.
Int J Radiat Oncol Biol Phys 57:522-535, 2003. Medline Similar articles
16. Enzmann DR, Anglada PM, Haviley C, et al: Providing professional mammography services: Financial analysis. Radiology
219:467-473, 2001. Also online at http://radiology.rsnajnls.org/cgi/content/full/219/2/467 Medline Similar articles
17. Nisenbaum HL, Birnbaum BA, Myers MM, et al: The costs of CT procedures in an academic radiology department
determined by an activity-based costing (ABC) method. J Comput Assist Tomogr 24:813-823, 2000. Medline
Similar articles
18. Laurila J, Suramo I, Brommels M, et al: Activity-based costing in radiology: Application in a pediatric radiological unit. Acta
SUGGESTED READING
Cokins G: Activity-Based Cost Management: An Executive's Guide. Columbus, OH: McGraw-Hill, 2001.
Cooper R, Kaplan RS: How cost accounting distorts product costs. Management Accounting April 1988a, 21(4):20-27, 1988a.
Cooper R: The rise of activity-based costing-Part Three: How many cost drivers do you need, and how do you select them?"
Journal of Cost Management, Winter 1989, 16(2):34-46, 1939.
Cooper R: Cost classification in unit-based and activity manufacturing cost system, Journal of Cost Management 4:4-14, 1990.
Cooper R, Kaplan RS: Cost & Effect: Using Integrated Cost Systems to Drive Profitability and Performance. Boston: Harvard
Business School Press, 1997.
Cooper R, Kaplan RS: The promise and peril of integrated cost systems. Harvard Business Review 76:109-119, 1998.
Forrest E: Activity-Based Management: A Comprehensive Implementation Guide. New York, NY: McGraw-Hill, 1996.
Hicks DT: Activity-Based Costing: Making it Work for Small and Mid-Sized Companies. Indianapolis, IN: John Wiley & Sons,
2002.
Johnson HT, Kaplan RS: Relevance lost: The rise and fall of management accounting. Boston: Harvard Business School
Press, 1987.
Kaplan RS, Cooper R: Cost and Effect: Using Integrated Cost Systems to Drive Profitability and Performance. Boston:
Harvard Business School Press, 1998.
3 of 4 15-04-2010 22:07
Roztocki N, Valenzuela JF, Porter JD, et al: A procedure for smooth implementation of activity based costing in small
companies. ASEM National Conference Proceedings, October 21-23, 1999, Virginia Beach, pp 279-288. Also online at
http://www2.newpaltz.edu/~roztockn/virginia99.htm
Schuneman P: Master the 'ABCs' of activity-based costing: How do you find out whether a given capitation rate will be
profitable for your practice? Activity-based cost accounting is one tool that can help. An accountant explains how it works.
Managed Care 6:43-53, 1997. Also online at http://www.managedcaremag.com/archives/9705/9705.accounting.shtml
ONLINE RESOURCES
Activity-Based Costing (ABC) and Economic Value Added (EVA) References. Narcyz Roztocki, http://www.pitt.edu/~roztocki
/abc/abcrefer.htm
Activity-Based Costing (ABC) Economic Value Added Internet Website Guide. Narcyz Roztocki, http://www.pitt.edu/~roztocki
/abc/abc.htm
Search Radiology Online for the phrase "activity based costing": http://radiology.rsnajnls.org/cgi/search?sendit=Search&
fulltext=activity+based+costing&andorexactfulltext=phrase&journalcode=radiology
page 691
page 692
© 2010 Elsevier
4 of 4 15-04-2010 22:07
EART AND ASCULAR YSTEM

page 693
page 694
page 694
page 695
AGNETIC ESONANCE NGIOGRAPHY
David Saloner
Daisy Chien
Oliver M. Weber
Charles M. Anderson
Ralph E. Lee
Robert R. Edelman
Magnetic resonance angiography (MRA) methods have steadily improved over the past decade. These
advances have derived from improvements in hardware and technique: most importantly from higher
performance magnetic field gradients that permit rapid acquisition; from the application of contrast
agents that permit improved coverage with reduced motion complications; and from an improved
understanding of what the most efficacious approach is for a given clinical question. In certain
applications, the vascular resolution of MRA begins to approach that of conventional X-ray
angiography. Although images from catheter-injected X-ray angiography have higher contrast and
better dynamic information than MRA, MRA has a number of other advantages. Importantly, it is
noninvasive or minimally invasive. It routinely provides three-dimensional information and can be
obtained in conjunction with information on the surrounding soft-tissue. MRA also goes beyond the
mere depiction of vascular anatomy; it provides insight into underlying function.
The focus of this chapter is on the basic principles and methodologies of flow imaging using magnetic
resonance. More detailed coverage of specific MRA topics is provided in other chapters, including flow
quantification (Chapter 28, Basic Principles and Clinical Applications of Flow Quantification) and
contrast-enhanced MRA (Chapter 29, Principles and Optimization of Contrast-Enhanced Three-
Dimensional MRA).
page 695
page 696
Add to lightbox
Add to lightbox
Figure 27-1 Schematic diagrams of streamline flow showing plug flow (A) and laminar flow (B).
1 of 2 19-04-2010 23:54
© 2010 Elsevier
2 of 2 19-04-2010 23:54
BEHAVIOR OF BLOOD FLOW

The appearance of images in MRA is strongly influenced by blood motion. Therefore, a good
understanding of the fundamental properties of flow is useful for selecting the best protocol for MRA.
Streamlines and Flow Profiles

Blood flow is complex and highly variable in vivo. It can, however, be described by simple flow models.
In fluid dynamics, streamlines and flow profiles are often used to provide a graphic description of flow
in vessels. Common examples of streamline flow include plug flow and laminar flow (Fig. 27-1). In plug
flow, all fluid particles move forward in parallel lines with the same speed. Plug flow has a
characteristic blunt profile and is often observed in the descending thoracic aorta. Laminar flow, on the
other hand, has a parabolic flow profile with the fastest moving fluid particles at the center of the
lumen. It is called laminar because the particles move along in concentric sheets or laminae. 1 The flow
velocity at any radial location in steady laminar flow is precisely described by the following equation:
where V(r) is the velocity at distance r from the center of the lumen, R is the radius of the vessel, and
Vmax is the maximal flow velocity. Note that the velocity is maximum (Vmax) for particles at the center of
the lumen (r = 0), whereas the velocity is zero for fluid particles at the vessel wall (r = R). Friction and
the resulting drag at the vessel wall are responsible for this zero velocity.
Flow profiles have a significant impact on the flow contrast obtained in MRA. In addition, they can
affect the interpretation of measured flow velocity in different blood vessels. For example, the average
velocity in laminar flow is exactly 50% of the peak velocity at the center of the lumen, whereas the
average velocity in plug flow is equal to the peak velocity.
Vessel Geometry and Entrance Effect

Geometry plays an important role in fluid dynamics. Variations in vessel geometry, such as vessel
tortuosity, stenoses, and bifurcations, can alter the appearance of flow in MRA. For example, signal
heterogeneity in MRA of the carotid artery siphons is due to local vessel curvature that causes the
2
fastest moving fluid particles to swing toward the outer curve of the vessel (as a result of inertia).
The local flow profile often changes when the geometry of a blood vessel deviates from a long
cylinder. For example, the U-shape of the aortic arch results in blood traveling with a helical flow
pattern. In the vicinity of a bifurcation, flow separation occurs. Flow separation is the formation of a
local fluid recirculation zone that does not move with the main streamlines. This can be observed in the
carotid bulb. It can also occur adjacent to a stenosis (Fig. 27-2). The separation zone often appears
dark in MRA and is due to saturation, which is described in greater detail in this chapter.
Another important flow phenomenon occurs when laminar flow proceeds from a larger vessel or fluid
cavity into a smaller vessel. It is known as the entrance effect. Immediately at the entrance of the
smaller vessel, the flow profile is blunt and it takes a certain distance for flow to develop fully to the
3
parabolic flow profile (Fig. 27-3). As a result of the entrance effect, the measured flow profile and
velocity distribution can vary depending on the location of the measurement.
Add to lightbox
page 696
page 697
Figure 27-2 Schematic diagram of the change in streamline and flow profile caused by a vessel
1 of 3 19-04-2010 23:55
stenosis. Immediately distal to the stenosis is a flow separation (S) and recirculation (R) zone.
Add to lightbox
Figure 27-3 Schematic diagram of the flow entrance effect.
Flow Behavior in the Human Vasculature

Blood flow in the major arteries is highly pulsatile, with flow velocity varying from greater than 100 cm/s
during systole to almost no flow during diastole. Flow quantification by MR has shown that blood flows
in the ascending aorta with a skewed velocity profile during systole (with an axis of skew symmetric
about the plane of the aortic arch) and flows in the descending aorta with a plug profile with minimal
4
skew. During diastole, blood flow can reverse for a short time in medium to large arteries, such as the
aorta. This flow reversal typically occurs during early diastole. Flow reversal can occur when a
compliant vessel relaxes after absorbing the systolic output or from reflection of the pulse pressure
wave at the distal vessels. Wave reflection at the peripheral vessels also results in the distinctive
triphasic (forward-backward-forward) waveform in the popliteal artery observed by MR flow
quantification. This triphasic waveform may be lost when significant atherosclerotic disease is present.
There is always resistance to blood flow in a vessel. Flow resistance depends on the fluid viscosity as
well as the shape of the vessel. It has a strong dependence on the vessel diameter (it decreases with
the fourth power of the radius). This is because most of the resistance occurs at the vessel wall. As a
result, flow encounters increasing resistance as it goes from larger vessels to smaller ones. Because
of vessel tapering and higher and higher resistance to flow at the distal vessels, the arterial tree has
considerable damping of cardiac pulsations as blood travels from the arteries to the veins.
Viscosity and Non-Newtonian Properties of Blood

Viscosity is the internal friction of a fluid. Unlike flow resistance, it depends only on the nature of the
fluid and is relatively independent of the vessel geometry. Viscosity of blood is approximately three
times higher than that of water as a result of the red blood cells exerting frictional drag on neighboring
cells and against the wall of the vessel. A fluid that has a viscosity that is constant regardless of the
shear rate (steepness of the velocity profile), is termed a Newtonian fluid. In larger vessels with
relatively brisk flow, there is little tendency for red blood cells to aggregate, and blood has a
Newtonian behavior. However, at low shear rates, such as in regions of slow recirculation, the viscosity
of blood increases as the red blood cells adhere to each other and the viscosity is non-Newtonian.
Another condition of nonconstant viscosity occurs when blood passes through the capillary bed where
red blood cells pass through the center of the capillaries leaving a low concentration at the capillary
walls, effectively reducing the blood viscosity, an effect termed plasma skimming.
Turbulence and the Reynolds Number

In laminar flow in a straight tube, the fluid moves down the length of the tube parallel to the walls. As
the flow velocity increases, there is a transition to turbulence, where fluid particles begin to have
components of motion at right angles to the principal direction of motion. One of the few manifestations
of turbulence in vivo occurs distal to a stenosis where the flow jets into a large chamber and generates
2 of 3 19-04-2010 23:55
vortices or eddies. The onset of turbulence is determined by the Reynolds number (Re), which is given
by:
where ρ is the density of the fluid, v is the average velocity, D is the diameter of the vessel, and μ is
the viscosity. Flow becomes turbulent when the Reynolds number is 2000 and above. As shown in
Equation 27-2, the Reynolds number increases with the diameter of the tube and the average velocity,
and decreases with the viscosity. In the human aorta, the Reynolds number is about 1500. True
turbulence, which is accompanied by energy dissipation and can be heard as an audible bruit, is
seldom encountered in vivo. Even in the absence of turbulence, regions with complex flow,
recirculation, or rapid changes in velocity, can lead to phase incoherence and loss of signal in MRA.
As can be seen from this brief discussion of flow dynamics, different flow conditions can have a
significant impact on the results of MRA. With this in mind, optimal flow contrast can be achieved with
an understanding of both the flow behavior and the principles of MRA.
© 2010 Elsevier
3 of 3 19-04-2010 23:55
TECHNIQUES FOR MAGNETIC RESONANCE ANGIOGRAPHY

A wide variety of MRA techniques have been proposed and evaluated over the years.5,6 It would be
beyond the scope of this chapter to describe all the methods that can be used to depict flow. Rather,
this chapter focuses on the techniques that have established applicability and are used routinely in a
clinical setting for diagnosis of vascular abnormalities.
There are three major approaches to acquiring MR angiographic images. The first two rely on the
behavior of the intrinsic magnetization. They are the time-of-flight (TOF) effect, first observed more
than four decades ago, and the phase effect, also discovered in the early days of MR. 7-9 The third
method, contrast-enhanced MRA (CE-MRA), exploits the reduction in T1 of intravascular blood that
results from the injection of contrast agents (such as gadolinium diethylenetriaminepentaacetic acid
[Gd-DTPA]). Although the three MRA methods are profoundly different, they reflect the power and
flexibility of MRA methods in that they are all built on variants of one specific type of pulse sequence,
the gradient recalled echo (GRE) sequence. A combination of these methods can be used to generate
MR angiograms, evaluate vessel stenosis, detect flow direction, and quantify blood flow.
© 2010 Elsevier
1 of 1 19-04-2010 23:56
TIME-OF-FLIGHT
Time-of-flight methods have been successfully used in clinical applications for longer than the other
MRA methods. Time-of-flight imaging has been used to depict the intracranial, extracranial, thoracic,
abdominal, and peripheral arteries and veins.10-12 Clinical studies of TOF are numerous13-21 and are
testimony to the flexibility of this approach. These methods can be applied to answer many important
clinical questions without the cost or inconvenience of using contrast agents.
Basic Mechanism
Time-of-flight imaging typically consists of 1. suppression of background tissue signal; and 2. retention
of a high signal from flowing blood.
Background Tissue Suppression

Time-of-flight methods use rapid slice selective radiofrequency (RF) excitation pulses applied so
rapidly (using a repetition time [TR] that is short compared with the T1 of the stationary tissue) that
spins in stationary material do not have enough time to regain their longitudinal magnetization. This
reduction in longitudinal magnetization that results from repeated RF excitations is referred to as
saturation. Saturated spins produce a dim signal, whereas unsaturated spins produce a bright signal.
Intentional saturation of stationary tissues makes them appear dark.
Add to lightbox
Figure 27-4 Left, Schematic diagram showing increased inflow of fresh spins into the imaging slice
with increasing flow velocity. Right, Plot of the flow signal intensity versus velocity for gradient-echo
imaging. The flow signal increases linearly with velocity until all the spins within the slice are replaced
by fresh spins entering with full magnetization.
Bright Inflow Signal

Blood outside the imaging slice is not affected by the slice-selective RF pulses. This unsaturated blood
flows into the volume with full magnetization strength, producing a bright signal. The vascular signal in
TOF increases with the velocity of blood flow. The increase in signal with flow is linear until the blood is
completely replaced with each TR, that is, all the spins in the imaging volume are entirely replenished
22
after each pulse. For flow orthogonal to the imaging slice, complete inflow occurs when:
where TR is the repetition time, th is the slice thickness, and v is the blood velocity. For a given TR,
total replenishment of spins occurs at lower velocities for a thin slice than a thick one. The actual
amount of flow enhancement depends on the number of spins entering the slice with full magnetization.
For TOF imaging, the flow signal increases linearly until it reaches a plateau at which all the spins
within the slice are replaced with fully magnetized ones (Fig. 27-4).
1 of 9 19-04-2010 23:57
Effects of Time-of-Flight Acquisition Parameters on Vascular Contrast

Flow contrast in TOF varies with the choice of a number of image parameters, including TR, echo time
(TE), flip angle (FA), size of the imaging volume, pixel size, and slice orientation. Angiographic contrast
also depends on flow conditions such as velocity, pulsatility, and vessel tortuosity. A good
understanding of how the sequence parameters affect vascular contrast enables one to tailor
sequences for specific applications.
page 698
page 699
Add to lightbox
Figure 27-5 A, Three-dimensional time-of-flight (3D TOF) image of the circle of Willis (TR/TE/FA =
40/6/25°). B, Acquisition repeated with a larger FA (TR/TE/FA = 40/6/45°) demonstrates background
suppression. Distal flow is better retained with smaller FA (arrows).
Repetition Time, Flip Angle, and Inflow

To suppress background signal, RF pulses are delivered at a TR much shorter than the T1 of tissue to
cause effective spin saturation. Typical TR values range between 25 and 70 ms. A shorter TR also
results in a shorter acquisition time, therefore, there is further incentive to select short TR values.
However, TR values must not be so short that there is insufficient time for the inflow of fresh spins.
Effective saturation is also achieved by using large FA values. The same RF pulses that reduce the
stationary tissue signal, however, can also affect the blood signal. This is particularly true for
slow-moving blood that spends a relatively long time within the imaging volume, which results in a
diminished signal. To maximize the blood signal, a long TR and a small FA should be used to minimize
spin saturation of blood. Unfortunately, the background signal is then relatively high and flow contrast
(the difference in signal between flowing blood and stationary tissue) is low. Therefore, the selection of
TR and FA is a compromise between saturating background tissue and not saturating blood. The
optimal values of TR and FA depend on the velocity of blood and the anatomic distances to be
covered. Typical FA values range between 20° and 35° for three-dimensional (3D) TOF and between
30° and 90° for two-dimensional (2D) TOF where full replenishment of unsaturated magnetization
throughout the thin slice is more easily accomplished than for a thick 3D slab. As FA increases, the
more distal portions of vessels become saturated and are poorly visualized (Fig. 27-5).
Slice Orientation
2 of 9 19-04-2010 23:57
Add to lightbox
Figure 27-6 Pulse sequence diagram of a 2D time-of-flight (TOF) gradient-echo sequence with flow
compensation along the read and the slice-select directions.
When performing 2D acquisition, inflow can be maximized by positioning the imaging slice
perpendicular to the direction of flow. This provides effective replenishment of fresh spins into the slice
and allows the use of a large FA excitation for maximal flow contrast. A drawback of a large FA
excitation is an increase in ghost artifacts resulting from pulsatile flow. Placing the imaging slice in the
plane of the vessel is undesirable because slow-moving spins become progressively saturated; this
effect worsens as the FA is increased. This is known as in-plane saturation.
Flow Compensation and Echo Time

In GRE sequences, RF pulses create transverse magnetization, and magnetic field gradients act to
change the magnetization phase: the orientation of the magnetization vector in space. Gradient
recalled echo sequences dispense with RF-refocusing pulses and rely on appropriate design of
gradient waveforms to generate the echo that is the basis for image formation. In particular,
conventional GRE sequences have bi-lobed gradient waveforms along the slice select and frequency
encoding axes to rephase all spins-i.e., to bring all magnetization vectors within the slice into alignment
at the time of signal readout. That strategy works perfectly well for stationary spins but fails for moving
spins that accumulate a phase that depends on their velocity. Flow compensation works by achieving
phase coherence of both the stationary and moving spins at the time of the echo.23 At least three
gradient lobes are needed to have velocity compensation (also known as first-order flow
compensation) (Fig. 27-6).
page 699
page 700
Unwanted loss of flow signal occurs when there is phase dispersion caused by turbulent flow. This is
discussed in greater detail later in this chapter. The amount of phase dispersion can be reduced by
using flow compensation with the shortest TE.24 Higher order motions, such as acceleration, are not
refocused by first-order flow compensation and are best dealt with by using a short TE. The TOF
method typically uses a gradient-echo sequence with a short TE and velocity compensation.25-28 Figure
27-6 shows a pulse sequence diagram of a 2D-TOF sequence that is flow compensated along the
read and slice-select directions. A short TE minimizes phase dispersion related to higher order
motions. Furthermore, one can sample the echo earlier in time by using an asymmetric echo. This
shortens the time the spins are allowed to dephase in the presence of the readout gradient, at the
expense of slight blurring in the read direction.
Another important impact of the TE in MRA is its effect on the signal intensity of fat. An unwanted
bright fat signal can be a problem for MRA of the thoracic, abdominal, and peripheral vasculature. The
presence of fat is troublesome for MRA particularly when projections of the vasculature are generated
from a stack of 2D or 3D slices. The bright fat signal can mask the region of interest and confound the
results. To minimize the fat signal, MRA is performed with a TE such that protons in fat are out of
phase with protons in water. On a 1.5-T machine, this occurs at approximately 7, 12, or 17 ms. The
3 of 9 19-04-2010 23:57
out-of-phase time is inversely proportional to the field strength of the instrument. For example, the TE
for fat and water out of phase is 11 ms at 1 T.
Voxel Size
All evaluations of vascular disease benefit from increased spatial resolution. Reducing voxel size in
MRA also has an additional benefit. Incoherent flow leads to phase cancellation and, therefore, loss of
signal. This signal loss can be reduced by using smaller voxels (i.e., higher spatial resolution) to
diminish the amount of intravoxel dephasing. Therefore, a small voxel is often used to better visualize
small vessels, given a sufficient signal-to-noise ratio (SNR). Three-dimentional sequences usually give
a better SNR for small voxels than do 2D sequences.
Add to lightbox
Figure 27-7 Three-dimensional time-of-flight (3D TOF) images of the circle of Willis (TR/TE/FA =
35/7/20°) A, with and B, without magnetization transfer (MT) (frequency offset = 1500 Hz). Note that
MT increases background suppression and improves depiction of smaller vessels (arrows).
Background Tissue Suppression

Stationary tissue can be selectively suppressed by using magnetization transfer (MT) pulses. An MT
pulse is an off-resonance RF pulse that saturates the protons in bound water, which have a broad
resonance peak compared with the sharp, narrow peak of protons in free water. These pulses
selectively remove signal from stationary tissue while causing only slight attenuation of the blood
29
signal (Fig. 27-7). Because of the diffusion of water, saturated bound protons exchange with
free-water protons causing some of the free water to become saturated as well. Magnetization
transfer tissue suppression is most effective when the ratio of bound to free water in a tissue is large.
It has been particularly effective in intracranial angiography, in which the large number of bound
protons in brain parenchyma contrasts with the mostly free water in blood serum.
Presaturation
Arteries and veins are often companion structures. An unobscured view of the arteries, therefore,
requires removal of venous signal and vice versa. This may be readily achieved by saturating the
unwanted vessel upstream from the region of interest, using a presaturation band (Fig. 27-8). For most
applications, the presaturation band should be close to the acquired slice. A presaturation band that
moves in tandem with the acquired slice in sequential 2D TOF is termed a walking or traveling
saturation band.
4 of 9 19-04-2010 23:57
Add to lightbox
page 700
page 701
Figure 27-8 A coronal projection from axial 2D time-of-flight (TOF) through the popliteal trifurcation in a
patient (TR/TE = 35/10) acquired A, without and B, with venous saturation to demonstrate the use of
spatial presaturation to eliminate venous signal. Without venous saturation, the veins obscure the
patient's arteries.
Saturation bands may also be used to reduce flow artifacts or artifacts resulting from respiratory
30
and/or cardiac motion. For example, a saturation band can be applied to the anterior chest wall to
minimize ghost artifacts related to breathing.31,32 Saturation bands can be used to good effect to
determine the feeding vessels of an arteriovenous malformation, the cervical vessels supplying a
33
middle cerebral artery, or the direction of flow. When a saturation band is placed over the vessel, all
branches downstream from it disappear from the angiogram.
Two- and Three-Dimensional Time-of-Flight

Time-of-flight may be implemented as either a 2D or a 3D sequence. Two-dimensional TOF consists of
individual, thin sections usually acquired in a sequential fashion.34 As pointed out earlier, these sections
are often positioned perpendicular to the direction of the blood flow to maximize inflow effects. After
the first section is acquired, a second acquisition is made adjacent to the first, followed by a third, and
so on. Note that this method of acquisition, in which slices are acquired sequentially, is different from
the familiar multislice 2D acquisition used in spin-echo imaging, in which the slices are all acquired at
the same time. Sequential 2D acquisition maximizes the inflow effect.
Three-dimensional TOF has substantially better spatial resolution than 2D TOF and has many useful
clinical applications.35,36 As with any 3D sequence, the volume is subdivided into smaller sections
called partitions. Also, as with any 3D sequence, the partitions are typically quite thin (<1 mm) and are
contiguous, with no interslice gap. The increase in resolution, however, comes at a price. Blood must
enter a larger volume and travel a longer distance within the imaging volume. The inflow effect persists
for only a short distance into the volume before the signal from blood is lost and the vessel intensity
fades. The challenge is, therefore, to provide RF pulses that adequately suppress the background
tissue without noticeably diminishing the blood signal. The selection of an optimal RF pulse is especially
difficult for imaging vessels with slow flow. Use of 3D TOF is, therefore, limited to vessels with
moderately high rates of flow.
The progressive saturation of blood as it flows across a thick 3D-TOF slab often results in high blood
signal intensity at the entry side of the slab and lower signal intensity at the exit side. This gradient in
blood intensity may be partly corrected by gradually increasing the FA across the slab, a technique
called tilted optimized nonselective excitation (TONE).37 Lower FA values are applied at the entry side
5 of 9 19-04-2010 23:57
of the imaging volume to minimize saturation effects, and higher FA values are applied at the exit side
to maximize blood signal more distally. In a TONE 3D TOF axial acquisition of the circle of Willis, for
instance, vascular structures at the inflow end of the slab (e.g., carotid siphons) can be made as bright
as structures at the outflow end (e.g., small distal vessels). Although this technique is effective at
evening out the angiographic intensity across a 3D volume, it does not eliminate the tendency for
vessels with unusually slow flow to become saturated. It also presumes that flow crosses the slab in
one direction, which may not be the case for tortuous vessels such as the middle cerebral arteries.
Comparison of Two-Dimensional and Three-Dimensional Time-of-Flight
Add to lightbox
page 701
page 702
Figure 27-9 A pair of images comparing the degree of spin saturation in 2D versus 3D MRA. A,
Coronal projection from an axial 3D acquisition in the area of the popliteal artery. Note the rapid loss
of vascular signal. The 3D acquisition was acquired with TR/TE/FA = 40/7/20°, effective slice
thickness = 2 mm, and 64 partitions. B, Coronal projection of 2D axial acquisitions (TR/TE/FA =
44/10/40°, and slice thickness = 3 mm). The degree of spin saturation is reduced in the 2D compared
with the 3D acquisition.
6 of 9 19-04-2010 23:57
Add to lightbox
Add to lightbox
Figure 27-10 A pair of images comparing the extent of intravoxel dephasing in 2D versus 3D MRA of
a patient with proximal internal carotid artery stenosis. A, Sagittal projection from 2D axial acquisitions
7 of 9 19-04-2010 23:57
(TR/TE/FA = 35/9/35°, slice thickness = 2 mm). Note the complete loss of vascular signal in the area
of the stenosis (arrow). B, Sagittal projection from a 3D axial acquisition from the same patient
(TR/TE/FA = 35/7/25°, effective slice thickness = 1 mm). Reduction of signal loss caused by
turbulence in the 3D acquisition resulted in a more accurate depiction of the true vessel lumen (arrow).
Add to lightbox
Figure 27-11 Multiple overlapping thin-slab acquisition (MOTSA) acquisition: a coronal projection of
four overlapping axial 3D thin slabs of the intracranial circulation. This multislab sequential 3D
technique offers coverage of an extended volume of interest.
Some applications are best performed with 2D TOF, others with 3D TOF. Seldom are these
sequences interchangeable. The choice of 2D or 3D techniques is made on the basis of several
factors:
2D TOF is sensitive to slow blood flow, whereas 3D TOF often cannot detect vessels with
slowly flowing blood such as veins and peripheral arteries (Fig. 27-9). This results from the
difference in thickness between the volumes into which blood must enter: a thin slice in the case
of 2D, a thick slab in 3D.
2D TOF can be acquired with breath-holding to minimize respiratory motion artifacts. This
makes 2D a favorite sequence for abdominal venous studies, because the anatomy may be
covered by a series of breath-holds. 3D TOF requires a longer scan time and is typically not
compatible with breath-holding. Most applications that require breath-holding and 3D coverage
are best addressed with CE-MRA methods, described later.
3D TOF provides higher spatial resolution than 2D TOF.38 The partitions of 3D acquisition are
usually less than 1 mm thick, which compares favorably with the 2-mm or thicker slices of 2D
acquisition.24
3D TOF is less susceptible to turbulent signal loss (e.g., within a stenotic segment) than 2D
TOF (Fig. 27-10). Since 2D methods require thinner slices, they typically require stronger slice
selective gradients than do 3D methods and, consequently, longer echo times are needed to
attain flow compensation. The shorter echo time and the smaller voxel size in 3D TOF makes
that method less sensitive to disordered flow effects. Thus, one can select a 3D TOF
acquisition to grade the percent stenosis within an internal carotid artery lesion but use a 2D
TOF sequence to reliably visualize slow flow beyond a stenosis.
A technique that seeks to combine the advantages of 2D and 3D acquisition is the 3D multislab
approach, also known as multiple overlapping thin-slab acquisition (MOTSA).39 As the name indicates,
this consists of a series of thin-slab 3D acquisitions (Fig. 27-11). Thin slabs are chosen to reduce
blood saturation and improve sensitivity to slow flow. The slabs are acquired with overlap because the
slices at the end of a 3D slab often have low signal intensity. In regions of overlap, the partition with
the brightest signal is retained. Multiple overlapping thin-slab acquisition is often used when
high-resolution images are required over a large anatomic area, for instance, the aortic arch, carotid,
and intracranial arteries.
8 of 9 19-04-2010 23:57
© 2010 Elsevier
9 of 9 19-04-2010 23:57
SEGMENTED k-SPACE ACQUISITIONS

Magnetic resonance angiography of the thorax poses a greater challenge than intracranial MRA. This
is due to motion artifacts caused by cardiac pulsation and respiration. Removal of such ghost artifacts
using TOF methods requires fast acquisitions, for instance, high-speed cardiac-triggered MRA with a
segmented k-space data acquisition scheme. With segmented k-space, the entire cardiac-triggered
acquisition can be completed within a single breath-hold, and such a technique is especially effective
40,41
for thoracic studies.
The substantial reduction in scan time is achieved by acquiring N phase-encode lines per segment (i.e.,
multiple phase-encode lines are acquired in each heartbeat). The saving in scan time is determined by
N. For example, for N = 16, the scan time for a triggered acquisition is decreased by a factor of 16.
Faster acquisition is achieved at the expense of precise electrocardiographic triggering to a single time
point within the cardiac cycle. When performing segmented acquisitions, the time interval over which
the N echoes are acquired spans a portion of the cardiac cycle. Typically, for TR = 10 ms, the
acquisition window spans 160 ms for N = 6. This time interval increases with increasing N. When
imaging moving objects, care should be taken to find the optimal N because there is more time for
blurring to occur as N is increased. An acquisition window of 300 ms is acceptable for renal artery
imaging but not for coronary artery imaging, where cardiac motion is a problem.
page 702
page 703
Add to lightbox
1 of 4 19-04-2010 23:58
Add to lightbox
Figure 27-12 A and B are two different volume-rendered views of a whole-heart coronary MRA data
set, showing the left anterior descending, the left circumflex, and the right coronary arteries. The data
set was acquired without administration of contrast agent. Contrast was obtained by applying
prepulses saturating the signal of fat and myocardium, but leaving unaffected the blood signal. Image
acquisition (steady-state free precession, TR = 5.0 ms; TE = 2.5 ms; flip angle; 110°, matrix; 256 ×
256; 160 slices; spatial resolution; 1.0 × 1.0 × 1.5 mm3) was limited to a short acquisition window
(140 ms) in end-diastole to limit the effects of cardiac motion. A navigator beam, placed on the right
hemi-diaphragm, provided prospective correction of respiratory motion and facilitated free-breathing
during the 12-minute acquisition.
By using segmented acquisitions, cardiac-triggered acquisitions that previously took 3 to 5 minutes can
now be completed within 20 seconds. When segmented acquisition is combined with breath-holding,
both cardiac pulsations and respiratory motion are eliminated. Segmented k-space has been applied to
the acquisition of excellent MR angiograms of the coronary artery. The major advantages of
segmented data acquisitions include 1. speed; 2. elimination of physiologic motions; and 3. syncopated
data acquisition for greater inflow (syncopation is discussed in the next paragraph).
Triggering and Syncopation

Acquisitions are syncopated in segmented k-space acquisition; that is, RF pulses are applied in
clusters with a relatively long pause between clusters. For example, with N = 8, eight RF pulses are
applied with a TR of 4 to 8 ms; then the next set of RF pulses are applied in the following heartbeat,
which is approximately 1 second later. The pause between the RF pulses allows flow of fresh spins
into the imaging slice. As a result, there is more inflow for segmented data acquisition than for
standard TOF imaging. Pulmonary angiography, for example, has been achieved with a series of
segments separated by a delay time to allow wash-in.55 This use of segmentation, triggered to the
heartbeat or acquired at a constant delay independent of the heart rate, is called syncopated
acquisition.
Magnetization Preparation
To tailor the desired image contrast, magnetization preparation pulses are used in combination with
segmented k-space sequences. These preparation pulses are applied before each data segment to
achieve certain purposes. For example, fat saturation pulses are used to suppress adipose tissue. An
2 of 4 19-04-2010 23:58
inversion pulse with an appropriate delay time can also be used to suppress the background tissue.42
In approaching coronary artery imaging, the practitioner must make use of many of the tools that are
43,44
available in MRI (Fig. 27-12). Preparation pulses are useful for eliminating the adjacent signal from
fat and from myocardium. In order to reduce motion problems, cardiac triggering is used, together with
navigator pulses that monitor the excursions of the diaphragm and enable data acquisition only when
the diaphragm is in a specific position.
Dark Blood Imaging

page 703
page 704
Add to lightbox
Add to lightbox
Add to lightbox
Figure 27-13 Multicontrast image of carotid atherosclerosis. A, A maximal intensity projection (MIP) of
a 3D MR angiogram. B, A single multi-planar reformatted image in the plane of the bifurcation shows
3 of 4 19-04-2010 23:58
both the bright lumenal signal, and a dark line (arrows) along the outside of the vessel wall. C, An axial
fat saturated double inversion fast spin-echo black blood image of the same patient shows the
atherosclerotic plaque (arrow) surrounding the lumen.
Standard TOF MRA obtains images in which flow appears bright. Loss of flow signal can, however,
arise in the vicinity of a stenosis as phase incoherence results from disrupted flow. A way to overcome
this problem is to perform dark blood imaging, in which flow is made to appear dark while a high signal
intensity from stationary tissue is maintained45-50 (Fig. 27-13). Dark blood can be achieved using
presaturation bands or preparation pulses to null magnetization in blood before it enters the imaging
volume. The most effective preparation method is a double inversion technique where all magnetization
is inverted followed by rapid re-inversion of the slice of interest only. An inversion delay time then
follows that is designed such that the magnetization of blood is null when the imaging RF pulses are
applied. Flowing blood, therefore, appears dark and stationary tissue in the slice appears as bright as
it would if it never experienced any preparation pulses. Spin echo or fast spin echo using a thin slice,
and a dephasing gradient are particularly well suited to further eliminate signal from flowing blood while
retaining strong information on the vessel wall. The objective in doing dark blood imaging is to eliminate
the flow signal completely while maximizing the signal from plaque and surrounding tissue.
Dark blood imaging has not achieved wide use for angiographic type purposes. This arises from the
difficulty in rendering slow or recirculating flow dark, and because calcified plaque has low signal
intensity, which may be indistinguishable from flow. Nevertheless, it is an extremely powerful method
for examining the vessel wall.
© 2010 Elsevier
4 of 4 19-04-2010 23:58
PHASE-CONTRAST ANGIOGRAPHY
Add to lightbox
Figure 27-14 The transverse magnetization is characterized by both its magnitude (M) and phase (ϕ).
The phase is defined as the angle of rotation of the transverse magnetization from a reference axis.
Phase-contrast angiography (PCA) is based on the principle that moving spins develop a phase shift
with respect to stationary spins as they move in a pair of opposing magnetic field gradients. This
phase shift ϕ(t) is related to the rate of motion and can be expressed as:
where γ is the gyromagnetic ratio, x(t) is the time-varying
position along the gradient axis, and G(t) is the magnetic field gradient. The shift may be used to
indicate flow or calculate flow velocities, or it may be displayed as an angiogram. 51 Phase contrast
images are most commonly acquired as either thick-slice 2D PC MRA, used for scout views or rapid
projections that can be viewed from one direction only; 3D PC MRA, used for higher resolution and
rotatable angiograms; or cardiac-gated 2D PC flow velocity imaging, used for blood flow quantitation.
Basic Mechanisms
One simple principle underlies the use of PC imaging: for constant-velocity flow, the amount of phase
shift increases linearly with flow velocity. Spin magnetization is a vector that has both magnitude and
direction. The phase (angle) is given by the rotation of the transverse magnetization relative to a
reference axis. As a spin moves along a magnetic field gradient, it precesses faster when it
experiences a higher magnetic field and its phase changes. As a result, the amount of motion is
52,53
encoded in the phase. Unlike TOF imaging, which displays only the magnitude component of the
magnetization, PC utilizes the phase information to obtain MR angiograms, images of directional flow,
and flow quantification.
In the presence of a pair of bipolar gradient lobes, stationary spins develop no net phase shift,
whereas spins moving along the direction of the gradients develop a net phase shift (Fig. 27-14). This
net phase shift ϕ is given by:
where γ is the gyromagnetic ratio, V is the velocity along the direction of the gradient, G is the gradient
strength, τ is the gradient duration, and T is the time interval between the gradient lobes. As indicated
by the equation, the phase shift is proportional to the velocity, the gradient strength, and the gradient
duration.
page 704
page 705
Flow velocity can be readily quantified based on the linear relationship between velocity and phase
54
shift. To generate the phase information, a magnetic field gradient is intentionally applied. Such
flow-encoding gradients can be applied along any axis, such as the frequency, phase-encode, or slice-
select direction. Only motion along the flow-encoding direction, however, results in a net phase shift.
Magnetic resonance flow capabilities are a powerful tool for evaluating the physiological significance of
a diseased vessel (Fig. 27-15).
1 of 12 19-04-2010 23:59
Sensitivity of the Velocity-Encoding Gradient

The phase (angle) is measured in units of degrees or radians. Its range is finite: the maximal encoded
phase shift is +180° for forward flow and -180° for reverse flow (when symmetric flow encoding is
used). That is, if a velocity-encoding range (Venc) of 80 cm/s is chosen, spins flowing forward at a
velocity approaching 80 cm/s accumulate 180° phase shifts and are assigned the brightest signal
intensity and backward flow approaching -80 cm/s is assigned the darkest signal intensity. The Venc of
the pulse sequence can be adjusted by varying the gradient strength and duration. The sequence can
be tailored to image fast flow or slow flow. The Venc is particularly important when imaging slow flow
(Fig. 27-16). When the vessel of interest is anticipated to have a peak velocity of 15 cm/s, reducing the
Venc from 100 to 20 cm/s improves the dynamic range of the study. This is true for both MRA and flow
quantification. Choosing the proper Venc results in better visibility of vessels in PC MRA.
One should always choose a Venc that is slightly higher than the maximal anticipated velocity in the
vessel of interest. When the peak velocity exceeds the Venc, velocity aliasing results and artifacts are
produced within the vessel of interest. For example, when the chosen Venc is 20 cm/s and the blood is
traveling at 21 cm/s, the image of blood is not assigned to the brightest signal plus one. Rather, the
signal is wrapped around and has a low intensity in the image. This gives a misleading impression that
flow is going in the opposite direction. Velocity aliasing, when present, can be readily detected in the
image. Normal phase images show a gradual transition of phase within the vessel lumen. Velocity
aliasing, on the other hand, shows abrupt transition of phase from positive to negative within a vessel.
When that occurs, the study should be repeated with the correct Venc.
Correction of Phase Errors

Magnetic field inhomogeneities and eddy currents can induce local phase shifts. When these are
present, stationary spins no longer have zero phase. To achieve background suppression, a
subtraction of two phase images is performed: two phase images are acquired, one with flow
encoding and one with flow compensation. Stationary spins have the same phase in both images. A
subtraction of the two phase images produces zero phase shift for the stationary material.55,56 Flowing
spins, on the other hand, have a phase shift in the flow-encoded image but not in the
flow-compensated image. As a result, a subtraction of the two images gives a net phase shift for the
flowing spins and each pixel is then assigned a signal intensity that increases with the phase shift. An
alternative way is to subtract two images, one acquired with a pair of bipolar gradient lobes and the
second acquired with the polarity of the gradient lobes reversed. This also produces a net phase shift
for flow while subtracting out the background (Fig. 27-17).
Based on the linear relationship between phase shift and flow, fast-flowing spins are depicted with a
brighter signal than are slower moving spins. The advantage of the first technique is that the same
acquisition without encoding may be subtracted from each of the three encoded acquisitions, assuming
three-directional sensitivity, thereby reducing the total acquisitions from six to four.
Speed, Complex Difference, and Phase Difference Display

Once the phase shifts resulting from flow encoding have been acquired, they may be displayed in
several ways.
Net phase shift (phase difference) is used for flow quantification. The phase-difference image is
computed from the difference in phase between the positive-encoded and negative-encoded
acquisitions. The value of the phase difference is proportional to velocity; therefore, this image is also
called a velocity map and is used for velocity and flow volume calculations (Fig. 27-18). Flow encoding
is performed in one direction only.
Complex difference is used mainly for 2D scout imaging. The complex difference image is the
magnitude of the vector difference (rather than the phase difference) between the positive-encoded
and negative-encoded acquisitions. Rather than being simply a reflection of velocity, it is also
proportional to the magnitude of the magnetization. As with speed images, there is no indication of
2 of 12 19-04-2010 23:59
whether flow is along the positive or negative direction of the flow-encoding gradient. When the peak
velocity exceeds the Venc, the flow signal remains bright. As with phase difference images, flow
encoding is performed in one direction only. Complex difference is used in 2D PC acquisition (when
special background spoiler gradients are employed) because it is less susceptible to phase artifacts
than are phase difference images. The complex difference obtained in these images, however, no
longer follows a simple linear relationship with velocity. As a result, velocity quantification is not so
straightforward in the complex difference image.
page 705
page 706
Add to lightbox
Add to lightbox
3 of 12 19-04-2010 23:59
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
page 706
page 707
Figure 27-15 Phase mapping used to evaluate functional significance of a coarctation. A,
Dual-inversion black-blood turbo-spin echo MRI of the aorta shows coarctation of the descending
aorta. Phase-contrast flow measurements were performed immediately downstream of the coarctation
(B, modulus image; D, phase [velocity] image) as well as at the level of the diaphragm (C, modulus
4 of 12 19-04-2010 23:59
image; E, phase [velocity] image). F, Time-versus-flow curve demonstrates delayed pulse wave arrival
and lower flow values at the level of the diaphragm. Flow was determined to be 33 ml/s at the
coarctation, and 29 ml/s at the diaphragm level. These values indicate absence of collateral flow; the
coarctation was thus diagnosed not to be flow limiting. However, aortic anatomy causes considerable
backflow below the coarctation, as seen by the bright signal intensity in the descending aorta. At the
level of the diaphragm, flow is directed towards the feet throughout the aorta (dark signal in D and E).
Add to lightbox
Figure 27-16 Sagittal 2D thick slab of the head with Venc = 30 cm/s. The use of a reduced Venc
improves depiction of structures with slow moving blood such as the sagittal sinus.
5 of 12 19-04-2010 23:59
Add to lightbox
page 707
page 708
Figure 27-17 Diagram showing the use of bipolar gradient lobes in phase-contrast flow quantification
to obtain zero phase shift for the stationary tissue and a net phase shift for flow.
Add to lightbox
6 of 12 19-04-2010 23:59
Add to lightbox
Add to lightbox
7 of 12 19-04-2010 23:59
Add to lightbox
Figure 27-18 Flow measured by cine phase contrast of a patient with occlusion of the upper left
pulmonary artery. MRA (A) and dual-inversion black-blood TSE (B and C) show complete occlusion of
the upper left pulmonary artery. The right pulmonary tree is normal. D, Flow measurements in the left
and right pulmonary arteries show dramatically reduced flow on the left side.
Speed information is often used to generate an angiogram. The speed image shows simply the speed
of flow. Speed is calculated at each pixel based on √(vx2 + vy2 + vz2), where vx, vy, and vz are
velocities along the x, y, and z directions. In the speed image, there is no information about the
direction of flow. Aliasing is less of a problem with speed display, because the peak positive-encoded
flow is identical in intensity to the peak negative-encoded flow. There is no sudden change from bright
to dark as the Venc is exceeded. Speed images are ideal for anatomic depictions of vascular anatomy.
Magnetization vectors within pixels that have little or no signal, such as those representing air and bone
cortex, have random and meaningless phases. Air might therefore contribute a speckled appearance
on the angiogram, obscuring the true vascular signal. This can be removed by setting pixels
corresponding to small magnetic vectors to the value of the background signal. This is called applying a
mask.
Thick-Slab Two-Dimensional Phase Contrast

page 708
page 709
8 of 12 19-04-2010 23:59
Add to lightbox
Add to lightbox
Figure 27-19 Maximum intensity projections at two different obliquities, A and B, of a 3D
high-resolution SENSE phase-contrast MRA of a whole brain using an 8-element head-coil. No
contrast agent was used. Acquisition: Complex-difference phase-contrast MRA with 350 slices; FOV,
256 × 180; matrix, 256 × 256; TR = 20 ms; TE = 5.1 ms; flip angle, 15°; Venc = 40 cm/s. Scan
resolution 1.0 × 1.0 × 1.2 mm, reconstructed to 0.5 × 0.5 × 0.6 mm. SENSE reduction factor of 6,
reducing the original scantime from 42 minutes to 7 minutes. (Courtesy of Romhild M. Hoogeveen,
Ph.D., Philips Medical Systems, Best, the Netherlands.)
Thick-slab 2D PC offers a quick assessment of flow and is often used to obtain a projective view of the
vasculature.57,58 This can be performed rapidly, unlike 3D PC (Fig. 27-19). The slab thickness ranges
from 2 to 10 cm. Because of the thicker slice, partial voluming does occur and the spatial resolution
may sometimes be suboptimal. Most importantly, when two vessels cross within the slice, the acquired
phase is the sum of the two, so various phase artifacts may be present that are worse with increasing
slice thickness. When pulsatile flow is present, multiple acquisitions are acquired to reduce respiratory
and pulsatility artifacts. Alternatively, one may use electrocardiographic triggering.
9 of 12 19-04-2010 23:59
Directional Flow Imaging

A simple and rapid application of 2D PCA is to determine direction of flow (Fig. 27-20). An image is
acquired with flow encoding in one direction. Based on image intensity, one can readily determine
whether flow is in the positive or negative direction along the flow-encoding axis.59 For example, if the
Venc is 100 cm/s, the normal convention is to assign white to blood that is flowing at 100 cm/s in the
positive direction, black to flow at -100 cm/s, and shades of gray to velocities in between. Stationary
tissue is midgray. Note that if flow in a vessel is perpendicular to the encoding axis, there is no
evidence of motion as long as there is no motion component in the encoded direction.
Flow Velocity Quantification and Cine Phase Contrast
Add to lightbox
Figure 27-20 Two-dimensional phase contrast acquisitions through the basilar artery and its distal
branch vessels. Flow encoding along the head-foot direction. Flow toward the top of the head is
encoded as white.
60
Flow quantification is a major application for PC acquisitions. Typically, a pair of phase images are
acquired (with and without flow-encoding gradients) as described earlier. To remove inhomogeneities
of the static field, the net phase shift is obtained by subtracting the flow-sensitive phase image from
the flow-compensated phase image.56 By using the linear relationship between flow velocity and the
net phase shift (Eq. 27-5), the velocity can be calculated.
page 709
page 710
Cine PC allows the measurement of pulsatile flow velocity61-63 (see Fig. 27-18). This is done by
acquiring data at multiple time points throughout the cardiac cycle (known as a cine measurement)
using electrocardiographic triggering. The time resolution of the series is given by TR, which is typically
30 to 40 ms. Multiple pairs of flow-sensitive and flow-compensated images are acquired in a triggered
study. A time-velocity curve can be obtained by subtraction of the images and determination of the net
phase shift at each time point.64 In addition, volumetric flow rates can be calculated.65-67 The
volumetric flow rate is given by the product of the measured velocity and the cross-sectional area of
the vessel lumen, which can be determined.
Rapid flow quantification has been achieved that enables measurements to be completed within a
10 of 12 19-04-2010 23:59
single breath-hold. High-speed acquisitions are particularly important for flow quantification in the
thorax, for example, measurements in the coronary arteries.68
The velocity-encoding range is an important parameter in phase acquisitions, particularly for flow
quantification. It must be chosen with care otherwise aliasing can occur, which produces misleading
results. For example, consider an acquisition in which the Venc is 20 cm/s. For symmetric velocity
encoding, a forward flow of 20 cm/s would accumulate a +180° phase shift whereas a backward flow
of -20 cm/s would accumulate a -180° phase shift. Flow that exceeds this Venc would have a phase
shift larger than 180°. For example, flow at 60 cm/s would accrue a 540° phase shift, which is
equivalent to a net phase shift of 180°, and this would erroneously suggest a flow velocity of 20 cm/s.
As another example, flow at 30 cm/s would have a phase shift of 270° (which is equivalent to -90°) and
the velocity measurement would incorrectly indicate a flow velocity of -10 cm/s (i.e., 10 cm/s in the
opposite direction). The problem of aliasing can be readily corrected by choosing a Venc that is larger
than the maximal flow velocity.
Because the measured velocity varies substantially over the pulsatile cycle, one may wish to use a
69
smaller Venc during diastole and a larger Venc during systole. This ensures increased dynamic range
70
for all the cardiac phases. Care must be taken to avoid velocity aliasing at any time over the pulsatile
cycle. It is therefore helpful to have an estimate of the anticipated time-velocity waveform when
specifying the variable Venc.
The following are the causes and remedies for the more common artifacts that can occur with MR PC
velocity quantification:
1. Velocity aliasing (phase wrapping). The peak velocity exceeds the Venc. Remedy: increase the
Venc to have it larger than the maximal velocity.
2. Local signal loss. Phase incoherence is due to higher order motions or turbulent flow. Remedy:
use a short TE and a small voxel to minimize the amount of intravoxel dephasing.
3. Background phase inhomogeneity. Local variations of phase shift are caused by magnetic field
inhomogeneity. Remedy: do careful shimming and phase map correction using a static
reference.
Three-Dimensional Phase Contrast

Three-dimensional PC MRA, like 3D TOF, generates a volumetric data set that can be reformatted
and displayed in any chosen orientation by using the maximal intensity projection (MIP) algorithm. In
addition to MR angiograms, flow directional images can be generated to show flow along different
axes. Repeated acquisitions are needed to encode flow along all three orthogonal directions. Cardiac
triggering is not used with 3D PC because of prohibitive acquisition times. As a result, the flow velocity
obtained is averaged over the cardiac cycle. Common applications of 3D PC are in cerebral
angiography, renal and other visceral angiography, and studies of distal extremities. Until recently, the
time required to collect 3D phase-contrast angiograms was prohibitively long. The implementation of
parallel acquisition strategies has helped to substantially reduce the total acquisition time needed.
Applying this strategy together with complex difference subtractions methods, acquisition times can be
reduced to of the order of 7 minutes71 (see Fig. 27-19).
Effects of Pulse Sequence Parameters

Typically, both 2D PC and 3D PC use a short TR (approximately 30 to 40 ms), a short TE (8 ms or
less), and a small FA of 20° to 30°. Saturation of blood occurs with PCA, just as it does with TOF.
However, because the image intensity is related to the phase rather than the magnitude of the
magnetization, the vessel remains bright until the magnetization is quite small or falls below the level of
the mask.
Phase incoherence related to turbulent or disorderly flow patterns can lead to signal loss, just as in
TOF. In TOF, such phase incoherence is minimized by using flow-compensated gradients. In PCA,
11 of 12 19-04-2010 23:59
however, the flow-encoding portion of the acquisition does not have flow compensation. To minimize
phase incoherence in the flow-sensitive image, a small voxel size is used, which also reduces the
effects of partial voluming.
Advantages and Limitations of Phase-Contrast Angiography

Phase-contrast angiography has several qualities that distinguish it from TOF. First, the values
portrayed in an image are phase shifts rather than magnitudes. Second, PCA is a reflection of blood
motion along a gradient rather than inflow into a slice. Third, nearly complete background elimination is
achieved, whereas in TOF the vessel wall and other structures are faintly visible.
page 710
page 711
Phase-contrast angiography enjoys the following strengths. First, it improves the detection of small
vessels by virtue of its excellent background elimination. All stationary materials, even those with a
short T1 (such as fat and methemoglobin), are eliminated. As a result, hematomas are not mistaken
for vascular structures, as may happen with TOF. Phase-contrast angiography can often provide an
72
MIP of larger volumes than TOF because of its superior background suppression. Second, PCA is
well suited for imaging of slow flow, as in vein structures or large aneurysms. By adjusting the gradient
strength and duration, one can make a pulse sequence that is sensitive even to velocities as slow as
that of cerebrospinal fluid motion. Third, cine PCA can be used to perform flow quantification and to
obtain the velocity-time curve in pulsatile flow.
Phase-contrast angiography also has limitations. First, PC acquisitions, particularly in 3D PCA, take
approximately two to three times longer than TOF acquisitions of comparable resolution. This is
because flow-encoded data along each of the three axes are separately acquired. Second, PC MRA is
sensitive to instrument imperfections. The presence of eddy currents or field instabilities degrades the
background suppression. The use of self-shielded gradient coils and careful eddy current
compensation alleviate these problems. Having to partially dephase flowing spins to obtain flow
sensitivity makes PC images more sensitive than TOF images to artifacts resulting from complex flow,
which may occur distal to a stenosis or in some aneurysms.
© 2010 Elsevier
12 of 12 19-04-2010 23:59
CONTRAST-ENHANCED MAGNETIC RESONANCE ANGIOGRAPHY

In recent years, the development of CE-MRA methods has burgeoned. The great majority of this
development has been achieved using the extravascular paramagnetic contrast agents commonly used
in clinical practice, such as Gd-DTPA. The principle mechanism that is exploited with the use of
contrast agents is the dramatic reduction (nearly an order of magnitude) in the T1 relaxation time of
blood that is attained.73-76
The successful accomplishment of excellent MRA results using contrast agents relies on careful timing
to capture the high magnetization strength that is associated with the first pass of the injected contrast
agent, and to avoid the enhancement of venous signal that can obscure the arteries of interest. This in
turn requires the acquisition of a 3D volume in the relatively short time when those conditions are
met-typically of the order of 20 s. In order to accomplish this, sequences with very short TRs, of the
order of 5 ms or less, and correspondingly short TEs, of the order of 1 ms, are needed. In order to
achieve these extremely short echo times, it is necessary to make the gradient waveforms as short in
duration as possible, and motion compensation is discarded to accomplish this. The conventional
assumption is that the echo times used in CE-MRA are sufficiently short that little dephasing from
motion can occur. There is, however, some indication that this is not always the case and that signal
loss can occur, particularly when the longest duration gradient waveform, the frequency encoding
gradient, is aligned with the direction of high flow velocities.28
It is only relatively recently that clinical scanners have been widely available that are capable of the
high-gradient performance needed for these applications. Subtraction of pre- from post-contrast
images in the absence of motion of the patient gives excellent background suppression and better
vessel conspicuity than are possible using standard MRA techniques. Contrast-enhanced vessels
appear bright regardless of the presence of slow flow.
Acquisition Timing
Add to lightbox
1 of 7 20-04-2010 00:00
Add to lightbox
page 711
page 712
Figure 27-21 Importance of timing in contrast-enhanced MRA. Maximal intensity projections of two
consecutive 25 s acquisitions of the carotid arteries. A, Appropriate timing with strong arterial and low
venous signal. B, Arterial and venous signal are similar intensity and difficult to differentiate.
Add to lightbox
2 of 7 20-04-2010 00:00
Add to lightbox
Figure 27-22 Importance of timing in contrast-enhanced MRA. Two consecutive MRA runs after
intravenously administered contrast agent. A, The first dynamic maximal intensity projection (MIP)
shows mainly the true lumen, and only faintly the origin of the false lumen. B, The later dynamic MIP
(acquired 30 s after A) shows remaining contrast in the false lumen, and recirculating contrast agent in
the true descending aorta.
Suitable timing is essential for successful CE-MRA. A number of strategies have been adopted for
ensuring the correct timing interval between injection of the contrast agent (typically into the antecubital
vein), and initiation of the MR acquisition sequence. The first strategy is to use a test bolus of
approximately 2 ml and to use a rapid T1-weighted sequence to scan repeatedly at a single level at a
76,77
rate of about one image per sequence to determine the delay between injection and arrival. The
full contrast injection (of the order of 20 ml) is then initiated and acquisition started using the delay
interval determined from the test bolus run. With correct timing, a strong arterial signal is noted. If the
sequence initiation is delayed, there is a drop in arterial signal intensity and an increase in venous
signal (Fig. 27-21). In order to most effectively capture the peak signal intensity, a common practice is
to order the collection of k-space to acquire the central lines of k-space at the beginning of image
acquisition when magnetization strength is highest-a technique referred to as elliptic-centric
reordering.78-81 Timing to avoid venous contamination is important because the strategy of
presaturation that is so effective for TOF imaging is ineffective in eliminating the signal from contrast-
enhanced venous blood because of its very short T1.
A second strategy is to again use a rapid acquisition, typically with a thick 2D slab over the vessels of
interest. The full injection of contrast material is delivered, and, when arrival of contrast is detected at
80,82
the target area, the full 3D CE-MRA sequence is initiated. This approach is more efficient than the
bolus timing approach as only one acquisition is used. The method also is based on the true arrival of
the contrast agent rather than on the arrival of a small test bolus. However, at present most scanners
have a latency period of 2 to 3 s, and there is a short interval when magnetization enhancement occurs
where that signal is not being sampled. The decision to initiate the scanning sequence is also a
judgment call, and if an error is made there is no option to repeat the study.
The dynamic passage of contrast material also offers possibilities to examine vascular abnormalities
such as delayed filling of vessels using repeated scans. This is particularly useful in conditions such as
3 of 7 20-04-2010 00:00
aortic dissection (Fig. 27-22).
The requirement that data acquisition be completed before there is substantial filling of venous blood
imposes a limit on the total duration of the injection. Because this duration is relatively short (of the
order of 20 s), a compromise must be reached whereby image resolution is somewhat lower than in
conventional TOF MRA. This loss in resolution is generally set along the direction of the slab partitions.
Data interpolation is imposed along that direction to compensate for some of this resolution loss. 83
Insensitivity to Saturation
One of the major advantages of CE-MRA is that, following administration of the contrast, the
magnetization of blood recovers so rapidly that it is virtually unaffected by the number of RF pulses
that it is subjected to. This is in noted distinction to TOF MRA. The insensitivity to magnetization
saturation permits acquisition of images where the imaging volume covers vessels over a large field of
view. This feature is extremely powerful in applications such as runoff studies of the vessels of the
lower extremities (Fig. 27-23). A contrast-enhanced acquisition can be used in combination with a
moving table to chase the injected bolus from the pelvis down through the legs.84-86 Total coverage of
those vessels can be achieved with acquisitions at three stations to complete the full study in under 2
minutes (compared with over 20 minutes for a successful TOF acquisition).
page 712
page 713
Add to lightbox
4 of 7 20-04-2010 00:00
Add to lightbox
Figure 27-23 A comparison of A, contrast-enhanced (CE) MRA and B, 2D time-of-flight (TOF) in a
run-off examination of the lower extremities. Both acquisitions show the important regions of
pathology. The 2D TOF study was acquired with triggering for inflow enhancement and took 25
minutes. The CE-MRA study (with slightly better coverage) took only 2 minutes.
In regions with extremely slow or recirculating flow, conventional TOF methods result in substantial
87
saturation and are unsatisfactory for visualizing structures such as aneurysms. Again, the rapid
relaxation of contrast-enhanced blood permits excellent visualization of these structures (Fig. 27-24).
Signal Enhancement
5 of 7 20-04-2010 00:00
Add to lightbox
Figure 27-24 Maximal intensity projection (MIP) of contrast-enhanced (CE) MRA of a giant fusiform
aneurysm of the basilar artery. Flow in these aneurysms is characterized by slowly rotating vortices
that saturate in time-of-flight acquisition but is well depicted with a 23 s CE study.
Add to lightbox
6 of 7 20-04-2010 00:00
Add to lightbox
page 713
page 714
Figure 27-25 Comparison of acquisition from carotid arteries using A, standard vendor-supplied coil,
and B, custom-designed phased-array carotid coil providing an improvement of a factor of 3 in the
signal-to-noise ratio at the carotid bifurcation.
Many features of CE-MRA sequences are not conducive to high SNRs. These include a high receiver
bandwidth to reduce echo time and the reduction in total number of phase-encoding steps to reduce
total acquisition time. Like any other MRI study, MRA methods benefit from receiver coils that have
increased sensitivity. Such coils have been used to good advantage in studies of the extracranial
88
carotid arteries (Fig. 27-25). Further, the recent innovations in parallel imaging strategies, which use
multiple coil elements to increase acquisition speed, have been used to particularly good benefit in
studies of dynamic processes such as passage of a contrast bolus. Strategies such as SENSE and
SMASH provide the ability to substantially speed up acquisitions with benefits either in signal strength
or in spatial resolution.
© 2010 Elsevier
7 of 7 20-04-2010 00:00
TEMPORAL INTERPOLATION SCHEMES

The dynamic passage of contrast material through the vascular system provides a powerful means to
assess different territories at different times, and to assess issues such as relative delays in passage
of blood through a given region. A number of innovative methods have been designed to provide
multiple 3D data sets at rapid temporal intervals. In order to reconstruct a 3D data set, it is necessary
to collect a full 3D k-space data set. A number of strategies have been devised that have a relatively
rapid resampling of certain portions of k-space (typically the central portion, which is most important in
determining overall signal contrast).89-94 The interval between consecutive samplings of the center of
k-space then defines the time between different reconstructed 3D image data sets. At any one of
these specific sampling times, only the central portion of k-space and a small portion of the remaining
k-space are sampled. All remaining segments of k-space needed for reconstruction at a given time
point are derived by interpolating from the corresponding k-space data acquired at neighboring time
points. A variety of different specific implementations of this approach have been pursued, including
time resolved imaging of contrast kinetics (TRICKS), or the keyhole approach (Fig. 27-26). Using a
combination of temporal interpolation and parallel imaging methods, 3D data sets can be acquired in a
few seconds.
© 2010 Elsevier
1 of 1 20-04-2010 00:00
DISPLAY AND INTERPRETATION

The introduction of the MIP algorithm first allowed realistic angiograms to be calculated and
displayed.95 There is a wealth of information available, and clear visualization of anatomy is provided
by the ability to view data from different perspectives, and also to restrict the data viewing to specific
volumes of interest (Fig. 27-27). The MIP allows individual slices of blood vessels, acquired over a 3D
96
volume, to be combined and rendered in a 2D projection (Fig. 27-28). Projection angiograms can be
generated from either a series of 2D images or a 3D data set.97,98 The method has become so
universal that it is nearly synonymous with MRA. Yet MIP is not always the best technique for vascular
display. Here, the most important features of these methods are summarized.
Maximal Intensity Projection

Blood vessels course across many slices in a typical angiographic acquisition. To display the entire
vessel, one must combine the slices. If one simply sums the intensities on the slices, an immediate
problem presents itself-the background signal added from the many slices exceeds that of the vessels,
with the result that vessels are no longer visible. The MIP algorithm was designed to overcome this
problem. Rather than adding intensities, it uses only the largest single intensity at each position in the
projection in the combined image.
First, the direction of projection is chosen; the computer algorithm then goes through the 3D data along
that direction and automatically picks out and displays the highest signal intensity from the slices.
Because the selection is based on signal intensity alone, the program does not differentiate between
blood and background tissue and display of blood vessels can, therefore, be uneven.
Although MIP is much better than simple pixel addition, weak vascular features can be lost in the
projection because the vessel signal intensity is exceeded by the background signal intensity. Vessels
may appear narrowed or absent on an MIP. Therefore, when interpreting a study, the original images
should be consulted to determine the true vessel width.
Vessel depiction can be improved by targeted MIP, in which a selected volume of the 3D data is
projected (Fig. 27-29). By limiting the projection volume, signals from other vessels or bright structures
are removed, and the chance that vessel intensity is exceeded by background intensity is reduced.
Another approach that can be helpful is to first perform multiplanar reconstructions of the 3D data and
then obtain an MIP of these reconstructions.
Vessel Tracking
In vessel tracking one tries to identify the signal corresponding to blood vessels and then remove all
nonvascular voxels. This is done by picking out an object with contiguous voxels that have signal
intensity over a chosen threshold. With the program, a volume is separated into different contiguous
objects, many of which are vessels.99 The nonvascular objects are then edited out by the operator.
The algorithm may be assisted by "seeding," in which a point in the vessel is selected by the operator.
The computer then looks for bright voxels contiguous with that seed. These voxels in turn are used as
seeds for the next iteration. The object grows outward until it reaches the vessel wall. The progression
of the seed is based on signal comparison between neighboring voxels and is governed by criteria set
by the user.
Surface Rendering
Once the voxels corresponding to a vessel are identified, and assuming the angiogram is of adequate
resolution, one may calculate the position of the outer margin of the lumen. This can be rendered into a
realistic surface with shading and perspective, as if a model had been made of the vessel. 100 Although
still uncommon, this method of presentation is effective for displaying complex anatomy (Fig. 27-30).
page 714
page 715
1 of 7 20-04-2010 00:01
Add to lightbox
Add to lightbox
Add to lightbox
2 of 7 20-04-2010 00:01
Add to lightbox
Add to lightbox
Add to lightbox
3 of 7 20-04-2010 00:01
Add to lightbox
Add to lightbox
Add to lightbox
page 715
page 716
Figure 27-26 Time-resolved MRA using keyhole imaging for accelerated dynamic scanning. With this
technique, only the center of k-space, which defines contrast behavior, is updated for each dynamic
data set. The outer regions of k-space, which define edge definition and spatial resolution, are
sampled only once (during the last dynamic) and copied to the other data sets. Keyhole technique
allowed a time-resolved dynamic scan time of 2.5 s (instead of 7.5 s without keyhole) for a full 3D data
set (matrix, 288 × 216; 50 partitions; FOV, 475 mm; TR = 2.8 ms; TE = 0.9 ms; flip angle, 20°; SENSE
= 2.0 in phase-encoding direction, and SENSE = 1.5 in slice-selection direction). A-I, The passage of
the contrast agent is shown through the right heart, the pulmonary system, the left heart, the aorta, and
4 of 7 20-04-2010 00:01
the arterial system.
Add to lightbox
Add to lightbox
5 of 7 20-04-2010 00:01
Add to lightbox
Add to lightbox
Figure 27-27 A and C are orthogonal maximum intensity projections (MIPs) of a contrast-enhanced
neck and brain MRA. Frontal projection shows a stenosis in the right internal carotid artery. B, The
stenosis is obliterated in the left-right projection due to signal originating from the left internal carotid
artery. D, MIP of the right system only clearly shows the stenosis.
Source Display
The original source images are important for determination of vessel widths and the degree of
101
stenosis. The true lumen size is better evaluated in the cross-sectional views than in the projections.
Furthermore, vessels with slow flow are always better visualized on the original images. It is prudent to
photograph both MIP and selected source images when interpreting a study.
6 of 7 20-04-2010 00:01
© 2010 Elsevier
7 of 7 20-04-2010 00:01
PATTERNS OF BLOOD FLOW AND THEIR APPEARANCE ON MR

ANGIOGRAMS
page 716
page 717
Add to lightbox
Add to lightbox
Add to lightbox
1 of 4 20-04-2010 00:02
Add to lightbox
Figure 27-28 A set of four images showing three source images of the intracranial vessels (A-C) and
(D) the MIP.
Blood flow can sometimes be disorderly, turbulent, or highly pulsatile. Furthermore, blood can follow a
tortuous course. In some instances, the depiction of the vessel lumen in MRA may not resemble the
true shape or size of the lumen. The image may suggest disease when none is present or lead to an
overestimation of disease. These abnormal patterns are predictable and can be recognized when they
are present.
Turbulence and Phase Dispersion

The term turbulent flow in the context of MR means chaotic or incoherent flow in which there is
considerable mixing of spins, which no longer travel along straight streamlines. Such a flow pattern
often occurs just beyond a tight vascular stenosis, where blood is decelerating and mixing with much
lower velocity blood. Turbulent flow, or more precisely incoherent flow, often results in an unwanted
loss of signal caused by phase incoherence.
Add to lightbox
Figure 27-29 Selected volumes for performing maximum intensity projections (MIP) processing.
Volumes were selected from the full set shown in Fig. 27-28. Left, a lateral view of the right internal
carotid artery. Right, an anteroposterior view of the vertebral and basilar arteries.
2 of 4 20-04-2010 00:02
Add to lightbox
Figure 27-30 Surface rendered view of a 3D contrast-enhanced MRA study of the carotid bifurcation.
Within regions of turbulent flow, acceleration and higher order motions introduce variations in phase
within each voxel that cannot be corrected by flow compensation. As a result, the magnetization
vectors of spins in the voxel point in different directions and cancel each other. The intensity of the
voxel becomes reduced. This phenomenon can be readily observed at a vascular stenosis or near
cardiac valves as a focal dropout in signal. The degree of signal loss depends on the amount of
turbulence. Sometimes this phenomenon is seen as a flame-shaped dark jet beyond the lesion orifice,
sometimes as an exaggeration of the degree of stenosis, and sometimes as an apparent complete
interruption in the vessel. Note that only transverse magnetization is affected by this phenomenon-
beyond the turbulence, the blood again appears bright because longitudinal magnetization is still
available to be excited.
page 717
page 718
Several measures can be taken to mitigate this loss of signal. One can acquire small voxels so that
fewer spins are placed together in a voxel, as described earlier in this chapter. One can select a short
TE with an asymmetric echo so that the duration of the frequency-encoding gradient is minimized. This
gradient is thought to be the greatest contributor to phase dispersion. One can employ 3D rather than
thin-slice 2D acquisition, which reduces the size of the slice-select gradient. Most importantly, one can
inspect the individual, unprojected sections when assessing the degree of stenosis. Often, residual,
uncanceled signal within the stenosis may be faintly visible.
Pulsatile Flow
Arterial flow is periodic in nature. The variation in the wash-in rate of spins between systole and
diastole creates alterations in the amplitude of echoes acquired during the course of the acquisition.
This "view-to-view" variation results in ghost artifacts along the phase-encoding direction of TOF
3 of 4 20-04-2010 00:02
images. The amount of ghosting is made worse by using large FA values. These ghost artifacts can be
selectively removed by using presaturation bands to eliminate the arterial signal when performing MR
venography.30
The variation in velocity over the cardiac cycle also causes modulations of phase in both TOF and PCA
acquisition, which cannot be eliminated by first-order flow compensation. Cardiac triggering can be
used to synchronize data collection to the cardiac cycle to address this problem. When triggering is not
used, the blood signal can be variable and ghost artifacts can be seen. Pulsatile artifacts can be quite
complex in 3D imaging.102 With 3D images, there are two phase-encoding directions (in plane and slice
select), and ghost artifacts resulting from respiration and pulsatile flow can propagate in both
directions. Artifacts in the slice-select direction may cause slice-to-slice variations in signal intensity.
Flow Separation and Recirculation

Flow saturation can be pronounced at regions of a vessel where blood flow circles in one place
(recirculation) or is stagnant. The most common example of this is the carotid bulb, where blood
moves forward in the anterior portion of the proximal internal carotid artery and in the reverse direction
in the posterior portion. The laminar streamlines of flow separate from the vessel wall at that point, a
phenomenon called flow separation. Blood in the posterior bulb spends a long time in the acquired
volume and appears dark, which might simulate the appearance of a plaque. Flow separation may also
occur beyond a stenosis or within an ulcer or aneurysm, again with the result that those areas appear
more dark.
Flow Misregistration
A blood signal often seems to arise from a point outside the vessel, as if the image of the blood and
the image of the background were misregistered. This artifact occurs as a result of flow during the
time delay between the phase-encoding and the frequency-encoding portions of the sequence.103,104 In
other words, the position on the phase-encoding axis is determined at one point in time and the
position along the frequency-encoding axis is determined at a slightly later point in time. Blood flowing
at an angle to the phase-encoding axis appears displaced along the frequency-encoding axis.
Displacement artifacts and signal "pileup" are often seen in tortuous vessels such as the carotid
siphon.105 The presence of misregistration can be discovered by repeating the MRA sequence with the
frequency and phase axes swapped. If the vessel takes on a new shape or position, a displacement
artifact is present.
The degree of misregistration depends on the blood velocity, the angle of the vessel with respect to
the phase-encoding direction, and the time delay between phase-encoding and frequency encoding.
The displacement is in the same direction as the flow. This artifact differs from a chemical-shift artifact,
which also occurs along the frequency-encoding direction. Reduction of this artifact requires changes in
pulse sequence design so that the time between phase-encoding and frequency encoding is minimized.
© 2010 Elsevier
4 of 4 20-04-2010 00:02
SUMMARY
MR angiography continues to be an important clinical tool for assessing cardiovascular disorders.
Time-of-flight has been extensively applied to nearly every part of the body and is effective in
addressing numerous clinical questions. Phase-contrast MRA is also useful in many clinical
applications. Although PC MRA is slightly more complex to use than TOF, it provides excellent tissue
background suppression and flexibility in flow sensitivity. As a result, PC MRA provides better depiction
of small vessels and slow flow. Moreover, it enables flow quantification and the generation of
directional flow maps. The rapid advent of CE MRA has enabled the effective and rapid study of a
wide range of territories. Problems such as breathing artifacts and saturation effects can be minimized
with CE MRA.
Higher field systems will provide improved capabilities with higher resolution, and these are already
being realized at 3.0 T and higher. Specialized coils will be available for higher SNR for specific
applications. Magnetic resonance angiography methods will continue to be challenged by advances in
other modalities. Rotational angiography, multislice CT, and improved ultrasound methods will all have
their role to play in the clinical arena. The noninvasive nature of MRA, its ability to quantify flow, and its
strength in visualizing surrounding soft-tissue and the end organ, will ensure that MRA remains a
powerful contributor in the study of cardiovascular disease.
REFERENCES
1. Caro C, Pedley J, Schroter R, Seed W: The Mechanics of Circulation. Oxford: Oxford University Press, 1978.
2. McDonald D: Blood Flow in Arteries. London: Edward Arnold, 1974.
3. Milnor W: Hemodynamics. Baltimore: Williams and Wilkins, 1989.
page 718
page 719
4. Kilner PJ, Yang GZ, Mohiaddin RH, et al: Helical and retrograde secondary flow patterns in the aortic arch studied by three-
directional magnetic resonance velocity mapping. Circulation 88:2235-2247, 1993. Medline Similar articles
5. Edelman RR: Magnetic resonance angiography. An overview. Invest Radiol 28 Suppl 4:S43-46, 1993.
6. Saloner D, Chien D: Historical development of MRA. In Anderson C, Edelman RR, Turski P (eds): Clinical Magnetic
Resonance Angiography. New York: Raven Press, 1993.
7. Hahn E: Detection of sea water motion by nuclear precession. J Geophys Res 65:776-777, 1960.
8. Hahn E: Spin echoes. Phys Rev 80:580-594, 1960.
9. Suryan G: Nuclear resonance in flowing liquids. Proc Indian Acad Sci A33:107-113, 1951.
10. Masaryk TJ, Laub GA, Modic MT, et al: Carotid-CNS MR flow imaging. Magn Reson Med 14:308-314, 1990. Medline
Similar articles
11. Edelman RR, Manning WJ, Burstein D, Paulin S: Coronary arteries: breath-hold MR angiography. Radiology 181:641-643,
12. Manning WJ, Li W, Boyle NG, Edelman RR: Fat-suppressed breath-hold magnetic resonance coronary angiography.
13. Rother J, Wentz KU, Rautenberg W, et al: Magnetic resonance angiography in vertebrobasilar ischemia. Stroke
14. Durham JR, Hackworth CA, Tober JC, et al: Magnetic resonance angiography in the preoperative evaluation of abdominal
aortic aneurysms. Am J Surg 166:173-177, 1993 [discussion, pp177-178].
15. Anderson CM, Saloner D, Lee RE, et al: Assessment of carotid artery stenosis by MR angiography: comparison with x-ray
angiography and color-coded Doppler ultrasound. Am J Neuroradiol 13:989-1003, 1992 [discussion, pp1005-1008].
16. Lewis WD, Finn JP, Jenkins RL, et al: Use of magnetic resonance angiography in the pretransplant evaluation of portal vein
pathology. Transplantation 56:64-68, 1993. Medline Similar articles
17. Turnipseed WD, Kennell TW, Turski PA, et al: Combined use of duplex imaging and magnetic resonance angiography for
evaluation of patients with symptomatic ipsilateral high-grade carotid stenosis. J Vasc Surg 17:832-839, 1993 [discussion,
pp839-840].
18. Anson JA, Heiserman JE, Drayer BP, Spetzler RF: Surgical decisions on the basis of magnetic resonance angiography of
the carotid arteries. Neurosurgery 32:335-343, 1993 [discussion, p343].
19. Finn JP, Kane RA, Edelman RR, et al: Imaging of the portal venous system in patients with cirrhosis: MR angiography vs
duplex Doppler sonography. Am J Roentgenol 161:989-994, 1993.
20. Mohiaddin RH, Paz R, Theodoropoulos S, et al: Magnetic resonance characterization of pulmonary arterial blood flow after
single lung transplantation. J Thorac Cardiovasc Surg 101:1016-1023, 1991. Medline Similar articles
1 of 5 20-04-2010 00:03
21. Willinek WA, Born M, Simon B, et al: Time-of-flight MR angiography: comparison of 3.0-T imaging and 1.5-T imaging-initial
22. Wehrli FW: Time-of-flight effects in MR imaging of flow. Magn Reson Med 14:187-193, 1990. Medline
Similar articles
23. Laub GA, Kaiser WA: MR angiography with gradient motion refocusing. J Comput Assist Tomogr 12:377-382, 1988.
24. Schmalbrock P, Yuan C, Chakeres DW, et al: Volume MR angiography: methods to achieve very short echo times.
25. Wehrli FW, Shimakawa A, Gullberg GT, MacFall JR: Time-of-flight MR flow imaging: selective saturation recovery with
gradient refocusing. Radiology 160:781-785, 1986. Medline Similar articles
27. Nishimura DG: Time-of-flight MR angiography. Magn Reson Med 14:194-201, 1990. Medline Similar articles
28. Townsend TC, Saloner D, Pan XM, Rapp JH: Contrast material-enhanced MRA overestimates severity of carotid stenosis,
compared with 3D time-of-flight MRA. J Vasc Surg 38:36-40, 2003.
30. Ehman RL, Felmlee JP: Flow artifact reduction in MRI: a review of the roles of gradient moment nulling and spatial
presaturation. Magn Reson Med 14:293-307, 1990. Medline Similar articles
31. Felmlee JP, Ehman RL: Spatial presaturation: a method for suppressing flow artifacts and improving depiction of vascular
anatomy in MR imaging. Radiology 164:559-564, 1987. Medline Similar articles
32. Edelman RR, Atkinson DJ, Silver MS, et al: FRODO pulse sequences: a new means of eliminating motion, flow, and
wraparound artifacts. Radiology 166:231-236, 1988. Medline Similar articles
33. Edelman RR, Mattle HP, O'Reilly GV, et al: Magnetic resonance imaging of flow dynamics in the circle of Willis. Stroke
34. Gullberg GT, Wehrli FW, Shimakawa A, Simons MA: MR vascular imaging with a fast gradient refocusing pulse sequence
and reformatted images from transaxial sections. Radiology 165:241-246, 1987. Medline Similar articles
35. Lewin JS, Laub G, Hausmann R: Three-dimensional time-of-flight MR angiography: applications in the abdomen and
thorax. Radiology 179:261-264, 1991. Medline Similar articles
36. Ruggieri PM, Laub GA, Masaryk TJ, Modic MT: Intracranial circulation: pulse-sequence considerations in three-
dimensional (volume) MR angiography. Radiology 171:785-791, 1989. Medline Similar articles
37. Purdy D, Cadena G, Laub G: The design of a variable tip angle slab selection (TONE) pulse for improved 3D MRA. In:
Proceedings of the 11th annual meeting of the Society of Magnetic Resonance in Medicine. Berlin, The Society of Magnetic
Resonance in Medicine, 1992, p882.
38. Haacke EM, Masaryk TJ, Wielopolski PA, et al: Optimizing blood vessel contrast in fast three-dimensional MRI. Magn
39. Parker DL, Yuan C, Blatter DD: MR angiography by multiple thin slab 3D acquisition. Magn Reson Med 17:434-451, 1991.
40. Atkinson DJ, Edelman RR: Cineangiography of the heart in a single breath hold with a segmented turboFLASH sequence.
41. Chien D, Edelman RR: Ultrafast imaging using gradient echoes. Magn Reson Q 7:31-56, 1991. Medline
Similar articles
42. Edelman RR, Chien D, Atkinson DJ, Sandstrom J: Fast time-of-flight MR angiography with improved background
suppression. Radiology 179:867-870, 1991. Medline Similar articles
43. Weber OM, Martin AJ, Higgins CB: Whole-heart steady-state free precession coronary artery magnetic resonance
44. Kim WY, Danias PG, Stuber M, et al: Coronary magnetic resonance angiography for the detection of coronary stenoses. N
Engl J Med 345:1863-1869, 2001. Medline Similar articles
Similar articles
46. Chien D, Goldmann A, Edelman RR: High-speed black blood imaging of vessel stenosis in the presence of pulsatile flow. J
47. Edelman RR, Mattle HP, Wallner B, et al: Extracranial carotid arteries: evaluation with "black blood" MR angiography.
48. Kerwin W, Hooker A, Spilker M, et al: Quantitative magnetic resonance imaging analysis of neovasculature volume in
carotid atherosclerotic plaque. Circulation 107:851-856, 2003. Medline Similar articles
49. Yuan C, Miller ZE, Cai J, Hatsukami T: Carotid atherosclerotic wall imaging by MRI. Neuroimaging Clin N Am 12:391-401,
50. Yuan C, Zhao XQ, Hatsukami TS: Quantitative evaluation of carotid atherosclerotic plaques by magnetic resonance
2 of 5 20-04-2010 00:03
imaging. Curr Atheroscler Rep 4:351-357, 2002. Medline Similar articles

51. Wedeen VJ, Rosen BR, Chesler D, Brady TJ: MR velocity imaging by phase display. J Comput Assist Tomogr 9:530-536,
52. Moran PR: A flow velocity zeugmatographic interlace for NMR imaging in humans. Magn Reson Imaging 1:197-203, 1982.
53. von Schulthess GK, Higgins CB. Blood flow imaging with MR: spin-phase phenomena. Radiology 157:687-695, 1985.
54. Firmin DN, Nayler GL, Kilner PJ, Longmore DB: The application of phase shifts in NMR for flow measurement. Magn
55. Dumoulin CL, Souza SP, Walker MF, Wagle W: Three-dimensional phase contrast angiography. Magn Reson Med
56. Pelc NJ, Bernstein MA, Shimakawa A, Glover GH: Encoding strategies for three-direction phase-contrast MR imaging of
flow. J Magn Reson Imaging 1:405-413, 1991. Medline Similar articles
57. Turski PA, Korosec FR, Carroll TJ, et al: Contrast-Enhanced magnetic resonance angiography of the carotid bifurcation
using the time-resolved imaging of contrast kinetics (TRICKS) technique. Top Magn Reson Imaging 12:175-181, 2001.
58. Applegate GR, Talagala SL, Applegate LJ: MR angiography of the head and neck: value of two-dimensional phase-
contrast projection technique. Am J Roentgenol 159:369-374, 1992.
59. Pernicone JR, Siebert JE, Laird TA, et al: Determination of blood flow direction using velocity-phase image display with
3-D phase-contrast MR angiography. Am J Neuroradiol 13:1435-1438, 1992.
60. Underwood SR: Cine magnetic resonance imaging and flow measurements in the cardiovascular system. Br Med Bull
61. Enzmann DR, Ross MR, Marks MP, Pelc NJ: Blood flow in major cerebral arteries measured by phase-contrast cine MR.
62. Debatin JF, Ting RH, Wegmuller H, et al: Renal artery blood flow: quantitation with phase-contrast MR imaging with and
without breath holding. Radiology 190:371-378, 1994. Medline Similar articles
63. Applegate GR, Thaete FL, Meyers SP, et al: Blood flow in the portal vein: velocity quantitation with phase-contrast MR
64. Meier D, Maier S, Bosiger P: Quantitative flow measurements on phantoms and on blood vessels with MR. Magn Reson
65. Glover GH, Pelc NJ: A rapid-gated cine MRI technique. Magn Reson Ann 299-333, 1988.
66. Bogren HG, Klipstein RH, Firmin DN, et al: Quantitation of antegrade and retrograde blood flow in the human aorta by
magnetic resonance velocity mapping. Am Heart J 117:1214-1222, 1989. Medline Similar articles
67. Li KC, Whitney WS, McDonnell CH, et al: Chronic mesenteric ischemia: evaluation with phase-contrast cine MR imaging.
68. Edelman RR, Manning WJ, Gervino E, Li W: Flow velocity quantification in human coronary arteries with fast, breath-hold
MR angiography. J Magn Reson Imaging 3:699-703, 1993. Medline Similar articles
69. Korosec FR, Mistretta CA, Turski PA: ECG-optimized phase contrast line-scanned MR angiography. Magn Reson Med
page 719
page 720
70. Buonocore MH: Algorithms for improving calculated streamlines in 3-D phase contrast angiography. Magn Reson Med
31:22-30, 1994.
71. Iseda T, Nakano S, Miyahara D, et al: Poststenotic signal attenuation on 3D phase-contrast MR angiography: a useful
finding in haemodynamically significant carotid artery stenosis. Neuroradiology 42:868-873, 2000.
72. Steinberg FL, Yucel EK, Dumoulin CL, Souza SP: Peripheral vascular and abdominal applications of MR flow imaging
techniques. Magn Reson Med 14:315-320, 1990. Medline Similar articles
73. Creasy JL, Price RR, Presbrey T, et al: Gadolinium-enhanced MR angiography. Radiology 175:280-283, 1990.
74. Lin W, Haacke EM, Smith AS, Clampitt ME: Gadolinium-enhanced high-resolution MR angiography with adaptive vessel
tracking: preliminary results in the intracranial circulation. J Magn Reson Imaging 2:277-284, 1992. Medline
Similar articles
75. Saloner D: Determinants of image appearance in contrast-enhanced magnetic resonance angiography. A review. Invest
76. Kim JK, Farb RI, Wright GA: Test bolus examination in the carotid artery at dynamic gadolinium-enhanced MR
77. Earls JP, Rofsky NM, DeCorato DR, et al: Hepatic arterial-phase dynamic gadolinium-enhanced MR imaging: optimization
with a test examination and a power injector. Radiology 202:268-273, 1997. Medline Similar articles
3 of 5 20-04-2010 00:03
78. Wilman AH, Riederer SJ, Huston J 3rd, et al: Arterial phase carotid and vertebral artery imaging in 3D contrast-enhanced
MR angiography by combining fluoroscopic triggering with an elliptical centric acquisition order. Magn Reson Med 40:24-35,
1998.
79. Wilman AH, Riederer SJ: Performance of an elliptical centric view order for signal enhancement and motion artifact
suppression in breath-hold three-dimensional gradient echo imaging. Magn Reson Med 38:793-802, 1997. Medline
Similar articles
80. Wilman AH, Riederer SJ, King BF, et al: Fluoroscopically triggered contrast-enhanced three-dimensional MR angiography
with elliptical centric view order: application to the renal arteries. Radiology 205:137-146, 1997. Medline
Similar articles
81. Willinek WA, Gieseke J, Conrad R, et al: Randomly segmented central k-space ordering in high-spatial-resolution contrast-
enhanced MR angiography of the supraaortic arteries: initial experience. Radiology 225:583-588, 2002. Medline
Similar articles
82. Foo TK, Saranathan M, Prince MR, Chenevert TL: Automated detection of bolus arrival and initiation of data acquisition in
fast, three-dimensional, gadolinium-enhanced MR angiography. Radiology 203:275-280, 1997. Medline Similar articles
83. Du YP, Parker DL, Davis WL, Cao G: Reduction of partial-volume artifacts with zero-filled interpolation in three-dimensional
MR angiography. J Magn Reson Imaging 4:733-741, 1994. Medline Similar articles
84. Ruehm SG, Nanz D, Baumann A, et al: 3D contrast-enhanced MR angiography of the run-off vessels: value of image
subtraction. J Magn Reson Imaging 13:402-411, 2001. Medline Similar articles
85. Goyen M, Debatin JF, Ruehm SG: Peripheral magnetic resonance angiography. Top Magn Reson Imaging 12:327-335,
86. Herborn CU, Goyen M, Quick HH, et al: Whole-body 3D MR angiography of patients with peripheral arterial occlusive
disease. Am J Roentgenol 182:1427-1434, 2004.
87. Jou LD, Quick CM, Young WL, et al: Computational approach to quantifying hemodynamic forces in giant cerebral
aneurysms. Am J Neuroradiol 24:1804-1810, 2003. Medline Similar articles
88. Yuan C, Murakami JW, Hayes CE, et al: Phased-array magnetic resonance imaging of the carotid artery bifurcation:
preliminary results in healthy volunteers and a patient with atherosclerotic disease. J Magn Reson Imaging 5:561-565, 1995.
36:345-351, 1996.
90. Mistretta CA, Grist TM, Korosec FR, et al: 3D time-resolved contrast-enhanced MR DSA: advantages and tradeoffs. Magn
92. Du J, Carroll TJ, Wagner HJ, et al: Time-resolved, undersampled projection reconstruction imaging for high-resolution
CE-MRA of the distal runoff vessels. Magn Reson Med 48:516-522, 2002. Medline Similar articles
93. Melhem ER, Caruthers SD, Faddoul SG, et al: Use of three-dimensional MR angiography for tracking a contrast bolus in
the carotid artery. Am J Neuroradiol 20:263-266, 1999. Medline Similar articles
94. Tsuchiya K, Aoki C, Fujikawa A, Hachiya J: Three-dimensional MR digital subtraction angiography using parallel imaging
and keyhole data sampling in cerebrovascular diseases: initial experience. Eur Radiol 14:1494-1497, 2004. Medline
Similar articles
95. Rossnick S, Laub G, Braeckle R: Three dimensional display of blood vessels in MRI. New York: IEEE, 1990, pp193-195.
96. Keller PJ, Drayer BP, Fram EK, et al: MR angiography with two-dimensional acquisition and three-dimensional display.
Work in progress. Radiology 173:527-532, 1989. Medline Similar articles
97. Dixon WT, Du LN, Faul DD, et al: Projection angiograms of blood labeled by adiabatic fast passage. Magn Reson Med
98. Laub G: Displays for MR angiography. Magn Reson Med 14:222-229, 1990. Medline Similar articles
99. Saloner D, Hanson WA, Tsuruda JS, et al: Application of a connected-voxel algorithm to MR angiographic data. J Magn
100. Hu X, Alperin N, Levin DN, et al: Visualization of MR angiographic data with segmentation and volume-rendering
techniques. J Magn Reson Imaging 1:539-546, 1991. Medline Similar articles
101. Anderson CM, Saloner D, Tsuruda JS, et al: Artifacts in maximum-intensity-projection display of MR angiograms. Am J
Roentgenol 154:623-629, 1990.
102. Frank LR, Buxton RB, Kerber CW: Pulsatile flow artifacts in 3D magnetic resonance imaging. Magn Reson Med
30:296-304, 1993.
103. Nishimura DG, Jackson JI, Pauly JM: On the nature and reduction of the displacement artifact in flow images. Magn
104. Frank LR, Crawley AP, Buxton RB: Elimination of oblique flow artifacts in magnetic resonance imaging. Magn Reson Med
4 of 5 20-04-2010 00:03
105. van Tyen R, Saloner D, Jou LD, Berger S: MR imaging of flow through tortuous vessels: a numerical simulation. Magn
© 2010 Elsevier
5 of 5 20-04-2010 00:03
ASIC RINCIPLES AND LINICAL PPLICATIONS OF

LOW UANTIFICATION
David N. Firmin
Raad H. Mohiaddin
Peter D. Gatehouse
The nature of blood flow in different regions of the body is highly variable and complex. It is influenced
by many factors, including the cardiac function, the vascular geometry and compliance, and the state
of the capillary bed. Both the need to better understand the physiology and the clinical requirement for
a precise, flexible and noninvasive technique to quantify flow have motivated the application of
magnetic resonance (MR) to the measurement of in vivo blood flow. 1-5
The MR methods that have been developed over the years have used some variation on one of two
basic phenomena that affect the MR signal, these being the time-of-flight effects6-8 and flow-induced
phase shifts.9 Recently the method that has been established for clinical application has been based on
the latter and is generally known as phase contrast velocity mapping1-4,10-12 This method has been
relatively easy to implement on commercial scanners and studies have shown it to be both accurate
and robust. The following description will concentrate entirely on this technique.
Doppler ultrasound is the alternative and more widely used noninvasive approach to blood flow
measurement, with its advantages of being less expensive and more portable than MR. The
advantages of MR, however, are that it can acquire data in any orientation, unrestricted by windows of
access, and it allows choice of the direction in which velocities are measured with respect to the
imaging plane. Also, it has the ability to measure accurate velocities in pixels throughout the plane of
acquisition, resulting in the potential to accurately measure volume flow because of the simultaneous
acquisition of the mean velocity and the area of the vessel. Doppler echocardiography, on the other
hand, is accurate for measuring velocity but poor at assessing volume flow, particularly when the flow
pattern is complex. In addition, MR at present is the only technique with the potential to acquire
comprehensive 7D information (three spatial dimensions, three velocity components, and time) and is
13,14
well suited for studying spatial and temporal patterns of flow in the human cardiovascular system.
This is important for understanding blood flow physiology where complex multidirectional flows can be
found in the pumping cardiac chambers and in vessels that are curved or branching. The ability to
measure realtime flow information, once an advantage of Doppler, is now, with higher performance
cardiovascular MR scanners and the further development of realtime techniques, becoming possible
with MR along with other added advantages described above.
© 2010 Elsevier
1 of 1 20-04-2010 00:04
BASIC PRINCIPLES OF PHASE CONTRAST VELOCITY MAPPING

page 721
page 722
Add to lightbox
Figure 28-1 The principle behind phase contrast velocity encoding. At time 1 a magnetic field gradient is applied,
resulting in a frequency and associated phase shift that is equal for neighboring stationary and flowing spin
signals. At time point 2 an equal but opposite magnetic gradient is applied, the result of which is an equal but
opposite frequency and phase shift for the signal from stationary nuclei so that their phase returns to zero.
However, at this time the nuclei in the blood that is flowing have moved away from their original position to be in a
different strength of magnetic field during the gradient application. The result of this movement is that their signals
will accumulate a phase shift proportional to the distance moved and hence the velocity. (From Nagel E, van
Rossum AC, Fleck E: Kardiovaskulare Magnetresonanz-tomography-Methodenverstandnis und praktische
Anwendungen. Steinkopff-Verlag, Darmstadt, Germany, 2002, p 26, with permission.)
Phase contrast velocity mapping is based on the underlying principle that the signal from moving tissue will
undergo a phase shift, relative to that from stationary tissue, if a magnetic field gradient is applied in the direction
of motion. The received MR signal has three components, frequency, amplitude, and phase, which are all used to
reconstruct the image. Also, for each pixel on the final displayed image there is a magnitude and a phase although
often it is only the magnitude that is displayed. For a phase velocity map it is the phase that is displayed or, more
often, it is the difference between two phases from images acquired with slightly different sequences. A pixel with
a phase difference of +180° is displayed white whilst one with a phase difference of -179° is displayed black; 0° is
displayed mid gray. In order to introduce well-defined velocity-related phase shifts to the image, a temporally
bipolar magnetic field gradient is applied in the direction of motion. This means that a gradient is turned on for a
short period of time and then, after an interval, the reverse gradient is applied again for the same period.
An illustration of how a temporally bipolar gradient provides a velocity-related phase shift to the signal from
moving tissue is given in Figure 28-1. This figure compares the signal phases from initially neighboring stationary
and flowing tissues. To start with, at time 1 a positive field gradient is turned on for a short period of time,
resulting in an equal frequency shift to both tissues. When the field gradient is switched off the signals will have a
phase shift related to the distance of the tissues from the center of the scanner and the time that the gradient was
applied. At time 2 the reverse field gradient is applied which results in an equal but opposite frequency shift for the
stationary tissue. The flowing tissue, however, has moved to a different position, resulting in a different frequency
shift. Thus there is a phase difference between the flowing and stationary tissues and the size of the difference is
directly related to the distance moved in the time between the two field gradient pulses and hence the flow
velocity.
The temporal shape of gradient waveform normally used for velocity encoding is illustrated in Figure 28-2 and the
equation for calculating the velocity related phase shift is: ϕ = 2πγG∆δv where γ is the gyromagnetic ratio, a
constant. G, ∆ and δare the gradient amplitudes and times, are also constants for a particular sequence, meaning
that the phase shift ϕ is directly proportional to velocity.
1 of 13 20-04-2010 00:05
Add to lightbox
page 722
page 723
Figure 28-2 A bipolar gradient waveform designed to introduce a velocity-related phase shift to the signal.
Add to lightbox
Figure 28-3 The images that are acquired for phase contrast velocity mapping. On the left are the temporal
gradient waveforms for the reference at the top and the velocity encoded (= reference + bipolar gradient) below.
To the right of those are the magnitude images, from a horizontal long axis plane through the left ventricle,
acquired with the different waveforms. Very little difference is apparent between the two images. To the right of
the magnitude images are the phase images which are complicated by the many factors as well as flow that affect
the image phase. On the far right is the velocity map produced by phase subtraction of the two separate phase
images, which has the effect of removing all the phase variations that are common to both velocity-sensitive and
reference scans. This particular velocity map is timed during diastole and shows flow from the left atrium to
ventricle in lighter shades of gray. (From Nagel E, van Rossum AC, Fleck E: Kardiovaskulare Magnetresonanz-
tomography-Methodenverstandnis und praktische Anwendungen. Steinkopff-Verlag, Darmstadt, Germany,
2002, p 27, with permission.)
As the pixel phase is affected by many factors in addition to the flow velocity, two images are normally acquired
using temporal gradient waveforms that if subtracted from one another would be shown to differ by the described
bipolar shape and as such have a well-defined difference of velocity phase sensitivity. As stated earlier, after
acquisition, the phases are then subtracted on a pixel-by-pixel basis to remove the unwanted phase shifts that are
common to both images and leave only the velocity-related shifts. Figure 28-3 illustrates the dual magnitude and
phase images that are subtracted to produce the final phase velocity image.
The gradient waveform at the top left is typical for a so-called velocity-compensated readout and for that below
we have illustrated the addition of a bipolar velocity encoding gradient to introduce a well-defined velocity phase-
encoding. To the right are first the magnitude and then the phase reconstructions of the acquired images. The two
magnitude reconstructions of this horizontal long axis view of the heart look very similar because the relatively
small phase difference has little if any impact on the magnitude. Also, the phase reconstructions are so
complicated that it is difficult to identify the differences. Following the phase subtraction to the right, however, the
phase shifts can be clearly seen on the grayscale as lighter shades of gray, representing flow from the left atrium
to the left ventricle during the diastolic filling part of the cardiac cycle.
2 of 13 20-04-2010 00:05
Add to lightbox
Figure 28-4 Radiofrequency, slice selection and frequency-encoding waveforms, illustrating the inherent bipolar
nature of these (hashed section) for a simple gradient-echo sequence.
Velocity phase-encoding may be added to any of the three orthogonal gradient directions used during imaging.
Encoding in the slice or slab selection gradient direction enables measurement of through-plane velocities,
whereas encoding in either the frequency or phase-encode gradient directions enables measurement of
components of velocity either vertically or horizontally within the image plane. On a typical velocity map image
such as that illustrated to the right of Figure 28-3, stationary tissues which have zero phase shift are displayed as
midgray and flowing blood which has phase shift is displayed as either a darker or lighter shade depending on its
direction and velocity.
The velocity sensitivity of the sequence can be adjusted by altering the amplitude (G) duration (δ) or separation
(∆) of the bipolar waveform (Fig. 28-2). The term venc is used to define the sensitivity of the sequence, where venc
is the velocity that will result in a phase shift of π radians.
Signal Loss and Velocity Compensation

The earliest development of phase velocity mapping1,2 was achieved by using spin-echo sequences that were in
fact quite unsuitable for general blood flow measurement. There are three reasons for this unsuitability: first, the
sequence cannot be repeated rapidly to allow cine imaging; second, the standard spin-echo sequence will not
form its spin-echo signal unless the tissue or blood is within the slice for both the 90° and 180° pulses, and this is
often not the case for flowing blood; and finally the standard spin-echo sequence is inherently very phase sensitive
to motion which results in so-called flow-related signal loss. The first two of these problems were overcome by
using a gradient-echo sequence that only used one RF pulse per signal readout and could be repeated rapidly by
reducing the flip angle. However, the problem of flow-related signal loss was still apparent.
The basic spin-echo and gradient-echo sequences both have motion phase sensitivity because they contain
bipolar gradient waveforms in the slice selection and frequency encoding directions (as illustrated in Fig. 28-4).
page 723
page 724
Add to lightbox
Figure 28-5 An imaging voxel within a blood vessel containing a parabolic flow profile so that the voxel contains
blood with a range of blood flow velocities. Because of the velocity phase sensitivity of the sequence, the signals
from the voxel will contain a distribution of phases. If the distribution covers a complete cycle or more then
considerable phase cancellation and signal loss will occur.
If we consider an imaging voxel within a blood vessel containing, for example, a parabolic flow profile then the
3 of 13 20-04-2010 00:05
voxel will contain blood with a range of blood flow velocities (Fig. 28-5). The result of the velocity phase sensitivity
of the sequence is that the signals from the voxel will contain a distribution of phases and if the distribution covers
a complete cycle or more then considerable phase cancellation and signal loss will occur. Another consequence of
this phase sensitivity is that for a scan that requires several cardiac cycles to acquire the data, variations in flow
from one cardiac cycle to the next will result in different velocity-related phase shifts to the blood which will add to
the spatial phase-encoding and lead to any blood signal being mis-positioned and spread out in the phase-encode
direction as artifact (Fig. 28-6).
15
Long before the development of MRI, Carr and Purcell recognized the phenomenon known as even-echo
rephasing; in the presence of slow flow they noted that for a multiple spin-echo sequence the even spin-echoes
were larger than the odd ones due to more complete rephasing. In 1985 Nayler and colleagues reported the
3
incorporation of this even-echo rephasing phenomenon into a blood flow imaging sequence. The basic sequence
which is illustrated in Figure 28-7 involves a modification to the slice selection and frequency encoding gradient
waveforms. The modification is to add another bipolar section of waveform with the opposite gradient polarity.
The two bipolar waveforms are identified on Figure 28-7 by the different hashed sections.
Add to lightbox
Figure 28-6 Magnitude image illustrating blood signal artifact spread out in the vertical phase-encode direction.
This results from variations in flow velocities from one cardiac cycle to the next, giving rise to the flowing blood
having different velocity-related phase shifts which add to the spatial phase-encoding and result in blood signals
being mispositioned and spread out in the phase-encode direction as artifact.
Add to lightbox
Figure 28-7 Radiofrequency, slice selection and frequency-encoding waveforms, illustrating the additional
opposite polarity bipolar gradient waveform sections in a velocity-compensated gradient-echo sequence.
The velocity-related phase shift due to the first hashed section is given by the previously described equation: ϕ1 =
2πγG∆δv and the velocity-related phase shift due to the second hashed section is given by exactly the same
equation except that the gradient polarity is reversed: ϕ2 = 2πγ(-G)∆δv. The overall velocity-related phase shift for
the gradient waveform is given by ϕ1 + ϕ2 which can be seen to equate to zero. This sequence therefore has zero
4 of 13 20-04-2010 00:05
velocity phase sensitivity and is often referred to as velocity compensated. If we now consider the voxel within a
blood vessel containing parabolic flow, then whatever the distribution of velocities from within the voxel, the signal
phases will all be equal (Fig. 28-8) and produce a high blood signal. As discussed earlier, generally to measure
velocity two images are acquired, one with the velocity-compensated gradient waveforms and the other with a
relatively small and well-defined modification to add the required velocity sensitivity for the flow to be measured.
As a consequence of some cardiovascular diseases, however, the blood flow often becomes much more complex
and turbulent and contains higher orders of motion than simple velocity. These higher orders of flow motion can
introduce phase shifts even to a velocity-compensated sequence (Fig. 28-9) and the spatial scale of this motion is
such that this can also lead to significant phase dispersion and signal loss (Fig. 28-10).
page 724
page 725
Add to lightbox
Figure 28-8 An imaging voxel within a blood vessel containing a parabolic flow profile so that the voxel contains
blood with a range of blood flow velocities. Because the sequence is velocity compensated, the signals from the
voxel all have the same phase which will result in a high signal from the flowing blood.
Add to lightbox
Figure 28-9 An imaging voxel within a blood vessel containing complex turbulent flow so that the voxel contains
blood with a range of higher motion orders. The spatial scale of the motion is such that this can lead to significant
phase dispersion and signal loss.
5 of 13 20-04-2010 00:05
Add to lightbox
Figure 28-10 Coronal image showing the left ventricle and ascending aorta of a patient with a stenotic aortic
valve. At the systolic timing of this image the complex turbulent blood flow in the ascending aorta (arrowed)
resulted in intravoxel phase dispersion and complete signal loss.
Add to lightbox
Add to lightbox
Figure 28-11 Comparison of an acceleration sensitive breath-hold signal loss (A) and a conventional longer TE
(14 ms) (B) sequence in a patient with aortic valve stenosis. In this coronal image plane similar systolic signal loss
can be seen in the aorta on both images. However, the breath-hold image has considerably less motion artifact.
In fact, this type of image exhibiting complex flow-related signal loss has proved useful in the clinical setting to
16
help assess the severity and impact of stenoses and flow regurgitation, for example. However, it is of course
impossible to measure flow velocities if the flow signal is lost and this led to further study of its cause. It was soon
shown that the phase sensitivity to higher orders of motion was related to the duration of the velocity-
compensating gradient waveforms and by shortening these and in turn the echo time (TE) of the sequence, the
17,18
signal loss could be reduced considerably to enable flow measurements even in the most severe stenoses.
With hardware improvements and the recent move toward more rapid breath-hold acquisitions, the sequence
gradient waveforms and TEs are inherently very short and this form of signal loss is therefore rarely encountered.
In fact, Keegan and colleagues have described a breath-hold sequence with increased acceleration sensitivity to
mimic the older sequences (Fig. 28-11), whilst at the same time having the advantage of removing the problems
19
of respiratory artifact.
Velocity-to-Noise Ratio
6 of 13 20-04-2010 00:05
page 725
page 726
Add to lightbox
Figure 28-12 Phase contrast velocity map frame acquired in a transverse plane through the great vessels of a
healthy volunteer. The early systolic flow exhibiting high velocities in the ascending aorta has resulted in velocity
aliasing seen as dark shades of gray within the ascending aorta (A) and also illustrated on the velocity profile
below. This aliasing can be removed, in this case, by introducing a positive shift to the velocity window (B).
In 1990 Conturo and Smith20 showed there is an inverse relationship between phase noise σϕ in radians and the
image signal-to-noise ratio (SNR):
From this, the equation relating the velocity-to-noise ratio (VNR) to the SNR can be shown to be:
where v is the velocity being measured. This shows that the VNR is
approximately double the SNR if the venc is set to be equal to the velocity being measured. It also illustrates the
21
importance of using as small a venc as possible to maintain a high VNR. With this in mind, Buonocore suggested
using a smaller venc for the diastolic portion of the cardiac cycle to improve accuracy when the measured
velocities are lower. The problem with having too small a venc is that aliasing can occur (Fig. 28-12A) although if
the flow of interest is predominantly in one direction the phase velocity window can often be shifted to correct for
this (Fig. 28-12B). If shifting the phase velocity window is not an option then it might be possible to use image
22
processing phase unwrapping techniques to remove the aliasing.
Rapid Flow Imaging

Until relatively recently typical flow studies have required a number of minutes for the acquisition of the data
especially if two or three orthogonal directions of velocity encoding were required. Even now, with the fastest
clinical hardware, when using a segmented gradient-echo type acquisition, a long breath-hold is required and
considerable compromises have to be made to resolutions. The flow velocities displayed on the phase velocity
map represent a weighted average over the time of the acquisition. This averaging of velocities can be seen as an
advantage because any short-term variability of the flow does not seem to introduce errors to the measurement.
However, there are situations where very short-term variations in flow are of interest; for example, we might be
interested in the response to exercise, pharmacologic stress or a number of other factors that have dynamic
effects on the physiology of the body. In these types of cases, a more rapid acquisition method is required where
the image is acquired in a fraction of a second.
Over recent years there have been a number of approaches to rapidly acquiring the flow velocity data. The most
basic of these involves a further compromise in spatial resolution with the removal of some of the spatial phase-
23
encoding to speed up the scan. To reduce this compromise, a number of more rapid acquisition echo-planar
24,25
flow imaging methods have been described. Despite the use of specially designed velocity-compensating
26
gradient waveforms, a problem was that these more complicated and demanding imaging waveforms could, in
some types of flow, increase flow-related signal loss and show a generally more complex flow phase sensitivity
that introduced artifacts. These problems have been minimized by reducing the number of readout echoes and
7 of 13 20-04-2010 00:05
producing a hybrid sequence between echo-planar and a conventional gradient-echo.27
Spiral sequences have also been used for rapid readout with the advantage that with the readout starting at the
28,29
center of k-space, they are relatively insensitive to signal loss problems. It is a mistake to consider these
sequences completely flow compensated, however, as during the relatively long spiral readouts flow-related
phase errors can accumulate and lead to complex artifacts.30 Probably the most useful rapid imaging method has
been the zone selective approach where a 2D RF pulse has been used to excite signal only from a well-defined
column of tissue, which enables a greatly reduced number of phase-encoding steps and hence EPI echoes to be
used to rapidly read out the signal. This method is generally applicable to blood vessels because they are usually
small with respect to the imaging field of view. This zone selective method has been used to demonstrate and
31
measure blood flow changes due to reactive hyperemia in the femoral artery (Fig. 28-13).
Partial Volume Effects

Partial volume effects can become a major concern when acquiring images rapidly with a low spatial resolution or
when attempting to measure flow in small arteries such as the coronaries. Figure 28-14 illustrates a blood vessel
with a matrix of relatively large pixels. If we assume that fat is surrounding the vessel, as is often the case for the
coronary arteries for example, and a perfect fat saturation is applied then the first problem we can see is that on
the velocity map the vessel will appear larger than it actually is because any pixel either partially or completely
covered by the vessel will appear to contain the flow velocity. This problem can be corrected by more
sophisticated processing that incorporates the magnitude data but this is often not available.
page 726
page 727
Add to lightbox
Add to lightbox
Add to lightbox
Figure 28-13 (A,B) The magnitude images and phase contrast velocity maps of systolic frames showing the
femoral artery cross-section. These were acquired in realtime using a zone selective technique. The plot (C)
shows the changes in femoral artery mean velocity due to reactive hyperemia during a period of cuff inflation and
then release. (Reprinted from Mohiaddin RH, Gatehouse D, Moon JC et al: Assessment of reactive
hyperaemia using real time zonal echo-planar flow imaging. J Cardiovasc Magn Reson 4:283-287, 2002, by
courtesy of Marcel Dekker Inc., © 2002.)
If the tissue surrounding the vessel is not saturated then the measured velocities can become more complicated.
The signal magnitude and phase for that pixel will be made up from a combination of magnitude and phases from
the flowing blood and stationary surrounding tissue. At first sight it might appear that this would simply reduce the
8 of 13 20-04-2010 00:05
measured phase shift and therefore also the measured velocity. However, if the surrounding tissue is fat its
relative phase will depend on the TE and the effect of combining the signal becomes less clear.
It should also be remembered that these types of partial volume effects will also occur through the thickness of
the slice, particularly where in-plane velocities are being measured.
Background Phase Shifts
Add to lightbox
page 727
page 728
Figure 28-14 When the blood vessel (circle) is surrounded by no signal and the pixels are relatively large with
respect to it, then the apparent flow region (shaded) will appear larger than the vessel itself.
Phase differences for reasons other than the velocity-encoding pulses can cause the velocity values in phase
velocity maps to be offset from zero. This has the effect of making stationary tissues show an apparent velocity
and this offset also underlies the velocities measured in flow. The background phase shift varies gradually with
position across the image and it also varies with image plane orientation and other sequence parameters which
affect the gradient waveforms, such as Venc. Some distortion of the magnetic field gradients is unavoidable due to
fundamental electromagnetism and this is known in MRI as Maxwell or concomitant gradients. This is one cause
of the background phase shifts which become more significant when high-strength gradients are used and also
when imaging at a lower main field strength. However, with the latest gradient systems Maxwell gradient effects
are certainly a factor for phase contrast velocity mapping at 1.5 Tesla. The velocity maps can, however, be
32
corrected precisely and automatically in software with no user intervention required. A second common reason
for background phase shifts is the presence of small uncorrected side-effects of the gradient pulses in the
magnet, known as eddy currents. Software is sometimes provided which allows the user to place markers
identifying stationary tissues so that the background phase error can be calculated and removed from the entire
image.
Noise Pixels
In velocity maps, pixels without any signal in them have a random phase or show the phase of a weak ghost that
may not be visible on the magnitude image with normal brightness settings. Particularly for post-stenotic jet
images, it is important to check that the magnitude image pixels of the jet are not affected by signal loss. Avoiding
the inclusion of noise pixels can be problematic in regions of interest (ROIs) around the great vessels and ideally
the image analysis software enables the user to set the velocity to zero for pixels whose magnitude is below a
user-defined threshold.
Validation
Phase contrast velocity mapping has been shown to be accurate and reproducible, provided that the sequence
parameters are carefully chosen so that the potential errors and artifacts discussed earlier can be minimized or
9 of 13 20-04-2010 00:05
avoided. Validation has been reported in phantoms both by comparison with true measured flow and with
17,18,33
Doppler, in animal models by comparison with in vivo flow meter measurements34 and in humans by
35-37
comparison with methods such as Doppler ultrasound or catheterization. Perhaps the most convincing forms
of validation were those performed initially by comparing the aortic flow with the left ventricular stroke volume and
later flow in both the aorta and pulmonary artery with left and right ventricular stroke volumes in normal
4,38,39
subjects. For the latter, the four measurements should be the same apart from small differences due to
coronary and bronchial flow and it can be calculated that flow measurements in large vessels are accurate to
within 6%.
Another factor that could affect more recent studies is the impact of breath-holding on the flow measurement
itself. Sakuma and colleagues showed a significant change in both pulmonary and aortic cardiac output during a
40
large lung volume breath-hold. Conversely flows measured during a small volume breath-hold were found to be
similar to those measured during normal breathing.
Of course, other potential sources of error have been reported, including misalignment of the vessel with the
direction of velocity encoding and misregistration of flow signal due to flow between excitation and readout.
Because of these and the other sources of errors discussed earlier, care is required when setting up the scan
parameters, to minimize their impact.
Visualizing Flow and Flow Parameters

Phase contrast velocity maps contain only one velocity measurement per image pixel. These are generally
displayed using a grayscale such that flow in one direction tends towards black, flow in the opposite direction
tends towards white and stationary tissue is displayed midgray as illustrated earlier in Figure 28-3. If
multidirectional flow is of interest and velocities are measured in two or three orthogonal directions then more
sophisticated methods of display can be used. Figure 28-15 shows a streamline map of flow in the right atrium
during atrial filling. This systolic image shows a vortex pattern formed due to the combined inflow from the caval
veins. An alternative approach of representation has been to use the cine velocity images to calculate and display
41
the path of an imaginary "seed" or "seeds" drifting with the flow over time.
Add to lightbox
page 728
page 729
Figure 28-15 A streamline map of flow in the right atrium during atrial filling. This image shows a vortex pattern
formed by the combined inflow from the caval veins. (Reprinted with permission from Nature, Kilner PJ, Yang
GZ, Wilkes AJ, Mohiaddin RH, Firmin DN, Yacoub MH: Asymmetric redirection of flow through the heart.
Nature 404:759-761, 2000. Copyright 2000 Macmillian Magazines Limited.)
10 of 13 20-04-2010 00:05
Add to lightbox
Add to lightbox
11 of 13 20-04-2010 00:05
Add to lightbox
Add to lightbox
Add to lightbox
Figure 28-16 Flow pressures around the aortic arch derived using the Navier Stokes equations from cine phase
contrast velocity maps. The top and bottom rows are the same data with relative pressures at the top whilst at the
bottom the grayscale display is allowed to alias such that each band represents at 1 mmHg pressure range.
(From Yang GZ, Kilner PJ, Wood NB, Underwood SR, Firmin DN: Computation of flow pressure fields from
magnetic resonance velocity mapping. Magn Reson Med 36:520-526, 1996, with permission.)
The ability of MR to study flow in this relatively high detail and in any region of the body is unique. Interest has
been generated from those who wish to understand the physiology of blood flow, its interaction with blood vessels
42
and the cardiac chambers and its relationship to disease. A number of groups have investigated methods of
extracting a measure of the wall shear stress from the MR images despite what would be seen as a relatively
43 44
poor spatial resolution for this purpose. Oshinski et al and Oyre et al developed fitting methods to derive the
velocity profile at sub-pixel distances from the vessel wall. Frayne and Rutt suggested an alternative approach,
45
which potentially gave more information about the flow within a voxel that straddled the vessel wall. Their
method used Fourier velocity encoding to distinguish the distribution of flow velocities, so that only the spatial
location had to be considered.
Possibly a more realistic approach to producing the order of resolution required for studying wall shear is to
combine MR structural and blood flow measurements with computational flow modeling methods that are being
12 of 13 20-04-2010 00:05
developed to a point that is likely to make this feasible.46-49
The possibility of deriving pressure measurements from the MR images has generated considerable interest.
Urchuk et al considered the vessel compliance and the flow pulse wave to calculate the pressure waveform and
50
showed a good correlation with catheter pressure measurements made in a pig model. On the other hand, Yang
51
et al derived flow pressure maps from the cine phase contrast velocity maps using the Navier Stokes equations.
This method was used to measure the changing flow pressure around the aortic arch during the first half of the
cardiac cycle (Fig. 28-16). An interesting example from a patient who has had a dacron graft repair of an aortic
coarctation is shown in Figure 28-17. In contrast to the rest of the aorta, no pressure gradient can be seen in the
repaired region, possibly because of the reduced compliance.
Add to lightbox
page 729
page 730
Figure 28-17 The derived pressure variation in the descending aorta of a patient having had a dacron graft
aneurysm repair. It is thought that the lack of pressure gradient in the region of repair could be due to lack of
compliance of this region. The pressure image is displayed in a similar way to those on the bottom row of Fig.
28-16.
To fully visualize, measure and understand the flow parameters in blood vessels, a method of acquisition is
required which measures velocity in three directions and three dimensions over time, with both a high spatial and
temporal resolution. Kilner and colleagues extracted the flow patterns from multiple 2D image acquisitions to
52
describe the helical flow patterns in the aorta. However, even with the fastest gradient systems available today,
true 3D data acquisition presents a major problem in terms of acquisition time and data handling. Methods have
been suggested, however, using a combination of echo-planar and k-space view sharing that imply that the
acquisition times can be reduced to acceptable levels.53
© 2010 Elsevier
13 of 13 20-04-2010 00:05
Thoracic Aorta
The quantitative study of aortic flow using MR phase contrast velocity mapping has been a subject of
considerable interest in health and disease. This is in part because the aorta is relatively large and the
method offered an entirely novel way of measuring the detailed flow structure, initially within a 2D slice
and later within a volume. The blood flow in the ascending and descending thoracic aorta is phasic and
although the average resting flow measured by MR velocity mapping in the ascending and descending
thoracic aorta of normal subjects is 6.0 L/min and 3.9 L/min respectively, instantaneous peak systolic
flow in these vessels can reach over 40 L/min and 30 L/min respectively.54 The systolic flow velocity
profile in the ascending aorta is skewed, with higher velocities around the inside of the arch (Fig.
28-18). During early diastole the high velocities rotate around the perimeter and simultaneous forward
and reverse channels are formed.52,55
Add to lightbox
Add to lightbox
page 730
page 731
Figure 28-18 Selected cine phase contrast velocity map frames (A) acquired in a transverse plane
through the great vessels of a healthy volunteer. The early systolic flow velocity profile in the ascending
aorta is skewed, with higher velocities around the inside of the arch and later in systole and into
1 of 15 20-04-2010 00:07
diastole, a reverse flow channel forms. The plot (B) illustrates the instantaneous volume flow curves of
the ascending aorta (AA), main pulmonary artery (MPA), descending aorta (DA), and superior vena
cava (SVC).
Add to lightbox
Add to lightbox
Add to lightbox
Figure 28-19 Phase contrast velocity maps illustrating (A) forward systolic flow in darker shades of
gray (arrowed) and (B) diastolic regurgitant flow in lighter shades of gray (arrowed). Ascending aortic
flow volume curve (C) shows a large retrograde net flow during diastole.
In normal subjects, the early diastolic reverse flow channel is predominantly directed towards the
region of the left coronary sinus and it has been suggested that it may augment flow in the left
coronary artery by imparting momentum to the blood which is destined to enter it. In coronary artery
56
disease patients, the reverse flow is reduced and is more variable in direction. In aortic valve
regurgitation, the reverse flow is increased and both aortic or pulmonary regurgitation may be
quantified from the backflow of blood in the great vessels proximal to the respective valves (Fig.
2 of 15 20-04-2010 00:07
28-19).57,58 Abnormal patterns of flow with sustained vortices and larger than normal reverse flows
have been demonstrated in patients with an aortic aneurysm using reconstructed vector maps from
59
multidirectional MR phase velocity mapping.
Aortic Flow Wave Velocity

The flow pulse wave velocity can be measured by acquiring two slices of cine images simultaneously.
By measuring the delay between the leading edge and onset of the flow wave in the two slices and
dividing the distance between the slices by this, a measure for the pulse wave velocity is obtained. For
the thoracic aorta, only one slice is required positioned transverse to the ascending and descending
limbs and this enables a doubling of the temporal resolution.60,61 The flow pulse wave velocity is
considerably higher than the velocity of blood flow itself and is the velocity at which the wave of flow
and pressure is propagated down the aorta. This measurement is closely related to the vascular
compliance and function which may prove to be useful for the detection and monitoring of arterial
56,62-64
disease. The aortic pulse wave velocity is typically between 5 and 10 m/s.
Aortic Dissection
Aortic dissection is readily identified by spin-echo imaging and its extent and involvement of other
vessels can be defined. This method compares very favorably with transesophageal echocardiography
and X-ray computed tomography in the evaluation of aortic dissection. 65,66 However, a thin intimal flap
may not always be clearly visible in these black blood images unless static blood in the false lumen
provides natural contrast with the true lumen. Steady-state imaging can be a good alternative when the
false lumen is not completely thrombosed but differentiation between the true and the false lumen can
be difficult. Any doubt can be resolved by use of phase contrast velocity mapping, where the velocity
maps will confirm the diagnosis by demonstrating the differential flow velocities in each lumen (Fig.
67,68
28-20). Velocity mapping can also be helpful in detecting the site of the intimal tear because flow
between the true and false lumens is not always apparent from simple cine gradient-echo imaging.
Add to lightbox
3 of 15 20-04-2010 00:07
Add to lightbox
page 731
page 732
Figure 28-20 Black blood image (A) clearly identifying the intimal flap (arrowed) in a patient with an
aortic dissection. The phase contrast velocity map (B) confirms the diagnosis by demonstrating the
differential flow velocities in each lumen.
Pulmonary Arteries
Pulmonary blood flow can be difficult to measure by Doppler echocardiography because of the
retrosternal position of the central pulmonary arteries. MR phase contrast velocity mapping is not
constrained in such ways and is capable of demonstrating accurate and detailed velocity images in the
central pulmonary arteries. In normal subjects it can be seen that during most of systole the flow is
skewed towards the posterior two thirds of the artery. A small backflow channel develops posteriorly
towards the end of systole and backflow continues through early diastole. The spatial pulmonary flow
profiles have been studied relatively little in patients but phase contrast velocity mapping of temporal
waveforms has confirmed an abnormally early forward systolic peak and increased reverse diastolic
flow in patients with pulmonary hypertension.39,69,70 This abnormal pattern may be a result of stronger
reflected waves from the distal vasculature which has a high impedance.
Single lung transplant recipients are unique in that the right ventricular output is ejected into two
pulmonary vascular beds with different characteristics; the differential blood flow in these patients
depends on the relative resistance of each lung and both the flow volume and the temporal waveform
are affected (Fig. 28-21).71,72 In general, the ratio of blood flow in the transplanted to that in the native
lungs is about 3:1 and the flow profile in the artery of the transplanted lung shows forward flow during
systole and most of diastole, while that of the native lung shows a narrow early systolic peak and
reverse flow during most of diastole. The potential complications of pulmonary artery dysfunction and
vein anastomoses that can follow lung transplantation can be well monitored using a combination of
contrast-enhanced MRA and phase contrast velocity mapping.
Coronary Blood Flow

Noninvasive examination of the epicardial coronary arteries and measurement of blood flow in these
arteries are challenging but clinically important. The development of fast imaging techniques alongside
novel motion correction methods has improved the ability of MR to image directly the proximal portions
of both the left and right coronary arteries, although the resolution and reliability are both considerably
lower than for X-ray angiography.73,74 Although it is likely that new developments will reduce these
limitations, the ability of MRI to additionally measure proximal coronary flow velocities would be a very
useful adjunct to the anatomic assessment alone.
4 of 15 20-04-2010 00:07
Add to lightbox
Add to lightbox
Figure 28-21 Transverse image at the level of the pulmonary bifurcation (A) of a patient who has had a
left lung transplantation. The plot (B) shows the flow versus time curves for the main (MPA), left (LPA),
and right (RPA) pulmonary arteries. The flow profile in the LPA shows forward flow during systole and
most of diastole, while that of the RPA to the native lung shows a narrow early systolic peak and
reverse flow during diastole. The measured ratio of blood flow in the transplanted to that in the native
lung is about 3:1 in this case.
The feasibility of phase contrast velocity measurement of flow in the epicardial coronary arteries was
75,76
originally demonstrated in the early to mid 1990s. More recently groups have shown cine coronary
flow velocity measurements using breath-hold gradient-echo acquisitions.77 The method has been used
to measure coronary flow velocity reserve in patients with and without pharmacologic stress. The
combination of having to acquire the data within a breath-hold and using the relatively inefficient
gradient-echo sequence led to a low temporal resolution, despite incorporation of view sharing, along
with problems in incorporating fat suppression methods. Keegan et al have recently developed a
navigator-controlled interleaved spiral sequence enabling high spatial and temporal resolution along
78
with good fat suppression by use of a water selective excitation pulse. This sequence has shown
rapid temporal velocity variations that were not detected by the previous MR phase contrast methods
(Fig. 28-22). This is an important development and opens a new opportunity for cardiovascular MRI.
Coronary artery bypass grafts have also been assessed using MR velocity mapping.79 Flow
measurement within the grafts would be clinically valuable but is not always possible because of signal
loss due to sternal suture and clips left after surgery. Metallic objects create larger artifacts in field
echo sequences than in spin-echo imaging sequences.
5 of 15 20-04-2010 00:07
page 732
page 733
Add to lightbox
Add to lightbox
Add to lightbox
Figure 28-22 Magnitude image (A) and phase contrast velocity map (B) showing the right coronary
blood flow (arrowed). The acquisition used a navigator-controlled interleaved spiral sequence with a
water excitation pulse. This provided an in-plane spatial resolution of 0.9 mm and a temporal
resolution of 22 ms. The plot (C) shows the measured coronary flow through the cardiac cycle.
Caval Veins
The caval veins are also relatively large and despite the somewhat lower velocities, reliable velocity
maps and flow measurements have readily been obtained.80 Superior and inferior venae cavae flow
volumes are approximately 35% and 65% of cardiac output respectively.80 There are normally two
forward peaks in ventricular systole and diastole (see Fig. 28-18), but this pattern can be disturbed by
disease. Conditions that cause impaired filling of the right ventricle result in a reduction of the diastolic
peak, a pattern seen in both constrictive and restrictive cardiac disease (Fig. 28-23).80 The systolic
peak of caval flow is attenuated by tricuspid regurgitation sometimes to the extent that reverse flow
occurs.80 From a clinical point of view this is not especially helpful, because the severity of
regurgitation can be assessed by cine imaging or from a comparison of right and left ventricular stroke
volumes. Nevertheless, a normal systolic flow peak suggests that tricuspid regurgitation, if it is seen, is
not significant. MR is often requested for the clinical assessment of pericardial disease and the ability
to measure caval flow can be an important adjunct, providing a measure of the functional significance
of the disease. In patients with obstruction of the superior vena cava, absence of flow can be
confirmed and reverse flow in the azygous vein can be measured.78
6 of 15 20-04-2010 00:07
Pulmonary Veins
Add to lightbox
Add to lightbox
Figure 28-23 A transverse midventricular level spin-echo image (A) of a patient with constrictive
pericarditis exhibiting pericardial thickening (arrows). The plot (B) shows the superior vena caval flow
curve of the same patient measured from the complete cine velocity map acquisition throughout the
cardiac cycle. The diastolic peak is attenuated, which implies impaired right ventricular filling. (From
Mohiaddin RH, Longmore DB: The functional aspects of cardiovascular magnetic resonance
imaging: techniques and applications. Circulation 88:264-281, 1993, with permission.)
The temporal waveforms of normal pulmonary venous flows measured by phase contrast velocity
mapping show two forward flow peaks, one during ventricular systole and the other in diastole (Fig.
28-24).81,82 In addition a small backflow during atrial systole occurs, as has been demonstrated in the
pulmonary veins by transesophageal Doppler echocardiography and the transmitral "A" flow peak. 83 A
left ventricle which is noncompliant leads to high left atrial pressure during atrial systole and this causes
the retrograde flow in the pulmonary veins to become larger than the flow through the mitral valve. An
attenuated systolic forward flow peak has been demonstrated in patients with mitral valve regurgitation
and the degree of this attenuation correlates well with the severity of regurgitation. 37
Ventricular Filling
page 733
page 734
7 of 15 20-04-2010 00:07
Add to lightbox
Add to lightbox
Add to lightbox
Figure 28-24 Oblique plane phase contrast velocity maps illustrating forward pulmonary venous flows
during ventricular systole (A) and diastole (B). Plot (C) shows the temporal waveforms of normal
pulmonary venous flows measured by phase contrast velocity mapping. The flow has two forward flow
peaks, one during ventricular systole and the other in diastole.
An important parameter in the assessment of left and right ventricular diastolic function is measured
84
blood flow through the mitral and tricuspid valves. Normally this flow takes place in two phases which
can clearly be identified by phase contrast velocity mapping (Fig. 28-25).81 In early diastole the initial
flow (E wave) occurs as a consequence of the ventricles relaxing and allowing blood to flow from the
slightly higher pressure in the blood-filled atria. This flow normally takes place very quickly and there is
a mid-diastolic reduction or cessation of flow before a second phase of flow caused by atrial
contraction (A wave). Phase contrast velocity mapping of the E wave agrees well with those
measurements obtained by Doppler echocardiography, but MR underestimates A wave velocity. This
underestimation has probably been due to beat length variability, which occurs in the T-P time
8 of 15 20-04-2010 00:07
interval,85 combined with a fixed frame rate for prospectively triggered sequences. This may well be
resolved with use of more sophisticated retrospective cardiac gating. In patients with reduced
ventricular compliance, for example in ischemic heart disease, the A wave becomes more prominent
85
and the E/A ratio is reduced; these changes can be demonstrated using MRI.
Add to lightbox
Add to lightbox
Add to lightbox
Figure 28-25 Frames of a cine acquired in the long axis of the heart during systole (A), where flowing
blood can be seen as lighter shades of gray entering the aortic outflow tract, and during diastole (B)
where flowing blood can be seen as darker shades of gray entering the ventricle through the mitral
valve. Plot (C) illustrating the two forward peaks in normal mitral and tricuspid flows.
Reconstructed vector maps from multidirectional MR phase velocity mapping are well suited for
studying spatial and temporal patterns of flow in the cardiac chambers. In patients with myocardial
ischemia leading to a dilated left ventricle, the ventricular flow pattern is disturbed, with the inflow
directed toward the free wall as opposed to the apex, giving rise to a well-developed circular flow
pattern turning back towards the septum and outflow tract and persisting entirely through diastole to
9 of 15 20-04-2010 00:07
the next systolic phase (Fig. 28-26). This abnormal pattern of circular flow may be related to the
increased inertial mass of fluid in the dilated chamber combined with the unusual angle between the
mitral valve and the interventricular septum.14 The clinical significance of this abnormal flow pattern in
relation to left ventricular function and thrombosis is thus far unknown.
Valvular Disease
Valvular Stenosis
page 734
page 735
Add to lightbox
Figure 28-26 End-diastolic gradient-echo image (A) in the horizontal long axis plane of the left
ventricle in a patient with a dilated left ventricle (LV). The vector maps (B,C), calculated from diastolic
frames of cine phase velocity maps at 550 and 650 ms respectively, illustrate the well-developed
circular flow pattern normally found in these patients. (From Mohiaddin RH, Pennell DJ: MR blood
flow measurement. Clinical application in the heart and circulation. Cardiol Clin 16(2):161-187,
1998, with permission.)
Earlier Doppler studies have shown that a stenotic valve may be assessed by measuring the high flow
velocity in the jet of blood passing through and distal to the stenosis. For the same flow, the greater
the severity of the stenosis, the higher the velocity of flow will be through the orifice. The modified
Bernoulli equation is generally used to give the approximate relationship between the velocity of the jet
and the pressure difference across the stenosis:
where ∆P = pressure drop across the stenosis (often known as gradient)(mmHg) and v = velocity
(m/s).
Accurate velocity mapping of stenotic jets by MR requires the use of sequences that do not lose signal
in regions of complex flow. As described earlier, this can be achieved by shortening the gradient
durations which normally go hand in hand with TE.17,18 The velocity map may show the flow in-plane,
where the imaging plane is chosen to encompass the length of the jet, or through-plane, with the jet
passing perpendicularly through the chosen imaging plane (Fig. 28-27). The in-plane imaging gives a
10 of 15 20-04-2010 00:07
clearer representation of the jet with a greater number of pixels for analysis of velocity but if the jet is
small, in-plane imaging can be less reliable because of partial volume effects and possible movement
of the jet out of the imaging plane. Particular care is required in positioning the image plane and in
some circumstances it may be preferable to acquire data in both planes. In a recent development
Nayak and colleagues have described an interleaved short spiral acquisition that has higher scan
efficiency but which is robust against signal loss and other flow artifacts. 86
Add to lightbox
Add to lightbox
Figure 28-27 Magnitude image (A) transverse to the aortic root illustrating a stenosed bicuspid aortic
valve and the corresponding phase contrast velocity map (B) showing a jet (arrowed) with a peak flow
velocity of 3.8 m/s.
Valvular Regurgitation
Measurement of regurgitation is an important but difficult aspect of the assessment of valvular disease.
The need for intervention is determined partly by the severity of symptoms, but other measurements
such as the severity of regurgitation and ventricular function are important.
page 735
page 736
11 of 15 20-04-2010 00:07
Add to lightbox
Figure 28-28 Systolic frame of a horizontal long axis cine acquisition from a patient with an insufficient
mitral valve. There is a jet of signal loss in the left atrium indicating a moderate mitral valve
regurgitation.
As discussed earlier, turbulent blood flow in combination with certain sequence parameters can lead to
loss of signal and in this way the turbulent jet of regurgitation can easily be seen in the receiving
chamber (Fig. 28-28). The size of the signal void can be used as a semi-quantitative measure of
regurgitation16 although it is important to be aware of the influence of imaging parameters such as
echo time (TE). In some ways this is not dissimilar to color flow Doppler where technical factors such
as gain adjustment and filter setting are important. A more fundamental problem common to both is
that the size of the regurgitant jet is influenced by many factors in addition to the severity of
regurgitation, such as the shape and size of the regurgitant orifice, the size of the receiving chamber,
and whether the jet infringes on the chamber wall.
A correlation has been shown between the angiographic grading of mitral regurgitation and the length
and area of the regurgitant jet imaged by Doppler echocardiography. Doppler has also been shown to
correlate with cine magnetic resonance, although for the parameters used in this study magnetic
87
resonance shows a smaller area of flow disturbance. An advantage of magnetic resonance is that it
is able to sample the jet in any plane or potentially, if required, in a complete three-dimensional volume
and so may provide a more accurate assessment of the volume occupied by the regurgitant jet.
For further quantification, if only a single valve is regurgitant, the regurgitant volume and the regurgitant
fraction can be calculated from the difference between left and right ventricular stroke volumes
measured from anatomic images without use of phase contrast velocity imaging. If single valves on
both sides of the heart are regurgitant, the method can be extended by comparing ventricular stroke
volumes with great vessel flow measured by phase contrast velocity mapping. When using this
approach, the regurgitant fraction has been shown to compare well with the regurgitant grade
88
assessed by Doppler echocardiography. This method still fails if both valves on one side of the heart
are regurgitant. However, phase contrast velocity mapping in the proximal aorta or pulmonary artery
can be used to measure aortic or pulmonary regurgitation alone from the amount of retrograde
diastolic flow in the artery and it is then possible to assess regurgitant volumes of all four valves even
in the most complex cases.
Nayak and colleagues have described a display system where they have introduced a color overlay to
the MR magnitude images such that the end result is similar to color Doppler. This was reported to be
particularly useful for imaging flow through regurgitant valves. 89
An important development that will be particularly important for improving the accuracy of valvular flow
12 of 15 20-04-2010 00:07
studies is that of a moving slice cine acquisition. The potential of this has been demonstrated by
Kozerke and colleagues, when they used the method to study patients with aortic and mitral
regurgitation.90 Unfortunately, at present, no manufacturer has provided this as a clinical tool.
Abdominal Aorta
The abdominal aorta reduces gradually in cross-sectional area with distance in the caudal direction but
the blood flow velocity measured by MR remains relatively constant.91 Flow in the abdominal aorta has
a typical "plug" flow pattern often found in large arteries. It is likely that the helical flow described in the
normal thoracic aorta52 may persist to some level although this has not been studied. The blood flow is
closely coupled to the cardiac output so that maximum aortic flow is during midventricular systole and
then decreases towards the end of systole. During the majority of diastole, however, net aortic flow is
reversed below the origin of the renal arteries. This reversed flow component is predominantly
distributed along the posterior wall of the abdominal aorta. Superior mesenteric flow consists of a
predominant systolic peak of forward net flow, which nearly always remains forward during diastole.
This may well reflect the low resistance in the vascular bed of this artery and patients with chronic
mesenteric ischemia show abnormal flow.92 Total and differential renal blood flow has also been
measured by MR velocity mapping and the total renal blood flow has been shown to be accurate when
compared with conventional methods.93
Peripheral Arteries
page 736
page 737
Add to lightbox
Add to lightbox
13 of 15 20-04-2010 00:07
Add to lightbox
Figure 28-29 An X-ray angiogram (A) and a spin-echo MR image (B) showing multiple atheromatous
plaques causing stenoses of both the common iliac arteries and the origin of the left internal and
external iliac arteries (arrows). On the velocity map (C), acquired in the same plane as (B), there is an
increased peak velocity (whiter pixels) (arrowed) in both iliac arteries compared to the aorta. (From
Mohiaddin RH, Longmore DB: The functional aspects of cardiovascular magnetic resonance
imaging: techniques and applications. Circulation 88:264-281, 1993, with permission.)
Atherosclerosis in the more peripheral arteries produces clinical problems either by reducing blood flow
or by releasing emboli from ulcerated plaques. Stenoses can be detected and their severity assessed
using in-plane phase velocity mapping and measuring changes in the velocity profile across the
stenosis (Fig. 28-29). When the area of the vessel is decreased at a stenosis to accommodate the
same flow, the average velocity must increase before reducing again on the distal side. Using phase
contrast velocity mapping, it is also possible to measure the flow rate and the flow volume curve in a
vessel calculated from the mean velocity and the cross-sectional area of the vessel. The flow ratio in
paired vessels such as the iliac arteries can be a useful measure. This has been shown to be always
>0.85 in healthy volunteers and <0.85 in patients with tight stenosis of one iliac artery.94 In patients
studied before and after angioplasty, the ratio improved from 0.31 to 0.79. This finding correlated with
the clinical findings of an improvement in the walking distance and in the peripheral Doppler pressure
measurement (the posterior tibial/brachial artery index was right = 1, left = 0.88 on the preangioplasty
recordings and right = 1, left = 0.96 on the postangioplasty recordings).94
Sometimes of greater interest than flow ratios is the temporal flow curve which is altered in diseased
vessels when compared with the normal. This may be an important indicator, for example, when there
is disease of both iliac arteries as ischemia may be balanced and quantitative flow may be equal in the
two vessels.
Congenital Heart Disease

The excellent 3D structural information in combination with phase contrast velocity mapping provides an
95,96
excellent tool in both adolescent and adult patients with congenital heart disease. Transthoracic
echocardiography in these patients may be difficult because of the complexity and variability of the
disease and because ultrasonic access can be restricted, especially in patients who have undergone
surgery. Transesophageal echocardiography forms a complementary technique, 97 delineating
structures such as the atrial septum and any defects in it, and atrioventricular valves with their cordal
attachments. Magnetic resonance with its flow capability generally gives more information on the
structure and function of extracardiac vessels, notably anomalous venous connections, pulmonary
arteries, the aortic arch, and a patent duct or extracardiac conduit.
In infants and children with congenital heart disease MRI is less used, generally requiring sedation for
satisfactory imaging, whilst echocardiography is usually an effective investigative tool. Even in this
group of young patients, however, MRI with phase contrast velocity mapping can be valuable.
Examples include the investigation of patients following Fontan's operation and its modifications98 and
of patients with baffle obstruction following the Mustard procedure for transposition of the great
arteries.99,100
14 of 15 20-04-2010 00:07
Another important application of phase contrast velocity mapping in congenital heart disease patients is
in the quantification of intra- and extracardiac shunting which can be assessed in a number of ways.
Blood flowing directly through either atrial or ventricular defects can be visualized (Fig. 28-30);
however, this would be difficult to quantify due to cardiac motion and the best method has been to
measure pulmonary (Qp) to systemic (Qs) ratio directly from aortic and pulmonary flow.101,102 In
complex lesions, this is especially helpful, because the possibility of surgery in such patients depends
partly upon pulmonary flow which is difficult to measure by other techniques.
Phase contrast velocity mapping is also particularly valuable in patients with either native or repaired
aortic coarctation (Fig. 28-31) as these generally cannot be adequately assessed using ultrasound.
Cine gradient-echo imaging has been used to assess the severity of aortic coarctation from the
maximum area of signal loss in the descending aorta.103,104 However, it is probably more accurate to
use velocity mapping, which can be implemented in a number of ways. The most direct is to image the
jet at the site of the coarctation and from this to estimate the pressure "gradient" as described
earlier.105 Alternatively flow can be compared in the ascending aorta and the descending aorta beyond
the coarctation105 or by comparing descending aortic flow just distal to the coarctation with flow at the
level of the diaphragm.106 The latter technique gives an estimate of the importance of collateral flow to
the lower body, which may be of importance in guiding surgical intervention.
© 2010 Elsevier
15 of 15 20-04-2010 00:07
CONCLUSION
page 737
page 738
Add to lightbox
Add to lightbox
Add to lightbox
1 of 6 20-04-2010 00:07
Figure 28-30 Transverse magnitude image (A) showing a large atrial septal defect (asd) (arrowed).
The velocity map of the same slice (B) illustrates the blood flowing from the left to the right atrium in
lighter shades of gray (arrowed). Plot (C) illustrates the volume flow through the cardiac cycle
measured in the main pulmonary artery (MPA) and the aorta. The pulmonary to systemic flow ratio is
1.9. (A,B from Mohiaddin RH, Pennell DJ: MR blood flow measurement. Clinical application in the
heart and circulation. Cardiol Clin 16(2):161-187, 1998, with permission.)
Add to lightbox
Add to lightbox
Figure 28-31 An oblique spin-echo image (A) illustrating an aortic coarctation (arrowed). The velocity
map (B) shows the in-plane jet with a peak velocity of 3.5 m/s.
Phase contrast velocity mapping has been shown to be a robust, accurate, and flexible method of
obtaining functional information in the cardiovascular system. It can contribute significantly to clinical
decision-making and it should be a routine part of cardiovascular imaging whenever information on flow
is required. It also offers considerable potential for physiologic studies and furthering our understanding
of the relationships between blood flow and disease formation. Improvements in hardware alongside
the development of more rapid techniques have improved its usefulness. Whilst extraction of
quantitative flow information from the phase contrast velocity maps has in the past been tedious,
dedicated software is now becoming available and this will significantly extend the capabilities of the
technique.
Acknowledgments
The authors would particularly like to thank Cathy Hawkes, Lindsey Crowe, Sanjay Prasad, and Jenny
Keegan along with other members of the Cardiovascular Magnetic Resonance Unit at the Royal
Brompton Hospital/Imperial College, London, for their help with this text.
page 738
page 739
REFERENCES
1. Bryant DJ, Payne JA, Firmin DN, Longmore DB: Measurement of flow with NMR imaging using a gradient pulse and phase
2 of 6 20-04-2010 00:07
difference technique. J Comput Assist Tomogr 8:588-593, 1984. Medline Similar articles
2. van Dijk P: Direct cardiac NMR imaging of heart wall and blood flow velocity. J Comput Assist Tomogr 8:429-436, 1984.
3. Nayler GL, Firmin DN, Longmore DB: Blood flow imaging by cine magnetic resonance. J Comput Assist Tomogr
4. Firmin DN, Nayler GL, Klipstein RH, et al: In vivo validation of magnetic resonance velocity imaging. J Comput Assist
5. Walker MF, Souza SP, Dumoulin CL: Quantitative flow measurement in phase contrast MR angiography. J Comput Assist
6. Feinberg DA, Crooks L, Hoenninger J 3rd, et al: Pulsatile blood velocity in human arteries displayed by magnetic resonance
imaging. Radiology 153:177-180, 1984.
7. Edelman RR, Mattle HP, Kleefield J, Silver MS: Quantification of blood flow with dynamic MR imaging and presaturation
bolus tracking. Radiology 171:551-556, 1989. Medline Similar articles
8. Matthaei D, Haase A, Merboldt KD, et al: ECG-triggered arterial FLASH-MR flow measurement using an external standard.
9. Moran PR: A flow velocity zeugmatographic interlace for NMR imaging in humans. Magn Reson Imaging 1:197-203, 1982.
10. Mohiaddin RH, Longmore DB: The functional aspects of cardiovascular magnetic resonance imaging: techniques and
applications. Circulation 88:264-281, 1993. Medline Similar articles
11. Rebergen SA, van der Wall EE, Doornbos J, de Roos A: Magnetic resonance measurement of velocity and flow:
technique, validation and cardiovascular applications. Am Heart J 126:1439-1456, 1993. Medline Similar articles
12. Mostbeck GH, Caputo GR, Higgins CB: MR measurement of blood flow in the cardiovascular system. Am J Roentgenol
159:453-461, 1992.
13. Mohiaddin RH, Yang GZ, Kilner PJ: Visualization of flow by vector analysis of multidirectional cine magnetic resonance
velocity mapping. J Comput Assist Tomogr 18:383-392, 1994. Medline Similar articles
14. Mohiaddin RH: Flow patterns in the dilated ischaemic left ventricle studied by magnetic resonance imaging with velocity
vector mapping. J Magn Reson Imag 5:493-498, 1995.
15. Carr HY, Purcell EM: Effects of diffusion on free precession in nuclear magnetic resonance experiments. Phys Rev
94:630-638, 1954.
16. Schiebler M, Axel L, Reichek N, et al: Correlation of cine MR imaging with 2-dimensional pulsed Doppler
echocardiography in valvular insufficiency. J Comput Assist Tomogr 11:627-632, 1987.
17. Firmin DN, Nayler GL, Kilner PJ, Longmore DB: The application of phase shifts in NMR for flow measurement. Magn
18. Kilner PJ, Firmin DN, Mohiaddin RH, et al: Valve and great vessel stenosis: assessment with MR jet velocity mapping.
Radiology 178:229-235,1991. Medline Similar articles
19. Keegan J, Gatehouse PD, John AS, et al: Breath-hold signal-loss sequence for the qualitative assessment of flow
disturbances in cardiovascular MR. J Magn Reson Imag 18:496-501, 2003.
20. Conturo TE, Smith GD: Signal-to-noise in phase angle reconstruction: dynamic range extension using phase reference
offsets. Magn Reson Med 15:420-437, 1990. Medline Similar articles
21. Buonocore MH: Blood flow measurement using variable velocity encoding in the RR interval. Magn Reson Med
22. Yang GZ, Burger P, Kilner PJ, et al: Dynamic range extension of cine velocity measurements using motion-registered
spatiotemporal phase unwrapping. J Magn Reson Imag 6:495-502, 1996.
23. Irarrazabal P, Hu BS, Pauly JM, Nishimura DG: Spatially resolved and localized real-time velocity distribution. Magn Reson
24. Firmin DN, Klipstein RH, Hounsfield GL, et al: Echo-planar high-resolution flow velocity mapping. Magn Reson Med
25. Eichenberger AC, Schwitter J, McKinnon GC, et al: Phase-contrast echo-planar MR imaging: real-time quantification of
flow and velocity patterns in the thoracic vessels induced by Valsalva's maneuver. J Magn Reson Imag 5:648-655, 1995.
26. Duerk JL, Simonetti OP: Theoretical aspects of motion sensitivity and compensation in echo-planar imaging. J Magn
Reson Imag 1:643-650, 1991.
27. McKinnon GC, Debatin JF, Wetter DR, von Schulthess GK: Interleaved echo planar flow quantitation. Magn Reson Med
28. Gatehouse PD, Firmin DN, Collins S, Longmore DB: Real time blood flow imaging by spiral scan phase velocity mapping.
J Magn Res Med 31:504-512, 1994.
29. Mohiaddin RH, Gatehouse PD, Firmin DN: Exercise-related changes in aortic flow measured by magnetic resonance
spiral echo-planar phase-shift velocity mapping. J Magn Reson Imag 5:159-163, 1995.
30. Gatehouse PD, Firmin DN: Flow distortion and signal loss in spiral imaging. Magn Reson Med 41:1023-1031, 1999.
3 of 6 20-04-2010 00:07

31. Mohiaddin RH, Gatehouse D, Moon JC, et al: Assessment of reactive hyperaemia using real time zonal echo-planar flow
imaging. J Cardiovasc Magn Reson 4:283-287, 2002. Medline Similar articles
32. Bernstein MA, Zhou XJ, Polzin JA, et al: Concomitant gradient terms in phase contrast MR: analysis and correction. Magn
33. Meier D, Maier S, Boesiger P: Quantitative flow measurements on phantoms and on blood vessels with MR. Magn Reson
34. Pettigrew RI, Dannels W, Galloway JR, et al: Quantitative phase-flow imaging in dogs by using standard sequences:
comparison with invivo-flow meter measurements. Am J Roentgenol 148:411-414, 1987.
35. Kilner PJ, Manzara CC, Mohiaddin RH, et al: Magnetic resonance jet velocity mapping in mitral and aortic valve stenosis.
36. Maier SE, Meier D, Boesiger P, et al: Human abdominal aorta: comparative measurements of blood flow with MR imaging
and multigated Doppler US. Radiology 171:487-492, 1989. Medline Similar articles
37. Van Rossum A, Sprenger KH, Peels FC, et al: In vivo validation of quantitative flow imaging in arteries and veins using
magnetic resonance phase shift techniques. Eur Heart J 12:117-126, 1991. Medline Similar articles
38. Bogren HG, Klipstein RH, Firmin DN, et al: Quantitation of antegrade and retrograde blood flow in the human aorta by
magnetic resonance. Am Heart J 117:1214-1222, 1989. Medline Similar articles
39. Bogren HG, Klipstein RH, Mohiaddin RH, et al: Pulmonary artery distensibility and blood flow patterns: a magnetic
resonance study of normal subjects and of patients with pulmonary arterial hypertension. Am Heart J 118:990-999, 1989.
40. Sakuma H, Kawada N, Kubo H, et al: Effect of breath holding on blood flow measurement using fast velocity encoded cine
41. Napel S, Lee DH, Frayne R, Rutt BK: Visualizing three-dimensional flow with simulated streamlines and three-dimensional
phase-contrast MR imaging. J Magn Reson Imag 2:143, 1992.
42. Kilner PJ, Yang GZ, Wilkes AJ, et al: Asymmetric redirection of flow through the heart. Nature 404:759-761, 2000.
43. Oshinski JN, Ku DN, Mukundan S Jr, et al: Determination of wall shear stress in the aorta with the use of MR phase velocity
mapping. J Magn Reson Imag 5:640, 1995.
44. Oyre S, Ringgaard S, Kozerke S, et al: Accurate noninvasive quantitation of blood flow, cross-sectional lumen vessel area
and wall shear stress by three-dimensional paraboloid modeling of magnetic resonance imaging velocity data. J Am Coll
Cardiol 32:128, 1998. Medline Similar articles
45. Frayne R, Rutt BK: Measurement of fluid-shear rate by Fourier-encoded velocity imaging. Magn Reson Med 34:378, 1995.
46. Zhao SZ, Papathanasopoulou P, Long Q, et al: Comparative study of magnetic resonance imaging and image-based
computational fluid dynamics for quantification of pulsatile flow in a carotid bifurcation phantom. Ann Biomed Eng 31:962-971,
47. Saber NR, Wood NB, Gosman AD, et al: Progress towards patient-specific computational flow modeling of the left heart
via combination of magnetic resonance imaging with computational fluid dynamics. Ann Biomed Eng 31:42-52, 2003.
48. Thomas JB, Milner JS, Rutt BK, Steinman DA: Reproducibility of image-based computational fluid dynamics models of the
human carotid bifurcation. Ann Biomed Eng 31:132-141, 2003. Medline Similar articles
49. Jin S, Oshinski J, Giddens DP: Effects of wall motion and compliance on flow patterns in the ascending aorta. J Biomech
Eng 125:347-354, 2003. Medline Similar articles
50. Urchuk SN, Fremes SE, Plewes DB: In vivo validation of MR pulse pressure measurement in an aortic flow model:
preliminary results. Magn Reson Med 38:215-223, 1997. Medline Similar articles
51. Yang GZ, Kilner PJ, Wood NB, et al: Computation of flow pressure fields from magnetic resonance velocity mapping. Magn
52. Kilner PJ, Yang GZ, Mohiaddin RH, et al: Helical and retrograde secondary flow patterns in the aortic arch studied by three-
directional magnetic resonance velocity mapping. Circulation 88:2235-2247, 1993. Medline Similar articles
53. Firmin DN, Gatehouse PD, Yang GZ, et al: A 7-dimensional echo-planar flow imaging technique using a novel k-space
sampling scheme with velocity compensation. Proceedings of the Fifth Annual Meeting of the International Society of Magnetic
Resonance in Medicine 1997, p. 118 (abstract).
54. Mohiaddin RH, Kilner PJ, Rees RSO, Longmore DB: Magnetic resonance volume flow and jet velocity mapping in aortic
coarctation. J Am Coll Cardiol 22:1515-1521, 1993. Medline Similar articles
55. Klipstein RH, Firmin DN, Underwood SR, et al: Blood flow patterns in the human aorta studied by magnetic resonance. Br
Heart J 58:316-323, 1987. Medline Similar articles
56. Bogren HG, Mohiaddin RH, Klipstein RH, et al: The function of the aorta in ischemic heart disease: a magnetic resonance
and angiographic study of aortic compliance and blood flow patterns. Am Heart J 118:234-247, 1989. Medline
Similar articles
4 of 6 20-04-2010 00:07
57. Dulce MC, Mostbeck GH, O'Sullivan RN, et al: Severity of aortic regurgitation: interstudy reproducibility of measurements
with velocity-encoded cine MR imaging. Radiology 185:235-240, 1992. Medline Similar articles
58. Rebergen SA, Chin JGL, Ottenkamp J, et al: Pulmonary regurgitation in the late post-operative follow-up of tetralogy of
Fallot: volumetric quantification by nuclear magnetic resonance velocity mapping. Circulation 88:2257-2266, 1993. Medline
Similar articles
59. Bogren HG, Mohiaddin RH, Yang GZ, et al: Magnetic resonance velocity vector mapping of blood flow in thoracic aortic
aneurysms and grafts. J Thorac Cardiovasc Surg 110:704-714, 1995. Medline Similar articles
60. Mohiaddin RH, Firmin DN, Longmore DB: Age-related changes of human aortic flow wave velocity measured
non-invasively by magnetic resonance imaging. J Applied Physiol 74:492-497, 1993.
61. Dumoulin CL, Doorly DJ, Caro CG: Quantitative measurement of velocity at multiple positions using comb excitation and
Fourier velocity encoding. Magn Reson Med 29:44-52, 1993. Medline Similar articles
page 739
page 740
62. Mohiaddin RH, Underwood SR, Bogren HG, et al: Regional aortic compliance studied by magnetic resonance imaging: the
effects of age, training, and coronary artery disease. Br Heart J 62:90-96, 1989. Medline Similar articles
63. Stefanadis C, Wooley CF, Bush AC, et al: Aortic distensibility abnormalities in coronary artery disease. Am J Cardiol
64. Dart AM, Lacombe F, Yeoh JK, et al: Aortic distensibility in patients with hypercholesterolaemia, coronary artery disease,
or cardiac transplantation. Lancet 338:270-273, 1991. Medline Similar articles
65. Goldman AP, Kotler MN, Scanlon MH, et al: The complementary role of magnetic resonance imaging, Doppler
echocardiography, and computed tomography in the diagnosis of dissecting thoracic aneurysms. Am Heart J 111:970-981,
66. Nienaber CA, Spielmann RP, von Kodolitsch Y, et al: Diagnosis of thoracic aortic dissection: magnetic resonance imaging
versus transoesophageal echocardiography. Circulation 85:434-447, 1992. Medline Similar articles
67. Bogren HG, Underwood SR, Firmin DN, et al: Magnetic resonance velocity mapping in aortic dissection. Br J Radiol
68. Chang JM, Friese K, Caputo GR, et al: MR measurement of blood flow in the true and false channel in chronic aortic
dissection. J Comput Assist Tomogr 15:418-423, 1991. Medline Similar articles
69. Kondo C, Caputo GR, Masui T, et al: Pulmonary hypertension: pulmonary flow quantification and flow profile analysis with
velocity-enhanced cine MR imaging. Radiology 183:751-758, 1992. Medline Similar articles
70. Frank H, Globits S, Neuhold A, et al: Detection and quantification of pulmonary artery hypertension with MR imaging: results
in 23 patients. Am J Roentgenol 161:27-31, 1993.
71. Mohiaddin RH, Paz R, Theodoropolus S, et al: Magnetic resonance characterization of pulmonary arterial blood flow
following single lung transplantation. J Thoracic Cardiovasc Surg 101:1016-1023, 1991.
72. Silverman JM, Julien PJ, Herfkens RJ, Pelc NJ: Quantitative differential pulmonary perfusion: MR imaging versus
radionuclide lung scanning. Radiology 189:699-701, 1993. Medline Similar articles
73. Edelman RR, Manning W, Burstein D, Paulin S: Coronary arteries: breath-hold MR angiography. Radiology 181:641-643,
74. Wang SJ, Nishimura DG, Macovski A: Fast angiography using fast selective inversion recovery. Mag Res Med
23:109-121, 1992.
75. Edelman R, Manning WJ, Gervino E, Li W: Flow velocity quantification in human coronary arteries with breath hold MR
angiography. J Magn Reson Imag 3:699-703, 1993.
76. Keegan J, Firmin DN, Gatehouse PD, Longmore DB: The application of breath hold phase velocity mapping technique to
the measurement of coronary artery blood flow: phantom data and initial in vivo results. Magn Reson Med 31:526-536, 1994.
77. Saito Y, Sakuma H, Shibata M, et al: Assessment of coronary flow velocity reserve using fast velocity-encoded cine MRI for
noninvasive detection of restenosis after coronary stent implantation. J Cardiovasc Magn Reson 3:209-214, 2001. Medline
Similar articles
78. Keegan J, Gatehouse PD, Mohiaddin RH, et al: Comparison of spiral and FLASH phase velocity mapping, with and
without breath-holding, for the assessment of left and right coronary artery blood flow velocity. J Magn Reson Imag 19:40-49,
2004.
79. Debatin JF, Strong JA, Sostman HD, et al: MR characterization of blood flow in native and grafted internal arteries. J Magn
Reson Imag 3:443-450, 1993.
80. Mohiaddin RH, Wann SL, Underwood SR, et al: Vena caval flow: assessment with cine MR velocity mapping. Radiology
81. Mohiaddin RH, Amanuma M, Kilner PJ, et al: Magnetic resonance phase-shift velocity mapping of mitral and pulmonary
venous flow. J Comput Assist Tomogr 15:237-243, 1991. Medline Similar articles
82. Mohiaddin RH, Amanuma M, Longmore DB: Magnetic resonance measurement of pulmonary venous flow and
distensibility. Am J Noninvasive Cardiol 6:13-18, 1992.
5 of 6 20-04-2010 00:07
83. Nishimura RA, Abel MD, Hatle LK, Tajik J: Relation of pulmonary vein to mitral flow velocities by transoesophageal Doppler
echocardiography: effect of different loading conditions. Circulation 81:1488-1497, 1990. Medline Similar articles
84. Appleton CP, Hatle LK, Popp RL: Relation of mitral flow velocity patterns to left ventricular diastolic function: new insight
from a combined haemodynamic and Doppler echocardiographic study. J Am Coll Cardiol 12:426-440, 1988. Medline
Similar articles
85. Karwatoski SP, Brecker SJD, Yang GZ, et al: Mitral valve flow measured with cine MR velocity mapping in patients with
ischaemic heart disease: comparison with Doppler echocardiography. J Magn Reson Imag 5:89-92, 1995.
86. Nayak KS, Hu BS, Nishimura DG: Rapid quantitation of high-speed flow jets. Magn Reson Med 50:366-372, 2003.
87. Aurigemma G, Reichek N, Schiebler M, Axel L: Evaluation of mitral regurgitation by cine magnetic resonance imaging. Am
J Cardiol 66:621-625, 1990. Medline Similar articles
88. Manzara CC, Underwood SR, Pennell DJ, et al: Magnetic resonance velocity mapping for the measurement of valvular
regurgitation. Eur Heart J 12:1786, 1991.
89. Nayak KS, Pauly JM, Kerr AB, et al: Real-time color flow MRI. Magn Reson Med 43:251-258, 2000. Medline
Similar articles
90. Kozerke S, Schwitter J, Pedersen EM, Boesiger P: Aortic and mitral regurgitation: quantification using moving slice
velocity mapping. J Magn Reson Imag 14:106-112, 2001.
91. Amanuma M, Mohiaddin RH, Hasegawa M, et al: Abdominal aorta: characterization of blood flow and measurement of its
regional distribution by cine magnetic resonance phase-shift velocity mapping. Eur J Radiol 2:559-564, 1992.
92. Li KCP, Whitney WS, McDonnell CH, et al: Chronic mesenteric ischemia: evaluation with phase-contrast cine MR imaging.
93. de Haan MW, Kouwenhoven M, Kessels AG, van Engelshoven JM: Renal artery blood flow: quantification with breath-hold
or respiratory triggered phase-contrast MR imaging. Eur Radiol 10:1133-1137, 2000. Medline Similar articles
94. Mohiaddin RH, Sampson C, Firmin DN, Longmore DB: Magnetic resonance morphological, chemical shift and flow
imaging in peripheral vascular disease. Eur J Vasc Surg 5:383-396, 1991. Medline Similar articles
95. Rees RSO, Firmin DN, Mohiaddin RH, et al: Application of flow measurements by magnetic resonance velocity mapping to
congenital heart disease. Am J Cardiol 64:953-956, 1989. Medline Similar articles
96. Martinez JE, Mohiaddin RH, Kilner PJ, et al: Obstruction in extracardiac ventriculopulmonary conduits: value of magnetic
resonance imaging with velocity mapping and Doppler echocardiography. J Am Coll Cardiol 20:338-344, 1992. Medline
Similar articles
97. Hirsch R, Kilner PJ, Connelly M, et al: Diagnosis in adolescents and adults with congenital heart disease. Prospective
assessment of the individual and combined roles of magnetic resonance imaging and transesophageal echocardiography.
98. Rebergen SA, Ottenkamp J, Doornbos J, et al: Postoperative pulmonary flow dynamic after Fontan surgery: assessment
with nuclear magnetic resonance velocity mapping. J Am Coll Cardiol 21:123-131, 1993. Medline Similar articles
99. Sampson C, Kilner PJ, Hirsch R, et al: Venoatrial pathways after the Mustard operation for transposition of the great
arteries: anatomic and functional MR imaging. Radiology 193:211-217, 1994. Medline Similar articles
100. Rebergen SA, Helbing WA, van der Wall EE, et al: MR velocity mapping of tricuspid flow in healthy children and in
patients who have undergone Mustard or Senning repair. Radiology 194:505-512, 1995. Medline Similar articles
101. Sieverding L, Jung WI, Klose U, Apitz J: Noninvasive blood flow measurement and quantification of shunt volume by cine
magnetic resonance in congenital heart disease. Preliminary results. Paediatr Radiol 22:48-54, 1992.
102. Mohiaddin RH, Underwood SR, Romeira L, et al: Comparison between cine magnetic resonance velocity mapping and
first pass radionuclide angiocardiography for quantitating intracardiac shunts. Am J Cardiol 75:529-532, 1995. Medline
Similar articles
103. Simpson IA, Chung KJ, Glass RF, et al: Cine magnetic resonance imaging for evaluation of anatomy and flow relations in
infants and children with coarctation of the aorta. Circulation 78:142-148, 1988. Medline Similar articles
104. Rees RSO, Somerville J, Ward C, et al: Coarctation of the aorta: MR imaging in late postoperative assessment.
105. Mohiaddin RH, Kilner PJ, Rees RSO, Longmore DB: Magnetic resonance volume flow and jet velocity mapping in aortic
coarctation. J Am Coll Cardiol 22:1515-1521, 1993. Medline Similar articles
106. Steffens JC, Bourne MW, Sakuma H, et al: Quantification of collateral blood flow in coarctation of the aorta by velocity
encoded cine magnetic resonance imaging. Circulation 90:937-943, 1994. Medline Similar articles
© 2010 Elsevier
6 of 6 20-04-2010 00:07
RINCIPLES AND PTIMIZATION OF ONTRAST

NHANCED HREE IMENSIONAL AGNETIC ESONANCE
NGIOGRAPHY
Jeffrey H. Maki
Michael V. Knopp
Honglei Zhang
Martin R. Prince
Advances in three-dimensional (3D) contrast-enhanced magnetic resonance angiography (CEMRA) are
making it increasingly the primary method of vascular imaging throughout the body. This chapter
describes the basic principles underlying 3D CEMRA and approaches to optimizing applications
throughout the body. Inherent advantages of 3D contrast MRA over "flow" based MRA techniques,
such as time of flight (TOF) and phase contrast (PC), are discussed. Techniques for improving spatial
and temporal resolution, reducing respiratory motion artifacts, decreasing contrast dosing, and
accurately timing the contrast bolus are also addressed.
© 2010 Elsevier
1 of 1 20-04-2010 00:08
CONTRAST-ENHANCED MAGNETIC RESONANCE ANGIOGRAPHY: THEORY

Three-dimensional CEMRA is similar to conventional contrast angiography or helical computed
tomography (CT). Instead of relying on blood motion to differentiate blood vessels from background
tissues, a contrast agent (e.g., gadolinium) is injected to shorten the T1 (spin lattice) relaxation time of
blood so that it is significantly less than the T1 of surrounding tissues. Blood can then be directly
imaged (using a T1-weighted sequence) irrespective of flow effects. This alleviates many of the
problems inherent to TOF and PC angiography. In particular, sensitivity to turbulence is dramatically
reduced and in-plane saturation effects are eliminated.1,2 The technique allows a small number of
slices to be oriented in the plane of the target vessels to quickly image an extensive region of vascular
anatomy (equal to the field of view). Imaging in the plane of the arteries also takes advantage of high,
in-plane resolution. Thus, 3D CEMRA is intrinsically fast, allowing high-quality breath-hold MR
angiograms.
© 2010 Elsevier
1 of 1 20-04-2010 00:09
GADOLINIUM CHELATES
page 741
page 742
At present, most contrast-enhanced MRA examinations are conducted using gadolinium-based MR

contrast agents. Gadolinium (Gd3+) is a paramagnetic metal ion that decreases both the spin-lattice
(T1) and spin-spin (T2) relaxation times.3 Because Gd3+ itself is biologically toxic, it is chelated with
ligands such as DTPA (gadopentetate dimeglumine ), HP-DO3A (gadoteridol ), gadoversetamide or
DTPA-BMA (gadodiamide ) to form low molecular weight contrast agents.4 These "extracellular"
agents rapidly redistribute from the intravascular compartment into the interstitial space in minutes.
Thus, 3D imaging of vascular structures must be performed rapidly.2,4 As compared with iodinated
contrast agents, gadolinium chelates have a very low rate of adverse events and no nephrotoxicity,
which is a significant advantage when evaluating patients with impaired renal function. 5-9
1,7
Each gadolinium chelate decreases the T1 of blood according to the equation:
where R1 is the field-dependent T1 relaxivity of the gadolinium chelate, [Gd] is the concentration of
gadolinium chelate, and 1200 (ms) is the T1 of blood without gadolinium (at 1.5 T). A similar equation
exists for T2 (which is always shorter than T1).
© 2010 Elsevier
1 of 1 20-04-2010 00:10
PULSE SEQUENCE
Dynamic CEMRA exploits the transient shortening in blood T1 following intravenous administration of a
contrast agent using a fast 3D spoiled gradient-echo sequence.1,10,11 We use the term "transient"
because this technique exploits the "first pass" of the contrast agent. Typically, 80% of gadolinium
12
chelate leaks into the intravascular space within 5 minutes. Repetition times (TR) for 3D T1-weighted
gradient-echo (SPGR, T1-FFE, FLASH) sequences used for CEMRA are generally less than 5 ms,
with echo times (TE) of 1-2 ms and total scan times in the 10 to 30 s range. With ultra-short TR and/or
sharing of peripheral k-space data, time-resolved CEMRA is possible at subsecond temporal
resolution. This type of 3D sequence is ideally suited to MRA, as it has high spatial resolution with thin
slices and intrinsically high signal-to-noise ratio (SNR). In addition, it is fast (can be performed in a
13-17
breath-hold) and can be reformatted into any desired plane. Also, because intravascular signal is
dependent on T1 relaxation rather than on inflow or phase accumulation, in-plane saturation and
turbulence-induced signal loss are not a problem.1
Add to lightbox
Figure 29-1 Relative signal intensity (SI) for different T1 values versus flip angle for a spoiled gradient
sequence. TR = 5 ms, TE <T2*.
The relative signal intensity (SI) for a 3D spoiled gradient-echo acquisition is given as: 18
where N(H), TR, TE, and α are the proton density, repetition time, echo time, and flip angle
respectively. Examining Equation 29-2 reveals that SI is maximized when TR is long or T1 is short,
TR/T1
since as (TR/T1) becomes large, e- approaches zero. Of these options, shortening T1 is the
primary method for increasing intravascular signal intensity, as lengthening TR increases scan time,
which is not desirable. The Ernst angle (αE) is defined as the flip angle maximizing the SI in Equation
29-2. It can be expressed as:
Figure 29-1 shows the relative signal intensity versus flip angle α for different T1 values (assuming a
TR of 5 ms with TE<T2*). Note the significant increase in signal intensity as T1 approaches zero.
© 2010 Elsevier
1 of 1 20-04-2010 00:10
CONTRAST DOSE AND T1

In order for blood to appear bright with respect to background tissues during the arterial phase of
dynamic 3D CEMRA, sufficient contrast material must be administered to transiently reduce the arterial
blood T1 to substantially less than that of the brightest background tissue, fat, which has a T1 of
approximately 270 ms. During the first-pass "arterial phase", blood T1 is more related to infusion rate
and contrast agent relaxivity than total contrast dose, as arterial phase imaging occurs well before
intravascular contrast equilibration. This dynamic approach has the advantage of nearly eliminating
steady-state background signal but at the same time (for nonblood pool agents) is essentially a
one-shot technique, meaning that once contrast is administered and dynamic imaging performed, the
process cannot be repeated without a second contrast bolus. If a second bolus is administered, one
must consider the expense and total patient contrast agent dose. In addition, residual soft-tissue
enhancement from the first contrast injection can obscure vascular detail (as the biological half-life of
gadolinium chelates is approximately 90 minutes).19 When using more than one injection, subtraction of
20
a precontrast mask may help reduce the effects of residual contrast for subsequent injections.
page 742
page 743
Add to lightbox
Figure 29-2 Blood T1 versus gadolinium infusion rate for changing cardiac output (CO), assuming no
recirculation.
Timing is extremely important in CEMRA. In order to appreciate timing issues, the relationship between
injection rate, blood T1, and resultant signal intensity must first be reviewed.
For times shorter than the recirculation time, contrast concentration in arterial blood during the bolus
phase can be approximated as1,2,21:
Although this simple equation ignores bolus dilution at leading and trailing edges, as well as extraction
of contrast in the lungs, a complex model originally developed for intravenous digital subtraction
angiography arrives at the same value for maximum contrast concentration. 22 Combining Equations
29-1, 29-2 and 29-4 and assuming no recirculation, blood T1 can be calculated for a given cardiac
output and injection rate. This is demonstrated in Figure 29-2 for cardiac outputs of 3, 5, and 7 L/min
using a standard gadolinium chelate preparation with a concentration of 0.5 mol/L.
Note from Figure 29-2 that for a "typical" cardiac output of 5 L/min, an injection rate of 1 mL/s yields a
blood T1 of 36 ms, while a 2 mL/s injection gives a blood T1 of 18 ms. These T1 values are much
shorter than the T1 of fat. Our own experience suggests that true T1 values are not quite this short;
typically they are nearer double this estimate due to dilution effects. Such a rapid injection rate,
however, need not necessarily be sustained over the entire acquisition time. For a typical 80 kg patient
1 of 2 20-04-2010 00:11
receiving a "single dose" (0.1 mmol/kg) of gadolinium, this equates to 16 mL of a 0.5 mol/L gadolinium
chelate preparation. Thus a 1 mL/s injection can be sustained for 16 s or a 2 mL/s injection sustained
for 8 s. Even though the acquisition time may be considerably greater than this, by properly timing the
bolus with respect to image acquisition, full advantage of the dynamic blood T1 shortening can be
obtained. This is due to the Fourier method by which MR data are acquired and because a contrast
bolus lengthens as it propagates from the venous to the arterial circulation.
© 2010 Elsevier
2 of 2 20-04-2010 00:11
FOURIER (k-SPACE) CONSIDERATIONS

MR imaging does not collect data pixel by pixel and it does not map spatial data linearly with respect
to time, as does conventional CT.23 With 3D MR imaging, the entire 3D "Fourier" or "k-space" data set
is collected before reconstructing individual slices. Because k-space maps spatial frequencies rather
than spatial data, k-space data do not directly correspond to image space. Instead, different regions
of k-space data determine different image features. For example, the center of k-space, or "low"
spatial frequencies, dominates image contrast whereas the periphery of k-space, or "high" spatial
frequencies, contributes more to fine details such as edges.24,25 Thus the state of the intravascular T1
in large vessels is captured at the moment corresponding to acquisition of central k-space.25
If central k-space is collected when arterial contrast concentration is peaking but has not yet reached
the venous structures (arterial phase), only the arteries will be bright. The fact that intravascular
contrast concentration may not be constant over the entire acquisition time may generate artifacts but
these have been shown to be relatively minor provided that the contrast is not rapidly changing over
the critical central portion of k-space and at least some "tail" of the contrast bolus is present during
collection of high spatial frequency data at the periphery of k-space.21,25 Thus with proper bolus timing,
gadolinium injection duration need not be as long as the entire acquisition time.
Most protocols use injection durations much shorter than the acquisition time, typically half the
26,27
acquisition time, with excellent results. For a given gadolinium dose, the injection strategy then
becomes a trade-off between a fast injection (shorter T1-more intravascular signal) and long injection
(more uniform T1-fewer artifacts). We have a general rule of thumb that advocates injecting the total
contrast dose over a duration of approximately 50% to 60% that of the acquisition for single station
examinations.
Another factor that must be considered is phase-encoding order. Traditionally, phase-encoding is

performed "sequentially" such that central k-space is acquired at the midpoint of the scan.
Alternatively, phase-encoding can be performed in a "centric" fashion such that central k-space is
acquired at the beginning of the scan.28,29 Although centric phase-encoding order has been shown to
be somewhat more prone to artifacts if the contrast concentration changes rapidly during acquisition of
central k-space, it greatly simplifies bolus timing (discussed below) and is less susceptible to artifact
from incomplete breath-holds.30,31 Centric encoding is generally most effective when performed as
29
elliptical centric, as described by Wilman and Riederer. Alternatively, a partial Fourier acquisition
beginning near the center of k-space may have the benefit of both centric and sequential ordering.
© 2010 Elsevier
1 of 1 20-04-2010 00:11
CONTRAST MATERIAL INJECTION RATE

Figure 29-2 shows that for injection rates greater than approximately 2 mL/s, there is diminishing return
with respect to T1 shortening. But based on Figure 29-1, even small decreases in T1 lead to
tremendous increases in signal intensity. Combining Equations 29-1 to 29-4, signal intensity (for TR=5,
TE =2) can be plotted versus injection rate for a fixed cardiac output (5 L/min) and flip angle (Ernst
angle) as shown in Figure 29-3. This is done both disregarding T2* effects and using the approximation
that T2* = 66% of T1.21 This graph demonstrates that signal intensity increases asymptotically as
injection rate increases, particularly when taking into account T2* effects, with negligible increases
32,33
seen beyond a rate of 4-5 mL/s.
Different authors have taken different approaches to optimizing the injection rate. Before the advent of
high-speed gradient systems which permitted total acquisition times of less than one minute (in the
realm of a breath-hold), imaging times were of the order of 3-5 minutes and injections were typically
performed so that a "double dose" (0.2 mmol/kg) was administered at a uniform rate over the entire
scan duration.11,34 For a 4-minute scan on a 70 kg patient, this amounted to an injection rate of just
over 0.1 mL/s. With the much shorter acquisition times currently possible, many investigators advocate
a "double dose" (0.2 mmol/kg or alternatively a set volume of 20 or 30 mL) of gadolinium chelate with
injection rates of 1-2 mL/s. Others use up to a "quadruple dose" (0.4 mmol/kg) and rates of 4-5 mL/s,
while a new trend now advocates much smaller doses.13,17,35 We personally feel that using a single 20
mL vial of gadolinium chelate or possibly 30 mL with an injection rate of 1.5-2.5 mL/s is generally a
good compromise between maximizing arterial signal and minimizing artifacts for breath-hold scans. 36
Multistation exams, large aneurysms or dissections, and pediatric patients with high cardiac output,
cases where venous evaluation is important are different and require higher doses, which will be
discussed shortly.
Add to lightbox
Figure 29-3 Relative signal intensity (SI) versus gadolinium injection rate for a spoiled gradient-echo
sequence. Flip angle = Ernst angle, cardiac output = 5 L/min.
© 2010 Elsevier
1 of 1 20-04-2010 00:12
BOLUS TIMING CONSIDERATIONS

As discussed above, the timing of the gadolinium bolus relative to acquisition of central k-space is
extremely important. The contrast travel time, defined as the time required for contrast to travel from
the injection site to the vascular territory of interest, is highly variable, as is the degree to which the
bolus "broadens" as it transits the heart and pulmonary vasculature. A mean transit time to the
37
abdominal aorta of 23 ± 5 s, with a range of 13-37 s, has been reported. An example of
time-resolved abdominal aortic SI (which is proportional to Gd) after IV injection of 9.5 mL Gd chelate
at 2 mL/s is shown in Figure 29-4.
In Figure 29-4, note first that arterial signal peaks relatively sharply. (If the bolus were longer, i.e.,
larger total dose, the arterial peak would be lengthened.) This observation has important implications.
Because intravascular signal intensity is determined by the gadolinium concentration at the time the
25
center of k-space is collected, whether central k-space is acquired at time t1, t2 or t3 (see Figure
29-4) has a dramatic influence over the resulting image. Perfect timing (t 2) yields maximum arterial
signal with minimal venous signal, as shown in Figure 29-5A. If, however, central k-space data is
acquired too early (t1-while arterial Gd is still rapidly increasing), "ringing" or "banding" artifacts of
variable severity can be generated, as subtly demonstrated in the iliac arteries in Figure 29-5B.25
These artifacts are also referred to as "ringing" or "leading edge" artifacts since they result from
imaging during the leading edge of bolus arrival. Acquiring central k-space data too late (t3) leads to
submaximal arterial signal intensity and venous enhancement. These two factors can degrade
evaluation of arterial structures. A mild example of such delayed timing is shown in Figure 29-5C.
Hence if one knows or can predict or measure the time course of intravascular Gd, the placement of
central k-space can be tailored to achieve the desired image characteristics, e.g., maximum arterial
signal, maximum portal venous signal, etc.
Add to lightbox
page 744
page 745
Figure 29-4 Relative signal intensity (SI) in the aorta versus time following a 9.5 mL injection of
gadolinium chelate at a rate of 2 mL/s.
1 of 3 20-04-2010 00:13
Add to lightbox
Add to lightbox
2 of 3 20-04-2010 00:13
Add to lightbox
Figure 29-5 MIP images from different MRA studies demonstrating (A) good arterial phase timing in
an abdominal aortic study, (B) late flow through an abdominal aorta with centric ordering of k-space
resulting in leading edge artifact (ringing) and decreased SNR in the arteries, and (C) a carotid MRA
with slightly delayed triggering leading to mild-moderate enhancement of the right jugular vein.
© 2010 Elsevier
3 of 3 20-04-2010 00:13
"BEST GUESS" TECHNIQUE

Several solutions to proper bolus timing have been offered. Perhaps the simplest is to just make an
educated "best guess." This involves estimating the contrast travel time from the site of injection to the
vascular structure of interest. While this time can be highly variable, an experienced MR angiographer
can reliably achieve good results by taking factors such as injection site, age, cardiac output, and
vascular anatomy into consideration. Sequential phase-encoding is preferred when using "best guess",
as this encoding strategy is more tolerant of timing errors.38 Using this technique, timing can be
17
calculated as follows:
where imaging delay is the delay between initiating the bolus injection and starting the scan. Equation
29-5 times the injection such that the midpoint of the bolus arrives at the desired vascular territory at
the midpoint (center of k-space) of the scan. As an example, for a scan time of 30 s with an estimated
contrast travel time of 20 s and a 20 mL gadolinium bolus at 2 mL/s (10 s injection), the imaging delay
would be 20 + 10/2 - 30/2 = 10 s (i.e., start the acquisition 10 s after initiating contrast injection). This
timing is designed so that the midpoint of the bolus arrives as the center of k-space is collected,
meaning that depending on injection duration, there may already be venous enhancement by the time
central k-space is collected.
Another timing strategy is:
where rise time is the expected time from arterial arrival to arterial peak, typically 3-5 s. This timing
formula places the center of k-space at the approximate peak of contrast concentration.
The most difficult aspect of this "best guess" technique is estimating the contrast travel time. If timing
is not perfect, it is better to image slightly late (increased venous phase) than slightly early
(ringing/banding artifact and decreased signal; see Fig. 29-5B). With this in mind, as a general
guideline for travel time from an antecubital vein to the abdominal aorta, we start with approximately 15
s for a young healthy patient. A young hypertensive or athletic patient would be a few seconds less. A
healthy elderly patient (over 70 years old) would be approximately 20-25 s. Patients with cardiac
disease or an aortic aneurysm would be in the 25-35 s range. Severe cardiac failure in conjunction with
aortic pathology might be up to 40-50 s. We add 3-4 seconds if the IV is in the hand. When imaging
the pulmonary arteries, the contrast travel time is less, 5-10 s. In babies and especially neonates,
contrast travel times are extremely fast, just a few seconds to the pulmonary artery and about 5 s to
the aorta.
© 2010 Elsevier
1 of 1 20-04-2010 00:14
TEST BOLUS TECHNIQUE

A better and only slightly more time-consuming way to estimate contrast travel time is through use of a
test bolus.37,39,40 With this technique, 1-2 mL of gadolinium (followed by 10-15 mL saline flush) is
injected at the same rate as planned for the actual injection. Multiple single-slice fast gradient-echo
images in the vascular region of interest are acquired as rapidly as possible (at least every 1-2 s) for
approximately 1 minute. In order to minimize time of flight effects, the 2D test bolus image should
either be oriented in the plane of the imaged vessel (i.e., sagittal or coronal for the aorta) or
alternatively, it can be relatively thick (greater than 1 cm) with a superior saturation band or a blood-
nulling inversion pre-pulse. The time of peak arterial enhancement (contrast travel time) is then
determined visually or using region of interest (ROI) analysis. Acquisition timing is determined from
Equation 29-6, bearing in mind that a longer bolus will take longer to peak than a test bolus, and
therefore a 3-4 s rise time should be included. Alternatively, if elliptical centric encoding is desired
31
(beneficial if the patient cannot breath-hold well ), the following equation should be used, again
anticipating a rise time in the 3-4 s range:
37
Using the test bolus technique for abdominal aortic imaging, Earls et al demonstrated a substantial
aortic SNR increase in 15 patients following a 1 mL bolus of gadopentetate dimeglumine .37 By
measuring contrast travel time (24.0 ± 12.6 s), they were 100% successful in obtaining a "pure"
arterial phase study (as compared to 80% success with "best guess" timing). Perfect timing with a test
bolus significantly increased aortic SNR (29.8 versus 20.5). Recently many centers have replaced the
test bolus with a time-related MRA scan with injecting a slightly larger dose of 5-10 mL Gd.
page 746
page 747
There are three main drawbacks to the test bolus technique. First, setting up, performing, and
analyzing the test bolus lengthens the overall examination time. In experienced hands, the average time
37
penalty was reported as less than 5 minutes. Second, the test bolus rapidly redistributes into the
interstitial space, thereby increasing background signal. For such small test doses, however, these
effects are negligible. Both of those disadvantages can be entirely eliminated by using ultrasound to
detect a test bolus of ultrasound contrast agent. One final issue is that there is no absolute guarantee
that the imaging bolus will behave identical to the test bolus. This is because of moment-to-moment
patient variables such as venous return and cardiac output, as well as the different total volume of
injection. It is especially important to be aware that a test bolus with the arm at the patient's side may
not be accurate if the arms are above the head, stretching and pinching the subclavian veins, for the
actual scan.
© 2010 Elsevier
1 of 1 20-04-2010 00:15
MAGNETIC RESONANCE FLUOROSCOPY

A third method of contrast bolus timing uses extremely rapid MR fluoroscopic imaging. 24,30,41,42 With
this technique, 2D gradient refocused images are rapidly (<1 s/image) obtained through the vascular
structure of interest, ideally using complex subtraction to improve contrast-to-noise ratio and decrease
artifacts. Images are generated in near realtime and updated at greater than one per second. The
operator watches the contrast bolus arrive and then switches over to elliptical centric 3D MRA when
the desired enhancement is detected.
This technique allows real-time, operator-dependent decision-making. This may be particularly

advantageous in cases with unusual or asymmetric flow patterns such as unilateral stenoses and
slow-filling aneurysms. As an example, an abdominal aortic aneurysm can be monitored such that
triggering is delayed to allow for complete enhancement of the iliac vessels, which may fill slowly (see
Fig. 29-5B). The ability to assess the vasculature "on the fly" under these circumstances is a great
asset, as are the time and contrast savings associated with avoiding a test injection.
A recent refinement is to electronically shift the MR fluoroscopic monitoring to a proximal region of

vascular anatomy to get more advance warning of contrast arrival.43 Using carotid CEMRA as an
example, MR fluoroscopic monitoring in the chest shows contrast arriving in the subclavian vein, then
right heart, pulmonary arteries, left heart, arch, and finally proximal great vessels. Tracking the bolus
over such a long period makes precise triggering easier. If the flow is very slow, the operator can
compensate to allow for greater target vessel enhancement before triggering. This same work also
demonstrates that fluoroscopic triggering is improved by recessing the absolute center of k-space a
few seconds from the beginning of the 3D acquisition, thus avoiding the previously described "leading
edge" artifact from early triggering. For systems that do not have recessed elliptical centric MRA it
may be preferable to trigger early and order k-space sequentially.
© 2010 Elsevier
1 of 1 20-04-2010 00:15
TEMPORALLY RESOLVED 3D CONTRAST MAGNETIC RESONANCE

ANGIOGRAPHY
Because proper bolus timing is so difficult and a timing mistake can ruin the study, some authors
advocate "temporally" resolved techniques whereby multiple 3D data sets are acquired (or
synthesized) extremely rapidly (typically 1-10 s per acquisition).44-48 With such techniques, bolus timing
is no longer a factor, as multiple vascular phases are obtained without any predetermined timing (i.e.,
inject and begin scanning simultaneously). The operator simply selects the desired image set: pure
arterial phase, maximum venous, etc. This is particularly useful in the carotid and pulmonary arteries,
where the venous phase is extremely rapid, and in the calf, where variable rate filling may occur due to
stenoses, occlusions or rapid AV shunting because of soft-tissue ulcers, cellulitis, and other
44,48,49
inflammatory disease.
The most straightforward way to accelerate acquisition time is simply by scanning faster. This can be
done using some combination of limiting the imaging volume, decreasing the resolution (e.g., fewer,
thicker slices and fewer phase-encoding steps), decreasing TR or using a parallel imaging technique
such as SENSE.50-53 Recent developments in gradient systems allow TRs of <2 ms on many systems,
a threefold improvement compared to just a couple of years ago. In the pulmonary arteries, Goyen et
al achieved 4 s temporal resolution using partial k-space filling (TR = 1.6 ms), although slices were
52
relatively thick at 5.5 mm reconstructed to 2.75 mm.
In another approach, taken by Finn et al, the aorta is imaged in two alternating planes (oblique sagittal
and coronal, TR = 1.6 ms) at approximately 1 s resolution per plane.51 This is in part possible due to a
thick slab acquisition (~20 mm interpolated to 10 mm), meaning the data are quasi-projectional. While
off-plane reformats in this case are poor, in-plane MIPs (two planes) are of high resolution. Finn also
used a 3D radial sampling scheme to further increase resolution.54 Once high-resolution temporal data
sets are obtained, even if the temporal resolution is insufficient for isolating the desired vascular phase,
post-processing techniques such as correlation analysis can be used to synthesize images of pure
pulmonary arterial versus pure pulmonary venous phase.55
page 747
page 748
Simultaneous high spatial resolution and high temporal resolution over a large field of view can be
obtained using 3D TRICKS (time-resolved imaging of contrast kinetics) technique and its
47,49,56,57
variants. In the most basic implementation of this technique, k-space is divided into multiple
blocks. Central k-space (which contributes most to overall image contrast) is repeatedly collected
every 2-8 seconds in an alternating fashion with other blocks of more peripheral k-space. Images are
then "synthesized" at high temporal resolution by piecing together each unique block of central k-space
with a linear interpolation of the remaining k-space blocks acquired in closest temporal proximity. This
technique allows greater temporal and spatial resolution/volume coverage than with a streamlined
conventional sequence. It also lends itself to "video" format, where passage of contrast material can
be directly viewed. The 3D TRICKS-type sequences are now finding their way into commercial
packages. Their biggest drawback, other than the tremendously large number of images generated, is
the large amount of processing time required to perform the multiple reconstructions, although this is
rapidly improving. Also, k-space discontinuities, in conjunction with varying intravascular gadolinium
concentrations, can potentially lead to artifacts. However, these have been shown to be small provided
the gadolinium bolus is not too compact.
A recent refinement in this strategy is to traverse k-space using radial or spiral projections, which allow
for sliding window reconstructions at very high temporal and spatial resolutions.58 One variation, known
59
as vastly undersampled isotropic projection reconstruction (VIPR), is particularly promising.
© 2010 Elsevier
1 of 1 20-04-2010 00:16
ARTIFACTS
k-Space Related
Maximizing arterial enhancement requires collecting the center of k-space when the arterial contrast
concentration is at a maximum. Acquiring central k-space data after arterial contrast has peaked leads
to suboptimal arterial signal and increased venous signal (see Figs. 29-4 and 29-5C). Acquiring central
k-space during the rising leading edge of the bolus causes "ringing" type leading edge artifacts (see
Fig. 29-5B). Even with optimal timing, similar artifacts can be generated with the use of very compact
boluses such that contrast falls rapidly while still acquiring relatively central k-space. If the bolus is too
short, such that the contrast concentration tapers too rapidly toward the end of the acquisition (high
25
k-space frequencies), blurring of the edges results. These artifacts are not as severe if k-space is
mapped in a sequential as opposed to a centric manner.25
Thus, when using "best guess" timing, sequential phase-encoding is more forgiving. Also, if using this
technique, we suggest slightly overestimating the contrast travel time so that the (inevitable) timing
errors are more likely to cause increased venous enhancement than "ringing" artifacts. When using the
test bolus technique such that contrast travel time is known, either elliptical centric or sequential phase-
encoding can be used. The advantages of elliptical centric are twofold: less degradation with an
incomplete breath-hold (see under Respiratory heading, below) and simplified operator logistics with
respect to timing (no need to count backwards to the middle of the scan to figure out when to inject.
Just inject, wait the proper delay, initiate breath-hold, and start imaging). For automated bolus
detection, elliptical centric phase-encoding is typically used. Because contrast rise times are not
instantaneous, the delay from "triggering" to initiation of data acquisition becomes crucial in order to
avoid artifacts. Examining Figure 29-4, if the 3D MRA triggers at time t1, the appropriate delay is
(t2-t1). This delay, of course, varies by patient and vascular territory. A delay of 5-6 seconds typically
works well for abdominal aortic imaging. In patients expected to have particularly slow flow (congestive
heart failure, aneurysm), this time should be lengthened to 8-10 s. In the carotid arteries, however,
which have a more rapid upslope and a very short arteriovenous window, shorter trigger delays of 2-3
41,60
s are more appropriate in order to avoid jugular venous enhancement. Recessing the absolute
center of k-space for a few seconds at the beginning of the scan improves consistency by providing
high arterial-phase contrast even if the trigger is premature.
Respiratory
Respiratory motion causes ghosting, blurring, and signal loss. 61,62 In 3D imaging, blurring occurs in the
direction of the motion, whereas ghosting is most pronounced in the slow phase-encoding direction.
For elliptical centric phase-encoding order, the ghosting is spread out in both phase-encoding
63
directions. Before the advent of fast imaging systems, 3D contrast MRA acquisition times were of
the order of 3 minutes, making breath-holding an impossibility. Thus early contrast MRA images
suffered from significant image degradation, particularly when evaluating small vessels such as renal
arteries.1 Figure 29-6 illustrates the difference between good and poor-quality breath-holding.
1 of 3 20-04-2010 00:16
Add to lightbox
Add to lightbox
page 748
page 749
Figure 29-6 Coronal subvolume MIPs from renal MRA studies demonstrating (A) good breath-holding
and (B) poor breath-holding.
There is no disagreement among researchers regarding the beneficial effects of breath-holding on

abdominal and thoracic 3D contrast MRA.13,14,17,56,64 Holland et al evaluated the same six patients
using both breath-hold and nonbreath-hold renal 3D contrast MRA.64 Non-breath-hold studies failed to
identify three of four peripheral renal artery abnormalities, missed one of two accessory renal arteries,
and were unable to visualize any of the renal artery bifurcations at the hilum. Breath-hold 3D contrast
MRA identified all abnormalities, all accessory renal arteries, and renal artery bifurcations at the hilum
in all patients. In the carotid arteries, breath-holding does not appear essential, although it does
improve visualization of the arch vessels.65 When imaging the aortic arch and carotid arteries, it is
important to anticipate that patients may arch their back and neck during breath-holding in inspiration
and this can cause carotid arteries to move anteriorly, out of the imaging volume.
2 of 3 20-04-2010 00:16
Maki et al analyzed SNR loss and blur associated with incomplete breath-holds. 31 Their work
demonstrated that breath-holding during the acquisition of the central 50% of k-space was the most
important for minimizing blur. This means that for centric phase-encoding, the majority of the benefit
from breath-holding can be obtained even if the patient begins breathing at the midpoint of the scan.
Data from Vasbinder et al shows that the more fully "centric" the phase-encoding order is, the less the
artifact from partial breath-holds, and thus the preference for an elliptical versus a linear centric phase-
encoding order.63 These data suggest that for centric phase-encoding orders, an optimistic
assessment of a patient's breath-hold ability is in order, as there is little to lose if they "don't quite
make it." Recent data, however, suggested that a large number of patients have diaphragmatic motion
63
even during breath-holding. This causes blurring and ringing artifacts within the visceral arteries,
obscuring small vessels and subtle abnormalities (such as occur with fibromuscular dysplasia). This
argues for the shortening of abdominal MRA acquisition times to well below potential breath-hold
times, making rapid acquisition techniques such as SENSE and the ultra-short TR discussed above
even more attractive. In anatomic regions that can remain truly motionless, such as the calves and
feet, longer acquisitions work well.26
We have found that patients are better able to hold their breath if the procedure and expectations are
described well in advance of the actual scan. The vast majority of ambulatory patients can hold their
breath for 30 s and many have endurance well beyond this. A test breath-hold before the scan or
watching a patient's respiratory pattern often gives you a clue to their breath-hold potential. In our
experience, patients that take slow deep breaths with relatively long expiratory pauses have no
problem completing a 20 to 30 second breath-hold twice in a row for arterial and venous enhancement
phase. Patients who breathe rapidly (more than 25 breaths per minute) often have trouble holding the
breath. Sick or postoperative inpatients often cannot or will not hold their breath at all. In these
patients, either high temporal resolution scanning or longer scans (approximately 1 minute) for
averaging over many respiratory cycles produces better quality MRA.
We recommend the following breathing strategy. For "best guess" or test bolus timing techniques, ask
the patient to pre-breathe for three or four cycles ("deep breath in …and breathe out …") timed such
that the last inspiration is held ("deep breath in and hold …") as the scan starts. Some authors
advocate the use of supplemental oxygen during this pre-breathing.17,63 For automatic or fluoroscopic
techniques, ask the patient to pre-breathe for three or four cycles, then have them relax, start the
fluoroscopic sequence, and wait for bolus arrival. Once the MRA sequence is triggered, use some
prearranged signal such as squeezing the arm in addition to a loud verbal "take in a deep breath and
hold it" to initiate the breath-hold. This needs to occur during the delay time, such that the patient is
motionless once actual data collection starts and the critical center of k-space is collected. Recessing
the absolute center of k-space in a few seconds from the beginning of centric acquisitions helps to
eliminate motion artifact when breath-holding does not start exactly on schedule.
© 2010 Elsevier
3 of 3 20-04-2010 00:16
DETERMINING IMAGING TIME

page 749
page 750
The duration of a 3D contrast MRA acquisition (ts) depends on several variables and can be
approximated as:
where TR is the repetition time, YRES and ZRES are the number of phase-encoding steps in y and z,
fractional y FOV is the asymmetric y field of view fraction, and the SF is the speed-up factor due to
parallel imaging techniques such as SENSE66 or fractional excitations. TR is itself a function of echo
time (TE), echo type (i.e., full or fractional), and bandwidth. Spatial resolution is determined by voxel
size, which is determined by:
where x, y, and z FOV refer to the x, y, and z fields of view, YRES and ZRES are the number of y and
z phase-encodings collected, and XRES is the number of data points digitized during readout. Using
fractional y field of view decreases the number of y phase-encodings, shortening scan time while
maintaining resolution but collecting a smaller (or "rectangular") FOV in the y direction. This is
particularly useful for large FOV studies where structures at the periphery of the y FOV are not
important and wraparound (aliasing) can be tolerated. Parallel imaging techniques, such as SENSE,
require the use of multiple (phased array) coils. These techniques collect fewer phase-encoding steps
than are prescribed (phase and/or slice direction), effectively reducing the FOV(s) with resultant
aliasing.50,66,67 This can decrease scan time by up to a factor of the number of coils in the phased
array, depending on the geometry. SENSE has been most useful for cardiac imaging and has great
potential with MRA as well.26,68 Using sensitivity data for each coil from a "reference" scan, a
mathematical algorithm can "unfold" the aliased data set, restoring the original prescribed FOV with
only a small geometry-related loss in SNR when coil designs are optimized. When using SENSE, the
prescribed FOV must be large enough to avoid aliasing for the unfolding algorithm to work.
Imaging parameters must be juggled to achieve the required volume coverage, spatial resolution, and
SNR while keeping the scan time ts within the range of the patient's anticipated breath-hold duration tb.
There are multiple trade-offs. Increasing resolution (YRES or ZRES with same overall slab thickness),
increasing volume coverage (ZRES with same slice thickness) or increasing SNR (decreasing
bandwidth or changing from fractional to full echo) all increase scan time (ts). Decreasing the
rectangular FOV or using a parallel imaging technique will decrease ts. For a fixed bandwidth, SNR
scales as the voxel size times the square root of scan time ts.18,21 One interesting property unique to
contrast MRA, however, is that, all other things being equal (e.g., not changing bandwidth), if the
contrast injection rate is increased proportionally with the decrease in scan duration (i.e., inject same
total contrast dose faster), the SNR is unchanged and CNR increases. Maki et al describe a
mathematical analysis specific to breath-hold Gd MRA to optimize SNR by considering these
variables.21
A useful technique for increasing reconstructed resolution with no associated time penalty is to zero fill
k-space in multiple directions.69 Most MR scanners have an option to zero fill both frequency and
phase to 256 or 512. In addition, a twofold zero fill in slice direction can be specified, yielding twice the
number of slices. While the slice thickness is unchanged, the slices now overlap by 50% or one-half a
slice thickness. The net result is that zero filling yields smoother reformations by reducing partial
volume effects.
The art, then, of 3D breath-hold contrast MRA boils down to a careful choice of the imaging
parameters in Equations 29-8 and 29-9. It is often an iterative process. As a general rule, for non-time-
resolved abdominal MRA, we start with a frequency resolution (XRES) of approximately 400, with zero
filling to 512 × 512 in frequency and phase and twofold in the phase direction. Then we consider the
maximum estimated breath-hold duration (tb) and desired slice thickness and increase ZRES to cover
1 of 2 20-04-2010 00:17
the required volume. If an asymmetric FOV can be used, we do so. Alternatively, we use SENSE with
a SF of 2-326,66 and then increase YRES until tb is reached. Keep in mind the voxel dimension in z
versus that in y. They should be roughly similar to achieve isotropic resolution so that multiplanar
reformats (MPRs) maintain resolution in all planes.
If ts cannot be made less than tb after these manipulations, then sacrifices must be made. Partial
Fourier techniques collect only a fraction of y phase encodes at the cost of decreased SNR.
Alternatively, bandwidth can be increased, again at the cost of decreased SNR. Slice thickness can be
increased at the cost of decreased spatial resolution. Higher SENSE factors (>2.5-3.0) can be
employed, although this also decreases SNR. Finally, incomplete breath-holding can be anticipated at
the cost of increased ghosting and blurring.
In cases where breath-holding is not as important, such as in the extremities or carotids, there is more
flexibility in choosing imaging parameters. In these circumstances, the limiting factor becomes contrast
bolus duration. A "compact" bolus (<50% of imaging duration) can cause artifacts such as ringing and
blurring. So, assuming a double dose (0.2 mmol/kg) of a gadolinium chelate (0.5 mol/L) for a 80 kg
patient administered at 2 mL/s, the injection duration is only 16 s. Assuming the bolus will broaden by
several seconds before reaching the target vessels, this is still less than approximately 20-22 s. To
avoid artifacts, this needs to be a substantial fraction of the scan time t s, and scan times greater than
approximately 60 s may be of little additional benefit. Nonetheless, this increase in imaging time allows
for increased SNR and spatial resolution. For carotids or peripheral arteries, the rapidity of venous
return must also be considered. This can impose limits on scan duration. For example, in instances
where increasing scan time slows down how fast the center of k-space is traversed (e.g., increased
TR or using multiple averages), central k-space will be closer to venous signal arrival. This can lead to
greater venous signal and "leading edge" venous ringing artifact. With elliptical centric sequences,
however, carotid scan times of up to 45 s and lower station peripheral scan times of well over 1 minute
have been shown to work extremely well, due to recirculation of contrast in the latter part of the
scan.26,41 Longer scan time for extremity MRA are facilitated by application of subsystolic compression
to 50-60 mm Hg which prevents early venous enhancement.
© 2010 Elsevier
2 of 2 20-04-2010 00:17
PATIENT PREPARATION
The more relaxed and informed a patient is, the more easily they can remain still and perform a long
breath-hold. Reassurance and a brief description of the scan can help. In patients who are particularly
anxious, premedication with sedatives such as diazepam (5-10 mg PO) may be useful. This helps the
patient to relax and lie still and also decreases cardiac output. 1 This latter effect enhances image
quality because arterial phase Gd increases with decreasing cardiac output (see Equation 29-4 and
Fig. 29-2).
Since most 3D contrast MRA is performed in the coronal plane, arm position is important to allow for a
smaller FOV without aliasing, particularly when using SENSE. For carotid, large FOV aortic or
peripheral studies, the arms can remain by the patient's side, as aliasing is not a factor. For
pulmonary, small FOV aortic, renal, and mesenteric studies, the arms must either be elevated out of
the imaging plane using cushions, or extended over the head.17 Both of these positions in which the
arms are elevated have the added benefit of gravity-aided venous return from the IV injection site. This
ensures rapid delivery of the entire Gd dose to the central circulation as long as veins are not pinched.
page 750
page 751
We recommend placing the IV before the patient goes into the magnet. This alleviates anxiety and
possible shifts in patient position that may occur if it is placed while in the magnet. If the MRA is to be
performed with the arms by the patient's side, placing the IV in the antecubital vein is adequate. If the
arms will be over the head, it is preferable to place the IV below the antecubital fossa so it will not kink
and obstruct should the patient's elbow bend. A 22 gauge catheter is usually adequate, although for
smaller diameter catheters, warming the contrast to body temperature may decrease the viscosity
enough to allow for adequate injection rates. Use of a standardized IV tubing system is helpful to avoid
uncertainty about priming volume and to simplify flushing.
© 2010 Elsevier
1 of 1 20-04-2010 00:17
CONTRAST INJECTION
Total gadolinium dose and rate of delivery is a trade-off between maximizing intravascular signal
(higher injection rate) and minimizing artifacts (longer injection duration). In the ideal situation, a large
dose at a high rate would be optimal. This, however, must be weighed against safety, practicality, and
cost. For breath-hold contrast MRA with accurate bolus timing (test bolus, Smartprep or fluoroscopy),
20-30 mL (0.1-0.2 mmol/kg) administered at a rate of approximately 2 mL/s represents a good
balance between these factors, depending on total scan time. For "best guess" type techniques or
portal venous imaging, we prefer to use 40 mL at the same rate to create a longer bolus that is more
forgiving of timing errors. These recommendations can be modified. For example, in a particularly small
patient where the volume of a single dose is small, artifact from a compact bolus may be generated.
To remedy this, either the injection rate can be decreased (thereby lengthening the bolus at the
expense of intravascular signal) or the dose can be increased (at the expense of increased cost). If
cost is an overwhelming concern, the dose and injection rate can be decreased at the expense of
intravascular signal or the dose alone can be decreased at the expense of compact bolus artifact.
In experienced hands, we find that manual or hand injection works quite well. Since many smaller MR
centers do not have a power injector, this is often the only choice. Hand injection is also preferred for
pediatric patients (especially babies), when injecting central lines or when there is a tenuous IV line.
Manual injections are also simpler in some ways, as there is less "plumbing" and less opportunity for
mechanical difficulties. That said, in spite of the added complexity, a power injector offers a more
constant and standardized bolus delivery. Either way, the contrast injection must be immediately
followed by adequate saline flush to complete delivery of the bolus and help flush the arm veins.
17,37
Suggested flush volumes range from 15 to 50 mL, with most authors using 20 mL. We suggest 25
mL of saline flush at the same rate as the gadolinium bolus. For hand injections, a standardized tubing
set that allows rapid switch-over between gadolinium injection and saline flush eliminates many
potential injection pitfalls (SmartSet, TopSpins, Ann Arbor, MI).
© 2010 Elsevier
1 of 1 20-04-2010 00:18
IMAGING DIFFERENT VASCULAR PHASES

With most 3D contrast MRA studies, extensive efforts are made to optimally image the arterial phase.
Once the arterial phase data set is collected, the sequence can (and should) be repeated to obtain
venous and equilibrium phases. Later phases are useful in detecting and evaluating dissections, portal
venous or other venous structures, parenchymal enhancement patterns, tumors, and perhaps even
renal glomerular function.70,71 Temporal resolution is determined by a combination of scan time and the
time required for the patient to prepare for another breath-hold. Aside from techniques like 3D
47
TRICKS, the faster the scan, the better the temporal resolution. In cases where high temporal
resolution is desired (e.g., carotids or in cases of AV shunts), fast, multiphase techniques can be used
in a single breath-hold, although SNR and spatial resolution decrease. 48 In more "conventional"
breath-hold 3D contrast MRA, with a scan time of approximately 20-30 s, we advocate giving the
patient approximately 5-10 s to "catch their breath" before scanning again. In this manner, the second
"arteriovenous" phase occurs approximately 35 s after the peak arterial phase. This is usually well
timed for the portal venous phase and often adequate for evaluating venous structures and
parenchymal phases in the visceral organs. Sometimes a third phase may be desired, usually to
evaluate slow venous return or excretion of Gd into the renal collecting system.
Because these "later" phase examinations occur more in the equilibrium phase of gadolinium
distribution, signal is reduced as compared to arterial phase (see Fig. 29-5C). In order to maximize
signal in these later phases, the flip angle should, if possible, be reduced (see Fig. 29-1). Whereas the
optimum flip angle for the arterial phase may be 30-40° at a TR of 4-6 ms, the optimal flip angle
decreases to 15-25° in later phases. To image highly concentrated Gd excreted into the collecting
system and ureter, use a high flip angle (45-60°) with the widest bandwidth and shortest possible echo
time.
© 2010 Elsevier
1 of 1 20-04-2010 00:18
POST-PROCESSING AND DISPLAY

Three-dimensional Gd MRA produces a contiguous volume of image data. This data set is best viewed
interactively using a computer workstation allowing for thin multiplanar reformatting (MPR). 36,72 In this
manner, thin (typically 1.0-2.0 mm) slices can be viewed in axial, sagittal, coronal or oblique planes.
This eliminates overlapping structures and unfolds tortuous vessels. Because these reformatted slices
remain thin and are not projections (see below), they not only provide the best achievable contrast but
36,72,73
also minimize the chance of diagnostic error (see Fig. 29-7).
page 751
page 752
Add to lightbox
Add to lightbox
Figure 29-7 Demonstration of a potential MIP artifact. Subvolume axial MIP through the right renal
artery (A) looks normal. Careful review of the source coronal images (B) demonstrates how the MIP
volume (white lines) incorporates part of the ectatic aorta (black arrow), masking the high-grade right
renal artery stenosis.
Because thin reformations show only short vascular segments, it is advantageous to utilize the
maximum intensity projection (MIP) post-processing technique for combining information from multiple
slices to see longer segments of vessels. 74,75 Using this algorithm, the user specifies the subvolume
thickness and obliquity. The algorithm then generates rays perpendicular to the subvolume and uses
the maximum value of any voxel encountered along that ray for the corresponding pixel intensity in the
1 of 2 20-04-2010 00:19
output image. This technique is fast and works extremely well with Gd MRA, since vascular signal is
much greater than background signal. It provides projection images that are similar to conventional
angiograms. The MIP algorithm is demonstrated in Figure 29-8, which shows a full-thickness coronal
MIP of the thoracic aorta (Fig. 29-8A), a thin slice axial reformation (MPR) through the aorta at the
level of the pulmonary arteries (Fig. 29-8B), and an oblique subvolume MIP taken from the axial
reformat demonstrating the origins of the great vessels and a large Stanford type B aortic dissection
(Fig. 29-8C). This technique is helpful for displaying complex vascular anatomy, particularly when
vessels are not oriented along a single plane.
Despite its usefulness, the MIP algorithm is subject to artifacts. Perhaps the most common artifact
arises when stationary tissue has greater signal intensity than the vascular structures of interest, as
can occur in the presence of other vessel segments, fat, hemorrhage, metallic susceptibility artifacts or
motion artifacts. This in turn leads to the mapping of nonvascular signal into the projection image and
causes a discontinuity in vessel signal, potentially mimicking a stenosis or occlusion (see Figs. 29-7A
and B).72 This type of artifact is best overcome by minimizing the thickness of the MIP subvolume,
thereby excluding as much extraneous data as possible. Other artifacts inherent to the MIP technique
have been described.76 These mainly consist of underestimating vessel lumen and are more of a
problem with TOF or phase contrast techniques. Zebra stripe artifact from slices that are too thick can
be reduced using zerofilling in the slice direction. Because of these potential pitfalls, most authors
agree that the MIP images should be used as a roadmap, utilizing the source images for definitive
diagnosis.36,72,76 New work suggests that volume rendering may be more accurate than the MIP and
similar to using the MPR, but should still not replace careful evaluation of the source images and
MPRs.77
Subtraction techniques are also useful in image evaluation, particularly in vascular regions not
subjected to significant respiratory motion. These areas include the extremities, pelvic, and carotid
arteries. A "digital subtraction" MRA can most easily be accomplished by subtracting pre- from
post-contrast magnitude (reconstructed) images. Provided the patient maintains the same position on
both studies, subtraction will eliminate background signal and improve vessel conspicuity. This
20,78
technique has been used successfully in the extremities. An improvement on this technique,
particularly for thicker slice or 2D exams, involves complex subtraction of the pre- and post-contrast
"raw" (k-space) data sets. This overcomes phase differences in voxels that contain both stationary and
moving spins (e.g., very small vessels), allowing for increased vascular signal.79,80 It detects both the
phase and magnitude effects of Gd to obtain a larger effect from the same Gd.
© 2010 Elsevier
2 of 2 20-04-2010 00:19
DIRECT MAGNETIC RESONANCE VENOGRAPHY

Gadolinium-enhanced MR venograms can be performed in a manner similar to a conventional contrast
venogram.81,82 Injecting dilute gadolinium directly into the peripheral veins (arm or leg) during 3D image
acquisition enhances veins along the path of venous return to the heart, such as the subclavian veins,
SVC, and IVC (Fig. 29-9). This produces much greater venous enhancement than the venous phase of
an arterial study since the gadolinium is delivered to the veins in a much higher concentration (less
dilution of the bolus). The arterial enhancement is less than on a typical contrast MRA study.
page 752
page 753
Add to lightbox
Add to lightbox
1 of 4 20-04-2010 00:19
Add to lightbox
Figure 29-8 Stanford type B aortic dissection. A, Full volume oblique sagittal MIP. Note the bovine
arch. B, Subvolume MIP demonstrating the origin of the intimal flap. C, Axial MPR showing the two
lumens with mural thrombus in the false lumen.
Add to lightbox
2 of 4 20-04-2010 00:19
Add to lightbox
page 753
page 754
Figure 29-9 Early (A) and delayed (B) subvolume MIPs from a CEMRA study where dilute (1:12.5)
gadolinium was infused bilaterally into the antecubital veins. Note the complete occlusion of the left
innominate vein, with high-grade stenosis at the confluence of the right innominate vein and SVC. The
left subclavian vein signal void is secondary to slab positioning.
Table 29-1. Classification Scheme for "Vascular" MR Contrast Agents. The

Paramagnetic Gadolinium Chelates can be Classified According to their Degree
of Protein Interaction. The Ultra-small Iron Oxide Particles are "Blood Pool
Agents" which Demonstrate Long Intravascular Enhancement
Paramagnetic Superparamagnetic
Strong
No protein Weak protein protein
interaction interaction interaction Macromolecular USPIO
Gadopentetate Gadobenate MS-325 P792 SHU-555A
dimeglumine dimeglumine
Gd-DTPA; Gd-BOPTA; Resovist®
Magnevist® MultiHance®
Gadoteridol Ferumoxytol
Gd-HP-DO3A;
ProHance®
Gadodiamide
Gd-DTPA-BMA;
Omniscan®
Gadoversetamide B-22956 Gadomer-17 NC100150
Gd-DTPA-BMEA; Clariscan®
Optimark®
Gadoterate
meglumine
3 of 4 20-04-2010 00:19
Gd-DOTA;
Dotarem®
Gadobutrol
Gd-BT-Do3A;
Gadovist®
Diluting Gd is necessary because full-strength Gd shortens T2* so much that all signal is eliminated. A
20:1 dilution, 3 mL Gd into 57 mL saline in a 60 cc syringe, works well when injected into a peripheral
vein at 3 mL/s. Multiple veins can be injected simultaneously, as in Figure 29-9. Imaging delay is not
terribly important, as the dilute gadolinium solution can be delivered in large volumes. Volume coverage
and slice thickness must be tailored to the individual case and multiple phases should be obtained since
collateral pathways take a long time to fill when there is venous occlusion.
© 2010 Elsevier
4 of 4 20-04-2010 00:19
FUTURE DIRECTIONS
Contrast Agents
First-Pass Gadolinium Agents
The most widely used contrast agents for first-pass contrast-enhanced MRA are the "conventional"
gadolinium chelates (Table 29-1, column 1). These are available as 0.5 molar formulations and
83
possess T1 relaxivities (1.5 T) of between 4.3 and 5.0 mmol-s. These agents have similar vascular
imaging performance when injected at equal doses reflecting their similar concentrations and relaxation
properties.
Recently, gadolinium chelates have been developed which possess properties advantageous for
84
first-pass contrast-enhanced MRA. These include conventional gadolinium agents formulated at 1.0
molar rather than 0.5 molar (e.g., gadobutrol) and agents with higher in vivo relaxivity due to a capacity
for weak protein interaction (e.g., gadobenate dimeglumine MS-325). In the case of gadobutrol,
studies have shown that the higher concentration of gadolinium combined with a more rapid bolus
enables greater intravascular signal than is achievable with equivalent doses of traditional agents. 85-87
This may be useful for visualization of smaller arteries in vascular regions in which the SNR is limited,
such as in the calf and distal renal arteries. Improved visualization of smaller pelvic vessels has already
85
been reported.
Add to lightbox
page 754
page 755
Figure 29-10 The full-thickness MIP (0.1 mmol/kg Gd-BOPTA) shows occlusion of the left subclavian
artery (arrow) with collateralization from the left vertebral artery. The patient presented with typical
clinical symptoms and interventional therapy was performed. Due to the clear depiction of the size of
the narrowed lumen, an interventional procedure could be accurately planned. (Courtesy of Günther
Schneider, Department of Diagnostic Radiology, University Hospital, Homburg/Saar, Germany.)
1 of 7 20-04-2010 00:20
Add to lightbox
Add to lightbox
Figure 29-11 A 64-year-old patient post left side nephrectomy with newly developed arterial
hypertension. The full-thickness MIP image (A) reveals a proximal renal artery stenosis with slight
post-stenotic dilatation of the vessel. The corresponding DSA examination (B) performed during
interventional therapy confirms the stenosis. Note the excellent depiction of smaller vessels on the
MRA image and the good correlation with DSA. (Courtesy of Günther Schneider, Department of
Diagnostic Radiology, University Hospital, Homburg/Saar, Germany.)
Gadobenate dimeglumine (Table 29-1, column 2) also demonstrates increased intravascular signal as
2 of 7 20-04-2010 00:20
compared to conventional gadolinium chelates.85,88,89 This agent possesses a higher in vivo T1

relaxivity (9.7 mmol-s) due to weak and transient interactions between the Gd-BOPTA chelate and
83,90
serum proteins, particularly albumin. Clinical benefits from the increased relaxivity have been
demonstrated for all vascular territories including the peripheral vasculature.91-93 There is better
visualization of lesions in larger vessels (Fig. 29-10) as well as better depiction of smaller vessels (Fig.
29-11). Like the conventional nonprotein interacting gadolinium chelates, gadobenate dimeglumine has
94
an excellent safety profile.
"Blood Pool" Gadolinium Agents

Because rapid redistribution of the first-pass gadolinium chelates into the extracellular compartment
limits the time for MRA data acquisition, attempts have been made to formulate intravascular "blood
pool" contrast agents for 3D contrast MRA. Two types of intravascular agents based on gadolinium
are under development: agents with strong affinity for serum albumin (Table 29-1, column 3) and
macromolecular agents whose size precludes their rapid extravasation (Table 29-1, column 4).
Although no agents are yet approved for clinical use, two examples of agents with strong affinity for
serum proteins are currently undergoing clinical trials: MS-32595-97 (Fig. 29-12) and B2295698,99 (Fig.
29-13). Both of these agents are promising, with excellent safety profiles and the capacity to perform
conventional high-quality first-pass dynamic imaging in addition to delayed steady-state vascular
99-101
imaging. This may be particularly useful for coronary studies, for detecting internal bleeding, and
imaging stent graft endoleaks. Blood pool agents also offer opportunities to image multiple vascular
beds such as pulmonary arteries followed by peripheral MR venography in the blood pool phase.
Examples of gadolinium-based blood pool agents with macromolecular structures are gadomer-17102
and P792.103 Like MS-325 and B22956, preliminary results suggest these agents may be useful for
coronary MRA.104,105
Superparamagnetic Iron Oxide Agents

The second new category of contrast agents for MRA is the superparamagnetic group, which is based
on ultra-small particles of iron oxide (USPIO) (Table 29-1, column 5). Currently, SHU 555A is the only
clinically approved USPIO agent available for contrast-enhanced MRA (European Union), although few
studies have been conducted in vascular territories other than the upper abdomen. 106 Ferumoxytol is
another promising agent in this group, although again few studies have been conducted.107,108
Another USPIO agent is NC100150 (Fig. 29-14).109,110 As with the gadolinium blood pool agents, the
potential advantages of this agent include a long intravascular half-life with minimal leakage into the
interstitial space, thereby permitting steady-state vascular imaging. Preliminary studies on the use of
109
this agent for coronary MRA have already been reported and work is ongoing to establish its
potential for other applications such as renal perfusion. 111 Although this particluar contrast agent is no
longer in clinical development, other iron oxide blood pool agents, are moving along in clinical trials.
page 755
page 756
3 of 7 20-04-2010 00:20
Add to lightbox
Add to lightbox
Figure 29-12 CEMRA of the foot using the blood pool agent MS-325 in a patient with ischemic rest
pain. Arterial (A) and equilibrium (B) sagittal full volume MIPs, MS-325 dose 0.05 mmol/kg. The
3
arterial study is subtracted, 46 s in duration, with true voxel size of 0.9 × 0.9 × 1.8 mm . The
equilibrium study is fat suppressed at a delay of 5 minutes, 4:57 in duration, with a true voxel size 0.9 ×
0.9 × 0.8 mm3.
4 of 7 20-04-2010 00:20
Add to lightbox
Add to lightbox
page 756
page 757
Figure 29-13 B22956 in imaging of healthy right coronary artery. A, Unenhanced (T2Prep acquisition)
and B, B22956-enhanced (0.075 mmol/kg) image acquired at 25 min post contrast using 3D
5 of 7 20-04-2010 00:20
Navigator gated and corrected IR-FFE sequence. (Courtesy of Eckart Fleck and Ingo Paetsch,
German Heart Institute, Berlin, Germany; Matthias Stuber, Beth Israel Deaconess Medical Center,
Cardiovascular Division, Boston, MA, USA; and Friedrich Cavagna, Bracco Imaging SpA, Milano,
Italy.)
Add to lightbox
Figure 29-14 NC100150 (5.0 mg Fe/kg bodyweight, slow hand injection, third injection of an
accumulating dose scheme) reveals occlusion of the left renal artery and a 60-70% stenosis of the
right renal artery.
The high imaging quality achievable with first-pass agents has to some extent eliminated the initial
target role for intravascular agents. It has become apparent that real benefits will most likely only be
clinically achievable if the intravascular agents can be used for both first-pass and steady-state
imaging. Nevertheless, it is doubtful whether this type of blood pool agent will ever have any advantage
over the currently available gadolinium agents for dynamic, arterial phase MRA. Conceivably, this type
of agent may have benefits for ultra high-resolution imaging, assessment of vasculature during
intervention, and for realtime therapy monitoring. Other opportunities which may present themselves
include perfusion applications combined with first-pass imaging, internal bleeding such as in the GI
112 113
tract or in aortic stent graft endoleaks.
While MR angiography using blood pool agents may ultimately prove beneficial, all current clinical work
utilizes extracellular gadolinium chelates.
Hardware
Among the most important recent developments in MR hardware has been the introduction of
whole-body 3 Tesla scanners. The principal advantage of increasing the magnetic field strength is
6 of 7 20-04-2010 00:20
improved SNR, resulting in better resolution and conspicuity of vessels using similar acquisition
times.114 Although clinical experience with contrast-enhanced MRA at 3 T is limited at present, early
studies have indicated markedly improved performance of unenhanced TOF techniques, particularly in
114,115
imaging of the intracranial and cervical vasculature. Similarly, imaging at 3 T has been reported
to improve the resolution and delineation of small venous vessels in the brain116 and appears
117-119
advantageous for contrast-enhanced vascular and coronary imaging. It remains to be seen just
how the availability of 3 T machines will impact contrast-enhanced MRA. Several challenges remain to
be overcome, including developing new coils, overcoming SAR (RF power deposition) limitations, and
optimizing protocols. This aside, however, it appears likely that the SNR-derived improvement in spatial
and temporal resolution at 3 T will lead to significant improvements in CEMRA image quality. SNR
increases with field strength so 3 T scanners have more SNR than 1.5 T scanners. Recent FDA
approval at field strengths of up to 7 and 8 T for whole-body imaging opens up tremendous
opportunities for further SNR improvements.
One development still on the drawing board is a long magnet which can simultaneously image the
entire body. Potentially, this could provide time-resolved MRA data of the entire body with a single
contrast bolus injection.
There is a renaissance under way in the development of MR imaging coils to optimally implement
parallel imaging (SENSE and SMASH) techniques for accelerating MR data acquisition. All major
scanner manufacturers are engineering the capability to handle at least 32 channels of simultaneous
data acquisition. This, in theory, can speed up scanning by as much as 32-fold.
© 2010 Elsevier
7 of 7 20-04-2010 00:20
CONCLUSION
Contrast-enhanced MR angiography (CEMRA) provides a safe, fast, and cost-efficient alternative to
conventional diagnostic angiography throughout the vasculature. High-resolution breath-hold and
multistation imaging with optimal arterial phase contrast bolus timing is now routine. As the technique is
further refined, better coils, better contrast agents, and perhaps higher magnetic field strength will
allow for greater spatial and temporal resolution, further increasing the effectiveness and utility of
CEMRA.
REFERENCES
1. Prince MR: Gadolinium-enhanced MR aortography. Radiology 191:155-164, 1994. Medline Similar articles
2. Prince MR: Body MR angiography with gadolinium contrast agents. Magn Reson Imaging Clin North Am 4:11-24, 1996.
3. Weinmann HJ, Brasch RC, Press WR, Wesbey GE: Characteristics of gadolinium-DTPA complex: a potential NMR contrast
agent. Am J Roentgenol 142:619-624, 1984.
4. Schima W, Mukerjee A, Saini S: Contrast-enhanced MR imaging. Clin Radiol 51:235-244, 1996. Medline
Similar articles
5. Goldstein HA, Kashanian FK, Blumetti RF, et al: Safety assessment of gadopentetate dimeglumine in U.S. clinical trials.
page 757
page 758
7. Hohenschuh E, Watson A: Theory and mechanisms of contrast-enhancing agents. In Higgins CB, Hricak H, Helms CA (eds):
Magnetic Resonance Imaging of the Body. Philadelphia: Lippincott-Raven, 1997, pp. 1439-1464.
8. Niendorf HP, Haustein J, Cornelius I, et al: Safety of gadolinium-DTPA: extended clinical experience. Magn Reson Med
9. Prince MR, Arnoldus C, Frisoli JK: Nephrotoxicity of high-dose gadolinium compared with iodinated contrast. J Magn Reson
10. Runge VM, Kirsch JE, Lee C: Contrast-enhanced MR angiography. J Magn Reson Imaging 3:233-239, 1993. Medline
Similar articles
11. Prince MR, Yucel EK, Kaufman JA, et al: Dynamic gadolinium-enhanced three-dimensional abdominal MR arteriography. J
12. McNamara MT, Higgins CB, Ehman RL, et al: Acute myocardial ischemia: magnetic resonance contrast enhancement with
gadolinium-DTPA. Radiology 153:157-163, 1984. Medline Similar articles
13. Leung DA, McKinnon GC, Davis CP, et al: Breath-hold, contrast-enhanced, three-dimensional MR angiography. Radiology
14. Snidow JJ, Johnson MS, Harris VJ, et al: Three-dimensional gadolinium-enhanced MR angiography for aortoiliac inflow
assessment plus renal artery screening in a single breath hold. Radiology 198:725-732, 1996. Medline Similar articles
15. Moseley M, Sawyer A: Imaging techniques: pulse sequences. In Higgins CB, Hricak H, Helms CA (eds): Magnetic
Resonance Imaging of the Body. Philadelphia: Lippincott-Raven, 1997, pp. 43-69.
16. Wehrli F: Principles of magnetic resonance. In Stark DD, Bradley WG (eds): Magnetic Resonance Imaging. St Louis:
Mosby Yearbook, 1992, pp. 3-20.
17. Prince MR, Narasimham DL, Stanley JC, et al: Breath-hold gadolinium-enhanced MR angiography of the abdominal aorta
and its major branches. Radiology 197:785-792, 1995. Medline Similar articles
18. Hendrick R, Roff U: Image Contrast and Noise. Chicago: Mosby Yearbook, 1991.
19. Weinmann HJ, Laniado M, Mutzel W: Pharmacokinetics of GdDTPA/dimeglumine after intravenous injection into healthy
volunteers. Physiol Chem Phys Med NMR 16:167-172, 1984. Medline Similar articles
20. Lee VS, Flyer MA, Weinreb JC, et al: Image subtraction in gadolinium-enhanced MR imaging. Am J Roentgenol
167:1427-1432, 1996.
21. Maki JH, Prince MR, Chenevert TC: Optimizing three-dimensional gadolinium-enhanced magnetic resonance angiography.
Original investigation. Invest Radiol 33:528-537, 1998. Medline Similar articles
22. Verhoeven L: Digital subtraction angiography: the technique and an analysis of the physical factors influencing the image
quality. Delft: Technische Hogeschule, 1985.
23. Barnes G, Lakshminarayann A: Conventional and spiral computed tomography. In Lee J, Sagel S, Stanley R, Heiken J
(eds): Computed Body Tomography with MRI Correlation. Philadelphia: Lippincott-Raven, 1998, pp. 1-20.
24. Riederer SJ, Tasciyan T, Farzaneh F, et al: MR fluoroscopy: technical feasibility. Magn Reson Med 8:1-15, 1988.
1 of 5 20-04-2010 00:21
25. Maki JH, Prince MR, Londy FJ, Chenevert TL: The effects of time varying intravascular signal intensity and k-space
acquisition order on three-dimensional MR angiography image quality. J Magn Reson Imaging 6:642-651, 1996. Medline
Similar articles
26. Maki JH, Wilson GJ, Eubank WB, Hoogeveen RM: Utilizing SENSE to achieve lower station sub-millimeter isotropic
resolution and minimal venous enhancement in peripheral MR angiography. J Magn Reson Imaging 15:484-491, 2002.
27. Fain SB, Riederer SJ, Bernstein MA, Huston J: Theoretical limits of spatial resolution in elliptical-centric contrast-enhanced
3D-MRA. Magn Reson Med 42:1106-1116, 1999.
28. Bampton AE, Riederer SJ, Korin HW: Centric phase-encoding order in three-dimensional MP-RAGE sequences:
application to abdominal imaging. J Magn Reson Imaging 2:327-334, 1992. Medline Similar articles
29. Wilman AH, Riederer SJ: Improved centric phase encoding orders for three-dimensional magnetization-prepared MR
Similar articles
31. Maki JH, Chenevert TL, Prince MR: The effects of incomplete breath-holding on 3D MR image quality. J Magn Reson
Imaging 7:1132-1139, 1997.
32. Shetty AN, Bis KG, Kirsch M, et al: Contrast-enhanced breath-hold three-dimensional magnetic resonance angiography in
the evaluation of renal arteries: optimization of technique and pitfalls. J Magn Reson Imaging 12:912-923, 2000. Medline
Similar articles
33. Hany TF, Schmidt M, Hilfiker PR, et al: Optimization of contrast dosage for gadolinium-enhanced 3D MRA of the pulmonary
and renal arteries. Magn Reson Imaging Clin North Am 16:901-906, 1998.
34. Snidow JJ, Aisen AM, Harris VJ, et al: Iliac artery MR angiography: comparison of three-dimensional gadolinium-enhanced
and two-dimensional time-of-flight techniques. Radiology 196:371-378, 1995. Medline Similar articles
35. Finn JP, Baskaran V, Carr JC, et al: Thorax: low-dose contrast-enhanced three-dimensional MR angiography with
subsecond temporal resolution - initial results. Radiology 224:896-904, 2002. Medline Similar articles
36. Prince MR, Grist TM, Debatin JF: 3D Contrast MR Angiography. Berlin: Springer, 2003.
37. Earls JP, Rofsky NM, DeCorato DR, et al: Breath-hold single-dose gadolinium-enhanced three-dimensional MR
aortography: usefulness of a timing examination and MR power injector. Radiology 201:705-710, 1996. Medline
Similar articles
38. Maki J, Prince M, Londy F, Chenevert T: The effects of time varying intravascular signal intensity on three-dimensional MR
angiography image quality. J Magn Reson Imaging 6:642-651, 1996. Medline Similar articles
39. Kreitner KF, Kunz RP, Kalden P, et al: Contrast-enhanced three-dimensional MR angiography of the thoracic aorta:
experiences after 118 examinations with a standard dose contrast administration and different injection protocols. Eur Radiol
11:1355-1363, 2001.
40. Krinsky GA, Rofsky NM, Flyer M, et al: Gadolinium-enhanced three-dimensional MR angiography of acquired arch vessel
disease. Am J Roentgenol 167:981-987, 1996.
41. Huston J, Fain SB, Riederer SJ, et al: Carotid arteries: maximizing arterial to venous contrast in fluoroscopically triggered
contrast-enhanced MR angiography with elliptic centric view ordering. Radiology 211:265-273, 1999. Medline
Similar articles
42. Fellner FA, Fellner C, Wutke R, et al: Fluoroscopically triggered contrast-enhanced 3D MR DSA and 3D time-of-flight turbo
MRA of the carotid arteries: first clinical experiences in correlation with ultrasound, x-ray angiography, and endarterectomy
findings. Magn Reson Imaging Clin North Am 18:575-585, 2000.
43. Watts R, Wang Y, Redd B, et al: Recessed elliptical-centric view-ordering for contrast-enhanced 3D MR angiography of the
carotid arteries. Magn Reson Med 48:419-424, 2002.
44. Schoenberg SO, Essig M, Hallscheidt P, et al: Multiphase magnetic resonance angiography of the abdominal and pelvic
arteries: results of a bicenter multireader analysis. Invest Radiol 37:20-28, 2002. Medline Similar articles
45. Schoenberg SO, Bock M, Floemer F, et al: High-resolution pulmonary arterio- and venography using multiple-bolus
multiphase 3D-Gd-MRA. J Magn Reson Imaging 10:339-346, 1999.
46. Mistretta CA, Grist TM, Korosec FR, et al: 3D time-resolved contrast-enhanced MR DSA: advantages and tradeoffs. Magn
36:345-351, 1996.
48. Levy RA, Maki JH: Three-dimensional contrast-enhanced MR angiography of the extracranial carotid arteries: two
techniques. Am J Neuroradiol 19:688-690, 1998. Medline Similar articles
49. Stein B, DeMarco J, Zhou Y: 3D time-resolved MRA with elliptic centric view ordering. Initial clinical experience to evaluate
tibo-peroneal arteries. In: Proceedings of the 13th Annual International Workshop on MRA. Madison, WI, 2001, p. 107.
50. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P: SENSE: sensitivity encoding for fast MRI. Magn Reson Med
2 of 5 20-04-2010 00:21
51. Finn JP, Carr JR, Pereles FS, et al: Time-resolved, contrast-enhanced MR angiography of the thorax. In: XIIIth Annual
International Workshop on MRA. Madison, WI, 2001, p. 25.
52. Goyen M, Laub G, Ladd ME, et al: Dynamic 3D MR angiography of the pulmonary arteries in under four seconds. J Magn
Reson Imaging 13:372-377, 2001.
53. Hennig J, Scheffler K, Laubenberger J, Strecker R: Time-resolved projection angiography after bolus injection of contrast
agent. Magn Reson Med 37:341-345, 1997. Medline Similar articles
54. Finn JP, Larson A, Moore J, Simonetti O: Sub-second 3D contrast-enhanced MRA of the thorax with radial k-space
sampling. In: Proceedings of the 10th Scientific Meeting and Exhibiton of the International Society for Magnetic Resonance in
Medicine, Honolulu, HI, 2002.
55. Bock M, Schoenberg SO, Floemer F, Schad LR: Separation of arteries and veins in 3D MR angiography using correlation
analysis. Magn Reson Med 43:481-487, 2000.
56. Mazaheri Y, Carroll TJ, Du J, et al: Combined time-resolved and high-spatial-resolution 3D MRA using an extended
adaptive acquisition. J Magn Reson Imaging 15:291-301, 2002.
57. Turski PA, Korosec FR, Carroll TJ, et al: Contrast-enhanced magnetic resonance angiography of the carotid bifurcation
using the time-resolved imaging of contrast kinetics (TRICKS) technique. Top Magn Reson Imaging 12:175-181, 2001.
58. Peters DC, Korosec FR, Grist TM, et al: Undersampled projection reconstruction applied to MR angiography. Magn Reson
60. Willinek W, Gieseke J, Von Falkenhousen V: CD-3D MRA of the supraaortic arteries at 512 and 1024 matrix: the use of
randomly segmented central k-space ordering (CENTRA). In: Proceedings of the 10th Scientific Meeting and Exhibiton of the
International Society for Magnetic Resonance in Medicine, Honolulu, HI, 2002.
62. Wood ML, Runge VM, Henkelman RM: Overcoming motion in abdominal MR imaging. Am J Roentgenol 150:513-522,
1988.
63. Vasbinder GB, Maki JH, Nijenhuis RJ, et al: Motion of the distal renal artery during three-dimensional contrast-enhanced
breath-hold MRA. J Magn Reson Imaging 16:685-696, 2002. Medline Similar articles
64. Holland GA, Dougherty L, Carpenter JP, et al: Breath-hold ultrafast three-dimensional gadolinium-enhanced MR
angiography of the aorta and the renal and other visceral abdominal arteries. Am J Roentgenol 166:971-981, 1996.
65. Carr JC, Ma J, Desphande V, et al: High-resolution breath-hold contrast-enhanced MR angiography of the entire carotid
circulation. Am J Roentgenol 178:543-549, 2002.
66. Weiger M, Pruessmann KP, Kassner A, et al: Contrast-enhanced 3D MRA using SENSE. J Magn Reson Imaging
12:671-677, 2000.
page 758
page 759
68. Maki JH, Wilson GJ, Hoogeveen R: 3D Gd enhanced moving table MR angiography of the aorta and outflow vessels using
SENSE to achieve high resolution of the below knee vasculature. In: Proceedings of the Seventh Scientific Meeting of the
International Society for Magnetic Resonance in Medicine, Glasgow, Scotland, 2001.
69. Du Y, Parker DL, Davis WL, Blatter DD: Contrast-to-noise-ratio measurements in three-dimensional magnetic resonance
angiography. Invest Radiol 28:1004-1009, 1993. Medline Similar articles
70. Schoenberg SO, Bock M, Aumann S, et al: Quantitative recording of renal function with magnetic resonance tomography.
Radiologe 40:925-937, 2000. Medline Similar articles
71. Ros PR, Gauger J, Stoupis C, et al: Diagnosis of renal artery stenosis: feasibility of combining MR angiography, MR
renography, and gadopentetate-based measurements of glomerular filtration rate. Am J Roentgenol 165:1447-1451, 1995.
72. Saloner D. MRA: principles and display. In Higgins CB, Hricak H, Helms CA (eds): Magnetic Resonance Imaging of the
Body. Philadelphia: Lippincott-Raven, 1997.
73. De Marco JK, Nesbit GM, Wesbey GE, Richardson D: Prospective evaluation of extracranial carotid stenosis: MR
angiography with maximum-intensity projections and multiplanar reformation compared with conventional angiography. Am J
Roentgenol 163:1205-1212, 1994.
74. Rossnick S, Laub G, Braeckle R: Three dimensional display of blood vessels in MRI. In: Proceedings of the IEEE:
Computers in Cardiology. New York, 1986, pp. 193-195.
75. Laub G: Displays for MR angiography. Magn Reson Med 14:222-229, 1990. Medline Similar articles
76. Anderson CM, Saloner D, Tsuruda JS, et al: Artifacts in maximum-intensity-projection display of MR angiograms. Am J
Roentgenol 154:623-629, 1990.
77. Baskaran V, Pereles FS, Nemcek AA, et al: Gadolinium-enhanced 3D MR angiography of renal artery stenosis: a pilot
3 of 5 20-04-2010 00:21
Acad Radiol 9:50-59, 2002.

78. Douek PC, Revel D, Chazel S, et al: Fast MR angiography of the aortoiliac arteries and arteries of the lower extremity:
value of bolus-enhanced, whole-volume subtraction technique. Am J Roentgenol 165:431-437, 1995.
79. Wang Y, Winchester PA, Khilnani NM, et al: Contrast-enhanced peripheral MR angiography from the abdominal aorta to
the pedal arteries: combined dynamic two-dimensional and bolus-chase three-dimensional acquisitions. Invest Radiol
80. Wang Y, Johnston DL, Breen JF, et al: Dynamic MR digital subtraction angiography using contrast enhancement, fast data
acquisition, and complex subtraction. Magn Reson Med 36:551-556, 1996. Medline Similar articles
81. Li W, David V, Kaplan R, Edelman RR: Three-dimensional low dose gadolinium-enhanced peripheral MR venography. J
82. Ruehm SG, Zimny K, Debatin JF: Direct contrast-enhanced 3D MR venography. Eur Radiol 11:102-112, 2001.
83. de Haen C, Cabrini M, Akhnana L, et al: Gadobenate dimeglumine 0.5 M solution for injection (MultiHance) pharmaceutical
formulation and physicochemical properties of a new magnetic resonance imaging contrast medium. J Comput Assist Tomogr
Suppl 1:S161-168, 1999.
84. Knopp MV, von Tengg-Kobligk H, Floemer F, Schoenberg SO: Contrast agents for MRA: future directions. J Magn Reson
85. Herborn CU, Lauenstein TC, Ruehm SG, et al: Intraindividual comparison of gadopentetate dimeglumine , gadobenate
dimeglumine, and gadobutrol for pelvic 3D magnetic resonance angiography. Invest Radiol 38:27-33, 2003.
86. Goyen M, Lauenstein TC, Herborn CU, et al: 0.5 M Gd chelate (Magnevist) versus 1.0 M Gd chelate (Gadovist):
dose-independent effect on image quality of pelvic three-dimensional MR-angiography. J Magn Reson Imaging 14:602-607,
87. Tombach B, Reimer P, Prumer B, et al: Does a higher concentration of gadolinium chelates improve first-pass cardiac
signal changes? J Magn Reson Imaging 10:806-812, 1999. Medline Similar articles
88. Volk M, Strotzer M, Lenhart M, et al: Renal time-resolved MR angiography: quantitative comparison of gadobenate
dimeglumine and gadopentetate dimeglumine with different doses. Radiology 220:484-488, 2001. Medline
Similar articles
89. Knopp MV, Schoenberg SO, Rehm C, et al: Assessment of gadobenate dimeglumine for magnetic resonance
angiography: phase I studies. Invest Radiol 37:706-715, 2002. Medline Similar articles
90. Cavagna FM, Maggioni F, Castelli PM, et al: Gadolinium chelates with weak binding to serum proteins. A new class of
high-efficiency, general purpose contrast agents for magnetic resonance imaging. Invest Radiol 32:780-796, 1997.
91. Goyen M, Herborn CU, Lauenstein TC, et al: Optimization of contrast dosage for gadobenate dimeglumine-enhanced
high-resolution whole-body 3D magnetic resonance angiography. Invest Radiol 37:263-268, 2002.
92. Ruehm SG, Goyen M, Barkhausen J, et al: Rapid magnetic resonance angiography for detection of atherosclerosis. Lancet
93. Kroencke TJ, Wasser MN, Pattynama PM, et al: Gadobenate dimeglumine-enhanced MR angiography of the abdominal
aorta and renal arteries. Am J Roentgenol 179:1573-1582, 2002.
94. Kirchin MA, Pirovano G, Venetianer C, Spinazzi A: Safety assessment of gadobenate dimeglumine (MultiHance):
extended clinical experience from phase I studies to post-marketing surveillance. J Magn Reson Imaging 14:281-294, 2001.
95. Bluemke DA, Stillman AE, Bis KG, et al: Carotid MR angiography: phase II study of safety and efficacy for MS-325.
97. Grist TM, Korosec FR, Peters DC, et al: Steady-state and dynamic MR angiography with MS-325: initial experience in
humans. Radiology 207:539-544, 1998. Medline Similar articles
98. Zheng J, Li D, Cavagna FM, et al: Contrast-enhanced coronary MR angiography: relationship between coronary artery
delineation and blood T1. J Magn Reson Imaging 14:348-354, 2001. Medline Similar articles
99. La Noce A, Stoelben S, Scheffler K, et al: B22956/1, a new intravascular contrast agent for MRI: first administration to
humans-preliminary results. Acad Radiol Suppl 2:S404-406, 2002.
100. Huber ME, Paetsch I, Schnackenburg B, et al: Performance of a new gadolinium-based intravascular contrast agent in
free-breathing inversion-recovery 3D coronary MRA. Magn Reson Med 49:115-121, 2003.
101. Stuber M, Botnar RM, Danias PG, et al: Contrast agent-enhanced, free-breathing, three-dimensional coronary magnetic
resonance angiography. J Magn Reson Imaging 10:790-799, 1999. Medline Similar articles
102. Dong Q, Hurst DR, Weinmann HJ, et al: Magnetic resonance angiography with gadomer-17. An animal study original
investigation. Invest Radiol 33:699-708, 1998. Medline Similar articles
103. Gaillard S, Kubiak C, Stolz C, et al: Safety and pharmacokinetics of p792, a new blood-pool agent: results of clinical
testing in nonpatient volunteers. Invest Radiol 37:161-166, 2002. Medline Similar articles
104. Li D, Zheng J, Weinmann HJ: Contrast-enhanced MR imaging of coronary arteries: comparison of intra- and extravascular
4 of 5 20-04-2010 00:21
contrast agents in swine. Radiology 218:670-678, 2001. Medline Similar articles

105. Taupitz M, Schnorr J, Wagner S, et al: Coronary magnetic resonance angiography: experimental evaluation of the new
rapid clearance blood pool contrast medium P792. Magn Reson Med 46:932-938, 2001. Medline Similar articles
106. Reimer P, Allkemper T, Matuszewski L, Balzer T: Contrast-enhanced 3D-MRA of the upper abdomen with a bolus-
injectable SPIO (SHU 555 A). J Magn Reson Imaging 10:65-71, 1999.
107. Prince MR, Zhang HL, Chabra SG, et al: A pilot investigation of new superparamagnetic iron oxide (ferumoxytol) as a
contrast agent for cardiovascular MRI. J X-Ray Sci Technol 11:231-240, 2003.
108. Mayo-Smith WW, Saini S, Slater G, Kaufman JA, et al: MR contrast material for vascular enhancement: value of
superparamagnetic iron oxide. Am J Roentgenol 166:73-77, 1996.
109. Taylor AM, Panting JR, Keegan J, et al: Safety and preliminary findings with the intravascular contrast agent NC100150
injection for MR coronary angiography. J Magn Reson Imaging 9:220-227, 1999. Medline Similar articles
110. Weishaupt D, Ruhm SG, Binkert CA, et al: Equilibrium-phase MR angiography of the aortoiliac and renal arteries using a
blood pool contrast agent. Am J Roentgenol 175:189-195, 2000.
111. Bachmann R, Conrad R, Kreft B, et al: Evaluation of a new ultrasmall superparamagnetic iron oxide contrast agent
Clariscan, (NC100150) for MRI of renal perfusion: experimental study in an animal model. J Magn Reson Imaging 16:190-195,
112. Hilfiker PR, Zimmermann-Paul GG, Schmidt M, et al: Intestinal and peritoneal bleeding: detection with an intravascular
contrast agent and fast three-dimensional MR imaging-preliminary experience from an experimental study. Radiology
113. Ersoy HE, Prince MR, Kent KC: Detecting aortic stent graft endoleak with blood pool MRA. Am J Roentgenol (in press).
114. Campeau NG, Huston J, Bernstein MA, et al: Magnetic resonance angiography at 3.0 Tesla: initial clinical experience.
Top Magn Reson Imaging 12:183-204, 2001.
115. Al-Kwifi O, Emery DJ, Wilman AH: Vessel contrast at three Tesla in time-of-flight magnetic resonance angiography of the
intracranial and carotid arteries. Magn Reson Imaging Clin North Am 20:181-187, 2002.
116. Reichenbach JR, Barth M, Haacke EM, et al: High-resolution MR venography at 3.0 Tesla. J Comput Assist Tomogr
24:949-957, 2000.
117. Stuber M, Botnar RM, Fischer SE, et al: Preliminary report on in vivo coronary MRA at 3 Tesla in humans. Magn Reson
Med 48:425-429, 2002.
118. Leiner T, Vasbinder B, de Vries M, et al: Contrast-enhanced peripheral MRA at 3.0T: initial results. In: Proceedings of the
10th Scientific Meeting and Exhibition of the ISMRM, Honolulu, HI, 2002.
119. Hugg J, Rofsky NM, Stokar S, et al: Clinical whole body MRI at 3.0 T - initial experience. In: Proceedings of the 10th
Scientific Meeting and Exhibition of the ISMRM, Honolulu, HI, 2002.
© 2010 Elsevier
5 of 5 20-04-2010 00:21
AGNETIC ESONANCE NGIOGRAPHY
James P. Earls
Robert R. Edelman
In a relatively short period of time, body magnetic resonance angiography (MRA) has become the
accepted clinical standard for noninvasive imaging of the arteries of the chest, abdomen, pelvis, and
lower extremities. The initial clinical MR images of a human thorax were published in 19771 and
2-5
imaging of the cardiovascular system was an important goal of early MR researchers. During this
developmental period there was great hope that MRI could be used as an accurate noninvasive
method to replace diagnostic catheter-based X-ray angiography.
Body MRA applications were slow to develop and be accepted clinically, unlike neurologic MRA
methods which became widely used and accepted early in the development of MR. In the thorax and
abdomen, motion-related artifacts, cardiac pulsation, bowel peristalsis, and the need to image
relatively large areas of anatomy substantially complicate imaging. Many different motion reduction
methods and gating algorithms were developed to improve body MRA, with some success.
For many years, time-of-flight (TOF) techniques dominated clinical body MR applications. Phase
contrast (PC) techniques were used as an adjunct or occasionally as a stand-alone method. While
these techniques were valuable and scientifically proven, they had some inherent flaws that limited
more widespread application.
Currently contrast-enhanced MR angiography (CEMRA) has replaced TOF and PC techniques for
most applications outside the central nervous system. CEMRA is a fast, accurate, and flexible method
for noninvasive imaging of the arterial system and since its initial development has been continually
refined.6,7 In a short period of time it migrated from academic centers into the community and is now
the workhorse of body MRA.
CEMRA is augmented by two additional approaches used to image the vessels in cross-section.
"Black blood" techniques, such as spin-echo or the newer T2-weighted inversion recovery methods,
provide a high degree of contrast between the vessel wall and blood pool.8 The development of ECG
gating made SE technique especially useful by substantially reducing motion artifacts. 9 "Bright blood"
imaging yields both morphologic and functional data. Blood generates bright signal intensity and
multiple consecutive images are acquired that can be viewed dynamically to depict motion. The most
commonly used cine methods currently include segmented k-space fast GRE and steady-state free
precession (SSFP).
© 2010 Elsevier
1 of 1 20-04-2010 00:21
INDICATIONS AND COMMON APPLICATIONS

page 760
page 761
Add to lightbox
Figure 30-1 Contrast-enhanced 3D MR angiogram (CEMRA) depicts normal aorta and bilateral single
renal arteries without evidence of atherosclerosis.
Indications for MRA in the body are varied and widespread. MR is useful for imaging all the arteries in
the body from the carotids to the plantar arch in the feet. Clinically the most commonly imaged vessels,
outside the head and neck, are the thoracic and abdominal aorta and the renal arteries (Fig. 30-1).
The other major aortic branches, including the subclavian, celiac, and superior mesenteric arteries, are
also accurately imaged on MRI and the routine imaging of these vessels with MRA is increasing.
Dissections and aneurysms are the most frequently encountered indication for aortic MRA, in both the
chest and abdomen. MR can accurately depict the size, shape, location, and morphology of an
aneurysm, yielding adequate information for either surgical or interventional repair. Aortic dissections
are routinely depicted on MR studies and in many centers it is the first-line imaging modality for the
evaluation of known or suspected dissections and intramural hematomas.
Evaluation of the renal vascular system is accurate, fast, reliable, and routinely performed. While
suspected renal artery stenosis in the setting of uncontrolled or difficult to control hypertension is the
most common indication, there are many other uses in the renal vascular system. These include renal
stenosis causing renal insufficiency, fibromuscular dysplasia, renal artery aneurysms, renal
arteriovenous malformations and fistulas, and the evaluation of both potential renal donors as well as
transplanted kidneys.
The other abdominal arteries are also frequently imaged with MRA. These include the hepatic artery
and the superior mesenteric and celiac arteries. Suspected mesenteric ischemia is the most common
indication but other indications include the evaluation of suspected vasculitis, arterial malformations,
and the imaging of liver transplants and potential liver donors.
The peripheral or "run-off" arteries of the pelvis and lower extremities are accurately evaluated with
MRA. They present a challenge to image because the anatomic coverage needed to cover them, from
the renal arteries to the feet, can exceed 150 cm, far surpassing the maximal z-axis coverage of most
MR systems (normally 40-55 cm). Innovations such as elongated peripheral vascular coils and
"moving-table' multistation MRA have not only made high-quality imaging of the peripheral vessels
possible - it is now routine at many centers. Peripheral MRA has replaced diagnostic conventional
angiography in many hospitals and vascular surgeons and interventional radiologists currently rely on
1 of 2 20-04-2010 00:22
its accuracy to make therapeutic decisions on patients with peripheral vascular disease.
MRA has been compared favorably with other imaging techniques, including the recognized "gold
standard" of conventional angiography (CA). Individual comparison studies and specific clinical
validation will be discussed in the following sections.
© 2010 Elsevier
2 of 2 20-04-2010 00:22
RUNNING A SUCCESSFUL CLINICAL MAGNETIC RESONANCE

ANGIOGRAPHY SERVICE
Running a successful MRA service involves more than purchasing an MRI system with MRA
capabilities. It requires a combination of appropriate methods and techniques, good MR system
technology, up-to-date post-processing capability, optimal use of technicians and radiologists, and the
development of good lines of communications with referring physicians and surgeons.
The first step comes with actual performance of a good-quality exam that answers the clinical
question. Ideally, a state-of-the-art high-field MR system is used but practically speaking, most MR
systems of mid to high field strength are acceptable if sequence parameters are optimized. Most MR
systems made in the last 10 years have specific sequences available for either CEMRA studies or
nonenhanced studies with TOF or PC methods. The new scanners have faster acquisition times
because of shortened minimum attainable TEs and TRs, especially useful in CEMRA. In a practice that
has numerous available MR systems, the newest and most sophisticated system should be used to do
the majority of the more sophisticated MRA studies in order to optimize image quality.
Adequate training of technologists is essential for reliable exam quality. Many centers initially define
one or a small number of technologists as MRA specialty technologists ("MRA-techs"), to concentrate
experience and quicken the learning curve. Technologists can travel to academic or other centers for
formal on-site training, which they can bring back to their practice and spread to the other
technologists. As the "MRA-techs" gain experience they can train other technologists so that eventually
all technologists become equally proficient in MRA studies. Starting with a few technologists, however,
improves the quality of studies during the initial period when the number of exams performed may be
limited.
page 761
page 762
Investment in state-of-the-art hardware and software improves the quality and reliability of MRA
studies. For example, the use of a peripheral vascular coil has significant advantages for run-off
studies.10-14 Unfortunately, hardware like this is expensive and an adequate return on investment
cannot be guaranteed. Since the volume of MRA studies, even at the busiest centers, seldom exceeds
that which a single system can handle, only one MR system needs to be initially optimized for MRA.
Older and less capable scanners can perform less demanding work such as musculoskeletal and spine
imaging.
Like technologists, physicians should seek up-to-date training and knowledge in MRA. This information
is widely available through the literature, in commercial continuing medical education (CME) courses,
and by on-site "mini-fellowships" offered at many academic centers. Physicians should become
educated on state-of-the-art methods for performing and not rely on the technologist or MR system
vendor applications for this knowledge. Again, like the technologists, limiting the number of physicians
who read the MRA studies would initially concentrate the experience so that protocols for the critical
"early" cases during the growth phase of the service are determined correctly, data are acquired with
optimal quality, and the images are interpreted correctly.
Before setting the protocol for an MRA, the physicians must consider the study in the same way that
they would prepare for a conventional angiogram. A review of prior studies, surgical and clinical
history, and knowledge of the patient's anatomy (including the presence of stents and grafts) are
needed. A comprehensive study which demonstrates both the anatomy as well as possible inflow and
outflow problem areas is important if revascularization is a possibility. For example, in a patient
referred because of a suspected iliac stenosis, not only the iliac region but also the aorta need to be
evaluated for inflow anatomy and the run-off vessels of the legs scanned for downstream stenoses.
Often a request for a specific anatomic segment has to be modified, with referring physician approval,
into a different and more comprehensive exam. For instance, a run-off protocol which works well for an
average patient may not be sufficient for a particularly tall one, especially if one is trying to image from
the renal to pedal arteries. One might need to sacrifice imaging of one region or alternatively give an
additional contrast infusion to image that region.
1 of 2 20-04-2010 00:23
Once performed, studies are processed on 3D workstations (see Chapter 23). Using the MRI console
as a 3D processing station is not generally a good idea because it will reduce scanner throughput and
not be cost effective. There are numerous workstations available on the market today, ranging in price
from as little as $20,000 to well over $100,000. The MR vendors make workstations with some
proprietary software, all of which work very well for post-processing MRAs. Many third-party
workstations have either cost, ease of use or other advantages over the vendor-specific stations.
Traditionally, MR technologists have processed most of the MRAs performed. However, with the
growing complexity of the exams, as well as the many different vascular areas imaged today with
MRA, the amount of physician processing or on-line physician manipulation of technologist-processed
studies is growing. Even if they choose to allow technologists to process most of the data, physicians
should become comfortable with use of the 3D workstations to review cases. If physicians can sit
down and manipulate the data themselves on a 3D workstation their diagnostic confidence will improve
and they may make additional findings missed by a technologist who may not be as familiar with the
anatomy or pathology.
Output and communication of findings is an important component of MRA studies. Traditionally,

referring physicians only had access to typewritten reports describing the study and they were
required to obtain film-based images from the hospital or outpatient film library. Current picture archive
and communication systems (PACS) allow many referring physicians easy access to images either
within the hospital or remotely via web-based systems. In a non-PACS environment, some centers
provide high-quality paper prints of selected images from the studies with their written reports. This
allows physicians to immediately appreciate the pathology without the laborious process of getting the
printed and filed images from the film library. Paper prints on standard photo-quality paper can be
produced economically using off-the-shelf color PC printers.
© 2010 Elsevier
2 of 2 20-04-2010 00:23
PRINCIPLES OF BODY MAGNETIC RESONANCE ANGIOGRAPHY

The first clinical demonstration of MRA was published in 1985.304 It has only been within the last few years,
however, that MRA has become a widely applied diagnostic tool.
Numerous vascular imaging techniques have been described to date. To a large extent, the clinical success of MRA
is due to the advent of fast 3D gradient-echo imaging conjoined with vascular enhancement using a paramagnetic
contrast agent. The results are far more robust than with 2D TOF methods (Fig. 30-2). Contrast-enhanced 3D MRA
is routinely used nowadays for imaging of the aorta, peripheral arteries, carotid arteries, portal and systemic veins.
In some cases, it has supplanted X-ray angiography.
The basic idea of CEMRA is straightforward, i.e., shorten the T1 relaxation time of blood to the maximal extent
possible and image with a volumetric data acquisition that is insensitive to flow artifact. It is desirable to reduce the
T1 relaxation time of blood from its usual value of 1200 ms (at 1.5 tesla) to a value <50 ms. Blood with such a short
T1 will appear bright when imaged with a very short TR (e.g., <5 ms), whereas all other tissues, including fat,
appear dark. This condition is necessary for adequate image processing using a maximum intensity projection (MIP)
reconstruction. Imaging of veins is typically performed in a more delayed phase when the contrast agent has had
the opportunity to equilibrate.
Contrast Agents
page 762
page 763
Add to lightbox
1 of 14 20-04-2010 00:24
Add to lightbox
Figure 30-2 Peripheral MRA in a healthy subject. Comparison of 2D time of flight MRA (A), acquired in
approximately 5 minutes, with CEMRA (B), acquired in approximately 20 seconds. The CEMRA shows fewer
artifacts and superior detail.
Contrast-enhanced MR angiograms are acquired during the intravenous infusion of an extracellular gadolinium
chelate. Such agents only provide effective arterial enhancement during the first pass. A variety of contrast agents
313
(see Chapters 13 and 14) are under development for MRA, some with intravascular half-lives of the order of 30
314-332
minutes to many hours, and it is anticipated that at least some of these will come into routine use in the near
future. Potential benefits include: the ability to repeatedly image a vessel in multiple orientations, without the
limitation of imaging only during the first pass of contrast agent; ability to acquire data over many minutes, thereby
enabling MR angiograms to have much higher signal-to-noise ratios and spatial resolution; and improved imaging of
the portal and systemic venous systems. For instance gadofosveset trisodium (MS-325, EPIX Medical, Cambridge,
MA) consists of a gadolinium chelate with a strong affinity for albumin in human plasma, thereby prolonging its
residence time in the blood.315,316 This agent has several-fold higher relaxivity than standard extracellular gadolinium
chelates.
A blood pool CEMRA based on an iron particle currently in Phase II clinical trials (ferumoxytol, Advanced
Magnetics, Cambridge, MA) is illustrated in Figures 30-3 and 30-4. In addition to shortening T1, such agents may
prove useful because they also shorten T2 (r1=38 mM s and r2/r1~2.2 for ferumoxytol) and make blood appear
-1 -1
darker on T2-weighted images (Fig. 30-3B and C). In preliminary experiences using ferumoxytol, it has proven
essential to adjust the imaging protocol compared with using gadolinium chelates. If given undiluted, severe
susceptibility artifacts may occur. Moreover, the volume of contrast agent is only a few ccs, not sufficient for
prolonged enhancement during the first pass. Therefore, one can dilute the ferumoxytol (given at a dose of 4 mg
iron/kg bodyweight), typically by a factor of 4-8, in order to reduce the T2* effects during the first pass and
increase the administered volume. The agent is also taken up by phagocytic cells in lymph nodes, plaque, and other
regions; for vascular applications, these agents may eventually prove useful for characterization of vulnerable
plaque.
Acquisition Techniques and Parameter Selection

page 763
page 764
2 of 14 20-04-2010 00:24
Add to lightbox
3 of 14 20-04-2010 00:24
Add to lightbox
Add to lightbox
Figure 30-3 Blood pool CEMRA. A, Whole-body CEMRA obtained in under 10 minutes using ferumoxytol, an iron
oxide-based blood pool contrast agent (ferumoxytol, Advanced Magnetics, Cambridge, MA) with a half-life of 14
hours. B, Coronal single-shot FSE image of a patient with an abdominal aortic aneurysm shows intraluminal signal
due to slow flow. C, Administration of ferumoxytol eliminates the intraluminal signal because it shortens the T2
relaxation time of blood in addition to the T1 relaxation time.
The imaging technique is a spoiled 3D gradient-echo sequence with a very short TR and TE. Typical pulse
sequence parameters are given in the protocol appendix of this chapter. These include the use of a minimal TR and
TE, and large flip angle for RF excitation, e.g., TR/TE/flip angle = 3-5 ms/1-2 ms/20-40°. The flip angle can be
lower for very short TR in order to reduce the amount of power deposition, as otherwise limitations of the specific
absorption rate might preclude scanning or necessitate an increase in TR. Specific absorption rate (SAR) becomes
a particular challenge at 3 T, as discussed below. For body applications, slice thickness is of the order of 1-3 mm
and fields of view of the order of 30-40 cm, using a 128-256 × 256-512 acquisition matrix. Although the default
sampling bandwidth is typically of the order of 32 kHz, our experience is that higher bandwidths, of the order of 64
kHz or higher, are more effective in that they enable shorter TR and TE.
page 764
page 765
4 of 14 20-04-2010 00:24
Add to lightbox
Add to lightbox
Figure 30-4 Left calf venous malformation. A, Four phases from a TRICKS time-resolved first-pass CEMRA using
ferumoxytol demonstrates the leg arteries. B, Delayed imaging shows the venous malformation.
In order to achieve a very short TE, an asymmetric readout of the echo is typically applied. Additionally, it is
becoming routine to apply data interpolation (e.g., ZIP) to improve the quality of the MIP reconstructions.
Interpolation along the slice direction (e.g., ZIP 2) doubles the number of slices and effectively cuts the slice
thickness in half. One can use this property to improve the smoothness of the MIP images while keeping the scan
time constant or alternatively, cut the scan duration in half. Interpolation within the plane (e.g., ZIP 512) is helpful to
reduce partial volume averaging effects. Unfortunately, interpolation increases both the amount of storage space
occupied by the MRA images and the data reconstruction time.
Partial Fourier techniques shorten scan time at the expense of a reduction in the signal-to-noise ratio. At 1.5 T, we
routinely use a 3/4 Fourier acquisition for most of our breath-hold MRA studies, which speeds up the scan by nearly
25%. Acquisition of a reduced number of phase-encoding steps along the slice-select and in-plane directions, in
conjunction with an asymmetric application of the phase-encoding gradients, also can shorten scan time with only
minimal impact on spatial resolution.
Flow compensation (see Chapter 27) should be avoided for contrast-enhanced 3D acquisitions. It is better to use a
high bandwidth acquisition with a TE <2 ms, since this approach reduces signal loss from steady, pulsatile, and
turbulent flow, whereas flow compensation (which is helpful for TOF sequences) only eliminates artifacts from
5 of 14 20-04-2010 00:24
steady flow. The short TE also minimizes artifacts from magnetic susceptibility variations near clips or air-containing
bowel loops, and permits the use of a shorter TR so that scan time is reduced.
Timing for Contrast-Enhanced Three-Dimensional Magnetic Resonance

Angiography
page 765
page 766
Scan times can range from as little as 5-10 seconds for some carotid MRA to as long as 30 seconds for
high-resolution MRA of the peripheral arteries. The relative timing of contrast agent administration and data
acquisition is critical in order to ensure a consistent quality for MRA. There are several methods in use to determine
this relative timing, as discussed in greater detail in Chapter 29.
Test Bolus
A 1-2 cc bolus of contrast agent, immediately followed by a 10-20 cc saline flush, is given intravenously. Imaging is
repeatedly performed at 1-2 s intervals using an IR turboFLASH or fast T1-weighted spoiled gradient-echo
sequence (with presaturation to reduce the signal intensity of inflowing vascular spins). The time to peak arterial
enhancement (tp) for the test bolus is determined and then applied in the following formula:
delay time (from start of contrast agent infusion to start of 3D MRA acquisition) = tp + (duration of infusion)/2
- (duration of CEMRA acquisition)/2
One can simplify the formula by matching the duration of contrast agent infusion to the duration of the CEMRA
acquisition, so that delay time equals tp. Since the bolus spreads out as it passes through the venous system, the
contrast agent bolus persists within the target artery beyond the duration of the infusion. Therefore, one can
actually use a somewhat shorter duration for infusion than duration of acquisition without much effect on image
quality. Conversely, there is no point in infusing contrast for a longer duration than the acquisition. For instance, if
the scan time is cut in half using parallel imaging methods such as ASSET or SENSE (see Chapter 8), then the
contrast agent should be infused approximately twice as fast so that the duration of infusion is reduced
proportionately.
Automated Bolus Detection

One can position a "tracker" region of interest in the vessel, so as to detect the signal change that occurs with
arrival of the contrast agent (e.g., SmartPrep, CareBolus). Upon detection of the contrast agent, the measurement
control system immediately starts the 3D MRA acquisition.
Fluoroscopic Triggering
Our favored timing approach (on systems that possess the proper software) is to use a very fast, T1-weighted 2D
gradient-echo acquisition to image a vessel in real time. Upon arrival of the contrast bolus in the target vessel, an
elliptico-centric 3D acquisition is initiated. A drawback is that the method does not provide spare time to
hyperventilate the patient before the CEMRA begins. As a consequence, we routinely have the patient breathe in
and out several times while we wait for the target vessel to enhance, at which point the patient is instructed to
suspend respiration.
Best Guess
In the majority of patients (perhaps 80-90%), the contrast agent bolus will appear in the aorta at 20-25 seconds
after the start of infusion, assuming a satisfactory IV line has been positioned near the right antecubital fossa.
Delivery of contrast may be delayed when the IV line is on the left side, in the presence of venous obstruction. or
low cardiac output.
Time-Resolved Imaging
If one sequentially acquires 3D data with sufficient rapidity, then imaging during the peak of contrast enhancement
is assured. There are several approaches, including:
CEMRA using partial Fourier, minimum TR/TE, and a reduced number of slices and phase-encoding steps,
so that time resolution is of the order of 5-10 seconds; however, the penalty is a loss of spatial resolution
"keyhole" approaches such as time-resolved imaging of contrast kinetics (TRICKS), a method that acquires
305
the center portion of k-space more frequently than the outer portions (Figs. 30-4 and 30-5). The method is
particularly effective for peripheral and extracranial carotid MRA306,307
parallel imaging, a method that uses spatial information from the elements of a phased array coil to reduce
scan time
vastly undersampled isotropic projection reconstruction (VIPR), a projection reconstruction approach that
6 of 14 20-04-2010 00:24
308
skips some projections.
These methods are reviewed in greater depth in Chapters 7 and 8.
Contrast Agent Dosage

A single dose (i.e., 0.1 mmol/kg) of contrast agent is sufficient in most situations. For some high-resolution
applications, such as renal artery imaging, we use 0.15-0.2 mmol/kg. For venous imaging, we typically administer
0.2-0.3 mmol/kg of contrast agent since imaging is done during the equilibrium phase when the concentration of
agent in the blood is lower. Under certain circumstances one might repeat a contrast agent dose. For instance, we
typically perform peripheral MRA using a dual injection. A dose of 0.1 mmol/kg is administered for time-resolved
imaging of the calf/ankle/foot, followed by a moving-table acquisition during the slow infusion (e.g., 1 cc/s) of 0.2
mmol/kg of contrast agent. There are a few other circumstances when a repeat infusion is helpful (up to a
permissible total dose of 0.3 mmol/kg depending on the contrast agent), e.g., if the patient fails to properly suspend
respiration or there are other technical problems. A repeat infusion is also helpful for functional assessment, e.g.,
for thoracic outlet syndrome (Fig. 30-6); for this study, two CEMRAs are acquired with separate contrast infusions -
one with the arms elevated and a second with the arms down.
Phase-Encode Order
page 766
page 767
Add to lightbox
Add to lightbox
Figure 30-5 Multifocal peripheral vascular occlusions. A, Four phases from a TRICKS time-resolved first-pass
7 of 14 20-04-2010 00:24
CEMRA using gadolinium-DTPA distinguish arterial inflow from rapidly filling veins. B, Corresponding DSA.
Using a standard linear phase-encode order, it is desirable to maintain a nearly constant level of contrast
enhancement over the entire duration of the MRA acquisition. Alternatively, with the elliptical-centric phase-encoding
approach, one initially acquires the central phase-encoding lines in both the ky and kz directions. These phase-
encode lines determine image contrast. Afterwards, the outer k lines, which provide detail, are obtained. Only
vessels that are enhanced while the center of k-space is being acquired will appear bright. The main benefit of this
approach is that it permits selective imaging of arteries without venous overlap, even for lengthy acquisitions. The
drawback is that timing is more critical than for standard phase-encoding schemes - there is a greater risk of
missing the arterial phase of enhancement with the result being a non-diagnostic study.
Radiofrequency Coils
For most body MRA studies, the body coil transmits the RF energy and a phased array coil is used to detect the
MR signal. In addition, phased array coils are required for parallel imaging methods (see Chapter 8). Generally, the
size of the phased array coil should be matched to the body part (e.g., torso array coil for chest and abdomen,
peripheral coil for peripheral arteries, and so forth). Very small RF coils provide excellent signal-to-noise ratios near
the coil but with a restricted field of view; the sensitivity falls off rapidly with distance. In very obese subjects, it may
be more convenient to image with the body coil, compensating for the worsened signal-to-noise ratio by increasing
slice thickness and field of view.
Venography
page 767
page 768
Add to lightbox
8 of 14 20-04-2010 00:24
Add to lightbox
Figure 30-6 Patient with thoracic outlet syndrome. Volume-rendered images of CEMRA. Two injections of 0.1
mmol/kg were done, one with the arms down (A) and one with the arms above the head (B). The latter shows
functional occlusion of the right subclavian artery. (Courtesy of Larry Tanenbaum MD, Edison Imaging, NJ.)
Add to lightbox
9 of 14 20-04-2010 00:24
Add to lightbox
page 768
page 769
Figure 30-7 Contrast-enhanced MR venography. A, Portal hypertension. MIP from CEMRA of portal venous system
obtained approximately 50 seconds after infusion of 0.2 mmol/kg of gadolinium-DTPA. B, Using same technique
as (A), CEMRA in another patient showing portal vein occlusion with collateral formation.
MRA of veins can be acquired by direct or indirect methods (see Chapter 31). With the indirect method, a large
dose of contrast agent is given and delayed 3D imaging then done (Fig. 30-7). Both veins and arteries enhance,
which has the drawback that they superimpose on a MIP image (although arterial signal can be minimized by
subtracting the early phase from the late phase). With direct venography (Fig. 30-8A and B), a large volume (e.g.,
60 cc) of a solution of contrast agent diluted at least 1:20 in saline is administered over approximately 30 seconds
into a hand vein. Without dilution, the susceptibility artifact generated from concentrated gadolinium chelate is
sufficient to eliminate signal from both the infused vein and adjoining arteries (Fig. 30-8C).
Add to lightbox
10 of 14 20-04-2010 00:24
Add to lightbox
Add to lightbox
Figure 30-8 Direct venography. A, MIP from direct venography performed using 3 cc of gadolinium-DTPA diluted
1:20 in saline, administered at 2 cc/s into a right-hand vein. A tourniquet on the forearm helped to force contrast into
the deep venous system. B, Central venous occlusion shown by direct venography. C, Loss of signal from the
subclavian vein and adjoining artery on CEMRA caused by infusion of undiluted contrast agent.
Approximately 10 s after the start of the dilute infusion, a 3D acquisition is performed in a coronal plane. Only
directly infused veins are shown. Additional imaging with a 2D TOF sequence or high-dose CEMRA acquisition is
required to visualize the jugular veins and other veins that are not directly infused.
Peripheral Magnetic Resonance Angiography

page 769
page 770
As discussed further on, CEMRA is a viable alternative to X-ray angiography for imaging of the peripheral arteries,
particularly in patients with poor renal function or allergy to iodinated contrast agent. In some patients, MRA will
11 of 14 20-04-2010 00:24
detect run-off vessels that are missed by X-ray angiography. The main challenge for peripheral MRA is to cover a
large vascular territory, from renal arteries to pedal vessels, with adequate spatial resolution in a reasonably short
study time. Gated 2D TOF with tracking presaturation is moderately effective but excessively time consuming. A
moving-table approach with imaging at 2-3 stations during the slow infusion of contrast agent greatly reduces study
time. Subtraction of post-contrast from precontrast images eliminates the signal intensity of background tissues,
assuming that there is no patient motion between the scans.
High-Field Magnetic Resonance Angiography

Three tesla MRI systems are coming into standard use. They provide better signal-to-noise ratios (as much as
double) compared with 1.5 T systems. The higher signal-to-noise ratio directly benefits time-of-flight MRA. Because
T1 relaxation times increase with magnetic field strength, the change in T1 (and therefore in signal intensity)
produced by a contrast agent is higher at 3 T. This effect, along with the inherently greater signal-to-noise ratio,
improves image quality for CEMRA compared with 1.5 T. Alternatively, contrast agent dosage can be reduced. T2*
effects will be larger as well, resulting in more T2 effects for iron particles but also more problems with
susceptibility artifact in general.
There are several technical obstacles for high-field CEMRA. Partial Fourier methods at 3 T suffer more from
artifacts than at lower field strengths, presumably because of the greater off-resonance effects. The same
argument explains the greater tendency for artifacts with SSFP techniques (careful higher order shimming is needed
to avoid these), along with the fact that SAR limitations are more severe so that the minimum TR is lengthened.
Power deposition is approximately four times higher at 3 T compared with 1.5 T. Consequently, it can be difficult to
use a sufficiently large flip angle to maintain effective background suppression for CEMRA.
Nonetheless, with proper calibration of the SAR monitor and use of optimized pulse sequences, we have found that
high-quality contrast-enhanced 3D MRA images are routinely obtained at 3 T (see Fig. 30-30) that generally equal
or surpass the quality of studies at lower field strengths.
Low-Field Magnetic Resonance Angiography

Low field systems naturally suffer from reduced signal-to-noise ratios. Moreover, cost considerations may preclude
the availability of advanced MRA software, fast gradients or phased array coils. Nonetheless, state-of-the-art
systems permit TOF, SSFP, and contrast-enhanced 3D MRA. To compensate for the reduced signal-to-noise ratio,
the slice thickness and field of view should be increased. Although it would be desirable to acquire multiple signal
averages, this is not feasible for first-pass CEMRA. A reduced sampling bandwidth is helpful to improve the signal-
to-noise ratio for most low-field imaging procedures, but is problematic for CEMRA since it lengthens TE and
renders the images more sensitive to artifacts.
Artifacts
Contrast-enhanced MRA is quite robust but artifacts still occur (see Chapters 22 and 29). These can be
summarized as follows.
Signal dropout from clips. The artifacts worsen with increasing ferromagnetic content, static magnetic field
strength, voxel size, lower sampling bandwidth or longer TE.
Incorrect timing of acquisition relative to contrast agent infusion.
Wraparound artifact. This artifact is commonly encountered for CEMRA acquired in a coronal plane. For
renal artery imaging, elevation of the arms (e.g., with small pillows) out of the imaging volume is helpful.
Wraparound artifact also occurs along the slice direction, particularly if the patient is obese or a very short
RF pulse (which tends to have a worsened slice profile) is employed to keep the TR to a minimum. The use
of judicially placed presaturation pulses, or longer RF excitation pulse with optimized slice profile, can be
helpful. Oversampling (e.g., no phasewrap option) eliminates wraparound artifact but is problematic since it
increases scan time. Wraparound artifact is reduced as the field of view is increased but at the expense of
spatial resolution unless the acquisition matrix is similarly increased.
Insufficient dose of contrast agent, resulting in low signal-to-noise ratio. Moreover, severe artifacts may
occur if the duration of contrast agent infusion is much shorter (e.g., less than 60%) than the duration of the
data acquisition, since much of the data are then acquired when there is no vascular enhancement.
Susceptibility artifact from concentrated contrast agent. This problem occurs when an artery (e.g., the
subclavian artery) runs parallel to and nearby a vein containing undiluted contrast agent (see Fig. 30-8C). To
avoid this problem for subclavian artery MRA, the contrast agent can be infused into the opposite arm or into
a foot vein, or it can be diluted with saline prior to infusion.
Alternative Magnetic Resonance Angiography Methods

Single-shot balanced steady-state free precession techniques (e.g., FIESTA, trueFISP) make blood, fluids, and fat
12 of 14 20-04-2010 00:24
appear bright (Fig. 30-9). The approach is particularly useful as a scout sequence, since each image is acquired
sequentially in less than a second and an entire volume can be imaged within a breath-hold. Scout SSFP images
are helpful for depicting vascular anatomy, although they are less robust than contrast-enhanced 3D MRA.
page 770
page 771
Add to lightbox
Add to lightbox
Add to lightbox
Figure 30-9 Evaluation of pleural-based tumor by different pulse sequences. A, HASTE renders fluids bright and
blood dark. B, SSFP shows fluids, blood, and fat as bright tissues. C, Contrast-enhanced 3D FAME only makes
enhancing tissues and blood appear bright.
13 of 14 20-04-2010 00:24
Phase-contrast MRA (see Chapters 27 and 28) is the method of choice for measurement of flow; it can be
performed before or after administration of a contrast agent. It can be applied to detect turbulent flow in the renal
arteries (and other vessels), suggesting a pressure gradient across a stenosis.
Fat suppression improves the quality of MIPs by reducing the signal intensity of background tissue. However, the
background is usually dark anyway and the use of fat suppression lengthens scan time. Magnetization transfer (see
Chapter 7) is helpful for 3D time-of-flight MRA of the circle of Willis but is not usually indicated for body MRA.
Image-Processing Methods
These methods are covered in detail in Chapter 23. The MIP approach is most commonly used; it depicts the
brightest pixels along a series of user-defined views. Subvolume MIPs are particularly helpful for minimizing partial
volume averaging. Alternatively, one can use volume-rendering techniques which provide similar information. Volume
rendering, used more commonly for processing of CT angiograms, can occasionally provide better depiction of
vascular anatomy than MIP; proper adjustment of threshold levels is needed to avoid artifacts. Multiplanar
reconstructions (MPR), along with inspection of the source data, can provide more accurate depiction of stenoses
in some cases than the MIP approach and can also be used to reduce overlap with venous structures. Seed
growing methods extract an arterial tree from a complex mix of veins and arteries and have been used in research
settings in conjunction with equilibrium imaging using blood pool contrast agents.
© 2010 Elsevier
14 of 14 20-04-2010 00:24
APPLICATIONS
Chest Arteries
Imaging Technique
303
Magnetic resonance imaging and angiography accurately depict the thoracic blood vessels. Scout
images are acquired using balanced SSFP. Typical parameters include TR/TE/flip angle = 3 ms/1
ms/70°, 192 × 192 matrix, field of view 40 cm, axial orientation with 10 mm slices, single breath-hold
coverage of the entire chest. Next, a dark blood acquisition is performed. A single shot fast spin-echo
pulse sequence (e.g., HASTE), preferably with cardiac gating and dual inversion preparation to reduce
the intravascular signal intensity, is applied in an axial plane with TR/TE ~ R-R interval/60 ms. This
sequence renders the vessel lumen dark, making abnormalities of the vessel wall such as plaque or
intramural hematoma more conspicuous (Fig. 30-12).
page 771
page 772
For CEMRA, a single dose (e.g., 0.1 mmol/kg) of an extracellular gadolinium chelate is administered
intravenously at 2 cc/s; the pulse sequence is a spoiled 3D gradient-echo (e.g., SPGR, FLASH) with
typical parameters: TR/TE/flip angle ~ 4 ms/1 ms/30°, 160 × 256 matrix, zip 512, field of view 30-40
cm, slice thickness of 3 mm (zip 2 interpolation to 1.5 mm). Slice orientation is oblique sagittal for
studies of the aortic arch (i.e., aligned with the ascending and descending aorta as prescribed off an
axial scout image) or coronal for studies of the thoracic outlet vessels. If an oblique sagittal orientation
is used, then one can employ a rectangular field of view to reduce scan time; however, this may result
in an unacceptable amount of wraparound artifact for the coronal plane. The acquisition is repeated to
show patterns of delayed enhancement (e.g., late filling of false lumen of an aortic dissection).
Additional sequences are helpful in certain cases. For instance, cine imaging of the heart in a plane
aligned through the left ventricular outflow tract and aortic root detects the presence of aortic
insufficiency in type A dissection. Cine is necessary for congenital anomalies such as patent ductus
arteriosus and coarctation (see Chapter 38). Phase contrast shows vertebral artery flow direction in
patients with subclavian steal, measures flow in the false lumen of a dissection, and is essential for
measurement of the Qp/Qs ratio in patients with left-to-right shunts.
Thoracic Aortic Dissection and Aneurysm

Factors that weaken the aortic media increase wall stress, eventually resulting in aortic dilatation,
283
aneurysm or dissection. Risk factors include chronic hypertension; connective tissue disorders such
as Marfan's and Ehlers-Danlos syndrome as well as chromosomal abnormalities such as Turner's
syndrome, bicuspid aortic valve and coarctation, inflammatory disorders such as Takayasu's and giant
cell arteritis; deceleration injury, cocaine usage, and iatrogenic injury. Marfan's syndrome is the most
common etiology for dissection in patients under 40 years of age. The incidence of aortic dissection
ranges from 5 to 30 cases per million people per year with a strong male predisposition. Patients
typically present with chest pain, although other signs and symptoms related to complications such as
intestinal ischemia are common. The diagnosis can be missed in as many as a third of patients upon
initial presentation.292
According to the Stanford classification of dissection, type A (see Fig. 30-10) involves the ascending
aorta, whereas type B (Fig. 30-11) arises distal to the take-off of the left subclavian artery.284
Approximately 21% of acute (<2 weeks after onset of symptoms) patients with dissection die before
293
hospital admission. Acute type A dissections are a surgical emergency, with an early mortality rate
of 1% to 2% per hour.285 By comparison, the 30-day mortality rate for uncomplicated type B
dissections, which more often are managed conservatively, is only 10%; however, the mortality rate is
higher with complications such as leg ischemia or renal failure.
In most acute dissections an intimal flap develops and a tear is seen to connect the true and false
lumens.286 Extension may be antegrade or retrograde and involve side branches. Intramural
1 of 104 20-04-2010 00:28

hematoma, due to rupture of the vasa vasorum within the vessel media, is considered to be a
precursor of classic dissection.287 Like classic dissection, it may progress or regress; involvement of
the ascending aorta is considered a surgical indication. Penetrating atherosclerotic ulcers (Fig. 30-12)
infrequently progress to dissection; progression is more likely to result in aortic dilatation and aneurysm
288
formation.
The purpose of imaging is to define the presence and extent of the dissection, differentiate true and
false lumens, detect involvement of branch vessels, aortic valve or pericardium, evaluate left ventricular
function which may be compromised by decreased coronary artery flow, detect aneurysm formation,
presence of leakage or development of new lesions. Diagnosis can be made by a variety of tests
including plain films, X-ray angiography, transesophageal echocardiography (TEE), CT, and MRI.292
The relative utility for these imaging tests appears to depend on the prevalence of disease. For
high-risk patients, the positive predictive values are >85% for CT, MRI, TEE, and aortography. For
intermediate-risk patients, the positive predictive values are >90% for CT, MRI, and TEE but only 65%
for aortography. For low-risk patients (disease prevalence 1% or less), the positive predictive values
are <50% for CT scanning, TEE, and aortography but close to 100% for MRI. For all these patient
populations, the negative predictive values and accuracies are high, i.e., >85% for all four diagnostic
modalities.294
In experienced hands, TEE is highly accurate as a first-line test for acute patients. The main limitations
relate to its invasive nature (e.g., the presence of esophageal varices is a contraindication), lack of
objective documentation for following progression of the dissection over time, the presence of a blind
spot in the distal ascending thoracic aorta and proximal arch, and inability to follow the distal extent into
295
the abdomen. For stable patients CT is typically the first-line test, particularly for scanners with
multidetector capability. However, MR is the preferred test for patients who cannot tolerate iodinated
contrast or where radiation exposure is an issue. Contrast-enhanced MRA has proven superior to dark
blood imaging for detecting the presence or absence of intimal flaps and supra-aortic branch vessel
involvement, cine imaging demonstrates aortic regurgitation, and phase-contrast imaging permits
quantification of the degree of insufficiency.290 Flow is slower in the false lumen compared with the true
lumen. Phase-contrast MRI permits quantification of flow in the false lumen, which correlates with
291
aneurysm growth rate.
Anomalies of the Thoracic Vessels

Congenital anomalies of the heart and thoracic vessels are reviewed in Chapters 37 and 38. Both CT
and MRI are effective for depicting these anomalies in adult patients and often obviate the need for
diagnostic X-ray angiography.299 One example is congenital pulmonary venolobar syndrome, which
refers to a set of conditions including hypogenetic lung (including lobar agenesis, aplasia or
hypoplasia), partial anomalous pulmonary venous return, absence of pulmonary artery, pulmonary
sequestration, systemic arterialization of lung, absence of inferior vena cava, and accessory
296-298
diaphragm. An example of intralobar pulmonary sequestration with an anomalous vessel arising
off the descending thoracic aorta is given in Figure 30-13 and some other great vessel anomalies are
illustrated in Figures 30-14 to 30-17. Cine imaging helps to detect flow jets when a pressure gradient
exists across an anomalous vessel (see Figs. 30-15 and 30-16).
page 772
page 773
2 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
3 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
page 773
4 of 104 20-04-2010 00:28

page 774
Figure 30-10 Type A dissections of thoracic aorta. A, Axial HASTE shows the flap; the false lumen
appears brighter than the true lumen due to slower flow. B, Oblique sagittal image from CEMRA
shows the origin of the flap near the aortic root. Some blurring occurs near the heart, so that cine
sequences may be helpful. C, HASTE image in different patient shows the flap (arrow). D, CEMRA
shows the entry point (arrow) between the true and false lumens.
Add to lightbox
5 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
page 774
page 775
Figure 30-11 Type B dissections of the thoracic aorta. A, Volume-rendered image shows the extent of
the dissection (arrows) (courtesy of Larry Tanenbaum MD, Edison Imaging, NJ). B, In a different
patient, CEMRA better shows the true lumen and flap originating just distal to the left subclavian artery.
The individual image shows the flap better than MIP or volume-rendered views, where the flap may be
partially obscured. C, Delayed CEMRA shows late filling of the false lumen.
6 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
7 of 104 20-04-2010 00:28

Add to lightbox
8 of 104 20-04-2010 00:28

Add to lightbox
page 775
page 776
Figure 30-12 A,B, Penetrating ulcer. A, Axial HASTE image shows a complex appearance to the
penetrating ulcer, due to blood flow, plaque, and hemorrhage of differing ages. It is difficult to
distinguish signal from slow flow and thrombus. B, CEMRA better delineates the full extent of the ulcer
(arrow). C,D, Patient with postoperative mediastinitis. MIP (C) and individual image (D) show
penetration of anterior wall of ascending aorta with pseudoaneurysm formation.
Add to lightbox
Figure 30-13 Intralobar pulmonary sequestration. A 19-year-old male presented with recurrent left
lower lobe pneumonia. CEMRA shows an anomalous vessel (arrow) arising off the inferior portion of
the descending aorta.
9 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
page 776
page 777
Figure 30-14 Double aortic arch. A, Maximum intensity projections. B, The vascular ring is better
shown on a delayed axial 3D FAME acquisition.
10 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
Figure 30-15 Patent ductus arteriosus. A, CEMRA shows a connection (arrow) between the inferior
aspect of the aortic arch and the left pulmonary artery. B, Cine imaging demonstrates a low-intensity
flow jet (arrow) in the patent ductus.
11 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
page 777
page 778
Figure 30-16 Aortic coarctation in an adult. A, CEMRA demonstrates the narrowed isthmus and
collaterals. B, Cine imaging shows a flow jet (arrow) suggesting turbulent flow with a pressure gradient
(courtesy of Scott Pereles MD, Northwestern Memorial Hospital).
12 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-17 Pseudocoarctation of thoracic aorta. The aorta is kinked (arrow) near the origin of the
ligamentum arteriosus. Unlike a true coarctation, there is no pressure gradient across the buckled
segment, nor any stenosis or collateral circulation.
Subclavian Steal
Subclavian steal is characterized by reversal of flow in a vertebral artery associated with a proximal
obstruction of the subclavian artery. The condition can be diagnosed using a combination of CEMRA to
show the subclavian artery obstruction and vertebral artery origins, along with an axial 2D phase-
contrast acquisition (sensitized to flow in the slice direction) to demonstrate the direction of flow within
302
the vertebral arteries. Alternatively, presaturation pulses can be alternately applied above and below
the slice to show flow direction (Fig. 30-18).
Thoracic Outlet Syndrome

The signs and symptoms of thoracic outlet syndrome are caused by neurovascular compression
provoked by structural abnormalities and/or postural maneuvers.300 Diagnosis is facilitated by a
301
combination of clinical and imaging criteria. Both CT or MRA (see Fig. 30-6) have the ability to
demonstrate functional abnormalities of the thoracic outlet; CT is preferred because of its superior
ability to depict bony abnormalities such as cervical rib and greater ease of patient positioning for
provocative maneuvers.
Abdominal Aorta
General Approach
Prince et al initially described CEMRA for imaging the abdominal aorta. 7 By today's standards, their
early technique appears inefficient; a 5-minute long injection of 0.2 mmol/kg gadolinium chelate, no
breath-holding, and a 4-5 minute acquisition time. Despite this, image quality was dramatic and their
conclusion that "dynamic imaging during the injection of gadopentetate dimeglumine is a promising
7
technique for evaluation of the abdominal aorta and branch vessels" was unmistakably correct.
13 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
Add to lightbox
page 778
page 779
Figure 30-18 Subclavian steal syndrome. A, CEMRA shows a proximal occlusion of the left subclavian
artery. The distal artery is reconstituted by reversed flow through the left vertebral artery. B, Axial 2D
TOF image through the neck shows flow in both vertebral arteries. C, With superior presaturation, the
left vertebral artery (arrow) disappears, indicating reversed direction of blood flow.
A torso phased array coil helps boost signal to noise and can reduce acquisition times. A typical
abdominal MRA protocol consists of a combination of several basic pulse sequences. The abdomen is
14 of 104 20-04-2010 00:28

scouted with a fast sequence; a single-shot fast spin-echo (half-Fourier SSFSE) scout in the coronal
plane can be used as both an anatomic scout and a diagnostic T2-weighted sequence useful for
characterization of incidental pathology. Axial pre- and post-contrast T1-weighted gradient-echo (GRE)
images are used to depict the entire vessel wall (including plaques and thrombus that are incompletely
evaluated on the CEMRA) as well as visceral organs and the retroperitoneum (Fig. 30-19). A
multiphase CEMRA completes the routine protocol.
The CEMRA study is optimally acquired in the coronal or coronal oblique plane, which best depicts the
abdominal aorta and all its major branches. The renal, celiac, superior mesenteric, inferior mesenteric,
and iliac arteries are all included in the imaging volume. Care must be taken when positioning the 3D
slab so that the anterior wall of the aorta (or of an aortic aneurysm) is not inadvertently excluded from
the study. The use of a rectangular field of view (though not in a coronal plane, where wraparound
artifact will occur) and parallel imaging techniques will speed data acquisition; this can be used to
reduce the breath-hold duration, increase resolution by acquiring a larger number of phase-encoding
lines, or a combination of both.
Add to lightbox
15 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-19 Abdominal aortic aneurysm (AAA) depicted on 3D VR CEMRA (A) and 2D multiplanar
GRE imaging (B) performed following gadolinium administration. While the 3D study (A) depicts the
morphology of the aortic lumen and branch arteries accurately, the 2D GRE study gives a more
precise depiction of the aortic wall and peripheral thrombus (arrow).
If atheromata, intramural hematoma, aortitis or dissection flaps are suspected, optional additional
sequences are used, including T2-weighted FSE with multiple inversion pulses or cine bright blood
imaging (SSFP or segmented GRE). CEMRA images accurately depict the vascular lumen and are
used to diagnose areas of stenosis or aneurysm. Other sequences, including gradient-echo, spin-echo,
inversion recovery, and SSFP imaging, are used to further investigate abnormalities of the aortic wall
and of adjacent structures.
A quick review of the patient's chief complaint and medical history prior to the exam will insure that an
adequate study is performed and that rescanning is minimized. The vast majority of patients are
adequately imaged with the basic protocol. If there is suspicion of a dissection, large atheromata or
intramural hematoma, then the additional black and bright blood techniques are used.
Abdominal Aortic Aneurysms
The vast majority of aortic aneurysms are caused by atherosclerosis. Less common causes are
infection, inflammation, syphilis, and cystic medial necrosis. Atherosclerotic aneurysms are typically
fusiform, although focal eccentric aneurysms due to atherosclerosis are occasionally encountered.
One of the most common indications for abdominal MRA is the evaluation of known or suspected
abdominal aortic aneurysms. A MRA can provide a substantial amount of information about the
abdominal aortic aneurysm.15,16 Important imaging features of aortic aneurysms are the maximum
diameter, the length, angulation, and involvement of major branch vessels. All these features can be
identified and characterized with MR imaging.
page 779
page 780
The infrarenal abdominal aorta is the most common site of atherosclerotic aneurysms. CEMRA allows
assessment of involvement of the iliac, mesenteric, and renal arteries in addition to the aorta itself. In
addition to atherosclerotic aneurysms, mycotic aneurysms arise when a focal bacterial infection
weakens the vessel wall. This results in a saccular outpouching (see Fig. 30-12C and D). This usually
16 of 104 20-04-2010 00:28

involves the suprarenal portion of the aorta. While CEMRA demonstrates the aneurysm itself,
post-contrast T1-weighted GRE imaging may demonstrate enhancement in and around the vessel wall
itself.
The vascular lumen is best depicted on black blood T2-weighted inversion recovery (IR), bright blood
cine or CEMRA. Accurate lumen dimensions and depiction of larger atheromata, ulceration, and other
filling defects can be provided for on any of the three standard sequences. Imaging of the vessel wall
itself requires a 2D sequence; normally the IR or cine images are used. Thrombus is generally low in
signal intensity unless there has been an acute bleed into the thrombus itself.
Patients with aneurysms who require surgical resection often need to have the artery of Adamkiewicz
imaged and identified. This is a small vessel that can be the sole source of blood supply to portions of
the spinal cord. The artery of Adamkiewicz anatomically ascends from the dorsal branch of the T8 to
L1 intercostal or lumbar artery to the anterior midsagittal surface of the spinal cord. In some cases,
ligation of the artery during surgery can cause ischemic cord injury, resulting in paralysis. To prevent
ischemia, intercostal or lumbar arteries that serve as the origin of the artery of Adamkiewicz are
reattached to the aortic grafts to preserve cord blood flow. In the past, conventional angiography has
been used to attempt to identify this critical vessel. Recently CEMRA techniques were used
17
successfully to preoperatively identify the vessel.
Endovascular stent graft implantation is an alternative to conventional open surgery in the treatment of
aortic aneurysm, especially in high-risk patients. Use of endovascular grafts results in less blood loss,
shorter stays in intensive care units and subsequent hospitalization, and quicker recovery. Patients
being considered for this therapy and those who have had this performed both benefit from noninvasive
MR or CT angiography.
Initially, conventional digital subtraction angiography (DSA) or CT angiography (CTA) was used in the
preoperative assessment of the abdominal aorta in patients being considered for endovascular repair
with a stent graft. Recently, MR has been shown to be a useful tool for the preoperative assessment
of patients who are being considered for endovascular repair.18-21 MR imaging accurately determines
the size of the aneurysm, length, amount of thrombus, relationship with branch vessels, degree of
angulation, and the size and status of access vessels (common femoral and iliac arteries). All of these
are critical in the determination of a candidate's suitability for graft placement.
22
Following graft placement, MR can be useful in the evaluation of potential complications. Common
complications include endoleak, graft thrombosis, and graft kinking; more rare complications include
pseudoaneurysm caused by graft infection, graft occlusion, shower embolism, perforation of mural
thrombus by means of inadvertent penetration of the delivery system, colon necrosis, aortic dissection,
and hematoma at the arteriotomy site.
Endoleak, or leakage of blood (and contrast) through the graft into the native aneurysm, is the most
common complication after stent graft implantation. Rates of leakage after endovascular repair of
23
aortic aneurysm are 2.4% to 45.5%. Leakage is classified according to the site of origin as proximal,
distal or middle graft.24,25 The cause of proximal or distal endoleak is incomplete fixation of the stent
graft to the aortic wall while middle graft endoleak is caused by graft defects or retrograde blood flow
via patent arteries. Most endoleaks can be treated successfully using endovascular techniques without
resorting to surgical explantation of the device.
CT has been the standard of reference for detecting and classifying endoleaks; 24 findings on the basis
of the configuration and location of the leakage in relation to the stent graft and the aneurysm indicate
the cause of leakage. MR angiography is an alternative imaging modality to CT because the
development of improved techniques now permits satisfactory visualization of both the arterial and
venous vascular systems. For patients with renal insufficiency an important advantage of MR imaging
is the low toxicity of its contrast agents. Since some patients may require many follow-up imaging
tests, especially if the graft is placed in a young patient, the cumulative amount of radiation exposure
associated must be considered, another clear advantage for MR.
17 of 104 20-04-2010 00:28

MR angiography has been shown to be more sensitive than CT angiography for detection of type II
endoleak.26 In a series of 52 patients with angiography as the gold standard, MR angiography had no
false-negative and 13.5% false-positive results for endoleak detection. The use of persistent blood
pool contrast agents may further enhance the accuracy for endoleaks, because one can delay imaging
282
for many hours to allow sufficient time for extravasation of the contrast agent.
Aortitis
The aorta and other arteries are affected by a diverse group of conditions characterized by
inflammation and necrosis of vessel walls. They have many causes and pathogenetic mechanisms,
although these are not completely understood. The American College of Rheumatology published an
approach to classifying vasculitides, defining seven clinically distinct diseases.27 These included
polyarteritis nodosa, Churg-Strauss syndrome, Wegener's granulomatosis, hypersensitivity vasculitis,
Henoch-Schönlein syndrome, giant cell arteritis, and Takayasu's arteritis. The disease affects vessels
of different sizes; giant cell arteritis and Takayasu's arteritis affect large vessels such as the aorta and
the largest arterial branches directed toward major body regions.
page 780
page 781
Add to lightbox
Add to lightbox
18 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-20 Takayasu's arteritis in a 32-year-old female with multiple vascular stenoses. CEMRA
depicts a renal artery stenosis (arrow) (A), SMA and celiac axis narrowing (arrow) (B), and a
high-grade near occlusion of the distal abdominal aorta and proximal iliac arteries (arrow) (C).
Conventional X-ray angiography has traditionally played a prominent role in the radiologic evaluation of
large vessel vasculitis. Recently, MR methods have received much attention for their potential to
complement and possibly supplant conventional angiography. In particular, promising advances have
been made in the areas of MR angiography and physiologic imaging. MR provides high-resolution
anatomic information, including lumen configuration and vascular wall thickness, and physiologic data,
such as measurements of the degree of wall enhancement and the presence of edema.
Takayasu's arteritis is a disease of the large arteries involving the aorta, the pulmonary artery, and the
major branches of the aorta (Fig. 30-20).312 It typically affects young women and decreased or absent
pulses in the arms is a frequent manifestation of the disease. The diagnosis of the disease is difficult
because only nonspecific systemic symptoms, such as fever and general weakness, may be present
during the early phase and because symptoms of arterial stenosis or occlusion are late-phase
manifestations. The current diagnostic criteria focus mainly on the late manifestations of Takayasu's
arteritis such as stenosis or occlusion of the aorta and its branches. 28
Although Takayasu's arteritis classically involves the aortic arch, in 32% of cases it also affects the
remainder of the aorta and its branches and in 12% it is limited to the descending thoracic and
abdominal aorta.29 In most cases, the gross morphologic changes consist of irregular thickening of the
aortic wall with intimal wrinkling. In general, the possibility of underlying arteritis should be considered
when arteriosclerotic changes in the aorta are seen in young or middle-aged patients, especially Asian
and female patients, and when these changes are either segmental or occur at an unusual site. The
diagnosis is confirmed by a characteristic arteriographic pattern that includes irregular vessel walls,
stenosis, post-stenotic dilatation, aneurysm formation, occlusion, and evidence of increased collateral
circulation.
MRI shows stenosis and dilatation of large arteries, as well as thickening, edema, and contrast
enhancement of the vessel wall.312 In a group of patients with Takayasu's arteritis, CEMRA showed
increased enhancement of the thickened aortic wall as compared with normal myocardium, suggesting
30
active disease. Determination of disease activity using contrast-enhanced MR imaging was
concordant with clinical findings in 23 patients (88.5%). Contrast-enhanced MR findings were
concordant with laboratory findings (elevated erythrocyte sedimentation rate and C-reactive protein) in
most patients.
Aortic Ulceration and Intramural Hematoma

Penetrating atherosclerotic ulcer is characterized by ulceration of an atherosclerotic plaque that
penetrates through the intima into the media of the aortic wall.31 It typically affects elderly individuals
19 of 104 20-04-2010 00:28

with hypertension and extensive aortic atherosclerosis. It is seen as an outpouching extending beyond
the contour of the aortic lumen and can become quite large. A penetrating atherosclerotic ulcer is
typically located in the descending aorta and can be associated with a variable degree of hematoma
within the aortic wall.32 Diagnosis can be difficult when the presentation overlaps with that of atypical
aortic dissection. Placement of an endovascular stent graft is becoming a popular method of treating
this entity, given that the disease tends to occur in elderly patients with co-morbid conditions that put
them at high surgical risk.
page 781
page 782
Add to lightbox
Figure 30-21 Aortic ulceration in a 74-year-old male who presented with emboli to the lower
extremities. CEMRA depicts diffuse aortic atherosclerosis, a distal aortic aneurysm, and a shallow 4
cm ulcer along the posterolateral aorta (arrow).
The MR appearance was initially described by Yucel in 1990.33 MR findings include intramural
hematoma and focal aortic wall ulceration that may be associated with aortic rupture. Intramural
hematoma has increased signal intensity on T1- and T2-weighted MR images. MR imaging is superior
to angiography in depicting the extent of intramural thrombus because it can image the aorta in cross-
section, rather than simply depicting the internal lumen. The actual crater in cases of aortic ulceration
can be best depicted on 3D views using CEMRA (Fig. 30-21).
Atherosclerotic ulcers may progress to form an intramural hematoma of the aorta. Intramural
hematoma is an atypical form of dissection without flow in the false lumen or a discrete intraluminal
flap. Intramural hematoma may also develop as a result of blunt trauma with aortic wall injury or in
hypertensive patients. It is thought to begin with the rupture of the vasa vasorum, the blood vessels
that penetrate the outer half of the aortic media from the adventitia and supply the aortic wall itself.
20 of 104 20-04-2010 00:28

The hematoma propagates along the media layer of the aorta and weakens the aorta - this may
progress either to outward rupture of the aortic wall or to inward disruption of the intima, the latter
leading to an aortic dissection.34-36
It can be identified as crescentic thickening of the aortic wall with increased intramural signal intensity
on T1-weighted images. Precontrast fat saturation images can be helpful in differentiating intramural
hematoma from surrounding mediastinal fat. Intramural hematoma most frequently involves the
ascending aorta but can also be seen in the thoracic and, less commonly, the abdominal aorta.
Intramural hematoma is distinguished from mural thrombus by identification of the intima. Mural
thrombus is located on top of the intima, which is frequently calcified, while intramural hematoma is
below the intima. Within a dissection, MR imaging can aid in the distinction between slow flow in the
false lumen and no flow in an intramural hematoma. Fast cine GRE or SSFP images show cyclical
flow-related enhancement in the false lumen of aortic dissection, whereas images of intramural
hematoma show no signal intensity change. Dynamic phase-contrast MR imaging is more sensitive
than gradient-echo sequences for excluding slow flow in the thickened aortic wall that would indicate
aortic dissection rather than intramural hematoma.37
A retrospective analysis of 65 symptomatic patients with aortic intramural hematoma showed that it is
very important to determine if the hematoma is associated with a penetrating aortic ulcer or not.32
Forty-eight percent of patients with associated penetrating aortic ulcers had significant disease
progression compared with only 8% of those without an associated ulceration. Disease progression
was also correlated with the maximum diameter and maximum depth of the penetrating ulcer.
Aortic Dissection
Aortic dissections arising in the ascending (Stanford type A) or descending (Stanford type B) thoracic
aorta commonly extend into the abdominal aorta. Occasionally, a patient with a dissection limited to
the abdominal aorta will present; these may be spontaneous, post-traumatic, or following interventional
catheter-based procedures. In either case, evaluation of the extent of the dissection within the
abdomen, to include an evaluation of the renal, splenic and iliac vessels, is critical for diagnostic and
therapeutic purposes.
MR evaluates the extent of the dissection, the sizes of the true and false lumina, the patency of the
false lumen, and branch vessel involvement. Normally, imaging extends from the arch to the aortic
bifurcation to obtain a comprehensive view of the entire aorta. CEMRA is a rapid and accurate imaging
modality in diagnosis of dissections.38 CEMRA can sometimes be used as the sole MR technique if
patients are unstable or cannot tolerate a prolonged exam. However, a combination of black blood,
bright blood cine and CEMRA images is used for a comprehensive evaluation of the aorta when time
permits.
page 782
page 783
21 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
Add to lightbox
Figure 30-22 Dissection of the abdominal aorta with extension into the branch arteries. MIP image of
22 of 104 20-04-2010 00:28

a CEMRA (A) depicts a flap in the central portion of the aorta (arrow). Axial MPRs of the aorta depict
extension of the low signal intensity flap into the SMA (arrow) (B) and left renal artery (arrow) (C).
Approximately 5% of patients with aortic dissection develop mesenteric ischemia as a complication of

the dissection process.39,40 MRA is an excellent modality to assess these patients and 2D TOF and
gadolinium-enhanced techniques are used successfully in this setting. 41 MRA classifies the dissection,
defines entry and re-entry points, differentiates thrombus from slow flow, and evaluates branch vessel
involvement (Fig. 30-22). Isolated dissections of the SMA, either in association with cystic
degeneration or as a complication of catheter angiography, are rare.42
Because of the extensive anatomy required to perform a comprehensive aortic exam, a section
thickness of 8-10 mm with gaps may be necessary to achieve this coverage in a reasonable time. Axial
T1-weighted images are often obtained with cardiac gating and breath-holding to prevent linear
artifacts mimicking the intimal flap. However, a reliance on ECG triggering may result in long
examinations and poor image quality in patients with irregular cardiac rhythms. Half-Fourier rapid
acquisition with relaxation enhancement (RARE) sequence provides black blood imaging with
T2-weighted contrast that is less sensitive to cardiac and respiratory motion and is therefore useful for
43
the evaluation of aortic dissections.
Post-processing of the CEMRA data has traditionally been done using a MIP algorithm. However, MIP
images may not show the intimal flap because the image is based upon the high signal intensity value
voxels, not the presence of signal void as seen with a flap or other filling defect. Therefore, review of
raw data and multiplanar reformatted (MPR) images is required for visualization of the intimal tear and
re-entry sites. One additional pitfall of CEMRA is that it may be relatively insensitive to the presence of
38
intramural hematoma. Additional cross-section sequences should therefore also be reviewed in these
cases.
Renal Arteries
Renal MRA is now one of the most widely utilized MR angiographic applications.11,44-90 It allows for
accurate evaluation of patients suspected of having renal artery stenosis, the most common curable
cause of hypertension. Other uses of renal MRA include depicting anatomy of prior surgery or
intervention, evaluation of renal transplant vessels and of potential renal donors, staging vascular
involvement by renal tumors, and evaluating renal arteriovenous malformations. Numerous pulse
sequences are available to provide information regarding kidney morphology, arterial anatomy, blood
flow, and renal function and excretion. CEMRA is routinely used as part of a comprehensive approach
to renal MR angiography.
page 783
page 784
23 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
24 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-23 Bilateral renal artery stenosis with conventional angiographic correlation (CA). Volume-
rendered CEMRA (A) demonstrates high-grade bilateral renal stenosis (arrow). Selective left (B) and
right (C) conventional angiograms confirm the stenoses (arrow) just prior to intravascular stent
placement.
Renal artery stenosis is the leading cause of curable hypertension. Estimates suggest that the
prevalence of renovascular disease as a cause of hypertension ranges from 0.5% to 5% in the general
91-93
population to as high as 45% in selected patients with suggestive clinical features. Renal artery
stenosis leading to relative renal ischemia may cause an elevation of blood pressure that is often
difficult to control with medical therapy. Eventually, the stenosis progresses in severity and leads to
occlusion and a permanent reduction in renal function.
Other techniques, including ultrasonography (US) and computed tomography (CT), have been
considered promising for noninvasive renal vascular imaging. However, US for renal stenosis is limited
by the need for experienced ultrasonographers, has a relatively high failure rate due to large body
habitus and other implements, and controversy over the interpretation of the flow information. CT
angiography, especially with multidetector row scanners, is also promising but is limited somewhat by
the risks associated with iodinated contrast material and ionizing radiation and can be degraded by
artifacts associated with densely calcified plaques.
MRA is an accurate and safe technique for evaluating the main renal artery and vein. Renal CEMRA
has reported sensitivities and specificities of more than 90% for detection of renal artery stenosis
(>50%) as compared to conventional angiography (Fig. 30-23).* Grading of renal artery stenosis is
commonly done using a qualitative, broad classification scheme (normal, mild, moderate, high-grade,
severe, occluded) (Fig. 30-24). Current resolution limitations preclude routine determination of a
diameter stenosis percentage as performed elsewhere. In addition to an evaluation of morphologic
stenosis, the hemodynamic and functional significance of the stenosis has also been assessed,
although with mixed success.66,75,94,96,97 Newer techniques have also used MR to perform a
quantitative evaluation of renal function. 66,67,81,97-99
Techniques
CEMRA is the preferred noninvasive angiographic imaging method for the renal arteries.
Non-gadolinium based MR techniques for imaging the renal arteries, such as time of flight or phase
contrast, have had limited success.100-111 Limitations to these techniques include signal loss due to
turbulence at the stenoses, in-plane saturation, non-visualization of small vessels such as accessory
25 of 104 20-04-2010 00:28

arteries and the distal renal arteries themselves, and poor quality due to slow flow in patients with
cardiac disease or aortic aneurysm or older patients.
45,48,50,51,53,56,57,59,60,62,64,70,74,76,78,86-88,94,95
*(See references )
page 784
page 785
Add to lightbox
Figure 30-24 Spectrum of renal artery stenosis as depicted on CEMRA. The commonly used grading
classification on MR for renal artery stenosis includes (from top left) normal, mild, moderate, high
grade, and severe stenoses.
Renal CEMRA performed following a rapid bolus of gadolinium with a 3D gradient-echo sequence is
reliable and relatively easy to perform given the automation and speed of available MR systems today.
Renal MR has a high sensitivity and a high negative predictive value, making it an ideal screening
examination for the detection of renal vascular disease.* Unlike iodinated contrast agents, gadolinium
chelates can be used safely, even at high doses, in patients with renal failure.112 CEMRA depicts the
renal arteries along with the entire abdominal aorta, iliac arteries, and mesenteric arteries in a 10-30-
second acquisition performed during a breath-hold. The renal vein and inferior vena cava can be
evaluated by repeating the examination during the venous and equilibrium phases.
CEMRA gives a morphologic image of the renal artery but says little about the flow hemodynamics and
other quantitative factors that affect renal function. It is desirable to determine the hemodynamic
significance of an identified stenosis so that the potential benefit of revascularization can be assessed.
Many MR techniques have been proposed for evaluating the hemodynamic significance of renal artery
stenosis but no one technique has yet to be accepted clinically. The reported techniques include
measurement of blood flow with 2D cine PC, depicting turbulence-induced spin dephasing with 3D PC,
evaluating temporal enhancement, differential excretion of gadolinium, and the effect of angiotensin-
converting enzyme (ACE) inhibition on flow measurement with MR imaging or gadolinium clearance
47,75,79,94
rates. Many of these techniques are difficult to implement, require proprietary software or are
unreliable for routine clinical use. The 3D PC pulse approach has probably been used most widely but
is still not routinely performed at many centers.
Imaging Protocol
26 of 104 20-04-2010 00:28

A 3D spoiled gradient-echo (SPGR) volume, which includes the entire abdominal aorta, renal arteries,
and proximal iliac arteries, is used. The parameters for this sequence are optimized to attain the
highest quality images. In general, faster is better for data acquisition with 3D CEMRA and minimum
available TR and TE values are usually selected. Too much spin dephasing can cause signal loss at
stenoses, although this can be reduced or eliminated by selection of a TE less than 2 ms. Faster data
acquisitions allow the gadolinium contrast material to be injected with a faster injection rate, producing
a higher arterial gadolinium concentration and optimized enhancement of the renal arteries, branch
vessels, and accessory arteries. The high arterial signal-to-noise ratio may then compensate for
reduced T1 relaxation and signal averaging.
45,48,50,51,53,56,57,59,60,62,64,70,74,76,78,86-88,94,95
*(See references )
page 785
page 786
Add to lightbox
27 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-25 Accessory renal arteries. A, Accessory left renal artery arising from the left common iliac
artery (arrow). The imaging volume normally includes the entire abdominal aorta and proximal iliac
vessels in order to depict all accessory renal arteries. B, Crossed fused renal ectopia. CEMRA shows
a single large kidney supplied by multiple renal arteries.
Fast data acquisition minimizes motion artifacts and makes it easier for patients to successfully
suspend breathing for the entire data acquisition window. In addition to minimizing TR and TE, the
smallest number of sections sufficient to cover the arterial anatomy are normally prescribed to keep
the acquisition time to a minimum. Widening the bandwidth also makes the acquisition faster but may
also reduce signal-to-noise ratio. The signal of background tissue is also reduced or eliminated by
obtaining a 3D data set "mask" before gadolinium administration to use for digital subtraction. 113
The flip angle of the 3D sequence is optimized for T1 contrast on the basis of the repetition time and
expected blood gadolinium concentration. In practice, a flip angle of 30-45° works well in nearly all
cases. The flip angle could be larger for higher doses of contrast material and longer repetition time or
smaller for lower doses of contrast material and shorter repetition time.
Prescribe the image volume to include the abdominal aorta and the anterior two-thirds of the kidneys
by using 1-3 mm thick sections. Zero filling in the slice direction (zip 2 interpolation) is useful because it
doubles the number of sections, improving MIP quality without increasing imaging time. A field of view
(FOV) of 28-36 cm, usually just smaller than the width of the patient, minimizes the amount of
wraparound artifact but maintains adequate resolution. It is also useful to elevate the arms with
cushions or elevate them over the chest or head to prevent wraparound of the arms into the imaging
volume. The field of view usually includes the proximal iliac arteries because accessory renal arteries
may arise there (Fig. 30-25).
Correct timing co-ordination of the bolus injection with peak arterial enhancement during acquisition of
central k-space data is essential for good-quality studies. As discussed earlier, best guess is the
simplest, automated detection is the most "hands off",114 and a timing run is most reliable.115
Fluoroscopic triggering offers the most robust and fastest method for achieving optimal
28 of 104 20-04-2010 00:28

enhancement.53,77,116 We routinely use approximately 30 cc per adult patient; this dose represents
0.15-0.25 mmol/kg for an average adult in the United States. Contrast is usually bolus infused at a rate
of 1-2 cc/s and is followed immediately by a 10-15 cc saline flush.
For a crude characterization of renal function, the degree of contrast enhancement, representing
contrast transit, can be determined. Acquiring several data sets including the arterial, venous, and
equilibrium phases with three separate breath-holds can be accomplished on most MR systems (Fig.
30-26). Alternatively, several 3D data sets can be acquired in a single breath-hold with fast multiphase
3D gadolinium-enhanced MR angiography. With this technique, the acquisition time for a single 3D data
set is reduced to just a few seconds, thus allowing demonstration of minor changes in the temporal
evolution of renal enhancement. Unilateral delayed enhancement is often seen with focal high-grade
renal stenoses and occlusions (Fig. 30-27). Additional information regarding MR analysis of renal
function is available in Chapter 16.
page 786
page 787
Add to lightbox
Add to lightbox
29 of 104 20-04-2010 00:28

Figure 30-26 Diffuse high-grade left renal artery stenosis and delayed left renal enhancement.
CEMRA (A) demonstrates a diffusely narrowed left renal artery (arrow). MPR of the kidney
parenchyma (B) shows decreased left renal enhancement as compared to the right.
Add to lightbox
Add to lightbox
page 787
page 788
Figure 30-27 Left renal infarction. A, CEMRA shows partial non-opacification of the left kidney. B,
Axial T1-weighted post-contrast image confirms lack of opacification of the anterior pole due to an
embolus.
30 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-28 Volume-rendered image of bilateral renal artery stenosis (arrows) depicted by CEMRA
in a 63-year-old woman with refractory hypertension.
3D PC following gadolinium is the most commonly performed method to try and extract information
regarding the hemodynamic significance of a stenosis. The 3D PC images are acquired immediately
after CEMRA to evaluate the functional significance of renal artery stenoses. A velocity encoding
(Venc) should be set at 30-60 cm/s depending on age and estimated cardiac output. The presence of
a signal void on the 3D PC study in the presence of a morphologic stenosis has been correlated with
the presence of a hemodynamically significant stenosis.75
Renal Artery Stenosis
Renal artery stenosis is the most frequent pathologic condition affecting the renal arteries and is most
commonly due to atherosclerosis. In most cases atherosclerosis of the renal arteries is a symptom of
generalized atherosclerosis associated with disease elsewhere. Renal artery stenosis leads to
renovascular hypertension and renal insufficiency. Atherosclerotic plaque will often compromise the
ostium or the very proximal renal arteries although in rare cases it is isolated to the distal renal artery
or renal artery branches. If untreated, renal stenosis progresses to renal artery occlusion and
permanent loss of the renal parenchyma. For this reason, in patients with renovascular hypertension or
renal insufficiency, it is important to detect and treat renal artery stenosis early.
Table 30-1. Sensitivity and Specificity of CEMRA for Renal Artery for Stenoses
Authors Year # Patients Sensitivity (%) Specificity (%)
74 1995 19 100 93
Prince et al
Grist56 1996 28 88 88
62 1996 63 100 100
Holland et al
86 1996 47 100 89
Snidow et al
Steffens et al87 1997 50 96 95
Hany et al59 1997 39 93 98
31 of 104 20-04-2010 00:28

De Cobelli et al50 1997 55 100 97
Rieumont et al78 1997 30 100 71
Hany et al60 1998 103 93 90
Bakker et al45 1998 50 97 92

94 1999 26 100 95
Schoenberg et al
57 1999 22 91 79
Hahn et al
DeCobelli et al51 2000 45 93 95
Korst et al64 2000 38 100 85
Bongers et al48 2000 43 100 94
Volk et al88 2000 40 93 83

95 2000 51 96 92
Shetty et al
53 2001 25 97 92
Fain et al
Masunaga et al70 2001 39 100 100
Qanadli et al76 2001 41 90 80
CEMRA is an excellent test for diagnosing renal artery stenosis (Fig. 30-28).* The sensitivity and
specificity of CEMRA in the detection of renal artery stenosis are shown in Table 30-1. All these
studies use conventional digital subtraction angiography (DSA) as the gold standard. The sensitivity is
relatively high but the specificity is lower. The majority of renal artery stenoses are located within the
first several centimeters of the renal artery origin from the aorta. MR angiography is most accurate in
this region and is less accurate in the distal renal artery and segmental arteries. Recent improvements
in spatial resolution however, have allowed for depiction of some distal and segmental stenoses (Fig.
30-29). Imaging at 3 T offers the potential for improved image quality and therefore greater specificity
because of the higher intrinsic signal-to-noise ratio (Fig. 30-30). Because power deposition is four
times higher, the TR may need to be lengthened or the flip angle reduced compared with 1.5 T to
avoid exceeding the limits for specific absorption rates.
45,48,50,51,53,56,57,59,60,62,64,70,74,76,78,86-88,94,95
*(See references )
page 788
page 789
32 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-29 Segmental renal artery stenosis (arrow) depicted on CEMRA of a 32-year-old female
with hypertension.
Add to lightbox
33 of 104 20-04-2010 00:28

Add to lightbox
page 789
page 790
Figure 30-30 CEMRA of renal arteries at 3 T. A, Normal arteries. B, Bilateral disease.
Add to lightbox
34 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-31 Fibromuscular dysplasia. A, CEMRA is degraded by motion artifact. B, CTA done with
16 detector MDCT offers superior spatial resolution and less sensitivity to motion. Unlike the CEMRA,
calcium is seen which can obscure underlying stenosis.
As compared with DSA, some stenoses are over- or underdiagnosed with CEMRA for any of several
reasons. The spatial resolution of CEMRA, even with recent innovations, is substantially lower than
DSA and is generally lower than CTA as well (Fig. 30-31). However, DSA itself is an imperfect
reference standard, especially in cases of eccentric stenoses or tortuous vessels, in which assessment
of the exact morphology of the stenosis requires a high level of operator experience as well as multiple
views at various angles to image the stenosis in-plane. These stenoses are more easily depicted on
CEMRA because of its inherent 3D nature.
The recent introduction of parallel acquisition techniques allows the spatial resolution to be increased
by a factor of two or more in the same scan time.117,118 Isotropic spatial resolution of less than one
cubic millimeter can be acquired in reasonable breath-hold times of less than 20-25 seconds. The use
of multiplanar reformatted (MPR) images aids in depicting the exact morphology of the stenosis in any
plane that can help reduce misinterpretation of the degree of stenosis. Despite these differences,
agreement between the techniques remains high and CEMRA for the renal arteries is widely accepted
as a reliable and accurate examination.
Accurate interpretation of renal CEMRA requires interactive manipulation of the 3D data sets. It is not
possible to rely solely on the source images because they are subject to partial volume effects. 3D
reconstruction techniques generate 2D images representing 3D volumetric data. The reconstructed
images greatly enhance diagnostic confidence. In the past the most widely used post-processing
technique for CEMRA was MIP. The diagnostic accuracy of contrast-enhanced MR angiography using
the MIP technique is well described and accepted clinically.*
However, recent studies have shown the volume-rendering (VR) technique to have significant
46,119
advantages for renal CEMRA (Fig. 30-32). Renal artery stenoses determined with MIPs are
statistically greater than those determined using VR and MIPs have the largest mean difference from
DSA stenosis estimates.46 Mallouhi et al119 found that VR performed slightly better than MIP for
quantification of renal stenoses greater than 50% and significantly better for severe stenoses. VR also
35 of 104 20-04-2010 00:28

had a substantial improvement in positive predictive value and renal vascular delineation on VR images
was significantly better
The MIP algorithm selects only the voxel with the highest attenuation along a ray projected through the
data set; volume-averaged voxels may be erroneously excluded from the final image, resulting in
overestimation of stenosis. Volume rendering is based on the percentage classification technique,
which is used to estimate the probability of a material being homogeneously present in a voxel.119 This
provides an accurate determination of the amounts of materials when the voxel consists of two or more
materials, which are volume averaged. VR enables the volume-averaged voxels to be included in the
final image because it calculates a weighted sum of data from all voxels along a ray projected through
the data set.
45,50,51,56,57,59,60,62,74,78,86,87,94
*(See references )
page 790
page 791
Add to lightbox
Add to lightbox
Figure 30-32 Comparison of MIP (A) and VR (B) images in a 65-year-old male with bilateral renal
artery stenosis. MIP image overestimates the degree of stenoses (arrows) as compared to the VR
image (arrows), which was an accurate representation of the stenoses found at conventional
36 of 104 20-04-2010 00:28

angiography.
A combined morphologic and functional MR imaging protocol for detection and grading of renal artery
83
stenoses using CEMRA and PC flow measurements has been proposed by some authors. This
approach allows both grading of the morphologic stenosis but also an assessment of the hemodynamic
significance of the stenosis by means of time-resolved velocity curves in the renal artery. Studies have
shown agreement between the morphologic degree of stenosis and changes in the pattern of the flow
profile.81,94 The flow measurement technique also provides a functional grading of the degree of
stenosis independent of the accurate assessment of stenosis morphology. The two techniques of flow
and CEMRA can be used for a combined interpretation of renal artery stenosis. A multicenter trial has
shown that this combined approach allows a significant reduction in interobserver variability and an
improvement of overall accuracy compared with DSA, with sensitivities and specificities exceeding
84
95%.
In addition to the morphologic depiction of a renal stenosis, investigators have also identified functional
changes that indicate the severity of stenosis. These have included differences in parenchymal
enhancement and cortical thickness,75,96 signal dropout on 3D PC angiograms,75,96 reduction of mean
flow and the early systolic peak on cine PC imaging,94 changes in the gadolinium extraction fraction
and glomerular filtration rate at MR imaging, 120 and changes at captopril-sensitized dynamic MR
imaging in patients with renovascular hypertension. 55 None of these has achieved widespread clinical
use but research continues.
Accessory renal arteries are relatively common, occurring in approximately 15% to 35% of
121-125
kidneys. The reported sensitivity and specificity of MRA for the depiction of accessory renal
arteries vary greatly based upon the technique used and the experience of the reader.58,95,122 Because
of their variable size and location, no single imaging technique, including conventional angiography, is
122-124
100% accurate for their depiction.
While the majority of accessory renal arteries can be depicted with CEMRA, accurate diagnosis of an
accessory renal artery stenosis can be very difficult (Figs. 30-33 and 30-34). The caliber of these
accessory vessels often approaches 1 mm. This small size surpasses the actual resolution of many
MRA techniques. Accurate grading is not possible because of this resolution limitation. Unfortunately,
isolated accessory renal artery stenosis, with patent main renal arteries, can produce renovascular
123,126
hypertension. This is potentially a pitfall for the use of CEMRA as a screening tool for patients
with suspected renovascular hypertension.
Add to lightbox
page 791
page 792
37 of 104 20-04-2010 00:28

Figure 30-33 CEMRA depicts renal artery stenosis in three of the four renal arteries (arrows) in this
58-year-old male with worsening renal function and hypertension.
Add to lightbox
Add to lightbox
Figure 30-34 Accessory renal artery stenosis (arrows) depicted on two different patients using
CEMRA. Because of resolution limitations, one study (A) proved to be false positive, while the second
(B) was true positive, when correlated with conventional angiography.
While there are no large studies with respect to grading of accessory renal artery stenoses, one
recent publication found the prevalence of a hemodynamically significant accessory renal artery
stenosis in a group of patients with proven renovascular hypertension was only 1.5% (1 out of 68
127
subjects). Given that renovascular hypertension itself occurs in only 5% of all hypertensive patients,
if only 1.5% of these are due to an isolated accessory renal artery stenosis, then the actual number of
hypertensive patients with an isolated accessory stenosis as a cause is extremely low (less than
0.1%). The authors concluded that failure to detect accessory renal arteries should not unduly affect
the utility of a noninvasive test for detecting renovascular hypertension.127
38 of 104 20-04-2010 00:28

Recent developments have combined high-resolution CEMRA techniques with automatic table
movement, allowing for assessment of the renal arteries and the lower extremity arteries at the same
study using a single contrast medium injection.128 Patients with severe peripheral vascular disease
129
frequently have concurrent renal artery stenoses.
The recent introduction of multidetector row CT (MDCT) has substantially improved CT angiography of
the renal arteries (see Fig. 30-31) and this is being used clinically in many centers. MDCT has shorter
acquisition times, increased volume coverage, lower and improved spatial resolution as compared with
older CTA techniques.130,131 However, as compared to MRA, MDCT angiography has several
drawbacks that need to be considered. It uses ionizing radiation and a nephrotoxic iodinated contrast
agent. Post-processing can at times be more time consuming than MRA. There may also be some
difficulty assessing arterial luminal stenosis when there are dense vessel wall calcifications
present.132,133
Despite these limitations, a recent direct comparison of MDCT and CEMRA in the same patients using
89
DSA as the standard of reference found good agreement between the techniques. For detection of
hemodynamically significant renal arterial stenosis, the sensitivity of MRA was 86% to 100% and
specificity was 99% to 100% while the sensitivity of MDCT angiography was 86% to 93% and
specificity was 99% to 100% (differences not significant). Interobserver and intermodality agreement
was excellent (κ= 0.88-0.90). They did note that the time for performance of 3D reconstruction and
image analysis of CT data sets was significantly longer than that for MR data sets (P < .001).
Preprocedural planning with CEMRA significantly reduces the iodinated contrast material requirement
during percutaneous renal artery interventions.85 It can also significantly shorten procedure duration by
preoperative identification of additional stenoses and other anatomy that can then be accounted for
prior to the start of the intervention (Fig. 30-35).
Renal Fibromuscular Dysplasia

page 792
page 793
Add to lightbox
39 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-35 High-grade left renal artery stenosis (arrow) (A) depicted on a 76-year-old woman with
significant atherosclerosis. The MR angiogram also depicted a high-grade left common iliac stenosis
(arrow) (B), early identification of which provided for appropriate preprocedural planning and eventual
intervention.
Following atherosclerosis, the second most common cause of renal artery stenosis is fibromuscular
dysplasia (FMD). FMD is a non-atheromatous vascular lesion in medium and small arteries frequently
affecting the renal, carotid, and intracerebral arteries. The majority of patients are female and FMD
usually presents before 40 years of age. FMD lesions are classified as intimal fibroplasia, medial
fibromuscular dysplasia or adventitial fibroplasia. Medial FMD is most common and has angiographic
findings of a "string of beads" appearance with web-like stenoses alternating with small areas of
dilatation. In most cases the distal two-thirds of the main renal artery is involved, sometimes with
extension into segmental vessels. Bilateral involvement is common.
UPDATE Date Added: 29 January 2007
Robert R. Edelman
Editorial Comment: Renal fibromuscular dysplasia
Percutaneous transluminal angioplasty is highly effective in fibromuscular dysplasia, with technical

success and clinical success for renovascular hypertension of 95% and 87.9%.
de Fraissinette B, Garcier JM, Dieu V, et al: Percutaneous transluminal angioplasty of dysplastic stenoses of the renal artery:
Results on 70 adults. Cardiovasc Intervent Radiol 26(1):46-51, 2003.
In the past, MR angiography has not always been reliable for making a diagnosis of FMD. At times the
irregularities of the distal main renal arteries associated with FMD can be very subtle and they may not
be depicted on MR because greater spatial resolution is required.125 At other times, however, the
beaded appearance of the arteries can be very obvious and the characteristic appearance may be
demonstrated on MR (Fig. 30-36).
A recognized pitfall of CEMRA involves FMD and the "stair-stepping artifact" that sometimes occurs
with post-processing of CEMRA studies, especially if thicker partitions were utilized. This can lead to a
false-positive diagnosis of renal artery stenosis (Fig. 30-37). This artifact can be minimized by using
thinner partitions and zero filling, and by the use of VR techniques rather than maximal intensity
40 of 104 20-04-2010 00:28

projections. Even with these, the accuracy of MR angiography in diagnosis of fibromuscular dysplasia
is not established.95,125
UPDATE Date Added: 06 February 2007
Robert R. Edelman
Editorial Comment: Renal fibromuscular dysplasia
A recent retrospective study indicates that contrast-enhanced magnetic resonance angiography can be
an accurate test for fibromuscular dysplasia (FMD) involving the main renal arteries. Real-time bolus
tracking was combined with ellipticocentric phase-encode order to optimize arterial opacification and
minimize enhancement of the renal veins. In this study, 50 main renal arteries were analyzed in 25
patients. The sensitivity and specificity of contrast-enhanced magnetic resonance angiography for the
diagnosis of FMD were 97% and 93%. Sensitivity was 68%, 95%, and 100% for the diagnosis of
stenosis, string of pearls, and aneurysm. Although accuracy for accessory renal arteries would be
anticipated to be lower given the smaller vessel size, two of three such lesions were detected.
Willoteaux S, Faivre-Pierret M, Moranne O, et al: Fibromuscular dysplasia of the main renal arteries: Comparison of contrast-
enhanced MR angiography with digital subtraction angiography. Radiology 241(3):922-929, 2006.
Renal Artery Dissection

Renal artery blood flow can be reduced in patients with aortic dissections. Dissections involve the renal
arteries in one or more of several ways. Type A and B dissections may extend into the abdominal
aorta or extend directly into a renal artery. The flap may occlude the renal arterial origin or it may
extend into the main and segmental renal arteries interrupting renal blood flow. Even if the flap does
not extend directly into the vessel lumen, a normal renal artery arising from the true lumen may have
reduced flow due to either complete collapse of the true lumen or a substantially reduced perfusion
pressure.134
The renal arteries are well depicted on most CEMRA studies of the aorta, although the resolution may
be reduced compared to dedicated renal MRA studies because of the large FOV used to cover the
entire aorta. In cases of direct flap extension, the renal artery will have a dual lumen appearance
characteristic of dissection (see Fig. 30-22). One of the two lumens may thrombose due to slow or
stagnant blood flow, resulting in what appears to be a narrowed renal artery. Reduced enhancement
or partial enhancement of the kidney is a useful secondary finding.
Isolated spontaneous renal artery dissection is a rare condition that involves only the renal artery and
its branches.135 It can result in renal parenchymal loss and severe hypertension. Although several risk
factors have been identified in association with renal artery dissection, the natural history is not well
defined. The rarity and nonspecific presentation of the disease often lead to diagnostic delay. MRA
and CTA can potentially identify the entity and there is some reported success with percutaneous
intervention.136
Renal Artery Aneurysms

page 793
page 794
41 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
42 of 104 20-04-2010 00:28

Figure 30-36 Characteristic "string-of-beads" appearance of renal fibromuscular dysplasia (FMD) of

the mid and distal right renal artery (arrow) depicted on CEMRA (A). Conventional angiographic
correlation (B) depicts the web-like stenoses alternating with small areas of dilatation often seen in the
distal two-thirds of the renal arteries with FMD.
Add to lightbox
Add to lightbox
page 794
page 795
Figure 30-37 False-positive findings of FMD (arrow) seen on CEMRA of the renal arteries (A).
Conventional angiography revealed normal vessels (B). The "stair-stepping artifact" occurs with
post-processing if thicker partitions were utilized. This artifact can be minimized by using thinner
partitions and zero filling, and by the use of VR techniques rather than maximal intensity projections.
43 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-38 Bilateral segmental renal artery aneurysms (arrows) depicted using CEMRA in an
asymptomatic 78-year-old man being evaluated for suspected renovascular hypertension.
Most renal artery aneurysms have been found in older patients and they are most commonly due to
atherosclerosis.137 MR angiography is helpful in evaluating the size, orientation, and morphology of a
renal artery aneurysm and can accurately determine its relationship to aorta, renal veins, and other
vascular structures (Fig. 30-38).52 Less common causes of renal aneurysms include medial fibroplasia,
pregnancy, and mesenchymal diseases such as neurofibromatosis and Ehlers-Danlos syndrome. Renal
pseudoaneurysms are usually post-traumatic or inflammatory (Fig. 30-39).117,138 The prevalence of
renal artery aneurysms is quite low, reported to be <0.1%.139,140
Most patients are asymptomatic and the aneurysm is often discovered incidentally.140 It may cause
140
infarcts but the risk of rupture is small. A ruptured renal artery aneurysm should be considered when
retroperitoneal hemorrhage occurs in pregnant women.141 Although hypertension is present in 70% of
52
patients with renal artery aneurysm, a causal relationship has not been well documented.
44 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
Add to lightbox
page 795
page 796
Figure 30-39 Large pseudoaneurysms developing following renal biopsy in an 85-year-old male with
renal insufficiency. Computed tomography performed because of increasing left abdominal pain
depicted a large blood-filled mass distorting the left renal parenchyma (arrow) (A). CEMRA (B) depicts
the large pseudoaneurysm (arrow) as well as the area of communication with the left renal artery
(arrow) (C).
Renal artery aneurysms can be classified into four types: saccular, fusiform, dissecting, and
52
intrarenal. Saccular aneurysms are usually present in the main renal branch near the first bifurcation.
Fusiform aneurysms are usually found in medial fibroplasia and are not calcified. Dissecting aneurysms
are usually traumatic, spontaneous or iatrogenic. Intrarenal aneurysms are frequently associated with
arteritis (e.g., polyarteritis nodosa, Wegener's granulomatosis) but can also be caused by
atherosclerosis, fibroplasia, trauma, vascular malformations, syphilis, tuberculosis or tumors.
45 of 104 20-04-2010 00:28

On MR angiography, renal artery aneurysms have a number of differing presentations and can be
either very obvious or quite subtle. Use of the 3D nature of MRA is very helpful in fully evaluating the
size, location, and anatomic relationships. In very large aneurysms, there may be insufficient filling of
the aneurysm sac on first-pass CEMRA, so one or more delayed data sets may be helpful.
Additionally, as with aortic aneurysms, the exact outside diameter is better determined with 2D cross-
sectional images, as the CEMRA may only depict the lumen diameter.
Renal Arteriovenous Malformations and Fistulas

Renal arteriovenous malformations (AVMs) are abnormal communications between the intrarenal
arterial and venous systems. Patients with AVMs are often asymptomatic but they may develop
hematuria or hypertension. These malformations are either congenital or acquired by trauma or via
iatrogenic means. A renal AVM usually refers to the congenital type of malformation. Two types of
congenital renal AVMs are described; the cirsoid AVM is the most common type and the cavernous
congenital AVM is less common.142 Acquired renal AVMs are usually called renal arteriovenous
fistulas. Idiopathic renal arteriovenous fistulas have the radiographic characteristics of acquired fistulas
but no cause can be identified. They are usually associated with renal artery aneurysms.
Arteriovenous fistulas comprise 70% to 80% of arteriovenous communications in the kidney.143

Arteriovenous fistulas can result from trauma, surgery, tumors, inflammation or erosion of an aneurysm
directly into a vein (idiopathic arteriovenous fistula). Arteriovenous fistulas typically have a single
feeding artery and a single draining vein, both of which are markedly enlarged. The most common
clinical manifestation of a renal arteriovenous fistula is an abnormal bruit. Cardiomegaly or congestive
143
heart failure occurs in one-half of symptomatic patients. Persistent or delayed hematuria is also
common. Ischemia in the renal parenchyma distal to the arteriovenous fistula may induce renin-
mediated hypertension and impaired renal function.
MR can depict the feeding artery and draining vein(s), which are normally engorged (Fig. 30-40).
Because of the fast arterial to venous flow, there is prompt filling of the draining veins as well as the
143,144
renal vein and IVC immediately after enhancement of the arteries. The AVM or arteriovenous
fistula itself may be very small or may present as a large vascular mass. Identification of the arterial
and venous anatomy is very helpful both in making the initial diagnosis and for planning intervention.145
Preoperative Evaluation of Renal Transplant Donors
Add to lightbox
46 of 104 20-04-2010 00:28

Add to lightbox
page 796
page 797
Figure 30-40 Renal AVM depicted on a CEMRA. The right renal artery and vein are enlarged (A) and
there is early filling of the renal vein and IVC because of the high flow. MPR (B) depicts the artery and
vein within the kidney and shows that the actual site of communication is a small AVM (arrow) in a
subcapsular location in the lower pole.
MR imaging can be used for assessment of the renal arteries and renal parenchyma in potential living
kidney donors. It has been proposed as the sole preoperative imaging modality but has not been
completely adopted for this use, many centers continuing to use CTA or conventional angiography.146
The renal arteries and renal veins have a varied anatomy and may consist of one or more vessels at
several levels with variable calibers and levels of branching. Surgeons need reliable anatomic
information before harvesting of the kidney, especially if a laparoscopic technique is used.
A comprehensive preoperative assessment includes imaging of the arterial and venous anatomy, the
collecting system, and the renal parenchyma. Failure to identify variations in the renal vasculature or
collecting system or to detect parenchymal abnormalities may complicate the harvesting procedure as
well as the transplantation procedure and may compromise the eventual outcome for the recipient. In
the past various combinations of digital subtraction angiography, intravenous urography (IVP), nuclear
scintigraphy, and ultrasonography were used preoperatively.147
Recently, both CTA and MRA have been used as alternative techniques for the imaging work-up of
living kidney donors.44,58,69,72,148-150 DSA with urography is known to be an accurate method but it
requires catheterization, the use of iodine-containing contrast material, and exposure of the patient to
ionizing radiation. DSA is expensive and has well-described risks of arterial dissection and other
catheter-based morbidity. Only limited information about the venous anatomy is obtained with DSA.
Patients benefit from the noninvasive nature of MRA because there is little to no morbidity and it can
be performed as an outpatient procedure. The cost effectiveness of this model has been studied with
68
mixed results.
Imaging protocols typically include T2-weighted FSE or SSFSE, CEMRA and MRV, and delayed
fat-saturated T1-weighted GRE imaging. Identification of small accessory renal arteries is important
because unrecognized accessory arteries can be inadvertently ligated during the harvesting procedure,
leading to perioperative hemorrhage and renal infarction in portions of the transplanted kidney. For this
reason meticulous assessment of the source images and post-processed images is required.
Identification of nephrolithiasis is also important and can be difficult to achieve on MR imaging;
therefore many centers using MRA also perform either unenhanced CT or plain radiographs to identify
small calcified renal stones.
47 of 104 20-04-2010 00:28

Renal Transplant MRA

Renal transplantation is a common surgical event, with approximately 11,000 procedures performed
annually in the United States.151 MR imaging is playing an increasing role in the evaluation of the
61
transplant recipient. The causes of transplant dysfunction can be categorized as vascular (stenosis
or occlusion of the transplant artery or vein), parenchymal (acute tubular necrosis, rejection, and
medication toxicity) or extrinsic (peritransplant fluid collections and ureteral obstruction).
Because of its lower cost and wide availability, US is usually performed for detection of peritransplant
collections and ureteral obstruction. Arterial and venous patency and flow can also be evaluated with
US but it is very operator dependent and the main transplant artery is often incompletely visualized.
DSA is frequently used to evaluate the transplant artery but its invasiveness and the nephrotoxic
effects of iodinated contrast are a concern in patients with renal dysfunction.
The usefulness of MRA in the evaluation of the transplant renal artery is well
52,54,63,71,152,153
established. These studies have shown the usefulness of MR imaging in the evaluation
of the renal artery and vein as well as the transplanted kidney itself and the peritransplant region. The
more commonly encountered entities include arterial kinking, transplant renal artery stenosis,
fibromuscular dysplasia, renal infarction, lymphocele, and hydronephrosis.
The imaging protocol generally includes axial fat-suppressed FSE T2-weighted MR images through the
transplanted kidney, pre- and delayed post-gadolinium fat-suppressed gradient-echo T1-weighted
sequence, and a coronal multiphase CEMRA. A pelvic or torso phased array coil increases signal to
noise and helps improve image resolution.
Add to lightbox
page 797
page 798
Figure 30-41 Stenosis of the proximal transplant renal artery (arrow) depicted on CEMRA in a
34-year-old with poor renal function following right pelvic kidney transplant.
48 of 104 20-04-2010 00:28

Add to lightbox
49 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-42 Failing renal transplant due to acute rejection. A, CEMRA shows a patent feeding artery
(arrow) but poor organ perfusion. B, Delayed imaging shows occlusion of the renal transplant vein
(arrow).
Arterial complications involve kinking and stenosis. If the donor artery is too long, it can become
kinked, resulting in turbulent blood flow and possible reduction in renal transplant blood. The artery can
also kink at the anastomosis itself, resulting in stenosis. 154 Transplant renal artery stenosis is reported
155
in up to 12% of patients but the prevalence may be as high as 25% (Fig. 30-41) Causes include
atherosclerotis, surgical trauma, poor suture technique, and immunologic factors. 154,155 Arterial
stenosis can occur at any time after transplantation and may cause hypertension or unexplained
impairment of renal function. In cases of suspected stenosis, MR is used as a screening study, with
DSA reserved for selected patients who may be amenable to angioplasty. Venous occlusion is also
readily detected (Fig. 30-42).
Lymphoceles, hematomas, urinomas or abscesses may all present at various times in the
post-transplant period. Large fluid collections can impair renal function and may require drainage.
Transplant infarcts appear as wedge-shaped areas of decreased enhancement. Early infarction can be
the result of preservation damage to the donor kidney or arterial complications related to surgery. 156
Later infarction can be caused by embolic phenomena. Ureteral stenosis has been reported in up to
10% of transplants.157 Significant hydronephrosis may require intervention.
MR imaging is useful for the evaluation of solid and cystic masses in the transplanted kidney.
Post-transplantation lymphoproliferative disorder is a lymphoma-like condition associated with the
50 of 104 20-04-2010 00:28

Epstein-Barr virus present in approximately 1% of solid organ transplant recipients. 158

Post-transplantation lymphoproliferative disorder affecting a transplanted kidney can appear as either
153
a hilar or intraparenchymal mass. The lesions are reported to be hypointense on both T1- and
T2-weighted images and demonstrate minimal enhancement.153
page 798
page 799
Add to lightbox
Add to lightbox
Figure 30-43 False-positive stenosis of the iliac artery and transplant renal artery (arrow) (A)
secondary to susceptibility artifact from an adjacent surgical clip. Axial GRE sequence depicts a
region of typical dark blooming susceptibility (arrow) surrounding the adjacent metal clip (B).
Metallic artifact from vascular clips in the surgical bed is a common pitfall with MR angiography (Fig.
30-43). The susceptibility artifact can cause signal loss in adjacent structures such as the renal artery
or ureter. This can be avoided by recognizing its characteristic MR appearance and correlating the
imaging findings with those at prior CT or conventional radiography. Review of the GRE images, where
the artifact is especially pronounced, is often useful.
Renal Vein and Caval Invasion by Renal Cell Carcinoma
51 of 104 20-04-2010 00:28

Advanced renal cell carcinoma may involve the renal vein and extend into the inferior vena cava, at
times reaching as far as the right side of the heart. While such extensive spread is obvious on most
imaging modalities, more subtle extension can be more difficult to detect with routine US or CT
imaging. MR has been used successfully to reliably depict venous invasion of renal cell
carcinoma.49,159,160
Kallman et al160 and Choyke et al49 have shown that MR imaging has a high sensitivity for detection of
tumor thrombus beyond the distal renal vein (Fig. 30-44). While both of these earlier studies relied on
T1- and T2-weighted spin-echo sequences combined with axial time-of-flight imaging, more current
techniques emphasize CEMRA for vascular analysis. An evaluation of both the arterial and venous
phases of 3D CEMRA yields a comprehensive analysis of potential renal vein extension. 159
MR can depict other information that may be of use for operative planning. Renal sparing
nephrectomies are performed with increasing frequency. The multiplanar capabilities of MRI can be
utilized to present coronal or oblique images of the renal parenchyma, renal cancer, collecting
systems, and vasculature. Appropriateness of renal sparing nephrectomies can be determined based
upon the relative positions of the tumor and the adjacent structures. Other important information, such
as depiction of renal stenosis in the contralateral kidney, can also be depicted and addressed
preoperatively (Fig. 30-45).
Hepatic Artery
General Approach
Imaging of the hepatic artery presents a challenge due to its small size, circuitous course, and
considerable anatomic variations. While the portal vein supplies the majority of blood to the liver, the
hepatic arterial supply remains important because of the increasing surgical and interventional
treatment options for patients with liver cirrhosis and malignant liver disease. 73,161-163 In order to be
clinically useful, hepatic MR angiography must be capable of reliably identifying the hepatic artery and
its branches as well as any regions of arterial stenoses.
Previously, the hepatic artery was only imaged with DSA but recent advances in MR make accurate
hepatic artery imaging feasible.163-170 Older techniques such as PC and time-of-flight angiography
were limited by in-plane saturation, phase dispersion, and long acquisition times, which precluded the
performance of an examination during respiratory arrest in a patient. The evaluation of the portal
73,171-177
venous system with time-of-flight and phase-contrast techniques has had good results.
As shown in other regions of the body, CEMRA is an accurate and reliable method for performing MR
angiography. It has been successfully adopted to image the hepatic artery.164-166,168-170,178 Kim et al
reported a 100% sensitivity and a 74% specificity for depicting narrowing of the hepatic artery
following transplantation.170
page 799
page 800
52 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
Figure 30-44 Renal vein extension of a large left upper pole renal cell carcinoma (arrow) (A) depicted
as a low signal intensity mass (arrow) on 2D GRE images (B).
53 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
page 800
page 801
Figure 30-45 CEMRA demonstrates that there is no evidence of renal vein extension of the mass
(arrow) in the upper pole of the left kidney in this patient who is being evaluated preoperatively for a
nephrectomy (A). However, a stenosis of the contralateral right renal artery (B) was detected (arrow),
which had to be addressed prior to left nephrectomy.
Box 30-1 Hepatic Arterial Anatomic Type according to

Michels' Classification179
Type I : conventional anatomy (proper hepatic artery divides into RHA and LHA)
Type II : LHA replaced to LGA
Type III : replaced RHA to SMA
Type IV : replaced RHA and LHA
Type V : accessory LHA replaced to LGA
54 of 104 20-04-2010 00:28

Type VI : accessory RHA replaced to SMA

Type VII : accessory RHA and LHA
Type VIII : replaced RHA and accessory LHA, or replaced LHA and accessory RHA
Type IX : proper hepatic artery arises from SMA
Type X : proper hepatic artery arises from LGA
Abbreviations:
SMA, superior mesenteric artery
LGA, left gastric artery
RHA, right hepatic artery
LHA, left hepatic artery.
In patients donating a portion of their liver, it is critical to accurately determine the arterial and venous
anatomy prior to harvesting, because specific vascular anomalies can disqualify a donor candidate.
There is a great deal of variability of the hepatic artery and its branches. The arterial anatomic types
179
are categorized according to Michels' classification, summarized in Box 30-1. Some of the
accessory arteries can approach 1-2 mm in diameter or less, a size that exceeds the minimum voxel
size of most current MR protocols. Despite this, most variants are identified and CEMRA remains a
useful tool for imaging the hepatic artery (Fig. 30-46). CEMRA detected 91% of anatomic variants
found by either surgery or DSA in one group of living liver donors.164
Specific Techniques
For hepatic arterial imaging there are several general approaches in clinical use. The most common
method uses a routine CEMRA technique utilized elsewhere in the body for arterial MR angiography.
164-166,168-170,178
This technique yields diagnostic hepatic artery angiograms, but does not assess the
hepatic parenchyma in great detail. Although hepatic arteriography using CEMRA was not routinely
feasible several years ago, because of recent advances in gradient technology, much shorter repetition
times are now achievable. This allows for acquisition of images with higher spatial resolution than was
previously achievable in a single breath-hold. Consequently, it is now feasible to demonstrate branching
of the hepatic arterial system to the segmental level (see Fig. 30-46).164
An alternative approach has been gaining favor because it simultaneously obtains data that can be
used to evaluate parenchymal enhancement, very important for evaluating liver lesions. This approach
uses a hybrid MR sequence that provides nearly isotropic voxels. The nearly isotropic resolution
facilitates reconstruction methods such as multiplanar reconstruction that generate 2D parenchymal
imaging sets and 3D angiographic post-processing for definition of vascular and segmental
anatomy.168,180
55 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-46 Replaced right hepatic artery (arrow) depicted on CEMRA. The right hepatic artery
arises from the superior mesenteric artery, one of the most common variants of the hepatic artery.
This technique has been termed volumetric interpolated breath-hold examination (VIBE)181 or fast
acquisition with multiphase EFGRE 3D (also called FAME or LAVA). With volumetric interpolated
imaging, data sets that have nearly isotropic resolution in three dimensions (of the order of 2 mm voxel
size) can be obtained while preserving adequate anatomic coverage and uniform fat saturation within a
breath-hold. On the basis of qualitative and quantitative measures, image quality for this 3D method
was shown to be comparable to or improved over conventional 2D gradient-echo imaging when used
for nonenhanced or delayed contrast-enhanced imaging of the abdomen.168,180,181 As a breath-hold
technique, VIBEs can be performed repeatedly following injection of 0.1 mmol of gadopentetate
dimeglumine per kilogram of bodyweight to obtain dynamic 3D images of the liver.
The final method recently advocated a combined approach using both a SSFP sequence (fast imaging
164
with steady-state precession [TrueFISP]) and CEMRA. As in the above techniques, this approach
relies on the arterial phase CEMRA for hepatic artery depiction. However, portal venous anatomy,
frequently an important concern in addition to the arterial anatomy, is not always optimally depicted on
CEMRA. CEMRA is typically timed to the arterial phase of enhancement and the second post-contrast
acquisition does not occur exactly during the portal venous phase of enhancement. TrueFISP images
are obtained in both axial and coronal orientations to depict the portal structures optimally.
Evaluation of Potential Liver Donors
page 801
page 802
There has been a chronic shortage of cadaveric organs available for liver transplantation and the
number of transplants performed annually has continued to increase. The number of patients in need of
livers far exceeds the availability of livers suitable for transplantation. This shortage has driven the
gradual adoption of living donors for liver transplantation. Since the 1980s, adult donors have
successfully donated portions of their left lobes to pediatric patients needing transplantation; recent
improvements in surgical technique have also permitted the safe performance of right lobe
hepatectomy for transplantation in adults.182-185 Living donor liver transplantation is now increasingly
56 of 104 20-04-2010 00:28

performed in both adults and children.
Safe harvesting and successful transplantation require especially careful selection of donors and
accurate preoperative planning. Preoperative evaluation of the donor is required to exclude focal or
diffuse liver disease, vascular abnormalities such as portal vein thrombosis, arterial anomalies, and
anomalies of the biliary tract that might complicate right hepatic lobe resection and transplantation.
Relatively common anatomic variants can be considered reasons for exclusion of donors. Donor safety
is of great importance so it is essential to accurately evaluate the hepatic vasculature preoperatively in
all potential liver donors.186,187 This is done to detect unexpected anatomic variants of the vasculature.
CEMRA has been used successfully to evaluate the liver and its blood supply in potential donors. Lee
et al found that MR could accurately exclude 9 of 25 potential donors; of the nine patients who
eventually underwent successful right hepatectomy, all MR imaging findings were corroborated
intraoperatively.168 Carr et al164 found that segmental branch vessels were visualized on the MRA in
the majority of cases; the segment 4 branch was identified in 96% of patients. It is important to identify
the dominant arterial branch to segment 4 in order to prevent its inadvertent removal at surgery.
Variant arterial anatomy was seen in 50% of patients; MRA detected 91% of arterial variants found at
surgery and angiography.
Evaluation for Complications Following Liver Transplant

As previously noted, advances in surgical techniques have contributed to the improved results of living
related liver transplantation. However, there remains a high risk of vascular complications following
hepatic transplantation because of complex reconstruction performed for the hepatic artery and portal
vein. A significant percentage of liver grafts are lost from vascular complications after living related liver
transplantation.185,188 Prompt and accurate diagnosis of potential vascular complications is crucial for
increasing the survival rate of the graft in living related liver transplantation since complications, such as
stenoses or thromboses, are treatable with interventional procedures. Untreated vascular
complications often progress to severe hepatic failure or overwhelming biliary sepsis, which can lead to
graft failure and death.
Magnetic resonance is an excellent method for evaluation of liver transplant patients (see Chapter 83).
CEMRA is now performed for the assessment of the hepatic vasculature in patients who have
undergone liver transplantation (Fig. 30-47).189-192 Glockner at al189 found that all cases of hepatic
artery thrombosis were accurately detected, while CEMRA depicted moderate to severe hepatic
artery stenosis (>50% narrowing as determined by conventional angiography) in six of seven cases,
although there were three false-positive interpretations. Kim et al190 reported that the sensitivity,
specificity, positive predictive value, negative predictive value, and accuracy of MR angiography for
detection of hepatic artery stenoses or thrombosis were 100%, 74%, 29%, 100%, and 77%,
respectively.
Celiac and Superior Mesenteric Arteries

General Approach
Previously digital subtraction angiography was the only method used routinely to evaluate the
mesenteric circulation. US is used for screening purposes and can at times give useful anatomic and
functional information but is hampered by its high operator dependency and the fact that the
mesenteric vasculature frequently is not accessible because of overlying bowel gas. CTA of the
mesenteric circulation has improved dramatically with the use of MDCT scanners but it still uses
ionizing radiation and potentially nephrotoxic contrast agents, two significant drawbacks. 131,193
MR has been used for the assessment of the mesenteric vasculature for more than 10
62,74,194-196
years. However, only recently has MRA begun to play an important clinical role in this
evaluation, especially for patients with suspected mesenteric ischemia. Early MR attempts using
time-of-flight or phase-contrast angiography had some success but suffered from prolonged
57 of 104 20-04-2010 00:28

examination times and motion-related artifacts.195 Longer acquisition times precluded acquiring single
data sets during a breath-hold, leading to substantial image blurring from motion artifacts.
More recent techniques, most importantly CEMRA, have finally enabled reliable and accurate
evaluation of the mesenteric vessels, including the celiac axis, superior mesenteric arteries, and inferior
mesenteric artery (Fig. 30-48).197,198 It is interesting that in one of the earliest studies of CEMRA,
Prince et al evaluated the mesenteric vessels in addition to the aorta and renal arteries.74 The 3D
nature of CEMRA makes possible evaluation of the often tortuous mesenteric vessels (Fig. 30-49).
Flow-dependent imaging techniques often have difficulty with these vessels because of their complex
anatomy.
Other noncontrast techniques remain useful, however, and may re-emerge as an adjunct to
CEMRA.199 These techniques rely on a variety of factors to determine quantitative blood flow as well
as information about the chemical properties of the blood to assess for physiologic changes in the
194,199-203
gut. Combining the morphologic information from CEMRA and flow-related information from
other techniques allows for a comprehensive analysis of mesenteric arterial anatomy and function. This
is an area where the flexibility of MR has a clear advantage over CT, even with the higher spatial
resolution now available with MDCT scanners.
page 802
page 803
Add to lightbox
58 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-47 Stenosis of the hepatic artery following liver transplantation secondary to unsuspected
celiac axis stenosis. Immediately following transplant, liver function was noted to be poor. A CEMRA
depicted a stenosis (arrow) at the site of the hepatic artery anastomosis (A) and an unexpected
stenosis (arrow) at the celiac origin (B). Angioplasty and stenting of the celiac lesion improved inflow
to the hepatic artery and the distal stenosis resolved without additional intervention.
Add to lightbox
Figure 30-48 High-grade celiac stenosis (arrow) depicted on CEMRA. There is significant narrowing
of the celiac axis associated with moderate post-stenotic dilatation.
59 of 104 20-04-2010 00:28

The evaluation of known or suspected mesenteric ischemia remains the most common indication for
mesenteric MR angiography. Other indications include evaluation of possible tumor encasement by
retroperitoneal masses, such as in pancreatic adenocarcinoma. Determination of accurate splanchnic
anatomy is important for surgical planning prior to liver resections, liver transplantation, surgical shunt
placement, and resection of other retroperitoneal masses.
Specific Techniques
The standard CEMRA technique is easily modified to evaluate the mesenteric vessels. A coronal
volume is positioned to cover the abdominal aorta, celiac, and superior mesenteric arteries. A phased
array coil increases signal to noise, helps speed scan acquisition time, and improves image resolution.
Multiple post-contrast data sets timed to coincide with the arterial and then venous phase of perfusion
allow for a study that yields useful arterial and venous information. In many cases, an analysis of both
the arterial and venous systems is desired. In cases of suspected mesenteric ischemia, causative
factors include arterial or venous stenosis as well as arterial or venous thrombosis. For this reason, a
study that gives an accurate depiction of both the arterial and venous anatomy is desirable.
page 803
page 804
Add to lightbox
60 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-49 CEMRA permits detailed evaluation of the often tortuous mesenteric vessels. What
appears to be a very confusing tangle of mesenteric arteries is clarified by using lateral (A) and
superior (B) projections from a CEMRA. There is a single common origin of the SMA (arrow), hepatic,
splenic, and other mesenteric vessels.
High-performance gradient systems allow the acquisition of high-resolution CEMRA scans within a
short breath-hold, reducing artifacts resulting from peristalsis. Most centers use 20 to 40 cc of
gadolinium chelate injected at a rate of 2.0 to 3.0 cc/s, using a power injector for all patients. 204 This
volume-based approach must be modified for patients with low bodyweights or in children so that the
clinically accepted dose of 0.3 mmol/kg is not exceeded. If this is the case a dose of contrast is
calculated using the actual bodyweight and a range of 0.1 to 0.2 mmol/kg is generally acceptable.
As with other types of CEMRA studies, there are numerous techniques to achieve optimal
co-ordination of timing contrast administration to coincide with data acquisition so that peak arterial
enhancement is achieved. A dual-phase post-contrast CEMRA is performed with the first phase timed
for maximum arterial enhancement and a second phase timed at mesenteric venous enhancement after
a time delay of 10 to 15 seconds between the two acquisitions. Images are post-processed in the
typical manner using multiplanar reformations, maximal intensity projections, and volume rendering.
Noncontrast methods are less optimal than CEMRA. However, they have some utility and there is a
body of literature supporting their use. Time-of-flight MRA is used only rarely for the evaluation of the
mesenteric vasculature. 3D TOF requires prolonged scan times, leading to misregistration artifacts. 205
TOF is also less sensitive to in-plane flow, generating false-positive areas of decreased signal due to
the tortuous nature of the mesenteric vasculature. Although TOF MRA has little utility in evaluation of
173
the mesenteric arteries, it has been used successfully for evaluation of the portal venous system.
PC MRA is more useful in the mesenteric vessels than TOF. It allows direct quantitative evaluation of
flow direction and velocity. Spins move in the encoded direction to acquire a phase shift, which is
quantitatively measured. Unlike other MRA techniques, optimal utilization of PC requires that the flow
velocities be estimated and an appropriate velocity-encoding gradient chosen prior to scanning. If the
velocity-encoding is set too low, severe aliasing artifacts can occur. This need to estimate the flow
velocity has greatly hindered widespread acceptance of PC techniques.
Li et al194,202 have had the greatest experience with using PC techniques and have reported that
simultaneous postprandial measurement of SMA and SMV flow with PC is useful in detecting and
understanding suspected mesenteric ischemia. Gated 3D PC MRA techniques also have been used
61 of 104 20-04-2010 00:28

successfully for the detection of stenoses within the proximal portions of the CA and the SMA.195 In
addition to the 3D approach, 2D PC MRA with and without breath-holding and 2D ECG-gated cine PC
have also been described as useful methods for the functional evaluation of the mesenteric
194,201,206
vasculature. These 2D approaches allow quantitative measurements of flow velocities and
flow volume.
page 804
page 805
Add to lightbox
Figure 30-50 Acute partial thrombosis of the celiac axis in a 27-year-old male with protein S
deficiency syndrome. The low-intensity thrombus (arrow) partially fills the celiac artery. Note near
occlusion of the SMA secondary to prior thrombosis.
Other techniques useful in the evaluation of the mesenteric vasculature include a blood oxygen level-
dependent (BOLD) method which is similar to that used for functional brain imaging. The BOLD
207
technique uses a heavily T2*-weighted sequence to depict changes in blood oxygenation. The BOLD
signal response can help identify reduced blood flow response in patients with mesenteric ischemia.
Ischemic bowel increases the quantity of deoxyhemoglobin (which has a reduced BOLD signal) in the
superior mesenteric vein.208
Mesenteric Ischemia
Mesenteric ischemia can develop acutely or be a chronic problem. Mesenteric MRA is playing an
40
important role in the diagnosis of these disorders. Because of the serious clinical status and urgent
need for a diagnosis and intervention, CT is usually the preferred tool at most centers for the diagnosis
of acute mesenteric ischemia (AMI). MRA is performed less commonly in this setting because
logistically CT is usually more readily available and can be a faster scan. MRA is playing a more
important role in patients with chronic mesenteric ischemia (CMI).
62 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-51 Severe chronic mesenteric ischemia secondary to SMA occlusion and a high-grade
celiac stenosis (arrow) depicted by CEMRA (left). Anterior projection shows a large IMA (arrow) that
supplies the entire abdomen via large mesenteric and retroperitoneal collaterals (right).
AMI is caused by an abrupt interruption of the blood supply to the small bowel or colon. It carries a
high morbidity and the mortality rate has been reported to exceed 60%. 209 A SMA embolus or
thrombosis, mesenteric venous thrombosis, and mesenteric vasoconstriction cause AMI most
commonly. Aortic dissection may either extend into the mesenteric vessels or compromise blood flow
to the vessels and is also a cause of AMI. Acute emboli to the SMA account for 40% to 50% of all
40,209
episodes of AMI and can cause abrupt termination of flow. The embolus is recognized on MRA by
the presence of a cut-off sign, if completely occlusive, or appears as filling defects if non-occlusive
(Fig. 30-50). In a porcine model Chan et al210 reported that CEMRA had a 91% sensitivity and 80%
specificity for detecting ischemic regions. They were also able to detect a significant drop in blood
oxygen saturation in the superior mesenteric vein using BOLD techniques.
Acute mesenteric thrombosis usually occurs in patients with significant atherosclerosis. Symptoms are
usually more insidious because many patients have already developed some collateral circulation
because of their chronic atherosclerosis. Unlike AMI, where the emboli frequently lodge more distally in
the artery, thrombotic occlusion of the SMA commonly occurs within the proximal 2 cm of the vessel
origin. CEMRA can show these findings in addition to the visualization of collateral vessels. 40
Chronic Mesenteric Ischemia

page 805
page 806
CMI is most commonly observed in patients with significant atherosclerosis. In most cases, at least
two of the three mesenteric arteries (SMA, celiac axis [CA] or IMA) must be occluded or severely
stenosed before the disease becomes clinically apparent (Fig. 30-51). A classic clinical triad of
postprandial abdominal pain, weight loss, and avoidance of food characterize CMI. While
atherosclerosis of the mesenteric vessels is frequent, CMI is relatively uncommon because of the rich
mesenteric, collateral circulation. This collateral network in the mesenteric system makes it difficult to
estimate the degree of mesenteric vascular stenosis necessary to cause intestinal angina.
CEMRA is the major noninvasive MR screening technique for patients suspected of having CMI. It is
well suited for the detection of atherosclerotic occlusive disease in the proximal CA and SMA. It has
limitations in that it is not reliable for assessing peripheral SMA branches and the IMA. Abnormal
morphologic findings depicted on CEMRA can also be corroborated with functional flow measurements
using noncontrast techniques.
63 of 104 20-04-2010 00:28

In most clinical cases CEMRA is currently utilized and has proven to be accurate and reliable for
stenosis of the proximal mesenteric arteries. Meaney et al211 evaluated CEMRA in patients with
suspected mesenteric ischemia and compared the results with catheter angiography or surgery. They
reported an overall sensitivity and specificity of 100% and 95%, respectively, although they did note
two errors caused by overgrading the severity of IMA disease. Because of its small size, the IMA is
212
more difficult to assess accurately, a fact also found by Carlos et al. They assessed accuracy and
interobserver variability of CEMRA in patients with CMI and reported cumulative accuracies for
detecting significant stenosis of 0.95 to 0.97. However, while interobserver agreement was 0.90 for
CA and 0.92 for SMA, it was only 0.48 for the IMA.
Unenhanced studies for morphologic assessment have been disappointing. One study assessed
systolically gated 3D PC MRA but only 66% of stenoses were detected and several false positives
were reported.195
Investigations of functional flow dynamics in patients with CMI have also been useful. Burkart et
al172,213 used cine cardiac-gated PC MRA to show that flow rates in the SMA and SMV correlate well
and that patients with CMI show a significantly reduced rate of post-prandial flow augmentation in the
201,202
SMV compared with controls. Li et al reported that the percentage flow change in the SMA 30
minutes after a meal provides the best discriminator between patients with and without CMI.
Investigators have also reported that BOLD measurement of blood oxygenation in the SMV shows that
a decrease in post-prandial oxygenation is a sensitive indicator of CMI. 201,210
Aneurysms
Aneurysms of proximal SMA and CA can occur in association with aortic aneurysms or they can be
isolated. Isolated aneurysms can also involve the splenic artery, hepatic, duodenal branches, gastric
and inferior mesenteric arteries. Splenic artery aneurysms are the most common and up to 10% of
these may eventually rupture and require intervention.214 These are more commonly seen in
post-partum patients or in patients following pancreatitis where the vascular wall has become
weakened. Mycotic aneurysms of the SMA may develop from infections arising in the adjacent bowel.
Celiac aneurysms are associated with post-stenotic dilatation of the vessels secondary to extrinsic
compression by the median arcuate ligament.
CEMRA can depict focal aneurysms of the visceral arteries. As with CEMRA elsewhere, the study
usually depicts only the vessel lumen and will not demonstrate any thrombus of other material outside
the blood pool. Cross-sectional (2D) analysis of the aneurysm with other pulse sequences is required
for an accurate overall dimension; this is the actual size that determines the need for intervention or
treatment.
Median Arcuate Ligament Syndrome

MRA has been used to diagnose the median arcuate ligament syndrome. The median arcuate ligament
of the diaphragm is formed by muscular fibers that connect the right and left crura of the diaphragm
and it defines the anterior margin of the aortic hiatus. Compression of the celiac axis by this ligament is
referred to as celiac artery compression syndrome or median arcuate ligament syndrome and it has
been reported to cause intestinal angina, most commonly in thin patients, 215,216 although this diagnosis
217
has been controversial. The clinical manifestations of celiac artery compression syndrome are often
vague and may include post-prandial pain and an abdominal bruit. Extrinsic compression by the median
arcuate ligament entity has also been reported to predispose patients who have undergone liver
transplantation to develop hepatic artery thrombosis.218 Surgical treatment can lead to persistent
clinical improvement in symptomatic patients.
The diagnosis of median arcuate ligament syndrome relies primarily on imaging; initially this was done
with DSA but currently MRA and CTA are used. However, differentiation between clinically relevant
celiac artery compression and incidental narrowing may be difficult. When additional findings such as
post-stenotic dilatation are present at imaging, they can be strongly suggestive of true celiac artery
64 of 104 20-04-2010 00:28

compression syndrome.219 Augmentation with functional imaging to further quantify blood flow may be
199-201
helpful.
A recognized pitfall of CEMRA is the false-positive finding of celiac artery compression (Fig. 30-52).219
CEMRA is typically performed during suspended respiration to minimize image degradation caused by
respiratory motion. If respiration is suspended at end expiration, there can be a false-positive finding of
extrinsic celiac compression. Lee at al219 measured the prevalence and degree of celiac artery
compression during breath-hold imaging at end inspiration and end expiration on CEMRA in a group of
asymptomatic patients. They found that 57% of patients had at least mild artery narrowing at end
expiration and that of these, 73% had less narrowing at end inspiration. In order to reduce false-
positive findings, if celiac compression is suspected imaging should be performed during inspiration.
Pancreatic Tumor Staging for Resectability

page 806
page 807
Add to lightbox
65 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-52 False-positive finding of celiac artery compression (arrow), a recognized pitfall of
CEMRA performed during end-expiratory breath-holding (A). If respiration is suspended at end
expiration, there can be a false-positive finding of extrinsic celiac compression by the median arcuate
ligament. Repeat CEMRA imaging performed during end inspiration depicts a normal artery (B).
Adenocarcinoma of the pancreas has a high mortality. Surgical resection is often the only possible
cure. As part of the preoperative evaluation, CT or MR imaging is performed to assess for metastases
220,221
and local invasion of the adjacent arteries and veins. At many centers MRA, pancreatic MRI, and
MR cholangiopancreatography (MRCP) reliably diagnose, stage, and assess the resectability of
pancreatic adenocarcinoma.
Using CEMRA, MRCP, and 2D MR imaging, Richter et al222 assessed surgical resectability in patients
with pancreatic cancer. They achieved a sensitivity of 96% and specificity of 89.5% (positive predictive
value [PPV], 92.3%; negative predictive value [NPV], 94.4%) in a series of 143 patients with benign
and malignant diseases of the pancreas with surgical correlation. Previous attempts using 2D inflow
and 3D PC MRA were considerably less successful.223,224
Signs of non-resectability include encasement of the superior mesenteric artery or vein and thrombosis
of the vessels. CT is more routinely used for this purpose than MR but there are often cases where
225
patients are unable to undergo contrast-enhanced CT, so MR remains an important option.
Peripheral Arteries
General Approach
Atherosclerotic peripheral vascular disease (ASPVD) of the lower extremities is a common disease
with a significant morbidity and mortality. Atherosclerosis is the leading cause of occlusive arterial
disease in older patients.226 Symptomatic ASPVD affects 7% of the older population and leads to
66 of 104 20-04-2010 00:28

more than 100,000 surgical procedures annually in the United States.227,228 Since it is widely
recognized that atherosclerosis is a systemic disease, patients with ASPVD are also at risk of
complications elsewhere, including myocardial infarction and ischemic stroke. Morbidity associated
with the disorder includes claudication, rest pain, tissue loss, and gangrene.
Atherosclerotic lesions arise most commonly in areas of elevated mechanical stress and complex flow.
Typical areas include the popliteal artery, common iliac bifurcation, and the distal superficial femoral
artery (SFA) in the adductor canal. Diabetics tend to get smaller vessel disease, such as in the calf
and foot, while smokers get more isolated proximal lesions (Fig. 30-53). Prior to angiography,
symptomatic patients normally undergo segmental pressure measurements and Doppler or
plethysmographic analysis of flow. These tests can give objective evidence of possible areas of flow
limitation that can then be compared to the patient's symptoms to determine the likelihood of actual
disease.
page 807
page 808
Once the suspicion of likely disease is established, arterial imaging becomes an important component
in the management and treatment of ASPVD. Accurate details regarding the number, length, and
severity of vascular lesions are essential for planning intervention or surgery for revascularization.
Despite the fact that many patients with ASPVD have renal insufficiency and poor arterial access for
catheter placement, traditionally DSA using iodinated contrast agents has been the standard of
reference for imaging patients with this disorder. In the past decade a substantial body of knowledge
has emerged supporting the use of MRA for the diagnosis of ASPVD. MR is accurate and noninvasive,
uses a non-nephrotoxic contrast media, and is relatively easy for the patient to tolerate. 311
When performing and interpreting peripheral MR angiography, it is important to know exactly what
anatomic and other information the treating physician needs to make clinical decisions on the patient.
Peripheral MRA is best done after a thorough surgical and clinical history has been obtained. Direct
communication with the referring surgeon or physician allows the radiologist to specifically tailor each
exam.
In general, not only is it important to characterize the morphology of an arterial stenosis but it is also
critical to assess the vascular inflow and outflow of a lesion. For example, a stenosis of the superficial
femoral artery may be treated differently if there is an inflow stenosis in the iliac artery. The integrity of
bypass grafts, vascular reconstructions, and intra-arterial stents depends on adequate inflow and
outflow. For this reason, a comprehensive analysis of the entire lower extremity system, from the
abdominal aorta to the feet, is the standard of care when assessing ASPVD.
Unlike the strict criteria available to guide management in the carotid arteries, treatment planning for
ASPVD patients is less restrictive. In order to help surgeons and other physicians make management
decisions, an imaging study must be able to characterize stenoses as being greater than or less than
229,230
50% diameter narrowing. Atherosclerotic lesions with more than 50% diameter stenosis are
usually considered hemodynamically significant.229 The length of the narrowing is also important;
diseased segments less than 10 cm in length are often amenable to angioplasty and stenting, while
longer stenoses or occlusions frequently require bypass grafting. The location of disease also
influences treatment; proximal and shorter segment (<3 cm in length) stenoses normally yield the best
231
results. Finally, the nature of the narrowing is important to the treating clinician. Atheromatous
lesions are treated with bypass, balloon angioplasty or stent. Thrombotic lesions are usually managed
with thrombolytic therapy and embolic lesions undergo embolectomy.
67 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-53 Patient with history of smoking and decreased pedal pulses. Moving-table CEMRA
shows aortic occlusion and distal reconstitution through collaterals.
MR angiography has become the standard diagnostic method for evaluation of ASPVD, although some
centers still rely on DSA and a few have recently turned to the use of MDCT. A decade ago, there was
much discussion regarding the use of TOF techniques for performing run-off MRA. However, artifacts
from this technique have resulted in an overestimation of both the length and degree of stenosis and
the exam times often exceeded an hour, which many patients could not tolerate.232-236 CEMRA has
proven to be a reliable method for evaluating the lower extremity arteries (Figs. 30-54 and 30-55). The
artifacts associated with inflow techniques are greatly reduced or eliminated with CEMRA, although it
has its own unique challenges, especially with contrast timing and resolution.
Specific Techniques
68 of 104 20-04-2010 00:28

Anatomic Coverage
page 808
page 809
Add to lightbox
69 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-54 Comparison of peripheral MRA using 2D time of flight (A) and moving-table CEMRA
(B). The quality of the CEMRA is greatly superior and the study is an order of magnitude faster.
Table 30-2. Sensitivity and Specificity of MR Angiography for Occlusive Disease

of the Lower Extremities
Year # Patients Technique Sensitivity (%) Specificity (%)
242 1992 23 2D TOF superior to DSA
Owen
Baum279 1995 155 2D TOF 82 84
Prince74 1995 43 3D Gd 94 98
Snidow86 1996 32 3D Gd 100 98

59 1997 39 3D Gd 93-96 96-100
Hany
70 of 104 20-04-2010 00:28

Ho237 1998 28 3D Gd 93 98
Meaney211 1998 20 3D Gd 81-89 91-95
Lee258 1998 23 2D Gd 94 91
Winchester280 1998 22 2D Gd 90 98
281 1999 67 3D Gd 100 83
Link
240 2003 45 3D Gd 99 97
Morasch
page 809
page 810
Add to lightbox
71 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-55 Comparison of peripheral MRA using 2D time of flight (A) and CEMRA (B) for the foot
and ankle. The quality of the CEMRA is again noted to be superior, with good depiction of the plantar
arch.
There are many unique challenges to performing MRA of the run-off vessels. Foremost is the
determination of anatomic coverage. The useable z-axis length of most MR systems is 40-55 cm, a
length that is insufficient for comprehensive imaging of the lower extremities. It is good clinical practice
to image from the abdominal aorta to the feet for routine lower extremity angiography. Although the
patient's symptoms may be caused by a segmental stenosis in one location, the "inflow" and "outflow"
of that anatomic segment must be evaluated prior to any intervention. A successfully placed surgical
bypass will eventually fail if there is a more proximal stenosis reducing inflow of blood to the graft.
Using DSA, it has been routine clinical practice to include the renal arteries. Atherosclerosis is a
systemic disease and renal artery atherosclerosis is frequently found in patients with symptomatic
ASPVD. In order to replace DSA successfully as a screening exam, most centers currently include the
renal arteries and abdominal aorta in CEMRA studies. 237-240 One drawback of this approach, however,
is that the FOV used for the first station of most CEMRA studies is substantially larger than that
routinely used for renal CEMRA exams, resulting in a lower resolution evaluation of the renal arteries.
If disease is found, patients may always return for a dedicated renal MRA at another time.
The distal portion of the study generally extends to include the ankles and feet. Inclusion of the pedal
vessels is critical for patients suffering from limb-threatening ischemia (Fig. 30-56) because a direct
72 of 104 20-04-2010 00:28

continuous flow of pulsatile blood flow needs to be established in most cases. When imaging patients
with claudication, this may be less important. With many MR systems, a three-station MR study will be
able to include the feet. When using MR systems with smaller z-axis or when imaging very tall patients,
three routine stations may not include the feet. An additional station, using either CEMRA or TOF
techniques, may be required. The clinical need to include the feet must be reviewed prior to study
protocol.
Examination Techniques
Depending on the MR systems used, the coils available, and the experience of the examiner, there are
many differing approaches to adequately performing a run-off MR angiogram. Each of the following
techniques has been validated by comparison with the gold standard, DSA. Each of the techniques has
merit and there has been much discussion as to which would emerge as the most beneficial clinical
approach.
Most clinicians now favor the multistation "bolus-chase" or "moving-table' approach of a single
prolonged contrast injection and multiple rapid 3D acquisitions performed successively over a period of
14,237-241
approximately 60 seconds with fast table increments between stations. This method is the
most rapid and has the added benefit of CEMRA at each station, in contrast to TOF with its
recognized limitations.
With all techniques, the patient is positioned on the MR system gantry in the supine position. A
comfortable position is important to prevent the patient from moving during the exam; this was
especially true when TOF was used and exam times stretched up to 2 hours. The legs are positioned
horizontally in such a manner that the arteries stay in the same plane to limit anterior-to-posterior
coverage. The legs are flexed slightly at the hip and knee, allowing the arteries to align in the same
horizontal plane. Newer scanners performing bolus-chase angiography allow prescription of multiple
oblique 3D volumes and in this case, the patient can be positioned less rigorously and in a somewhat
more comfortable position if necessary.
page 810
page 811
73 of 104 20-04-2010 00:28

Add to lightbox
74 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-56 Patient with limb-threatening ischemia. A, DSA fails to show a patent distal dorsalis
pedis for bypass. B, MRA shows that the distal dorsalis pedis is patent (arrow). (Courtesy of Barry
Stein MD.)
Careful planning of the localizer, or scout, is more critical for CEMRA than for most other MR
applications and this is especially so for moving-table MRA. The rapid three-plane GRE localizer used
for most image localization purposes does not delineate the arteries sufficiently well to allow close
tailoring of the 3D volume. So that the position of the arteries can be determined on the scout, many
centers perform a low-resolution axial 2D TOF or PC scout.239 This allows for accurate positioning of
the imaging volumes since the actual vessel positions can be determined from the image data. If a
multistation study is to be performed, all three stations are scouted prior to the diagnostic study. This
also allows for repositioning prior to the study, which can avoid inadvertent exclusion of critical
anatomy.
The presence of intra-arterial stents and extrinsic ferromagnetic materials may cause sufficient
susceptibility artifact to produce a false-positive stenosis or occlusion on MRA (Fig. 30-57). Therefore
an accurate clinical history covering the presence of such devices is required prior to planning the
study.
If optimal evaluation of the distal vessels is required, the moving table study should be preceded by a
time-resolved (e.g. TRICKS or SENSE) study extending from upper calf through mid-foot using a
peripheral vascular phased array coil and separate Gd injection (e.g. 0.1 mmol/lg at 2 cc/sec).
Time-of-Flight
233,242-244
Initially described methods used TOF for the entire exam. TOF is a widely available technique
75 of 104 20-04-2010 00:28

and in general its use requires less sophisticated MR capability and specialized lower extremity run-off
(multistation) coils were not needed. However, this approach is lengthy, requires numerous
repositionings of the coils used, and suffers from the well-described drawbacks of overestimation of
stenoses due to in-plane saturation and other artifacts.
page 811
page 812
There was much enthusiasm regarding the use of TOF angiography for the evaluation of peripheral
vascular disease.232,242,244-247 Two studies by Owen et al242,244 were widely reported and gave very
promising data. In the initial study, the authors found that TOF MRA detected all vessels identified by
CA but CA failed to detect 22% of run-off vessels identified by MRA. This altered the surgical
242
management in 17%. A second study found that blood vessels depicted on MRA, but not seen on
conventional arteriograms, altered surgical management in 16%.244 The implication was that MRA not
only was an adequate diagnostic tool in its own right but that it surpassed the reference standard for
identification of blood vessels that were surgically important.
Although the early data were very encouraging, acceptance of TOF MRA was hampered by several
substantial limitations. The examinations were extremely time consuming to perform, commonly taking
60 to 120 minutes to complete, which was not tolerated well by most patients. As stated previously,
the technique is prone to overestimation of the degree and length of stenosis that is due mostly to
intravoxel dephasing secondary to turbulent, slow and pulsatile flow. TOF studies also lead to
overestimation of the length of occlusion due to elimination of the retrograde component of flow in
patients with distally reconstituted vessels by the trailing inferior saturation pulses designed to
suppress venous flow.239 Thus, while promising, TOF methods have not had a major impact on clinical
practice.
Combination of CEMRA and TOF

Following the introduction of CEMRA, many direct studies of its use, as compared to TOF MRA and
DSA, were performed.248,249 A meta-analysis of 23 direct comparison studies confirmed the superiority
of CEMRA.249 Review of these studies by the author found a relative diagnostic odds ratio (DOR) for
CEMRA compared with TOF of 7.46 and this improved to 4.53 when review of source images or
MPRs was added to interpretation. The paper concluded that the diagnostic accuracy of 3D CEMRA is
superior to that of 2D TOF MRA.
Once CEMRA was shown to be a useful method for evaluation of peripheral vessels, it was
incrementally utilized as a portion of run-off examinations. Combination techniques, using CEMRA for
one or more stations and TOF for the remaining station(s), developed for two reasons. Manual and
automated bolus-chase techniques were not yet developed or were otherwise unavailable and
contemporary belief was that both CEMRA and TOF were better in specific anatomic locations. It was
hypothesized that the accuracy of MR angiography varied with anatomic level; 2D TOF was felt to be
most accurate below the knee while CEMRA was better in the more proximal, larger vessels. 229,243
The rationale behind this was the belief that distal vessels are more difficult to study, because the
concentration of contrast material is reduced or the arrival of the bolus may be delayed, particularly if
proximal vessels are diseased.
76 of 104 20-04-2010 00:28

Add to lightbox
Add to lightbox
Figure 30-57 Prosthetic hip generating a false-positive finding of right colon iliac artery stenosis. A
clear signal dropout (arrow) is depicted on the CEMRA simulating a focal occlusion (A). However,
review of the scout images reveals an artificial right hip generating a large area of susceptibility
artifact (arrow) (B).
Most combination approaches use two gadolinium injections to perform separate MRAs, one of the
distal aorta to common femoral artery (CFA) region, followed by a second of the CFA to popliteal
region. TOF imaging of the calf and feet is performed prior to gadolinium-enhanced 3D MR
angiography. If TOF imaging is obtained after gadolinium-enhanced 3D MR angiography, the presence
77 of 104 20-04-2010 00:28

of circulating contrast medium can limit the ability to effectively suppress the signal from veins. 250 The
use of two low-dose CEMRAs during a single examination has been a feasible approach to increase
anatomic coverage. One often-cited criticism of this dual-contrast infusion approach is that residual
contrast will decrease the image quality on the second CEMRA. However, this residual gadolinium
does not significantly change the diagnostic usefulness or overall image quality of the second
251
angiogram.
page 812
page 813
There are advantages to the combination approach. Each run-off study essentially consists of multiple
discrete scans, so scan parameters and image resolution can be optimized for each individual location,
unlike earlier bolus-chase techniques where the same imaging parameters had to be repeated at each
anatomic station. Disadvantages include a somewhat time-consuming approach, which can exceed one
hour depending on the center's experience. When multiple gadolinium doses are used, often a lower
dose is used for each station than may be optimal if performed separately.
Bolus Chase Strategies

The bolus-chase or moving-table concept of performing MR angiography of the lower extremity
arteries relies on imaging sequential anatomic sites, during and following a single prolonged gadolinium
infusion, while there is preferential concentration of contrast in the arteries. Using this approach, all
three stations, including the calves, are acquired using the gadolinium-enhanced approach (Fig. 30-58).
The patient is moved during the injection of contrast medium, in a manner similar to that used for
conventional angiograms. With improvements in spatial and temporal resolution, bolus-chase MR
angiography of the peripheral vasculature has evolved into a proven technique that is currently the most
11,237,238,252-254
common MR approach used clinically.
The bolus-chase approach to peripheral MR angiography extends the anatomic coverage that can be
acquired with a given dose of contrast medium and has the capacity to greatly reduce table times.
Many differing strategies for the rate, dose, and duration of contrast administration have been
proposed. At our center we currently use a biphasic approach of 1 cc/s for 10 seconds, followed by
0.6 cc/s for 50 sec, for a total volume of 40 cc.
Bolus-chase CEMRA is performed in one of three ways: manually moving or sliding the patient,
manually moving the table or using automated table movement. With manual table translation, the
255
imaging table is moved manually 40-50 cm between two acquisitions. On most scanners this "tricks"
the MR system into rapidly rescanning a new anatomic region by unlatching the table from its motor
drive and moving it far enough to image the next anatomic segment. Most MR systems would normally
perform a time-consuming prescan prior to acquiring the data once the patient is displaced along the
z-axis; however, this would allow the contrast bolus to degrade considerably and result in poorly
enhanced angiograms. By unlatching and moving the table, the MR system scans without the
time-consuming prescan and its associated adjustments. With practice, the interscan delay required for
movement can approach 2-3 seconds, a speed which is still faster than the more routinely performed
automated table movement method.
The same principle can be adopted for the manual patient motion approach, where the patient is slid
229,241
40-50 cm and the scan repeated rapidly. For this approach, patients are placed on a device that
can slide or roll over the MR table, which remains stationary.
78 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-58 CEMRA using a bolus chase allows for accurate rapid angiographic studies of the lower
extremities in a time-efficient manner. This 57-year-old male with recurrent right thigh pain has a large
right common iliac aneurysm (arrow), but has adequate outflow to the right lower extremity.
The automated table movement approach is simpler to perform, requires fewer MR system
technologists to be involved, and bestows precision in reproducing patient positions (useful for applying
subtraction techniques).237,238 Because the software to perform the procedure is now widely available,
this is the most common approach to bolus-chase CEMRA currently in use.
One of the advantages of bolus-chase MRA is that it addresses the issue of extended anatomic
coverage in the most time-efficient manner, routinely taking only 15-20 minutes for the entire exam to
be completed. It also makes efficient use of a single contrast dose. Disadvantages include the fact that
the automated technique requires additional software and hardware. Use of dedicated peripheral
vascular coils with bolus-chase strategies provides substantial improvements in the signal-to-noise
ratio, especially in the distal anatomic segments where the vessels are substantially smaller. 13,14,254
page 813
page 814
The most significant challenge to this approach is that acquisition of optimally high-resolution CEMRA
data at three consecutive locations takes substantially longer than the transit time of contrast from the
aorta to the feet. If the exam is protocolled so that there is higher resolution in the proximal stations
79 of 104 20-04-2010 00:28

and the acquisition time of the proximal stations increases, the likelihood of venous enhancement within
the legs on the third station will also increase and this will impair image quality. Prescribing the study
so that sufficiently fast data acquisition occurs at the first two locations (in order to eliminate the
chance of venous enhancement within the legs) sacrifices image resolution at these levels. A balance
between speed and adequate resolution in the proximal stations must be attained in order to "beat the
bolus" to the calf station.
One commonly performed method of minimizing venous contamination in the distal station is
manipulation of how k-space is filled during the examination. By mapping central k-space data as soon
as contrast arrives in the artery of interest, the risk of venous enhancement is reduced or eliminated,
even if venous enhancement appears late in the scan. Mapping can be performed using various
methods, referred to commonly as centric or elliptical centric view order k-space filling.90,256 This
effectively reduces the length of time available for venous enhancement by approximately one half of
the scan acquisition time. Centric or elliptical centric k-space filling is commonly used in the calf for this
reason.
Another drawback of bolus-chase MRA is that the method is based upon the assumption that there are
equal blood flow rates in both legs. If there are one or more stenoses in one leg, there may be
substantially differing flow rates from one side to the other. As the table moves, the arteries in one leg
may be substantially less well enhanced than those in the contralateral leg (Fig. 30-59).
A novel approach currently gaining clinical support is the performance of a second contrast timing bolus
and separate acquisitions for the calves and the pelvis, sometimes referred to as the "hybrid"
240
approach. This improves reliability and reduces venous contamination in the calf to levels that may
render conventional angiography obsolete. Morash et al240 used the combination dual-timing/dual-
injection technique and found an overall sensitivity of 99%, specificity of 97%, and accuracy of 98%.
They reported that venous contamination was virtually eliminated and that the diagnostic quality of calf
and foot vessels was significantly superior to conventional bolus-chase magnetic resonance techniques
(P < .01).
Other Approaches
80 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-59 Asymmetric enhancement of the lower extremities is a pitfall of bolus-chase MRA. A
distal left SFA stenosis (arrow) decreased the blood flow to the left calf on this MR angiogram. While
the right calf has mixed arterial and venous enhancement (arrow), the left calf arteries are not yet fully
enhanced.
There are other approaches to CEMRA apart from the standard 3D approach discussed above. The
TRICKS technique, which is highly effective for imaging the calf/ankle/foot region, has been discussed
earlier. Dynamic contrast-enhanced 2D imaging offers shorter scan times compared with 3D imaging.
2D imaging can achieve high in-plane spatial resolution in a short scan time but there is no
through-plane resolution.252,257,258 This technique is not directly applicable in the proximal anatomic
segment because the tortuosity of the arteries of the pelvis and thigh would create confusing
overlapping projections, unless multiple obliqued data sets were performed. In the calf, the 2D
technique has been shown to be useful.
Parallel Imaging
Parallel imaging techniques such as SENSE, SMASH, and GRAPPA are now applied to many MR
pulse sequences (see Chapter 8). They increase acquisition speed by utilization of multiple
phased-array coils with known sensitivities. Increased acquisition speed can be used to either reduce
the scan acquisition time or increase resolution scan by acquiring more phase lines in the same
acquisition time. A combination of reduced acquisition time and increased resolution can also be used.
Using parallel techniques, there is a well-described reduction in SNR proportional to the square root of
the time savings. At a factor of 2, this does not substantially affect the image quality of most
CEMRA.118
page 814
page 815
81 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-60 Parallel techniques reduced the acquisition time from 22 to 12 seconds for this pelvic
bolus-chase MRA. Excellent SNR and spatial resolution remain. The study depicts an arteriovenous
(AV) fistula (arrow) between the right common femoral artery and vein.
Parallel techniques have been used successfully with peripheral CEMRA (Fig. 30-60).118 Currently
scan time is typically reduced by a factor of two, although higher SENSE factors will soon be available
that may increase time savings by 4, 8 or even 16. SENSE is particularly attractive for the peripheral
vasculature, where most of the time savings are invested in faster imaging for the first two locations so
that imaging of the third station can occur prior to substantial venous enhancement. Parallel techniques
can then also be used to perform a higher resolution acquisition for the legs.
Patients with lower extremity ischemia can be divided into two distinct clinical groups: those with
intermittent claudication and those with limb-threatening ischemia. Although both conditions are almost
always due to atherosclerosis, they present and are managed differently. Patients with
259
limb-threatening ischemia have a worse prognosis and a much higher rate of amputation. Patients
with claudication have a relatively good prognosis, a low rate of amputation, and require surgical
intervention much less commonly.228,260,261
Intermittent Claudication
82 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-61 CEMRA in this patient complaining of bilateral thigh claudication depicts bilateral focal
high-grade common iliac artery stenoses (arrows). The focal high-grade stenoses are amenable to
percutaneous transluminal angioplasty (PTA) and stent placement.
Claudication is relatively common, reported in at least 10% of the population over the age of 70
years.262 Patients with claudication have pain in a functional muscle group that is reproducible with
exercise and relieved by cessation of exercise. Symptoms are determined by the level of the stenosis
and generally manifest one anatomic level below the lesion. For example, patients with iliac artery
stenosis present with buttock and thigh while those with superficial femoral artery (SFA) disease have
calf claudication.
The natural history of intermittent claudication is generally encouraging. Up to 80% of such patients
improve significantly with treatment and amputation rates are low, approaching 1% to 2% per
year.260,261,263 The vast majority of patients with claudication are treated conservatively because of this
relatively benign course and the risks associated with bypass surgery. Less invasive treatment, such
as angioplasty and stent placement, is now a therapeutic option for this group.231,264-268
The goal in imaging patients with claudication is to distinguish cases that are amenable to angioplasty
from those that are not. Generally, patients with clinically relevant, short-segment (5 cm) lesions that
are high grade and localized are referred for angioplasty and often stent placement (Fig. 30-61). In
general, success rates are highest for short segmental stenoses, while longer lesions are technically
more difficult but possible to treat.
Studies specifically comparing TOF to CEMRA uniformly demonstrate greater accuracy with
86,248,269
gadolinium-enhanced 3D MR angiography. Studies have confirmed that CEMRA has sufficient
diagnostic accuracy for treatment planning.230,248,269 As discussed above, TOF or phase-contrast MR
angiography was used alone or in combination with CEMRA of one or more anatomic
234,246,270
segments. After several years of some controversy, the bolus-chase technique is now the
favored diagnostic approach.237-239
page 815
page 816
83 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-62 Diffuse bilateral calf stenoses and occlusions are depicted on this high-resolution calf
station from a bolus-chase MRA.
There are several benefits to the use of CEMRA for patients with claudication. The study is performed
routinely as an outpatient procedure. Studies can be performed earlier in the course of the disease,
rather than delaying performance of an angiogram until symptoms were more limb threatening, as was
previously done when DSA was the only option. CEMRA allows for substantial reduction in both image
acquisition time and overall exam time compared to prior techniques. The gadolinium chelates also do
not have the nephrotoxic effects and other adverse reactions that may be encountered with iodinated
contrast media.112,271
Limb-Threatening Ischemia
Patients with limb-threatening ischemia have pain at rest, skin ulceration or gangrene. Limb-threatening
ischemia occurs because arterial blockages prevent adequate resting blood flow from meeting
baseline metabolic demands. Gangrene occurs when arterial flow is so poor that tissue becomes
necrotic. Treating physicians must combine the results of noninvasive studies and clinical findings to
determine the feasibility of surgical limb salvage. Urgent evaluation of the patient with limb-threatening
ischemia is essential and the patient with critical ischemia has a poor prognosis and a high probability
of amputation if timely intervention is withheld.
84 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-63 Two stations from a bolus-chase MRA depict a right popliteal to distal anterior tibial
artery bypass graft (arrow). While long distal grafts have a high failure rate, they may be the only
alternative to amputation in patients with limb-threatening ischemia.
Unlike claudicators, patients with limb-threatening ischemia typically have multiple areas of occlusive
disease, complicating both image interpretation and treatment (Fig. 30-62). The imaging study must
include a complete evaluation from the abdominal aorta to the foot. Inflow (proximal) stenoses are
important to identify and address to improve perfusion and enhance the blood flow into a distal bypass.
Surgically, it is crucial to restore pulsatile flow to the level of gangrene. A comprehensive imaging study
of the entire arterial system from the aorta to the pedal arch is therefore required so that all areas of
272
stenosis or occlusion can be addressed.
An understanding of the potential surgical interventions that are required for limb-threatening ischemia
is useful when interpreting MR angiograms of the lower extremities. The popliteal artery is an
especially important vessel because it is a frequent target for insertion of surgical bypass grafts. It is
useful to provide the surgeon with osseous landmarks when depicting the popliteal artery so that the
site of the surgical incision and the type of graft material are appropriate.273
page 816
page 817
85 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-64 Bolus-chase MRA depicts patent femorofemoral and right femoropopliteal bypass grafts
(arrows). Both grafts are widely patent without thrombosis or anastomotic stenosis.
Accurate imaging of the calf and foot vessels is also essential and will determine the type of proposed
bypass. In the case of multiple stenoses, a bypass graft would likely be inserted beyond the distal
occlusion in the leg so that there is direct-line flow to the gangrenous foot (Fig. 30-63).274 In cases of
severe disease, a bypass may be placed into a "blind" or isolated popliteal segment. The isolated
popliteal artery has no direct inflow or outflow.274,275 Bypass to such a vessel may be desirable when
the tibial arteries appear inadequate as outflow vessels.
In the recent past, the MR angiographic approach to the tibial vessels was dominated by the use of
86 of 104 20-04-2010 00:28

TOF techniques.233,242-245,276 Despite the recognized benefits of CEMRA in the more proximal vessels,
it was believed that a flow-dependent strategy was better than contrast-enhanced techniques. These
concerns arose from the fact that it was often a challenge to achieve sufficient concentration of
contrast to effectively demonstrate the distal arterial vessels, and image resolution in the calf was
limited with earlier techniques.
With current techniques, improved automated software, dedicated peripheral vascular coils, and
improved MR system hardware, routine imaging of the calf and foot is now possible using a
multistation CEMRA technique.237-240 CEMRA circumvents in-plane saturation effects that limit the
demonstration of the proximal anterior tibial artery and portions of the pedal arch with TOF techniques.
Post-Procedural Imaging
MR is useful following surgical intervention and treatment for ischemia. Patients who have previously
undergone surgical bypass and who require an anatomic assessment of a treated segment of disease
277
can benefit from MR angiography. CEMRA can be used to assess the surgical grafts and confirm a
suspicion of a failing graft (Fig. 30-64). Femoro-femoral crossover grafts are difficult to assess with
TOF techniques because their horizontal positioning would be very vulnerable to in-plane saturation
effects. However, femorofemoral crossover grafts and other long segment grafts are particularly well
suited to CEMRA with its larger field of view, sensitivity to flow in all planes, and relatively high spatial
resolution.
MR angiography has been shown to be useful for assessment of pelvic arteries following
278
angioplasty. However, once a stent has been placed, artifact is generated from the ferromagnetic
material in the stent wall and there is a risk of signal loss from susceptibility effects. Stent patency can
sometimes be determined but CEMRA is not suitable for routine stent evaluation owing to signal
intensity dropout. However, CEMRA can provide information about the vascular anatomic areas
proximal and distal to the stent that can often be useful.
Peripheral Aneurysms
Peripheral arterial aneurysms are usually located in the femoropopliteal region. They range in size from
1 cm to over 5 cm depending on location. Peripheral aneurysms can be painful and may lead to
complications such as distal emboli and thrombosis. Rupture is rare but may be life threatening. MR
allows depiction of aneurysms and of the precise site, dimensions, quantity, and quality of peripheral
thrombus. Unlike MRI, DSA cannot provide direct information about the wall of the vessel and about
the thrombus. MR is useful in the diagnosis of these lesions and can be very useful in the planning of
interventions (Figs. 30-65 and 30-66). It provides exact measurements of the vessels before and after
the aneurysm, which is essential to plan intervention.
Upper Extremity CEMRA

page 817
page 818
87 of 104 20-04-2010 00:28

Add to lightbox
88 of 104 20-04-2010 00:28

Add to lightbox
Figure 30-65 Thrombosed left popliteal artery aneurysm in a 75 year old with complaints of left calf
pain. MIP image of the bolus-chase study (A) depicts occlusion (arrow) of a 10 cm segment of the left
popliteal artery. MPR reveals a diffuse popliteal artery aneurysm filled with occluding thrombus (arrow)
(B).
The blood supply to the hand is quite variable, with the most common arrangement being deep and
superficial palmar arches arising respectively from the radial and ulnar arteries. CEMRA is an effective
means for depicting the arterial supply to the arm and hand.309 For evaluation of the distal hand
vessels, very high levels of spatial resolution are required, e.g., 256 × 256 matrix with FOV less than
16 cm and slice thickness less than 1 mm (Fig. 30-67). It can be problematic to obtain such high levels
of spatial resolution with fast temporal resolution (e.g., 5-10 seconds, even faster for dialysis fistulas).
Parallel imaging techniques may be helpful. Alternatively, one can administer two separate infusions of
contrast agent, one for a time-resolved acquisition with a large FOV and lower spatial resolution and a
second for a slower acquisition with high spatial resolution. A small surface coil designed for the hand
and wrist should be used for the high-resolution study to ensure an adequate signal-to-noise ratio,
although a larger coil may be needed to show the arm vessels. Clinical indications include arterial
mapping prior to reconstructive surgery, management of trauma, characterization of masses, vascular
malformations, aneurysms, and inflammatory disorders.310
page 818
page 819
89 of 104 20-04-2010 00:28

Add to lightbox
90 of 104 20-04-2010 00:28

Add to lightbox
page 819
page 820
Figure 30-66 Immunosuppressed patient with mycotic pseudoaneurysm of popliteal artery. A, CEMRA
shows proximal occlusion of the popliteal artery and faint filling of the pseudoaneurysm. B, The extent
of the pseudoaneurysm is better appreciated on an axial T1-weighted image.
Add to lightbox
Figure 30-67 Early (left) and delayed (right) phases from high-resolution CEMRA of the hand in a
patient with Raynaud's phenomenon. There is non-filling of the ulnar artery and poor filling of the digital
arteries distally.
REFERENCES
1. Damadian R, Goldsmith M, Minkoff L: NMR in cancer: XVI. FONAR image of the live human body. Physiol Chem Phys
9:97-100, 108, 1977. Medline Similar articles
2. Bottomley PA, Edelstein WA, Leue WM, et al: Head and body imaging by hydrogen nuclear magnetic resonance. Magn
3. Herfkens RJ, Higgins CB, Hricak H, et al: Nuclear magnetic resonance imaging of the cardiovascular system: normal and
91 of 104 20-04-2010 00:28

pathologic findings. Radiology 147:749-759, 1983. Medline Similar articles

4. Herfkens RJ, Higgins CB, Hricak H, et al: Nuclear magnetic resonance imaging of atherosclerotic disease. Radiology
5. Axel L, Herman GT, Udupa JK, et al: Three-dimensional display of nuclear magnetic resonance (NMR) cardiovascular
images. J Comput Assist Tomogr 7:172-174, 1983. Medline Similar articles
6. Prince MR, Yucel EK, Kaufman JA, et al: Dynamic gadolinium-enhanced three-dimensional abdominal MR arteriography. J
7. Prince MR: Gadolinium-enhanced MR aortography. Radiology 191:155-164, 1994. Medline Similar articles
Similar articles
9. Lanzer P, Botvinick EH, Schiller NB, et al: Cardiac imaging using gated magnetic resonance. Radiology 150:121-127,
10. Ruehm SG, Wiesner W, Debatin JF: Pelvic and lower extremity veins: contrast-enhanced three-dimensional MR
venography with a dedicated vascular coil-initial experience. Radiology 215:421-427, 2000. Medline Similar articles
11. Ho VB, Choyke PL, Foo TK, et al: Automated bolus chase peripheral MR angiography: initial practical experiences and
future directions of this work-in-progress. J Magn Reson Imaging 10:376-388, 1999. Medline Similar articles
12. Riederer SJ, Bernstein MA, Breen JF, et al: Three-dimensional contrast-enhanced MR angiography with real-time
fluoroscopic triggering: design specifications and technical reliability in 330 patient studies. Radiology 215:584-593, 2000.
13. Lenhart M, Herold T, Volk M, et al: [Contrast media-enhanced MR angiography of the lower extremity arteries using a
dedicated peripheral vascular coil system. First clinical results]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr
14. Huber A, Scheidler J, Wintersperger B, et al: Moving-table MR angiography of the peripheral run-off vessels: comparison of
body coil and dedicated phased array coil systems. Am J Roentgenol 180:1365-1373, 2003.
15. Anbarasu A, Harris PL, McWilliams RG: The role of gadolinium-enhanced MR imaging in the preoperative evaluation of
inflammatory abdominal aortic aneurysm. Eur Radiol 12 Suppl 3:S192-195, 2002.
16. Schoenberg SO, Wunsch C, Knopp MV, et al: Abdominal aortic aneurysm. Detection of multilevel vascular pathology by
time-resolved multiphase 3D gadolinium MR angiography: initial report. Invest Radiol 34:648-659, 1999.
17. Yamada N, Okita Y, Minatoya K, et al: Preoperative demonstration of the Adamkiewicz artery by magnetic resonance
angiography in patients with descending or thoracoabdominal aortic aneurysms. Eur J Cardiothorac Surg 18:104-111, 2000.
18. Nasim A, Thompson MM, Sayers RD, et al: Role of magnetic resonance angiography for assessment of abdominal aortic
aneurysm before endoluminal repair. Br J Surg 85:641-644, 1998. Medline Similar articles
19. Engellau L, Albrechtsson U, Dahlstrom N, et al: Measurements before endovascular repair of abdominal aortic aneurysms.
MR imaging with MRA vs. angiography and CT. Acta Radiol 44:177-184, 2003. Medline Similar articles
20. Lutz AM, Willmann JK, Pfammatter T, et al: Evaluation of aortoiliac aneurysm before endovascular repair: comparison of
contrast-enhanced magnetic resonance angiography with multidetector row computed tomographic angiography with an
automated analysis software tool. J Vasc Surg 37:619-627, 2003. Medline Similar articles
21. Ludman CN, Yusuf SW, Whitaker SC, et al: Feasibility of using dynamic contrast-enhanced magnetic resonance
angiography as the sole imaging modality prior to endovascular repair of abdominal aortic aneurysms. Eur J Vasc Endovasc
22. Faries PL, Agarwal G, Lookstein R, et al: Use of cine magnetic resonance angiography in quantifying aneurysm pulsatility
associated with endoleak. J Vasc Surg 38:652-656, 2003. Medline Similar articles
23. Gorich J, Rilinger N, Soldner J, et al: Endovascular repair of aortic aneurysms: treatment of complications. J Endovasc
24. Gorich J, Rilinger N, Sokiranski R, et al: Leakages after endovascular repair of aortic aneurysms: classification based on
findings at CT, angiography, and radiography. Radiology 213:767-772, 1999. Medline Similar articles
25. White GH, Yu W, May J, et al: Endoleak as a complication of endoluminal grafting of abdominal aortic aneurysms:
classification, incidence, diagnosis, and management. J Endovasc Surg 4:152-168, 1997. Medline Similar articles
26. Wicky S, Fan CM, Geller SC, et al: MR angiography of endoleak with inconclusive concomitant CT angiography. Am J
Roentgenol 181:736-738, 2003.
27. Bruce IN, Bell AL: A comparison of two nomenclature systems for primary systemic vasculitis. Br J Rheumatol 36:453-458,
28. Sharma BK, Jain S, Suri S, Numano F: Diagnostic criteria for Takayasu arteritis. Int J Cardiol 54 Suppl:S141-147, 1996.
29. Schoen F, Cotran R: Blood vessels. In Cotran R, Kumar V, Collins T (eds): Robbins Pathologic Basis of Disease, 6th ed.
Philadelphia: 1999, pp 493-542.
30. Choe YH, Han BK, Koh EM, et al: Takayasu's arteritis: assessment of disease activity with contrast-enhanced MR imaging.
Am J Roentgenol 175:505-511, 2000.
31. Coady MA, Rizzo JA, Hammond GL, et al: Penetrating ulcer of the thoracic aorta: what is it? How do we recognize it? How
92 of 104 20-04-2010 00:28

do we manage it? J Vasc Surg 27:1006-1015; discussion 1015-1006, 1998. Medline Similar articles
32. Ganaha F, Miller DC, Sugimoto K, et al: Prognosis of aortic intramural hematoma with and without penetrating
atherosclerotic ulcer: a clinical and radiological analysis. Circulation 106:342-348, 2002. Medline Similar articles
33. Yucel EK, Steinberg FL, Egglin TK, et al: Penetrating aortic ulcers: diagnosis with MR imaging. Radiology 177:779-781,
34. Vilacosta I, Roman JA: Acute aortic syndrome. Heart 85:365-368, 2001. Medline Similar articles
35. Choi SH, Choi SJ, Kim JH, et al: Useful CT findings for predicting the progression of aortic intramural hematoma to overt
aortic dissection. J Comput Assist Tomogr 25:295-299, 2001. Medline Similar articles
36. Hayashi H, Matsuoka Y, Sakamoto I, et al: Penetrating atherosclerotic ulcer of the aorta: imaging features and disease
concept. Radiographics 20:995-1005, 2000. Medline Similar articles
37. Murray JG, Manisali M, Flamm SD, et al: Intramural hematoma of the thoracic aorta: MR image findings and their
prognostic implications. Radiology 204:349-355, 1997. Medline Similar articles
38. Krinsky GA, Rofsky NM, DeCorato DR, et al: Thoracic aorta: comparison of gadolinium-enhanced three-dimensional MR
angiography with conventional MR imaging. Radiology 202:183-193, 1997. Medline Similar articles
39. Cambria RP, Brewster DC, Gertler J, et al: Vascular complications associated with spontaneous aortic dissection. J Vasc
40. Hagspiel KD, Leung DA, Angle JF, et al: MR angiography of the mesenteric vasculature. Radiol Clin North Am 40:867-886,
41. Bogaert J, Meyns B, Rademakers FE, et al: Follow-up of aortic dissection: contribution of MR angiography for evaluation of
the abdominal aorta and its branches. Eur Radiol 7:695-702, 1997. Medline Similar articles
42. Marzelle J, Gigou F, Cormier F, et al: [Isolated dissecting aneurysm of the superior mesenteric artery. Dysplasia, a rare
cause of mesenteric ischemia. Report of a case]. Presse Med 21:1157-1159, 1992. Medline Similar articles
43. Stemerman DH, Krinsky GA, Lee VS, et al: Thoracic aorta: rapid black-blood MR imaging with half-Fourier rapid
acquisition with relaxation enhancement with or without electrocardiographic triggering. Radiology 213:185-191, 1999.
44. Agildere AM, Tutar NU, Demirag A, et al: Renal magnetic resonance angiography with Gd-DTPA in living renal transplant
donors. Transplant Proc 31:3317-3319, 1999. Medline Similar articles
page 820
page 821
45. Bakker J, Beek FJ, Beutler JJ, et al: Renal artery stenosis and accessory renal arteries: accuracy of detection and
visualization with gadolinium-enhanced breath-hold MR angiography. Radiology 207:497-504, 1998. Medline
Similar articles
46. Baskaran V, Pereles FS, Nemcek AA Jr, et al: Gadolinium-enhanced 3D MR angiography of renal artery stenosis: a pilot
Acad Radiol 9:50-59, 2002.
47. Bass JC, Prince MR, Londy FJ, Chenevert TL: Effect of gadolinium on phase-contrast MR angiography of the renal
arteries. Am J Roentgenol 168:261-266, 1997.
48. Bongers V, Bakker J, Beutler JJ, et al: Assessment of renal artery stenosis: comparison of captopril renography and
gadolinium-enhanced breath-hold MR angiography. Clin Radiol 55:346-353, 2000. Medline Similar articles
49. Choyke PL, Walther MM, Wagner JR, et al: Renal cancer: preoperative evaluation with dual-phase three-dimensional MR
50. De Cobelli F, Vanzulli A, Sironi S, et al: Renal artery stenosis: evaluation with breath-hold, three-dimensional, dynamic,
gadolinium-enhanced versus three-dimensional, phase-contrast MR angiography. Radiology 205:689-695, 1997. Medline
Similar articles
51. De Cobelli F, Venturini M, Vanzulli A, et al: Renal arterial stenosis: prospective comparison of color Doppler US and
breath-hold, three-dimensional, dynamic, gadolinium-enhanced MR angiography. Radiology 214:373-380, 2000. Medline
Similar articles
52. Dong Q, Schoenberg SO, Carlos RC, et al: Diagnosis of renal vascular disease with MR angiography. Radiographics
53. Fain SB, King BF, Breen JF, et al: High-spatial-resolution contrast-enhanced MR angiography of the renal arteries: a
prospective comparison with digital subtraction angiography. Radiology 218:481-490, 2001. Medline Similar articles
54. Ferreiros J, Mendez R, Jorquera M, et al: Using gadolinium-enhanced three-dimensional MR angiography to assess
arterial inflow stenosis after kidney transplantation. Am J Roentgenol 172:751-757, 1999.
55. Grenier N, Trillaud H, Combe C, et al: Diagnosis of renovascular hypertension: feasibility of captopril-sensitized dynamic
MR imaging and comparison with captopril scintigraphy. Am J Roentgenol 166:835-843, 1996.
56. Grist T: MR angiography of the renal arteries during a breath-hold using gadolinium-enhanced 3D TOF with k-space
zero-filling and a contrast timing scan. In: Proceedings of the Fourth Meeting of the International Society for Magnetic
Resonance in Medicine. Berkeley, CA: 1996, p 163.
57. Hahn U, Miller S, Nagele T, et al: Renal MR angiography at 1.0 T: three-dimensional (3D) phase-contrast techniques versus
93 of 104 20-04-2010 00:28

gadolinium-enhanced 3D fast low-angle shot breath-hold imaging. Am J Roentgenol 172:1501-1508, 1999.

58. Halpern EJ, Mitchell DG, Wechsler RJ, et al: Preoperative evaluation of living renal donors: comparison of CT angiography
and MR angiography. Radiology 216:434-439, 2000. Medline Similar articles
59. Hany TF, Debatin JF, Leung DA, Pfammatter T: Evaluation of the aortoiliac and renal arteries: comparison of breath-hold,
contrast-enhanced, three-dimensional MR angiography with conventional catheter angiography. Radiology 204:357-362, 1997.
60. Hany TF, Leung DA, Pfammatter T, Debatin JF: Contrast-enhanced magnetic resonance angiography of the renal arteries.
Original investigation. Invest Radiol 33:653-659, 1998. Medline Similar articles
61. Hohenwalter MD, Skowlund CJ, Erickson SJ, et al: Renal transplant evaluation with MR angiography and MR imaging.
Radiographics 21:1505-1517, 2001. Medline Similar articles
62. Holland GA, Dougherty L, Carpenter JP, et al: Breath-hold ultrafast three-dimensional gadolinium-enhanced MR
angiography of the aorta and the renal and other visceral abdominal arteries. Am J Roentgenol 166:971-981, 1996.
63. Johnson DB, Lerner CA, Prince MR, et al: Gadolinium-enhanced magnetic resonance angiography of renal transplants.
64. Korst MB, Joosten FB, Postma CT, et al: Accuracy of normal-dose contrast-enhanced MR angiography in assessing renal
artery stenosis and accessory renal arteries. Am J Roentgenol 174:629-634, 2000.
65. Kroencke TJ, Wasser MN, Pattynama PM, et al: Gadobenate dimeglumine-enhanced MR angiography of the abdominal
aorta and renal arteries. Am J Roentgenol 179:1573-1582, 2002.
66. Lee VS, Rofsky NM, Ton AT, et al: Angiotensin-converting enzyme inhibitor-enhanced phase-contrast MR imaging to
measure renal artery velocity waveforms in patients with suspected renovascular hypertension. Am J Roentgenol 174:499-508,
2000.
67. Lee VS, Rusinek H, Johnson G, et al: MR renography with low-dose gadopentetate dimeglumine: feasibility. Radiology
68. Liem YS, Kock MC, Ijzermans JN, et al: Living renal donors: optimizing the imaging strategy - decision- and
cost-effectiveness analysis. Radiology 226:53-62, 2003. Medline Similar articles
69. Low RN, Martinez AG, Steinberg SM, et al: Potential renal transplant donors: evaluation with gadolinium-enhanced MR
angiography and MR urography. Radiology 207:165-172, 1998. Medline Similar articles
70. Masunaga H, Takehara Y, Isoda H, et al: Assessment of gadolinium-enhanced time-resolved three-dimensional MR
angiography for evaluating renal artery stenosis. Am J Roentgenol 176:1213-1219, 2001.
71. Neimatallah MA, Dong Q, Schoenberg SO, et al: Magnetic resonance imaging in renal transplantation. J Magn Reson
72. Nelson HA, Gilfeather M, Holman JM, et al: Gadolinium-enhanced breathhold three-dimensional time-of-flight renal MR
angiography in the evaluation of potential renal donors. J Vasc Interv Radiol 10:175-181, 1999. Medline Similar articles
73. Nghiem HV, Winter TC 3rd, Mountford MC, et al: Evaluation of the portal venous system before liver transplantation: value
of phase-contrast MR angiography. Am J Roentgenol 164:871-878, 1995.
74. Prince MR, Narasimham DL, Stanley JC, et al: Gadolinium-enhanced magnetic resonance angiography of abdominal
aortic aneurysms. J Vasc Surg 21:656-669, 1995. Medline Similar articles
75. Prince MR, Schoenberg SO, Ward JS, et al: Hemodynamically significant atherosclerotic renal artery stenosis: MR
angiographic features. Radiology 205:128-136, 1997. Medline Similar articles
76. Qanadli SD, Soulez G, Therasse E, et al: Detection of renal artery stenosis: prospective comparison of captopril-enhanced
Doppler sonography, captopril-enhanced scintigraphy, and MR angiography. Am J Roentgenol 177:1123-1129, 2001.
77. Riederer SJ, Fain SB, Kruger DG, Busse RF: Real-time imaging and triggering of 3D contrast-enhanced MR angiograms
using MR fluoroscopy. Magma 8:196-206, 1999.
78. Rieumont MJ, Kaufman JA, Geller SC, et al: Evaluation of renal artery stenosis with dynamic gadolinium-enhanced MR
79. Ros PR, Gauger J, Stoupis C, et al: Diagnosis of renal artery stenosis: feasibility of combining MR angiography, MR
renography, and gadopentetate-based measurements of glomerular filtration rate. Am J Roentgenol 165:1447-1451, 1995.
80. Ruehm SG, Weishaupt D, Debatin JF: Contrast-enhanced MR angiography in patients with aortic occlusion (Leriche
syndrome). J Magn Reson Imaging 11:401-410, 2000. Medline Similar articles
81. Schoenberg SO, Bock M, Kallinowski F, Just A: Correlation of hemodynamic impact and morphologic degree of renal
artery stenosis in a canine model. J Am Soc Nephrol 11:2190-2198, 2000. Medline Similar articles
82. Schoenberg SO, Bock M, Knopp MV, et al: Renal arteries: optimization of three-dimensional gadolinium-enhanced MR
angiography with bolus-timing-independent fast multiphase acquisition in a single breath hold. Radiology 211:667-679, 1999.
83. Schoenberg SO, Knopp MV, Bock M, et al: Combined morphologic and functional assessment of renal artery stenosis
using gadolinium enhanced magnetic resonance imaging. Nephrol Dial Transplant 13:2738-2742, 1998. Medline
Similar articles
84. Schoenberg SO, Knopp MV, Londy F, et al: Morphologic and functional magnetic resonance imaging of renal artery
stenosis: a multireader tricenter study. J Am Soc Nephrol 13:158-169, 2002. Medline Similar articles
94 of 104 20-04-2010 00:28

85. Sharafuddin MJ, Stolpen AH, Dixon BS, et al: Value of MR angiography before percutaneous transluminal renal artery
angioplasty and stent placement. J Vasc Interv Radiol 13:901-908, 2002. Medline Similar articles
86. Snidow JJ, Johnson MS, Harris VJ, et al: Three-dimensional gadolinium-enhanced MR angiography for aortoiliac inflow
assessment plus renal artery screening in a single breath hold. Radiology 198:725-732, 1996. Medline Similar articles
87. Steffens JC, Link J, Grassner J, et al: Contrast-enhanced, K-space-centered, breath-hold MR angiography of the renal
arteries and the abdominal aorta. J Magn Reson Imaging 7:617-622, 1997. Medline Similar articles
88. Volk M, Strotzer M, Lenhart M, et al: Time-resolved contrast-enhanced MR angiography of renal artery stenosis: diagnostic
accuracy and interobserver variability. Am J Roentgenol 174:1583-1588, 2000.
89. Willmann JK, Wildermuth S, Pfammatter T, et al: Aortoiliac and renal arteries: prospective intraindividual comparison of
contrast-enhanced three-dimensional MR angiography and multi-detector row CT angiography. Radiology 226:798-811, 2003.
Similar articles
91. Working Group on Renovascular Hypertension. Detection, evaluation, and treatment of renovascular hypertension. Final
report. Arch Intern Med 147:820-829, 1987. Medline Similar articles
92. Derkx FH, Schalekamp MA: Renal artery stenosis and hypertension. Lancet 344:237-239, 1994. Medline
Similar articles
93. Davis BA, Crook JE, Vestal RE, Oates JA: Prevalence of renovascular hypertension in patients with grade III or IV
hypertensive retinopathy. N Engl J Med 301:1273-1276, 1979. Medline Similar articles
94. Schoenberg SO, Knopp MV, Bock M, et al: Renal artery stenosis: grading of hemodynamic changes with cine phase-
contrast MR blood flow measurements. Radiology 203:45-53, 1997. Medline Similar articles
95. Shetty AN, Bis KG, Kirsch M, et al: Contrast-enhanced breath-hold three-dimensional magnetic resonance angiography in
the evaluation of renal arteries: optimization of technique and pitfalls. J Magn Reson Imaging 12:912-923, 2000. Medline
Similar articles
96. Wasser MN, Westenberg J, van der Hulst VP, et al: Hemodynamic significance of renal artery stenosis: digital subtraction
angiography versus systolically gated three-dimensional phase-contrast MR angiography. Radiology 202:333-338, 1997.
97. Huang AJ, Lee VS, Rusinek H: MR imaging of renal function. Radiol Clin North Am 41:1001-1017, 2003. Medline
Similar articles
page 821
page 822
98. Prasad PV, Goldfarb J, Sundaram C, et al: Captopril MR renography in a swine model: toward a comprehensive
evaluation of renal arterial stenosis. Radiology 217:813-818, 2000. Medline Similar articles
99. Lee VS, Rusinek H, Noz ME, et al: Dynamic three-dimensional MR renography for the measurement of single kidney
function: initial experience. Radiology 227:289-294, 2003. Medline Similar articles
100. Kim D, Edelman RR, Kent KC, et al: Abdominal aorta and renal artery stenosis: evaluation with MR angiography.
101. Debatin JF, Spritzer CE, Grist TM, et al: Imaging of the renal arteries: value of MR angiography. Am J Roentgenol
157:981-990, 1991.
102. Kent KC, Edelman RR, Kim D, et al: Magnetic resonance imaging: a reliable test for the evaluation of proximal
atherosclerotic renal arterial stenosis. J Vasc Surg 13:311-318, 1991. Medline Similar articles
103. Gedroyc WM, Negus R, al-Kutoubi A, et al: Magnetic resonance angiography of renal transplants. Lancet 339:789-791,
104. Grist TM: Magnetic resonance angiography of the aorta and renal arteries. Magn Reson Imaging Clin North Am
1:253-269, 1993.
105. Yucel EK, Kaufman JA, Prince M, et al: Time of flight renal MR angiography: utility in patients with renal insufficiency.
106. Hertz SM, Holland GA, Baum RA, et al: Evaluation of renal artery stenosis by magnetic resonance angiography. Am J
107. Loubeyre P, Cahen R, Grozel F, et al: Transplant renal artery stenosis. Evaluation of diagnosis with magnetic resonance
angiography compared with color duplex sonography and arteriography. Transplantation 62:446-450, 1996. Medline
Similar articles
108. Borrello JA, Li D, Vesely TM, et al: Renal arteries: clinical comparison of three-dimensional time-of-flight MR angiographic
sequences and radiographic angiography. Radiology 197:793-799, 1995. Medline Similar articles
109. de Haan MW, Kouwenhoven M, Thelissen RP, et al: Renovascular disease in patients with hypertension: detection with
systolic and diastolic gating in three-dimensional, phase-contrast MR angiography. Radiology198:449-456, 1996.
110. Duda SH, Schick F, Teufl F, et al: Phase-contrast MR angiography for detection of arteriosclerotic renal artery stenosis.
Acta Radiol 38:287-291, 1997. Medline Similar articles
95 of 104 20-04-2010 00:28

111. Gedroyc WM, Neerhut P, Negus R, et al: Magnetic resonance angiography of renal artery stenosis. Clin Radiol
112. Prince MR, Arnoldus C, Frisoli JK: Nephrotoxicity of high-dose gadolinium compared with iodinated contrast. J Magn
113. Lee VS, Flyer MA, Weinreb JC, et al: Image subtraction in gadolinium-enhanced MR imaging. Am J Roentgenol
167:1427-1432, 1996.
114. Foo TK, Saranathan M, Prince MR, Chenevert TL: Automated detection of bolus arrival and initiation of data acquisition in
fast, three-dimensional, gadolinium-enhanced MR angiography. Radiology 203:275-280, 1997. Medline Similar articles
115. Earls JP, Rofsky NM, DeCorato DR, et al: Breath-hold single-dose gadolinium-enhanced three-dimensional MR
aortography: usefulness of a timing examination and MR power injector. Radiology 201:705-710, 1996. Medline
Similar articles
116. Riederer SJ, Fain SB, Kruger DG, Busse RF: 3D contrast-enhanced MR angiography using fluoroscopic triggering and
an elliptical centric view order. Int J Card Imaging 15:117-129, 1999.
117. Deitch JS, Hansen KJ, Regan JD, et al: Infected renal artery pseudoaneurysm and mycotic aortic aneurysm after
percutaneous transluminal renal artery angioplasty and stent placement in a patient with a solitary kidney. J Vasc Surg
118. Weiger M, Pruessmann KP, Kassner A, et al: Contrast-enhanced 3D MRA using SENSE. J Magn Reson Imaging
12:671-677, 2000.
119. Mallouhi A, Schocke M, Judmaier W, et al: 3D MR angiography of renal arteries: comparison of volume rendering and
maximum intensity projection algorithms. Radiology 223:509-516, 2002. Medline Similar articles
120. Niendorf ER, Grist TM, Lee FT Jr, et al: Rapid in vivo measurement of single-kidney extraction fraction and glomerular
filtration rate with MR imaging. Radiology 206:791-798, 1998. Medline Similar articles
121. Geyer JR, Poutasse EF: Incidence of multiple renal arteries on aortography. Report of a series of 400 patients, 381 of
whom had arterial hypertension. JAMA 182:120-125, 1962.
122. Halpern EJ, Nazarian LN, Wechsler RJ, et al: US, CT, and MR evaluation of accessory renal arteries and proximal renal
arterial branches. Acad Radiol 6:299-304, 1999. Medline Similar articles
123. Berland LL, Koslin DB, Routh WD, Keller FS: Renal artery stenosis: prospective evaluation of diagnosis with color duplex
US compared with angiography. Work in progress. Radiology 174:421-423, 1990. Medline Similar articles
124. Desberg AL, Paushter DM, Lammert GK, et al: Renal artery stenosis: evaluation with color Doppler flow imaging.
125. Andreoni KA, Weeks SM, Gerber DA, et al: Incidence of donor renal fibromuscular dysplasia: does it justify routine
angiography? Transplantation 73:1112-1116, 2002. Medline Similar articles
126. Ram CV: Renovascular hypertension. Cardiol Clin 6:483-508, 1988. Medline Similar articles
127. Bude RO, Forauer AR, Caoili EM, Nghiem HV: Is it necessary to study accessory arteries when screening the renal
arteries for renovascular hypertension? Radiology 226:411-416, 2003. Medline Similar articles
128. Ruehm SG, Goyen M, Barkhausen J, et al: Rapid magnetic resonance angiography for detection of atherosclerosis.
Lancet 357:1086-1091, 2001. Medline Similar articles
129. Missouris CG, Buckenham T, Cappuccio FP, MacGregor GA: Renal artery stenosis: a common and important problem in
patients with peripheral vascular disease. Am J Med 96:10-14, 1994. Medline Similar articles
130. Rubin GD, Shiau MC, Leung AN, et al: Aorta and iliac arteries: single versus multiple detector-row helical CT
131. Horton KM, Fishman EK: 3D CT angiography of the celiac and superior mesenteric arteries with multidetector CT data
sets: preliminary observations. Abdom Imaging 25:523-525, 2000.
132. Rubin GD, Dake MD, Napel S, et al: Spiral CT of renal artery stenosis: comparison of three-dimensional rendering
techniques. Radiology 190:181-189, 1994. Medline Similar articles
133. Johnson PT, Halpern EJ, Kuszyk BS, et al: Renal artery stenosis: CT angiography - comparison of real-time volume-
rendering and maximum intensity projection algorithms. Radiology 211:337-343, 1999. Medline Similar articles
134. Williams DM, Lee DY, Hamilton BH, et al: The dissected aorta: part III. Anatomy and radiologic diagnosis of branch-vessel
compromise. Radiology 203:37-44, 1997. Medline Similar articles
135. Ramamoorthy SL, Vasquez JC, Taft PM, et al: Nonoperative management of acute spontaneous renal artery dissection.
Ann Vasc Surg 16:157-162, 2002. Medline Similar articles
136. Mudrick D, Arepally A, Geschwind JF, et al: Spontaneous renal artery dissection: treatment with coil embolization. J Vasc
Interv Radiol 14:497-500, 2003. Medline Similar articles
137. Lumsden AB, Salam TA, Walton KG: Renal artery aneurysm: a report of 28 cases. Cardiovasc Surg 4:185-189, 1996.
138. Sueyoshi E, Sakamoto I, Hayashi N, et al: Ruptured renal artery aneurysm due to Behçet's disease. Abdom Imaging
139. Bulbul MA, Farrow GA: Renal artery aneurysms. Urology 40:124-126, 1992. Medline Similar articles
140. Schorn B, Falk V, Dalichau H, Mohr FW: Kidney salvage in a case of ruptured renal artery aneurysm: case report and
96 of 104 20-04-2010 00:28

literature review. Cardiovasc Surg 5:134-136, 1997. Medline Similar articles

141. Yang JC, Hye RJ: Ruptured renal artery aneurysm during pregnancy. Ann Vasc Surg 10:370-372, 1996. Medline
Similar articles
142. Kawashima A, Sandler CM, Ernst RD, et al: CT evaluation of renovascular disease. Radiographics 20:1321-1340, 2000.
143. Crotty KL, Orihuela E, Warren MM: Recent advances in the diagnosis and treatment of renal arteriovenous malformations
and fistulas. J Urol 150:1355-1359, 1993. Medline Similar articles
144. Honda H, Onitsuka H, Naitou S, et al: Renal arteriovenous malformations: CT features. J Comput Assist Tomogr
145. McKeown BJ, Rankin SC: Aortocaval fistulae presenting with renal failure: CT diagnosis. Clin Radiol 49:570-572, 1994.
146. Hussain SM, Kock MCJM, Ijzermans JNM, et al: MR imaging: a "one-stop shop" modality for preoperative evaluation of
potential living kidney donors. Radiographics 23:505-520, 2004.
147. Shaffer D, Sahyoun AI, Madras PN, Monaco AP: Two hundred one consecutive living-donor nephrectomies. Arch Surg
148. Pozniak MA, Balison DJ, Lee FT Jr, et al: CT angiography of potential renal transplant donors. Radiographics
149. Del Pizzo JJ, Sklar GN, You-Cheong JW, et al: Helical computerized tomography arteriography for evaluation of live renal
donors undergoing laparoscopic nephrectomy. J Urol 162:31-34, 1999. Medline Similar articles
150. Kim TS, Chung JW, Park JH, et al: Renal artery evaluation: comparison of spiral CT angiography to intra-arterial DSA. J
Vasc Interv Radiol 9:553-559, 1998. Medline Similar articles
151. Cecka JM: The UNOS Scientific Renal Transplant Registry. Clin Transpl 1-21, 1999.
152. Omary RA, Baden JG, Becker BN, et al: Impact of MR angiography on the diagnosis and management of renal transplant
dysfunction. J Vasc Interv Radiol 11:991-996, 2000. Medline Similar articles
153. Ali MG, Coakley FV, Hricak H, Bretan PN: Complex posttransplantation abnormalities of renal allografts: evaluation with
154. Amante AJ, Kahan BD: Technical complications of renal transplantation. Surg Clin North Am 74:1117-1131, 1994.
155. Fervenza FC, Lafayette RA, Alfrey EJ, Petersen J: Renal artery stenosis in kidney transplants. Am J Kidney Dis
156. Cosio FG, Pesavento TE, Sedmak DD, et al: Clinical implications of the diagnosis of renal allograft infarction by
percutaneous biopsy. Transplantation 66:467-471, 1998. Medline Similar articles
157. Hobart MG, Streem SB, Gill IS: Renal transplant complications. Minimally invasive management. Urol Clin North Am
158. Newstead CG: Lymphoproliferative disease post-renal transplantation. Nephrol Dial Transplant 15:1913-1916, 2000.
159. Israel GM, Bosniak MA: Renal imaging for diagnosis and staging of renal cell carcinoma. Urol Clin North Am 30:499-514,
160. Kallman DA, King BF, Hattery RR, et al: Renal vein and inferior vena cava tumor thrombus in renal cell carcinoma: CT,
US, MRI and venacavography. J Comput Assist Tomogr 16:240-247, 1992. Medline Similar articles
161. Kraus BB, Ros PR, Abbitt PL, et al: Comparison of ultrasound, CT, and MR imaging in the evaluation of candidates for
TIPS. J Magn Reson Imaging 5:571-578, 1995. Medline Similar articles
162. Lewis WD, Finn JP, Jenkins RL, et al: Use of magnetic resonance angiography in the pretransplant evaluation of portal
vein pathology. Transplantation 56:64-68, 1993. Medline Similar articles
163. Winter TC 3rd, Freeny PC, Nghiem HV, et al: Hepatic arterial anatomy in transplantation candidates: evaluation with
three-dimensional CT arteriography. Radiology 195:363-370, 1995.
page 822
page 823
164. Carr JC, Nemcek AA Jr, Abecassis M, et al: Preoperative evaluation of the entire hepatic vasculature in living liver donors
with use of contrast-enhanced MR angiography and true fast imaging with steady-state precession. J Vasc Interv Radiol
165. Eubank WB, Wherry KL, Maki JH, et al: Preoperative evaluation of patients awaiting liver transplantation: comparison of
multiphasic contrast-enhanced 3D magnetic resonance to helical computed tomography examinations. J Magn Reson Imaging
16:565-575, 2002.
166. Vosshenrich R, Engeroff B, Obenauer S, Grabbe E: [Contrast-enhanced MR angiography of the arterial and portovenous
system of the liver with varying concentrations of contrast medium]. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr
167. Kopka L, Rodenwaldt J, Vosshenrich R, et al: Hepatic blood supply: comparison of optimized dual phase contrast-
enhanced three-dimensional MR angiography and digital subtraction angiography. Radiology 211:51-58, 1999. Medline
97 of 104 20-04-2010 00:28

Similar articles
168. Lee VS, Morgan GR, Teperman LW, et al: MR imaging as the sole preoperative imaging modality for right hepatectomy:
a prospective study of living adult-to-adult liver donor candidates. Am J Roentgenol 176:1475-1482, 2001.
169. Fulcher AS, Szucs RA, Bassignani MJ, Marcos A: Right lobe living donor liver transplantation: preoperative evaluation of
the donor with MR imaging. Am J Roentgenol 176:1483-1491, 2001.
170. Kim BS, Kim TK, Jung DJ, et al: Vascular complications after living related liver transplantation: evaluation with
gadolinium-enhanced three-dimensional MR angiography. Am J Roentgenol 181:467-474, 2003.
171. Applegate GR, Thaete FL, Meyers SP, et al: Blood flow in the portal vein: velocity quantitation with phase-contrast MR
172. Burkart DJ, Johnson CD, Morton MJ, Ehman RL: Phase-contrast cine MR angiography in chronic liver disease. Radiology
173. Finn JP, Kane RA, Edelman RR, et al: Imaging of the portal venous system in patients with cirrhosis: MR angiography vs
duplex Doppler sonography. Am J Roentgenol 161:989-994, 1993.
174. Nghiem HV, Freeny PC, Winter TC 3rd, et al: Phase-contrast MR angiography of the portal venous system: preoperative
findings in liver transplant recipients. Am J Roentgenol 163:445-450, 1994.
175. Suto Y, Ohuchi Y, Kimura T, et al: Three-dimensional black blood MR angiography of the liver during breath holding. A
comparison with two-dimensional time-of-flight MR angiography. Acta Radiol 35:131-134, 1994. Medline
Similar articles
176. Rodgers PM, Ward J, Baudouin CJ, et al: Dynamic contrast-enhanced MR imaging of the portal venous system:
comparison with X-ray angiography. Radiology 191:741-745, 1994. Medline Similar articles
177. Okumura A, Watanabe Y, Dohke M, et al: Contrast-enhanced three-dimensional MR portography. Radiographics
178. Lavelle MT, Lee VS, Rofsky NM, et al: Dynamic contrast-enhanced three-dimensional MR imaging of liver parenchyma:
source images and angiographic reconstructions to define hepatic arterial anatomy. Radiology 218:389-394, 2001.
179. Michels N: Blood Supply and Anatomy of the Upper Abdominal Organs with a Descriptive Atlas. Philadelphia: JB
Lippincott, 1955, pp 139-165.
180. Lee VS, Lavelle MT, Rofsky NM, et al: Hepatic MR imaging with a dynamic contrast-enhanced isotropic volumetric
interpolated breath-hold examination: feasibility, reproducibility, and technical quality. Radiology 215:365-372, 2000.
181. Rofsky NM, Lee VS, Laub G, et al: Abdominal MR imaging with a volumetric interpolated breath-hold examination.
182. Yamaoka Y, Washida M, Honda K, et al: Liver transplantation using a right lobe graft from a living related donor.
Transplantation 57:1127-1130, 1994. Medline Similar articles
183. Makuuchi M, Kawasaki S, Noguchi T, et al: Donor hepatectomy for living related partial liver transplantation. Surgery
184. de Santibanes E, McCormack L, Mattera J, et al: Partial left lateral segment transplant from a living donor. Liver Transpl
185. Marcos A, Fisher RA, Ham JM, et al: Selection and outcome of living donors for adult to adult right lobe transplantation.
Transplantation 69:2410-2415, 2000. Medline Similar articles
186. Todo S, Makowka L, Tzakis AG, et al: Hepatic artery in liver transplantation. Transplant Proc 19:2406-2411, 1987.
187. Rela M, McCall JL, Karani J, Heaton ND: Accessory right hepatic artery arising from the left: implications for split liver
transplantation. Transplantation 66:792-794, 1998. Medline Similar articles
188. Kawagishi N, Ohkohchi N, Fujimori K, et al: Safety of the donor operation in living-related liver transplantation: analysis of
22 donors. Transplant Proc 30:3279-3280, 1998.
189. Glockner JF, Forauer AR, Solomon H, et al: Three-dimensional gadolinium-enhanced MR angiography of vascular
complications after liver transplantation. Am J Roentgenol 174:1447-1453, 2000.
190. Kim BS, Kim TK, Jung DJ, et al: Vascular complications after living related liver transplantation: evaluation with
gadolinium-enhanced three-dimensional MR angiography. Am J Roentgenol 181:467-474, 2003.
191. Stafford-Johnson DB, Hamilton BH, Dong Q, et al: Vascular complications of liver transplantation: evaluation with
gadolinium-enhanced MR angiography. Radiology 207:153-160, 1998. Medline Similar articles
192. Pandharipande PV, Lee VS, Morgan GR, et al: Vascular and extravascular complications of liver transplantation:
comprehensive evaluation with three-dimensional contrast-enhanced volumetric MR imaging and MR
cholangiopancreatography. Am J Roentgenol 177:1101-1107, 2001.
193. Horton KM, Fishman EK: The current status of multidetector row CT and three-dimensional imaging of the small bowel.
Radiol Clin North Am 41:199-212, 2003. Medline Similar articles
194. Li KC, Whitney WS, McDonnell CH, et al: Chronic mesenteric ischemia: evaluation with phase-contrast cine MR imaging.
98 of 104 20-04-2010 00:28

195. Wasser MN, Geelkerken RH, Kouwenhoven M, et al: Systolically gated 3D phase contrast MRA of mesenteric arteries in
suspected mesenteric ischemia. J Comput Assist Tomogr 20:262-268, 1996.
196. Dalman RL, Li KC, Moon WK, et al: Diminished postprandial hyperemia in patients with aortic and mesenteric arterial
occlusive disease. Quantification by magnetic resonance flow imaging. Circulation 94:II206-210, 1996.
197. Ernst O, Asnar V, Sergent G, et al: Comparing contrast-enhanced breath-hold MR angiography and conventional
198. Baden JG, Racy DJ, Grist TM: Contrast-enhanced three-dimensional magnetic resonance angiography of the mesenteric
vasculature. J Magn Reson Imaging 10:369-375, 1999. Medline Similar articles
199. Li KC, Dalman RL, Wright GA: In vivo flow-independent T2 measurements of superior mesenteric vein blood in diagnosis
of chronic mesenteric ischemia: a preliminary evaluation. Acad Radiol 6:530-534, 1999. Medline Similar articles
200. Li KC: Mesenteric occlusive disease. Magn Reson Imaging Clin North Am 6:331-350, 1998.
201. Li KC, Dalman RL, Ch'en IY, et al: Chronic mesenteric ischemia: use of in vivo MR imaging measurements of blood
oxygen saturation in the superior mesenteric vein for diagnosis. Radiology 204:71-77, 1997. Medline Similar articles
202. Li KC, Hopkins KL, Dalman RL, Song CK: Simultaneous measurement of flow in the superior mesenteric vein and artery
with cine phase-contrast MR imaging: value in diagnosis of chronic mesenteric ischemia. Work in progress. Radiology
203. Li KC, Wright GA, Pelc LR, et al: Oxygen saturation of blood in the superior mesenteric vein: in vivo verification of MR
imaging measurements in a canine model. Work in progress. Radiology 194:321-325, 1995. Medline Similar articles
204. Kopka L, Vosshenrich R, Rodenwaldt J, Grabbe E: Differences in injection rates on contrast-enhanced breath-hold three-
dimensional MR angiography. Am J Roentgenol 170:345-348, 1998.
205. Miyazaki T, Yamashita Y, Shinzato J, et al: Two-dimensional time-of-flight magnetic resonance angiography in the coronal
plane for abdominal disease: its usefulness and comparison with conventional angiography. Br J Radiol 68:351-357, 1995.
206. Sadek AG, Mohamed FB, Outwater EK, et al: Respiratory and postprandial changes in portal flow rate: assessment by
phase contrast MR imaging. J Magn Reson Imaging 6:90-93, 1996. Medline Similar articles
207. Keogan MT, Edelman RR: Technologic advances in abdominal MR imaging. Radiology 220:310-320, 2001. Medline
Similar articles
208. Li KC, Pelc LR, Puvvala S, Wright GA: Mesenteric ischemia due to hemorrhagic shock: MR imaging diagnosis and
monitoring in a canine model. Radiology 206:219-225, 1998. Medline Similar articles
209. Hagspiel K, Angle J, Spinosa D: Mesenteric ischemia: angiography and endovascular interventions. In Longo W,
Peterson G, Jacobs D (eds): Intestinal Ischemia Disorders: Pathophysiology and Management. St Louis, MO: Quality Medical
Publishing, 1999, pp 105-154.
210. Chan FP, Li KC, Heiss SG, Razavi MK: A comprehensive approach using MR imaging to diagnose acute segmental
mesenteric ischemia in a porcine model. Am J Roentgenol 173:523-529, 1999.
211. Meaney JF, Prince MR, Nostrant TT, Stanley JC: Gadolinium-enhanced MR angiography of visceral arteries in patients
with suspected chronic mesenteric ischemia. J Magn Reson Imaging 7:171-176, 1997. Medline Similar articles
212. Carlos RC, Stanley JC, Stafford-Johnson D, Prince MR: Interobserver variability in the evaluation of chronic mesenteric
ischemia with gadolinium-enhanced MR angiography. Acad Radiol 8:879-887, 2001. Medline Similar articles
213. Burkart DJ, Johnson CD, Ehman RL: Correlation of arterial and venous blood flow in the mesenteric system based on MR
findings. 1993 ARRS Executive Council Award. Am J Roentgenol 161:1279-1282, 1993.
214. Probst P, Castaneda-Zuniga W, Gomes A, et al: Nonsurgical treatment of splenic-artery aneurysms. Radiology
215. Dunbar JD, Molnar W, Beman FF, Marable SA: Compression of the celiac trunk and abdominal angina. Am J Roentgenol
Radium Ther Nucl Med 95:731-744, 1965. Medline Similar articles
216. Reilly LM, Ammar AD, Stoney RJ, Ehrenfeld WK: Late results following operative repair for celiac artery compression
syndrome. J Vasc Surg 2:79-91, 1985. Medline Similar articles
217. Szilagyi DE, Rian RL, Elliott JP, Smith RF: The cardiac artery compression syndrome: does it exist? Surgery 72:849-863,
218. Jurim O, Shaked A, Kiai K, et al: Celiac compression syndrome and liver transplantation. Ann Surg 218:10-12, 1993.
219. Lee VS, Morgan JN, Tan AG, et al: Celiac artery compression by the median arcuate ligament: a pitfall of end-expiratory
220. McFarland EG, Kaufman JA, Saini S, et al: Preoperative staging of cancer of the pancreas: value of MR angiography
versus conventional angiography in detecting portal venous invasion. Am J Roentgenol 166:37-43, 1996.
221. Sironi S, De Cobelli F, Zerbi A, et al: Pancreatic adenocarcinoma: assessment of vascular invasion with high-field MR
imaging and a phased-array coil. Am J Roentgenol 167:997-1001, 1996.
page 823
page 824
222. Richter A, Gaa J, Niedergethmann M, et al: [Ultrafast magnetic resonance tomography changes the standard in pancreas
99 of 104 20-04-2010 00:28

diagnosis]. Chirurgie 72:697-703, 2001.

223. Catalano C, Pavone P, Laghi A, et al: Pancreatic adenocarcinoma: combination of MR imaging, MR angiography and
MR cholangiopancreatography for the diagnosis and assessment of resectability. Eur Radiol 8:428-434, 1998. Medline
Similar articles
224. Smedby O, Riesenfeld V, Karlson B, et al: Magnetic resonance angiography in the resectability assessment of suspected
pancreatic tumours. Eur Radiol 7:649-653, 1997. Medline Similar articles
225. Procacci C, Biasiutti C, Carbognin G, et al: Spiral computed tomography assessment of resectability of pancreatic ductal
adenocarcinoma: analysis of results. Dig Liver Dis 34:739-747, 2002. Medline Similar articles
226. Dormandy JA, Rutherford RB: Management of peripheral arterial disease (PAD). TASC Working Group. TransAtlantic
Inter-Society Concensus (TASC). J Vasc Surg 31:S1-S296, 2000.
227. Rutkow IM, Ernst CB: An analysis of vascular surgical manpower requirements and vascular surgical rates in the United
States. J Vasc Surg 3:74-83, 1986. Medline Similar articles
228. Criqui MH, Fronek A, Barrett-Connor E, et al: The prevalence of peripheral arterial disease in a defined population.
229. Rofsky NM, Adelman MA: MR angiography in the evaluation of atherosclerotic peripheral vascular disease. Radiology
230. Rofsky NM, Johnson G, Adelman MA, et al: Peripheral vascular disease evaluated with reduced-dose gadolinium-
enhanced MR angiography. Radiology 205:163-169, 1997. Medline Similar articles
231. Pentecost MJ, Criqui MH, Dorros G, et al: Guidelines for peripheral percutaneous transluminal angioplasty of the
abdominal aorta and lower extremity vessels. A statement for health professionals from a special writing group of the Councils
on Cardiovascular Radiology, Arteriosclerosis, Cardio-Thoracic and Vascular Surgery, Clinical Cardiology, and Epidemiology
and Prevention, the American Heart Association. Circulation 89:511-531, 1994. Medline Similar articles
232. Ho KY, de Haan MW, Oei TK, et al: MR angiography of the iliac and upper femoral arteries using four different inflow
techniques. Am J Roentgenol 169:45-53, 1997.
233. Hertz SM, Baum RA, Owen RS, et al: Comparison of magnetic resonance angiography and contrast arteriography in
peripheral arterial stenosis. Am J Surg 166:112-116; discussion 116, 1993. Medline Similar articles
234. Swan JS, Korosec FR, Heisey DM, Grist TM: Evaluation of presaturation pulse thickness and gap for lower extremity
peripheral two-dimensional time-of-flight MR angiography with a signa 1.5-T unit. J Vasc Interv Radiol 6:965-970, 1995.
235. Kaufman JA, McCarter D, Geller SC, Waltman AC: Two-dimensional time-of-flight MR angiography of the lower
extremities: artifacts and pitfalls. Am J Roentgenol 171:129-135, 1998.
236. Schnall MD, Holland GA, Baum RA, et al: MR angiography of the peripheral vasculature. Radiographics 13:920-930,
237. Ho KY, Leiner T, de Haan MW, et al: Peripheral vascular tree stenoses: evaluation with moving-bed infusion-tracking MR
238. Meaney JF, Ridgway JP, Chakraverty S, et al: Stepping-table gadolinium-enhanced digital subtraction MR angiography of
the aorta and lower extremity arteries: preliminary experience. Radiology 211:59-67, 1999. Medline Similar articles
239. Meaney JF: Magnetic resonance angiography of the peripheral arteries: current status. Eur Radiol 13:836-852, 2003.
240. Morasch MD, Collins J, Pereles FS, et al: Lower extremity stepping-table magnetic resonance angiography with multilevel
contrast timing and segmented contrast infusion. J Vasc Surg 37:62-71, 2003. Medline Similar articles
241. Pandharipande PV, Lee VS, Reuss PM, et al: Two-station bolus-chase MR angiography with a stationary table: a simple
alternative to automated-table techniques. Am J Roentgenol 179:1583-1589, 2002.
242. Owen RS, Carpenter JP, Baum RA, et al: Magnetic resonance imaging of angiographically occult run-off vessels in
peripheral arterial occlusive disease. N Engl J Med 326:1577-1581, 1992. Medline Similar articles
243. Carpenter JP, Owen RS, Baum RA, et al: Magnetic resonance angiography of peripheral run-off vessels. J Vasc Surg
16:807-813; discussion 813-805, 1992. Medline Similar articles
244. Owen RS, Baum RA, Carpenter JP, et al: Symptomatic peripheral vascular disease: selection of imaging parameters and
clinical evaluation with MR angiography. Radiology 187:627-635, 1993. Medline Similar articles
245. Yucel EK, Kaufman JA, Geller SC, Waltman AC: Atherosclerotic occlusive disease of the lower extremity: prospective
evaluation with two-dimensional time-of-flight MR angiography. Radiology 187:637-641, 1993. Medline Similar articles
246. Glickerman DJ, Obregon RG, Schmiedl UP, et al: Cardiac-gated MR angiography of the entire lower extremity: a
prospective comparison with conventional angiography. Am J Roentgenol 167:445-451, 1996.
247. Carpenter JP, Golden MA, Barker CF, et al: The fate of bypass grafts to angiographically occult run-off vessels detected
by magnetic resonance angiography. J Vasc Surg 23:483-489, 1996. Medline Similar articles
248. Poon E, Yucel EK, Pagan-Marin H, Kayne H: Iliac artery stenosis measurements: comparison of two-dimensional time-of-
flight and three-dimensional dynamic gadolinium-enhanced MR angiography. Am J Roentgenol 169:1139-1144, 1997.
249. Nelemans PJ, Leiner T, de Vet HC, van Engelshoven JM: Peripheral arterial disease: meta-analysis of the diagnostic
performance of MR angiography. Radiology 217:105-114, 2000. Medline Similar articles
100 of 104 20-04-2010 00:28

250. Rofsky NM, Purdy DE, Johnson G, et al: Suppression of venous signal in time-of-flight MR angiography of the lower
extremities after administration of gadopentetate dimeglumine . Radiology 202:177-182, 1997. Medline
Similar articles
251. Earls JP, Patel NH, Smith PA, et al: Gadolinium-enhanced three-dimensional MR angiography of the aorta and peripheral
arteries: evaluation of a multistation examination using two gadopentetate dimeglumine infusions. Am J Roentgenol
171:599-604, 1998.
252. Wang Y, Lee HM, Avakian R, et al: Timing algorithm for bolus chase MR digital subtraction angiography. Magn Reson
253. Sueyoshi E, Sakamoto I, Matsuoka Y, et al: Aortoiliac and lower extremity arteries: comparison of three-dimensional
dynamic contrast-enhanced subtraction MR angiography and conventional angiography. Radiology 210:683-688, 1999.
254. Ruehm SG, Hany TF, Pfammatter T, et al: Pelvic and lower extremity arterial imaging: diagnostic performance of three-
dimensional contrast-enhanced MR angiography. Am J Roentgenol 174:1127-1135, 2000.
255. Earls JP, DeSena S, Bluemke DA: Gadolinium-enhanced three-dimensional MR angiography of the entire aorta and iliac
arteries with dynamic manual table translation. Radiology 209:844-849, 1998. Medline Similar articles
256. Wilman AH, Riederer SJ: Improved centric phase encoding orders for three-dimensional magnetization-prepared MR
257. Wang Y, Winchester PA, Khilnani NM, et al: Contrast-enhanced peripheral MR angiography from the abdominal aorta to
the pedal arteries: combined dynamic two-dimensional and bolus-chase three-dimensional acquisitions. Invest Radiol
258. Lee HM, Wang Y, Sostman HD, et al: Distal lower extremity arteries: evaluation with two-dimensional MR digital
subtraction angiography. Radiology 207:505-512, 1998. Medline Similar articles
259. Humphrey LL, Palumbo PJ, Butters MA, et al: The contribution of non-insulin-dependent diabetes to lower-extremity
amputation in the community. Arch Intern Med 154:885-892, 1994. Medline Similar articles
260. Imparato AM, Kim GE, Davidson T, Crowley JG: Proceedings: intermittent claudication: its natural course. J Cardiovasc
Surg (Torino) 17:99-100, 1976.
261. Imparato AM, Kim GE, Davidson T, Crowley JG: Intermittent claudication: its natural course. Surgery 78:795-799, 1975.
262. Peabody CN, Kannel WB, McNamara PM: Intermittent claudication. Surgical significance. Arch Surg 109:693-697, 1974.
263. Jonason T, Ringqvist I: Factors of prognostic importance for subsequent rest pain in patients with intermittent claudication.
Acta Med Scand 218:27-33, 1985. Medline Similar articles
264. Wilson SE, Wolf GL, Cross AP: Percutaneous transluminal angioplasty versus operation for peripheral arteriosclerosis.
Report of a prospective randomized trial in a selected group of patients. J Vasc Surg 9:1-9, 1989. Medline
Similar articles
265. Murphy TP, Khwaja AA, Webb MS: Aortoiliac stent placement in patients treated for intermittent claudication. J Vasc
Interv Radiol 9:421-428, 1998. Medline Similar articles
266. Vorwerk D, Gunther RW, Schurmann K, Wendt G: Aortic and iliac stenoses: follow-up results of stent placement after
insufficient balloon angioplasty in 118 cases. Radiology 198:45-48, 1996.
267. Henry M, Amor M, Ethevenot G, et al: Palmaz stent placement in iliac and femoropopliteal arteries: primary and
secondary patency in 310 patients with 2-4-year follow-up. Radiology 197:167-174, 1995.
268. Martin EC, Katzen BT, Benenati JF, et al: Multicenter trial of the wallstent in the iliac and femoral arteries. J Vasc Interv
269. Quinn SF, Sheley RC, Szumowski J, Shimakawa A: Evaluation of the iliac arteries: comparison of two-dimensional time
of flight magnetic resonance angiography with cardiac compensated fast gradient recalled echo and contrast-enhanced three-
dimensional time of flight magnetic resonance angiography. J Magn Reson Imaging 7:197-203, 1997. Medline
Similar articles
270. Steffens JC, Link J, Muller-Hulsbeck S, et al: Cardiac-gated two-dimensional phase-contrast MR angiography of lower
extremity occlusive disease. Am J Roentgenol 169:749-754, 1997.
272. Dardik H, Ibrahim IM, Sussman B, et al: Morphologic structure of the pedal arch and its relationship to patency of crural
vascular reconstruction. Surg Gynecol Obstet 152:645-648, 1981. Medline Similar articles
273. Rofsky NM, Morana G, Adelman MA, et al: Improved gadolinium-enhanced subtraction MR angiography of the
femoropopliteal arteries: reintroduction of osseous anatomic landmarks. Am J Roentgenol 173:1009-1011, 1999.
274. Adelman MA, Jacobowitz GR: Body MR angiography: a surgeon's perspective. Magn Reson Imaging Clin North Am
6:397-416, 1998.
275. Brewster DC, Charlesworth PM, Monahan JE, et al: Isolated popliteal segment v tibial bypass. Comparison of
hemodynamic and clinical results. Arch Surg 119:775-779, 1984. Medline Similar articles
101 of 104 20-04-2010 00:28

276. Cortell ED, Kaufman JA, Geller SC, et al: MR angiography of tibial run-off vessels: imaging with the head coil compared
with conventional arteriography. Am J Roentgenol 167:147-151, 1996.
277. Krinsky G, Jacobowitz G, Rofsky N: Gadolinium-enhanced MR angiography of extraanatomic arterial bypass grafts. Am J
Roentgenol 170:735-741, 1998.
278. Davis CP, Schopke WD, Seifert B, et al: MR angiography of patients with peripheral arterial disease before and after
transluminal angioplasty. Am J Roentgenol 168:1027-1034, 1997.
page 824
page 825
279. Baum RA, Rutter CM, Sunshine JH, et al: Multicenter trial to evaluate vascular magnetic resonance angiography of the
lower extremity. American College of Radiology Rapid Technology Assessment Group. JAMA 274:875-880, 1995.
280. Winchester PA, Lee HM, Khilnani NM, et al: Comparison of two-dimensional MR digital subtraction angiography of the
lower extremity with X-ray angiography. J Vasc Interv Radiol 9:891-899; discussion 900, 1998. Medline Similar articles
281. Link J, Steffens JC, Brossmann J, et al: Iliofemoral arterial occlusive disease: contrast-enhanced MR angiography for
preinterventional evaluation and follow-up after stent placement. Radiology 212:371-377, 1999. Medline Similar articles
282. Ersoy H, Jacobs P, Kent CK, Prince MR: Blood pool MR angiography of aortic stent-graft endoleak. Am J Roentgenol
182:1181-1186, 2004.
283. Nienaber CA, Eagle, KA: Aortic dissection: new frontiers in diagnosis and management. Part I: from etiology to
diagnostic strategies. Circulation 108:528-635, 2003.
284. Dailey PO, Trueblood HW, Stinson EB, et al: Management of acute aortic dissections. Ann Thorac Surg 10: 237-246,
1970.
285. Hagan PG, Nienaber CA, Isselbacher EM, et al: The international registry of acute aortic dissection (IRAD): new insights
into an old disease. JAMA 283:897-903, 2000. Medline Similar articles
286. Pretre R, von Segesser LK: Aortic dissection. Lancet 349:1461-1464, 1997. Medline Similar articles
287. Nienaber CA, von Kodolitsch Y, Petersen B, et al: Intramural hemorrhage of the thoracic aorta: diagnostic and therapeutic
implications. Circulation 92:1465-1472, 1995. Medline Similar articles
288. Braverman AC: Penetrating atherosclerotic ulcers of the aorta. Curr Opin Cardiol 9:591-597, 1994. Medline
Similar articles
289. Nienaber CA, von Kodolitsch Y, Nicolas V, et al: The diagnosis of thoracic aortic dissection by non-invasive imaging
procedures. N Engl J Med 328:1-9, 1993. Medline Similar articles
290. Kunz RP, Oberholzer K, Kuroczynski W, et al: Assessment of chronic aortic dissection: contribution of different
ECG-gated breath-hold MRI techniques. Am J Roentgenol 182:1319-1326, 2004.
291. Inoue T, Watanabe S, Sakurada H, et al: Evaluation of flow volume and flow patterns in the patent false lumen of chronic
aortic dissections using velocity-encoded cine magnetic resonance imaging. Jpn Circ J 64:760-764, 2000. Medline
Similar articles
292. Khan IA, Nair CK: Clinical, diagnostic, and management perspectives of aortic dissection. Chest 122:311-328, 2002.
293. Meszaros I, Morocz J, Szlavi J, et al: Epidemiology and clinicopathology of aortic dissection. Chest 117:1271-1278,
294. Barbant S, Eisenberg M, Schiller N: Diagnostic value of imaging techniques for aortic dissection: Am Heart J
295. Keren A, Kim CB, Hu BS, et al: Accuracy of biplane and multiplane transesophageal echocardiography in diagnosis of
typical acute aortic dissection and intramural hematoma. J Am Coll Cardiol 28: 627-636, 1996.
296. Felson B: Chest Roentgenology. Philadelphia: WB Saunders, 1973, pp 87-92.
297. Woodring JH, Howard TA, Kanga JF: Congenital pulmonary venolobar syndrome revisited. Radiographics 14:349-369,
298. Konen E, Raviv-Zilka L, Cohen RA, et al: Congenital pulmonary venolobar syndrome: spectrum of helical CT findings with
emphasis on computerized reformatting. Radiographics 23:1175-1184, 2003. Medline Similar articles
299. Cho SY, Kim HC, Bae SH, et al: Demonstration of blood supply to pulmonary sequestration by MR and CT angiography. J
Comput Assist Tomogr 20:993-995, 1996. Medline Similar articles
300. Oates SD, Daley RA: Thoracic outlet syndrome. Hand Clin 12:705-718, 1996. Medline Similar articles
301. Remy-Jardin M, Remy J, Masson P, et al: Helical CT angiography of thoracic outlet syndrome: functional anatomy. Am J
Roentgenol 174:1667-1674, 2000.
302. Van Grimberge F, Dymarkowski S, Budts W, Bogaert J: Role of magnetic resonance in the diagnosis of subclavian steal
syndrome. J Magn Reson Imaging 12:39-42, 2000.
303. Matsunaga N, Hayashi K, Okada M, Sakamoto I: Magnetic resonance imaging features of aortic diseases. Top Magn
304. Wedeen VJ, Meuli RA, Edelman RR, et al: Projective imaging of pulsatile flow with magnetic resonance. Science
102 of 104 20-04-2010 00:28


36(3): 345-351, 1996.
306. Carroll TJ, Korosec FR, Petermann GM, et al: Carotid bifurcation: evaluation of time-resolved three-dimensional contrast-
enhanced MR angiography. Radiology 220(2):525-532, 2001.
309. Goldfarb JW, Hochman MG, Kim DS, Edelman RR: Contrast-enhanced MR angiography and perfusion imaging of the
hand. Am J Roentgenol 177:177-182, 2001.
310. Connell DA, Koulouris G, Thorn DA, Potter HG: Contrast-enhanced MR angiography of the hand. Radiographics
311. Tatli S, Lipton MJ, Davison BD, et al: MR imaging of aortic and peripheral vascular disease. Radiographics 23:59-78,
312. Nastri M, Baptista LPS, Baroni RH, et al: Gadolinium-enhanced three-dimensional MR angiography of Takayasu arteritis.
Radiographics 24:773-786, 2004. Medline Similar articles
313. Knopp MV, Tengg-Kobligk HV, Floemer F, Schoenberg SO: Contrast agents for MRA: future directions. J Magn Reson
314. Kroft LJM, Roos AD: Blood pool contrast agent for cardiovascular MR imaging. J Magn Reson Imaging 10:395-403,
315. Lauffer R, Parmlee D, Dunham S, et al: MS-325: albumin-targeted contrast agent for MR angiography. Radiology
207:529-538, 1997.
316. Grist TM, Korosec FR, Peters DC, et al: Steady-state and dynamic MR angiography with MS-325: initial experience in
humans. Radiology 207:539-544, 1998. Medline Similar articles
317. Mayo-Smith WW, Saini S, Slater G, et al: MR contrast material for vascular enhancement: value of superparamagnetic
iron oxide. Am J Roentgenol 166:73-77, 1996.
318. Louberyre P, Zhao S, Canet E, et al: Ultrasmall superparamagnetic iron oxide particles (AMI 227) as a blood pool
contrast agent for MR angiography: experimental study in rabbits. J Magn Reson Imaging 7:958-962, 1997.
319. Kellar KE, Fujii DK, Gunther WHH, et al: NC100150 injection, a preparation of optimized iron oxide nonoparticles for
positive-contrast MR angiography. J Magn Reson Imaging 11:488-494, 2000. Medline Similar articles
320. Bunce NH, Keegan J, Gatehouse PD, et al: Initial experience with the intravascular contrast agent NC100150-Injection
(Clariscan®) for breath-hold and navigator-gated magnetic resonance coronary artery imaging. J Magn Reson Imaging
10:404-410, 2002.
321. Stillman AE, Wilke N, Li D, et al: Ultra small superparamagnetic iron oxide to enhance MRA of the renal and coronary
arteries: studies in human patients. J Comput Assist Tomogr 20:51-55, 1996. Medline Similar articles
322. Anzai Y, Prince MR, Chenevert TL, et al: MR angiography with an ultrasmall superparamagnetic iron oxide blood pool
agent. J Magn Reson Imaging 7:209-214, 1997. Medline Similar articles
323. Loubeyre P, Zhao S, Canet E, et al: Ultrasmall superparamagnetic iron oxide particles (AMI 227) as a blood pool contrast
agent for MR angiography: experimental study in rabbits. J Magn Reson Imaging 7:958-962, 1997.
324. Knollmann FD, Böck JC, Teltenkötter S, et al: Evaluation of portal MR angiography using superparamagnetic iron oxide. J
325. Zheng J, Carr J, Harris K, et al: Three-dimensional MR pulmonary perfusion imaging and angiography with an injection of
a new blood pool contrast agent B-22956/1. J Magn Reson Imaging 14:425-432, 2001. Medline Similar articles
326. Hoffmann U, Loewe C, Bernhard C, et al: MRA of the lower extremities in patients with pulmonary embolism using a blood
pool contrast agent: initial experience. J Magn Reson Imaging 15:429-437, 2002. Medline Similar articles
327. Frank H, Weissleder R, Brandy TJ: Enhancement of MR angiography with iron oxide: preliminary studies in whole blood
phantom and in animals. Am J Roentgenol 162:209-213, 1994.
328. Taupitz M, Schnorr J, Abramjuk C, et al: New generation of monomer-stabilized very small superparamagnetic iron oxide
particles (VSOP) as contrast medium for MR angiography: preclinical results in rats and rabbits. J Magn Reson Imaging
329. Taupitz M, Schnorr J, Wagner S, et al: Coronary MR angiography: experimental results with a monomer-stabilized blood
pool contrast medium. Radiology 222:120-126, 2002. Medline Similar articles
330. Schnorr J, Wagner S, Abramjuk C, et al: Comparison of the iron-oxide-based blood-pool contrast medium VSOP-C184
with Gd-DTPA in first-pass MRA of the aorta and renal arteries in pigs. Proceedings of the 11th Annual Meeting of the ISMRM,
Toronto, Canada, 2003, p 1337.
331. Dirksen MS, Kaandorp TAM, Lamb HJ, et al: Three-dimensional navigator coronary MRA with the aid of a blood pool
agent in pigs: improved image quality with inclusion of the contrast agent first-pass. J Magn Reson Imaging 18:502-506, 2003.
103 of 104 20-04-2010 00:28

332. Prasad PV, Pierchala L, Chen Q, et al: Preliminary evaluation of code 7228 as an iron oxide based blood pool contrast
agent for CE-MRA. Proceedings of the 11th Annual Meeting of the ISMRM, Toronto, Canada, 2003, p 1338.
© 2010 Elsevier
104 of 104 20-04-2010 00:28

APPENDIX Protocols
Table 30-3. Abdominal Aorta MRA

System GE LXi Software version 9.0
POSITION IMAGING PARAMETERS
Position of Supine Plane Coronal
patient
Mode 3D
Patient Entry Feet first Pulse Vas TOF SPGR
Sequence
Coil Torso Options Variable bandwidth
Series EFGRE Fast, multiphase, ZIP
3D 512, ZIP 2
SCAN TIMING SCAN RANGE ACQUISITION
TIMING
TE Minimum FOV 28-36 Freq 512
Flip Angle 45 Slice 1.6-2.4 Phase 160-192
Thickness
Bandwidth 42-83 Locs Per Slab 32-40 NEX 0.5
Phase 0.9
FOV
ADDITIONAL PARAMETERS
User CVs Multiphase
Turbo 1 No Phases 3
Delay Minimum
CONTRAST 20-30 cc [commat] 2cc/s followed by NS 20 cc [commat] 2 cc/s
TIMING Acquisition delay after start of Gad = Time peak enhancement - (cc
Gad/2) + (Acq time/2)
Fluoroscopic triggering
Automated bolus detection
TIPS Arms above head or across abdomen to avoid wrap
FOV should be less than body width at level of renals
Do a second post gad run immediately after first, allowing for a quick 5 s
breath-hold
PROTOCOL Cor SSFSE Scout
Axial T1 FMP SPGR
Timing run
3D series, 1 pre and 2 post
Axial T1 FMP SPGR
page 826
page 827
1 of 5 20-04-2010 00:30
Table 30-4. Renal MRA

patient
Mode 3D
Sequence
3D 512, ZIP 2
TIMING
Flip Angle 45 Slice 1.6-2.4 Phase 160-192
Thickness
Bandwidth 42 Locs Per Slab 32-40 NEX 0.5
Phase 0.9
FOV
User CVs Multiphase
Turbo 1 No Phases 3
Delay Minimum
breath-hold
PROTOCOL Cor SSFSE Scout
Axial T1 FMP SPGR
Timing run
3D series, 1 pre and 2 post
Axial T1 FMP SPGR
Table 30-5. Hepatic Artery MRA

2 of 5 20-04-2010 00:30

patient
Mode 3D
Sequence
3D 512, ZIP 2
TIMING
Flip Angle 25-45 Slice 1.2-1.8 Phase 192-224
Thickness
Phase 0.9
FOV
User CVs Multiphase
Turbo 1 No Phases 3
Delay Minimum
breath-hold
page 827
page 828
Table 30-6. SMA Celiac MRA

patient
Mode 3D
Sequence
3D 512, ZIP 2
3 of 5 20-04-2010 00:30

TIMING
Flip Angle 25-45 Slice 1.2-2.4 Phase 160-192
Thickness
Phase 0.9
FOV
User CVs Multiphase
Turbo 1 No Phases 3
Delay Minimum
breath-hold
Table 30-7. Stepping Table Run-off

Position of Supine Plane Oblique
patient
Mode 3D
Sequence
Coil PV array Options Variable bandwidth
Series 3D upper Fast, multiphase, ZIP
512, ZIP 2
TIMING
TE Minimum FOV 44 Freq 512
Flip Angle 35 Slice 3.2 Phase 160
Thickness
Bandwidth 41.7 Locs Per Slab 32 NEX 0.57
Phase FOV 0.9
User CVs Multiphase Sat Bands
4 of 5 20-04-2010 00:30
#Stations 1 None
CONTRAST 40 cc [commat] 2cc/s for 10 s then 0.5 cc/s for 40 s followed by NS 20 cc
[commat] 2 cc/s
TIPS Use above for pelvic and thigh station
For calf, use elliptical centric (on user CV), increase matrix to 256 × 512,
reduce slice thickness to less than 2 mm
Use a low-resolution 2D TOF first as a scout
© 2010 Elsevier
5 of 5 20-04-2010 00:30
AGNETIC ESONANCE ENOGRAPHY OF THE ODY

John C. Salanitri
F. Scott Pereles
INTRODUCTION
Currently, venous pathology and anatomic variants are assessed utilizing a variety of imaging
modalities, each with limitations.
Conventional venography is currently considered the gold standard for the depiction of lower and upper
limb thrombosis. This technique requires: the cannulation of a dorsal foot or hand vein, which may be
problematic in obese or edematous limbs, injection of iodinated contrast medium which can be
nephrotoxic or thrombogenic and uses ionizing radiation. Only the venous territory downstream of the
cannulated vein can be imaged. Furthermore visualization of the inferior and superior venae cavae may
be suboptimal due to dilution of contrast, while the gonadal veins are typically not visualized.
Doppler ultrasound is widely used in the evaluation of venous thrombosis, particularly in the lower limbs
due to its noninvasive nature, ready availability and low cost. Concerns have been recently raised
regarding the accuracy of ultrasound, particularly in the calf veins and in cases of venous variants, e.g.
duplicated lower limb veins.1 Due to overlying bowel gas and bony structures, ultrasound is unreliable
for depicting the inferior vena cava, pelvic veins, brachiocephalic veins, and superior venae cava.
Multidetector row CT has also recently been used to evaluate the lower limb veins, particularly in
conjunction with a CT pulmonary angiogram study. The same concerns in the use of nephrotoxic
contrast medium and radiation exposure as encountered with conventional venography apply to CT
venography.
MRI is well suited to the evaluation of venous anatomy and pathology. There is no exposure to ionizing
radiation, the contrast media used has a wide safety profile without nephrotoxic or thrombogenic
side-effects. The multiplanar capabilities of MRI including three-dimensional (3D) data acquisition and
display aids in the delineation of sometimes complex venous anatomy. Cine sequences including
velocity encoded phase mapping can provide functional information regarding the direction and velocity
of venous blood flow.2
Compared to arterial imaging, veins are easier to image on MRI due to their larger size and slower,
more homogenous flow profiles. Generally, venous pathology (e.g. thrombosis) is more extensive than
arterial disease; thus less stringent techniques can be used for MRV, e.g. lower spatial and temporal
resolution, compared to MRA.3,4
© 2010 Elsevier
1 of 1 20-04-2010 00:31
IMAGING TECHNIQUES
MR venography techniques can be broadly classified as bright-blood techniques not utilizing gadolinium
contrast agents, bright-blood techniques utilizing gadolinium, including 3D contrast-enhanced MR
venography, and black-blood techniques (Box 31-1, Table 31-1).
page 829
page 830
Box 31-1 MR Venography Techniques

Bright-blood techniques not utilizing gadolinium contrast agents
1. Balanced gradient-echo recalled sequences with coherent steady-state

precession (SSFP):
True FISP (fast imaging with steady-state precession)
FFE (balanced fast field echoes)
FIESTA (fast imaging employing steady-state excitation)
2. Time-of-flight imaging (TOF):
2D TOF
3D TOF
3. Phase contrast imaging (PC)
Bright-blood techniques utilizing gadolinium contrast agents
1. 3D contrast-enhanced MR venography (3D CE MRV):

Direct MR venography
Indirect MR venography
2. 2D contrast-enhanced fat-suppressed T1-weighted spoiled gradient-echo:
FLASH (fast low angle shot)
SHARP (FLASH with shared prepulses)
Black-blood sequences
1. T1-weighted conventional spin-echo

2. Double inversion recovery spin-echo:
T1-weighted turbo spin-echo
T2 weighted turbo spin-echo
HASTE (half Fourier acquisition turbo spin-echo)
Non-Contrast-Enhanced "Bright-blood" Sequences

Balanced Gradient-Echo Recalled Sequences with Coherent Steady-State Free Precession
(SSFP)
These recently developed two- and three-dimensional gradient-echo sequences utilize a fully balanced
gradient waveform to recycle transverse magnetization in each TR period, thus increasing signal-
to-noise ratios (SNR) compared to fast spoiled gradient-echo techniques. 5-7 These sequences are
referred to by a variety of names given by different MR scanner manufacturer including true fast
imaging with steady-state precession (true FISP), balanced fast field echoes (FFE), and fast imaging
employing steady-state excitation (FIESTA).
Tissue contrast is a function of the T2 to T1 relaxation times in tissues with blood, fat, and water
having high signal intensities with this sequence.6,7 The blood pool signal intensity with SSFP is further
increased following gadolinium administration which, unlike spoiled gradient-echo techniques, persists
8
for a longer time period, even with decreasing blood pool concentrations of gadolinium over time.
High-quality single-shot images are rapidly acquired permitting complete coverage of the veins within
1 of 10 20-04-2010 00:32
the region of interest in a single breath-hold or alternatively as a non-breath-hold technique.6 Due to the
ultrashort repetition times (TR) utilized for SSFP, images without significant motion artifact can be
acquired during free breathing in patients unable or unwilling to breath-hold. As opposed to spoiled
gradient-echo methods, gadolinium contrast does not need to be administered with this sequence.
Intravenous thrombosis is detected as areas of lower signal intensity within the involved veins
compared to the high signal of flowing blood and may be either completely or partially occlusive. At our
institution, we have found true FISP sequences to provide a reliable and robust means of quickly
detecting venous thrombosis in the central chest, abdominal and pelvic veins, particularly in frail,
acutely ill patients.
Anecdotally, we have also found the addition of post-gadolinium contrast-enhanced T1-weighted

sequences to be helpful to confidently diagnose or exclude thrombus in thoracic, pelvic and extremity
veins (see below).
In addition, true FISP sequences have been found to be generally better in visualizing the portal vein
compared to contrast-enhanced 3D MR portal venography in prospective liver donors, and thus may
5
serve as a useful adjunct in the preoperative evaluation of these patients.
Table 31-1. MR Venography Sequences and Imaging Parameters, Siemens

Magnetom Sonata Scanner, 1.5 T
Sequence HASTE True FISP SHARP CE MRA/MRV
Description Turbo Balanced Spoiled Spoiled
spin-echo gradient-echo gradient-echo gradient-echo
Plane "Axial, "Axial, coronal" "Axial, coronal" Coronal
coronal"
Mode 2D 2D 2D 3D
TR (repetition 1000 3.62-3.72 224 2.89
time)
TE (echo time) 59-63 1.81-1.86 1.82 0.9
Flip angle 160 70 80 25
Echo train 256 - 1 -
Bandwidth 590 780 475 610
Matrix size 256 × 196 256 × 180 256 × 180 512 × 280
NEX (no. of 1 1 1 1
excitations)
% Phase FOV "Variable, "Variable, "Variable "Variable, 60-100%"
(field of view) 68-100%" 60-100%" 60-100%"
FOV "Variable, "Variable, "Variable "Variable, 300-400"
300-400" 300-400" 300-400"
Slice thickness 5 5 6 1.5-3
Interslice gap 0 0 0 0
Number of slices 24-40 24-40 24-32 60-80
Scan time <40 seconds <30 seconds <30 seconds <20
seconds/acquisition
Contrast No No Yes Yes
Fat saturation No No Yes Yes
Coil Body array Body array Body array Body array
page 830
2 of 10 20-04-2010 00:32
page 831
Time-of-Flight (TOF) Imaging (Fig. 31-1)

This flow-based MR technique is based upon the physical property of paradoxical enhancement of
blood due to inflow phenomenon.9,10 A rapid partial flip angle gradient-echo pulse sequence is used,
with repeated radiofrequency (RF) pulses being used to saturate the signal in stationary protons
subjected to multiple RF pulses during the imaging acquisition period. Protons in moving blood are only
subjected to a few RF pulses before being replaced by new inflowing blood protons, thus they have a
large signal intensity compared to the saturated soft tissues. TOF imaging can be performed as either
a two- or three-dimensional technique. Clinical applications of two-dimensional (2D) TOF MRV include
the evaluation of the superior vena cava, subclavian and brachiocephalic veins, portal vein and the
pelvic veins for thrombosis.
Gradient moment nulling is used for flow compensation; however the use of saturation prepulses may
result in nonvisualization of retrograde flow in veins that are incompetent or have a complex, tortuous
4
anatomic path with respect to the prescribed imaging plane. In-plane saturation of blood protons with
a predominately in-plane course may occur with use of thick 3D slabs.9 This artifact may be minimized
by prescribing a slab of thin 2D slices in the plane perpendicular to the vessel of interest using a long
TR time (30-40 ms), at the costs of increased overall imaging time and increased risk of patient
movement artifacts. Spin dephasing at sites of turbulence results in signal loss which may incorrectly
2
suggest vessel occlusion.
While 2D TOF MRV has been shown to be highly accurate in the diagnosis of deep venous thrombosis
of the femoral, popliteal and iliac veins and to be more accurate than either ultrasound or conventional
11
venography in the evaluation of gonadal vein thrombosis, the lengthy acquisition times associated
with this technique and inability to reliably demonstrate small calf veins have limited its clinical use. 12,13
With the advent of high-performance MR scanners with fast gradient systems, contrast-enhanced MRV
techniques have rendered TOF essentially obsolete.6 Although some advocate the use of 2D TOF as a
back-up in the event of failure of contrast-enhanced MRV to produce high-quality diagnostic images,4
our institutional experience has been that contrast-enhanced 3D MRV is an extremely robust technique
and that 2D TOF has no current practical benefit in the diagnosis of venous thrombus.
2D TOF is occasionally still used along with phase contrast techniques to determine the direction of
flow in the portal vein in patients with hepatic cirrhosis and portal hypertension.4,6 A coronal plane
through the portal vein is prescribed with the VENC (velocity encoding) set at a low level, e.g. 30 cm/s.
A saturation band at the level of the heart is used to null any arterial signal. A second thin saturation
band is then placed perpendicular to the extrahepatic portal vein to determine the direction of flow, i.e.
hepatopedal versus hepatofugal.
3 of 10 20-04-2010 00:32
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
page 831
page 832
Figure 31-1 TOF MRV. A, Coronal 2D TOF MRV of the pelvis demonstrating occlusion of the right
common iliac vein. B, Axial 2D TOF image demonstrating left common femoral vein DVT. C, 2D TOF
study with artifact (arrows) in inferior vena cava due to in-plane flow saturation. D, Artifact in the left
common femoral vein mimics DVT due to turbulent blood flow. Similar less pronounced artifact is also
present on the right.
Phase Contrast (PC) Imaging
4 of 10 20-04-2010 00:32
Protons in flowing blood experience phase shifts as they move along a gradient field.7,9,10 The PC
sequence uses bipolar gradients applied along a selected direction to amplify this effect with the
operator prescribing a velocity encoding (VENC) value (cm/s). At least two images are produced, a
magnitude image and a phase-encoded map of the direction and velocity of flow of blood within each
imaged voxel, which is used to derive functional information of blood flow within the imaged vessel. As
this sequence, like TOF, is a flow-based sequence, PC is susceptible to intravoxel dephasing and the
6
overestimation of stenosis or simulation of occlusion in vessels with turbulent flow.
With the advent of contrast-enhanced 3D MRV techniques, PC today has a limited role in the
assessment of venous disease. This technique is occasionally still used to differentiate very slow blood
flow from thrombus in equivocal cases.
Contrast-Enhanced "Bright-blood" Sequences

3D Contrast-Enhanced MR Venography
The development of MR scanners with stronger gradients and faster rise times (able to switch on and
off in shorter times) has allowed the development of 3D MRA and MRV techniques using very short TR
and TE times.2,5,6,7,9 Contrast-enhanced 3D MRA and MRV relies on the T1-shortening properties of
gadolinium-chelate contrast medium within the vessels, independent of flow related phenomena.9,10
The signal of the imaged vessels directly relates to the concentration of gadolinium within the vessel,
thus timing of the gadolinium bolus with respect to the injection time and the mode of k-space
acquisition (particularly central low spatial frequency lines) is paramount in arterial imaging. However,
as venous imaging generally has a significantly longer imaging window, timing of the k-space
acquisition is less of an issue.
Generally 3D spoiled gradient-echo sequences are used due to their high speed of acquisition, short
echo times and accentuated T1 contrast with suppression of background soft-tissue signal intensity,
enhancing image contrast without incurring an additional time penalty for imaging.4 Two techniques of
contrast-enhanced MRV have been described in the literature: direct MRV and indirect MRV.
Direct MR Venography
This technique has been utilized to evaluate the deep venous systems of both lower and upper
extremities in addition to the central thoracic veins (Fig. 31-2). Large volumes (60-120 mL) of dilute
gadolinium-normal saline (1:15 to 1:20) solution are continuously injected via a dorsal foot or hand vein
upstream from the vein of clinical interest, e.g. femoral or pelvic veins, brachiocephalic vein or superior
vena cava.1,4,12 A multichannel quadrature/phased-array vascular coil is used with four circular arrays
each covering a distance of 24 cm (i.e., total distance 96 cm) that can be activated separately or in
combination. With the patient in the supine position with the legs together or prone position with the
arms extended above the head, both limbs can be imaged with the flexible coils wrapped about them.
Generally 2 to 5 mm partitions are used along with the shortest TR and TE times (typically 3-5/1-2)
with flip angles of 30 to 40 degrees. Zero fill interpolation is used to palliate partial volume effects.
A tourniquet is placed about the ankle or wrist to ensure adequate filling of the deep venous system of
the extremity. A total volume of 120 mL of dilute gadolinium solution is injected into each limb at a rate
of 1 mL/s, equivalent to a dose of 8 mL undiluted gadolinium for each limb. Either one or both
extremities can be simultaneously injected. Data acquisition commences after the injection of the initial
40 mL of dilute gadolinium solution into the dorsal foot or hand vein with the pelvis and thighs or the
thorax and upper arms being imaged first. The injection is continued during the entire data acquisition
with sequential k-space mapping used, i.e., central k-space acquired in the middle of the acquisition.
5 of 10 20-04-2010 00:32
Add to lightbox
Add to lightbox
page 832
page 833
Figure 31-2 Thoracic MRV. A, Direct contrast-enhanced MRV (simultaneous dual injections)
demonstrating occlusion of the right axillary, subclavian and brachiocephalic veins. B, Subtracted
venous phase MIP from indirect contrast-enhanced MRV (arterial phase data set subtracted from the
mixed arterial-venous phase data set) demonstrating patent central thoracic veins.
After acquisition of the first data set, the table is manually moved and the calves or forearms are
imaged. If details regarding the superficial veins are needed, the tourniquets are removed and the data
acquisition repeated. Imaging times are less than 70 seconds with the total examination times less than
14
10 minutes compared to 2D TOF that may take up to 90 minutes. Maximum intensity projections
(MIPs) are routinely obtained for analysis.12 However as MIPs are less sensitive in the detection of
small nonocclusive thrombi, the source images should always be reviewed in addition to the MIPs.
The use of diluted gadolinium avoids the T2 shortening effects that occur with nondiluted gadolinium,
resulting in signal drop-out erroneously simulating vessel occlusion. Unlike iodinated contrast medium
used in conventional venography, there are greatly reduced risks of nephrotoxicity, anaphylaxis or
thrombogenesis with the near iso-osmolar diluted gadolinium-normal saline solution. 1,4,12 Due to the low
total administered dose of gadolinium for each limb, repeated injections of contrast and data
12
acquisitions with the tourniquet off or limb in different positions may be obtained. The contrast to
noise (CNR) obtained with this technique has been reported to be higher compared to equilibrium
phase images obtained from the indirect technique (see below), with more detailed depiction of venous
anatomy.4
Indirect MR Venography
6 of 10 20-04-2010 00:32
Images of the thoracic, abdominal and pelvic veins can be obtained with this technique in which higher
concentrations and volumes of nondiluted gadolinium contrast (compared to the direct MR venography
technique) are injected via an antecubital fossa vein.4 In a manner analogous to contrast-enhanced 3D
MRA, data sets of the region of interest may be acquired in the arterial phase, portal venous and
delayed equilibrium (arterial-venous) phases. The arterial phase data set may then be subtracted from
14
the equilibrium data set on a workstation to produce a pure venous phase data set for analysis.
A 3D spoiled gradient-echo sequence is used with minimum TR and TE values and a rectangular field-
of-view, in order to keep the required breath-hold times to a reasonable level, generally less than 18
seconds.6 The flip angle is generally the same as for arterial imaging, i.e., 25 to 40 degrees, however
some authors recommend the use of a lower flip angle for the equilibrium data set in order to maximize
the CNR in the opacified thoracic and abdominal veins.4,15 Slab volume and slice thicknesses are
generally identical to the arterial phase 3D data acquisition set. If required, thicker partitions (3 to 5
mm) with a 256 × 160 or 192 matrix may be used, due to the generally larger caliber of veins
compared to their accompanying arteries. For abdominal applications, fat saturation may be useful.
The time delay for the arterial phase data acquisition may be determined from a timing bolus injection
of 2 mL gadolinium with sequential 2D spoiled gradient-echo images of the aorta at the required level
obtained every 1 to 2 seconds or alternatively using semiautomatic bolus tracking techniques. The time
delay for either the portal venous or the delayed equilibrium phase data sets can be determined using
the arterial phase timing as a guide (i.e., addition of 30 seconds to the arrival of contrast in the
abdominal aorta for the portal venous phase). Alternatively, for portal venous MRV, a single acquisition
in the portal vein phase can be acquired at an empirically determined time delay (50 to 60 seconds
after start of contrast injection). In order to overcome the diminished gadolinium contrast effects in the
portal vein as a result of recirculation through the bowel and liver and partial uptake by capillary beds
elsewhere, a larger contrast dose (0.2 to 0.3 mmol/kg) may be administered at a slower rate (1
4
mL/s).
The acquired 3D data sets are generally viewed on soft copy workstations or PACS terminals. A
variety of post-processing techniques are employed, the most commonly used being maximum intensity
16
projection (MIP) and multiplanar reformatting (MPR). MIPs cast a set of parallel rays through
volumetric data sets at user defined angles with the brightest pixel along each ray mapped to a 2D
image, giving an end result similar to a projectional angiogram. As small nonocclusive venous thrombi
may be obscured on the MIPs, the source images of the nonsubtracted and subtracted data sets must
always be evaluated.10 Some studies have found no loss in diagnostic accuracy when using the
unsubtracted data sets with the residual signal in the arteries not impairing visualization of the venous
17
structures.
MPRs allow viewing of the data set in multiple user defined planes. While these images are generally
excellent to depict intraluminal thrombus, care must be used when processing these on the workstation
not to introduce false vessel occlusions or stenoses.6 A double subtraction algorithm has been recently
described for indirect lower limb MRV.18 In this technique two summed early measurements of the
arterial phase are subtracted from two summed late measurements of the venous phase in order to
produce images with higher signal intensities compared with single subtraction algorithms as used in
the chest and abdomen. Accurate diagnosis of iliac and femoral vein thrombosis was reported using
this subtraction algorithm, however the distal calf veins were not evaluated.18
2D Contrast-Enhanced Fat-Suppressed T1-Weighted Gradient-Echo Sequences

These sequences may be performed as a breath-hold following the 3D contrast-enhanced MRV,
typically in the axial and coronal planes 3 to 10 minutes after the initial infusion of contrast. The time
delay in acquisition is useful for complete opacification of the upper limb veins which have a longer
recirculation time compared to the head and neck veins and consequently may appear to have pseudo-
filling defects on the subtracted venous data set images obtained at the contrast-enhanced MRV.
Similarly, the infrarenal inferior vena cava and pelvic veins are usually well opacified and well visualized
7 of 10 20-04-2010 00:32
in the coronal plane on this sequence.
Together with the noncontrast single shot true FISP sequences, at our institution we have found the
post-contrast fat-suppressed T1-weighted gradient-echo sequences (SHARP-shared prepulses) to be
most useful in the evaluation of thoracic and pelvic venous thrombosis.
page 833
page 834
Add to lightbox
8 of 10 20-04-2010 00:32
Add to lightbox
Figure 31-3 A 26-year-old man who complained of arm swelling with left arm AV fistula. A, Sagittal
MIP image of an arteriovenous fistula in the left arm obtained using TRICKS and parallel imaging
techniques. The feeding basilar artery (best seen on the sagittal sequence) is ectatic and feeds the
dilated superficial brachial vein. No stenosis identified at the anastomosis site. The radial and ulnar
arteries are patent and there is venous filling on the delayed images. B, Sagittal unsubtracted source
image demonstrates the ectatic basilar artery and dilated subcutaneous basilic vein at the level of left
elbow.
"Black-blood" Sequences
T1-Weighted Conventional Spin-Echo
For a long time, this sequence was the "work-horse" for venous MR imaging with the dark central
lumens of vessels due to washout of magnetized spins in flowing blood from the imaging plane before
the time of echo sampling. However, if the blood flow was slow or the TE too short, the vein appeared
gray, thus obscuring intraluminal details. With the introduction of MR systems with high-performance
7
gradients, this technique is now of historical significance.
Double Inversion Recovery Fast Spin-Echo

In this sequence, the signal from flowing blood is preferentially eliminated by a preparatory pair of
nonselective and slice selective inversion recovery pulses combined with a readout time to coincide
with the null point (TI) of the flowing blood. Stationary tissues in the imaging slice experience two
inversion recovery pulses while the blood flowing into the slice at the readout time has only
experienced one pulse. By using a readout time equal to the blood TI time, the signal from the blood in
the image plane is nulled. For optimal images, this sequence is performed as a cardiac gated
segmented breath-hold sequence with limitations on the numbers of slices obtained, even when half
Fourier acquisitions, e.g. HASTE (half Fourier acquisition turbo spin-echo) are used. Subsequently,
respiratory motion and cardiac arrythmias may severely degrade the images obtained. Also, the signal
from slow flowing blood adjacent to vessel surfaces may be incompletely nulled, simulating thrombus. 7
9 of 10 20-04-2010 00:32
© 2010 Elsevier
10 of 10 20-04-2010 00:32
FUTURE DEVELOPMENTS
Parallel Imaging Techniques
Parallel imaging techniques are under development in order to obtain MRA and MRV images with
higher spatial resolution or with shorter breath-hold times. SMASH (simultaneous acquisition of spatial
harmonics) uses spatial information encoded in the detector coil array to construct multiple lines of
k-space, thus reducing the number of phase-encoding steps required for the image acquisition. The
result is a time saving in image acquisition albeit at a slightly decreased SNR without wrap-around
artifacts.
SENSE (sensitivity encoding) techniques allow decreased sampling of phase-encoding steps by

utilizing the sensitivity profile of the receiver coils. This encoding sensitivity profile is independent of the
examined object and can be used to localize the position of the detected signal relative to the coils.
Temporally Resolved MRA and MRV Techniques (Fig. 31-3)

A time resolved 3D contrast-enhanced MRA technique (TRICKS-temporally resolved imaging of
contrast kinetics) has been described in which multiple 3D image data sets are rapidly acquired from a
volume during the passage of a contrast agent bolus (approximately every 6 to 7 seconds). This
technique uses repeated sampling of critical central k-space data combined with temporal interpolation
of less frequently acquired peripheral k-space data to produce a series of time resolved images. The
need for a contrast timing bolus examination is obviated with this technique.
page 834
page 835
While this technique may be used to produce a pure arterial MRA data set, the potential exists to
image the later venous phase without significant arterial overlap, e.g., assessment of dialysis
arteriovenous fistulas and grafts. The combination of this technique with parallel imaging could
potentially result in further improvements in temporal resolution or better spatial resolution with the
same image acquisition frequencies. However, the utility of the TRICKS algorithm may be made
redundant by recent high-end MR scanner systems that have the capacity to dynamically and
sequentially fill 3D data sets with ample spatial and temporal resolution.
Blood Pool Contrast Agents

Currently, gadolinium compounds used in MRV are extracellular agents with small molecular size
allowing leakage through the capillaries and redistribution into the extracellular space. Blood pool (or
intravascular) agents are currently being developed with increased affinity to plasma proteins resulting
in slower elimination and increased half-lives permitting higher resolution imaging of the peripheral veins
in the equilibrium phase. These blood pool agents have a higher relaxivity than extracellular gadolinium
agents, with markedly increased and sustained intravascular T1 shortening effects.
In a recent study of MS-325 MR venography of the iliocaval veins, venous conspicuity was greater
19
compared with gadolinium-enhanced 3D MRV. The assessment of the abdominal and pelvic veins
was not limited by arterial-venous overlap on the multiplanar reformatted images viewed on a
workstation using MS-325. Thus blood pool agents may have an important role in MR venography in
the future.
© 2010 Elsevier
1 of 1 20-04-2010 00:33
PITFALLS AND ARTIFACTS (Box 31-2)

Time-of-Flight and Phase-Contrast Techniques
Both TOF and PC techniques were shown to have high accuracy in the diagnosis of pelvic and thigh
DVT, however, lengthy examination times, sensitivity to patient motion artifacts, and magnetic
susceptibility artifacts, e.g. from metallic stents and surgical clips, flow turbulence related signal voids
simulating occlusions, and the inability to reliably demonstrate small deep calf veins contributed to the
low clinical acceptance of these techniques.1,4,11,12
Direct 3D Contrast-Enhanced MRV

While excellent quality images of both upper and lower limb veins may be obtained with this technique,
it shares many of the same strengths and limitations of conventional venography.
For this technique, a dorsal foot or hand vein must be cannulated, which may not always be an easy
task especially if there is limb edema, which is frequently associated with deep venous thrombosis.
There is also the issue of applying a tourniquet to an already swollen limb.18 Similar to conventional
venography, only the vessels upstream of the injection site are imaged. Collateral pathways beyond
the site of obstruction may not be filled during image acquisition, resulting in a largely blank image. To
avoid this, several acquisitions can be collected during or after the contrast agent infusion or repeat the
4
image with a longer and larger contrast bolus. Alternatively, if the superficial veins or collateral venous
pathways are preferentially filled by the contrast injection rather than the deep veins, venous
thrombosis may be erroneously suggested.
Unopacified blood from the internal jugular or renal veins may dilute the contrast within the superior and
inferior vena cavae, simulating thrombus. T2 and T2* shortening effects due to insufficient dilution of
contrast may make the imaged veins completely dark, simulating a thrombus. The contrast must be
1,4,12
diluted at least by a factor of 10 to 100 to avoid this.
To image both upper limbs simultaneously, the patient must be prone with the arms extended above
the head, which is covered by the coils in order to image the upper thorax. Patient cooperation may be
an issue, as is the possibility of precipitating the symptoms of superior vena cava obstruction with the
elevated arm position.1 Spatial resolution limitations seen with this technique may make the
visualization of nonocclusive calf vein thrombus or thrombi associated with valve cusps a challenge. 12
Indirect MR Venography
Box 31-2 Pitfalls and Artifacts
Direct 3D CE MRV
1. Need to cannulate dorsal foot or hand vein

2. Only vessels upstream of cannulated vein imaged
3. Preferential filling of collateral or superficial veins
4. Dilution of contrast by unopacified renal venous/or Internal Jugular Vein
blood
5. T2/T2* shortening effects with insufficiently diluted gadolinium
6. Spatial resolution limitations
Indirect 3D CE MRV
1. Reduced SNR due to lower concentration of gadolinium in imaged veins

due to capillary dispersion and hepatic and renal uptake
2. Mixing of opacified and nonopacified blood simulating partially occlusive
thrombus
3. Misregistration errors (image blurring) introduced by subtraction
1 of 2 20-04-2010 00:33
techniques
Both techniques
1. Incomplete anatomic coverage thus not imaging veins containing thrombus

2. Sole reliance on MIPs potentially missing nonocclusive thrombus
page 835
page 836
The concentration of gadolinium within the central thoracic, portal and pelvic veins is generally lower
than the arterial phase due to dispersion of the gadolinium bolus as a result of partial uptake of
contrast by the capillary beds and extraction by the liver. Consequently, larger dose (0.2 to 0.3
mmol/kg) of gadolinium may need to be administered at a slower rate (1 mL/s) to prolong the venous
1,4
transit time and increase the SNR. Lower flip angles and thicker partition thicknesses may be utilized
for venous imaging.
The 3D data acquisition must be timed so that the center of k-space coincides with the venous phase
of the bolus. The suprarenal portion of the inferior vena cava shows preferential enhancement
compared to the infrarenal inferior vena cava, as do the head and neck veins compared to the more
distal upper limb veins with longer recirculation times. Consequently, the mixture of opacified and
nonopacified blood at these locations may simulate partially occlusive thrombus; however, examination
with true FISP or delayed contrast-enhanced T1-weighted spoiled gradient-echo sequences should
provide the correct diagnosis.
A pure venous data set is obtained by subtraction of the arterial data set from the equilibrium data
14
set. Misregistration errors due to image blurring, simulating venous pathology, may be introduced if
the patient is unable to accurately reproduce the breath-holds for each acquisition.4 Again, such errors
in diagnosis can be avoided by examining the unsubtracted source images.
Incomplete Coverage
If the image volume is misplaced or not large enough to encompass the entire venous anatomy, the
images may be incomplete. This can occur with any imaging sequence. Venous anatomy tends to be
more variable and less predictable compared to arteries. In particular, tortuous collateral vessels may
be problematic. The localizer images and the precontrast data sets should be used to determine the
correct volume of coverage prior to performing the contrast-enhanced sequences.
© 2010 Elsevier
2 of 2 20-04-2010 00:33
NORMAL ANATOMY AND ANATOMIC VARIANTS

In the extremities, the venous drainage may be categorized into two separate systems: superficial and
deep.
In the lower extremity, the deep calf veins are commonly paired (venae comitantes) and the popliteal
and femoral veins may also be paired as normal variants. The superficially located long and short
saphenous veins typically drain into the superficial femoral vein and popliteal veins respectively.
Perforating veins connect the deep and superficial veins of the lower limb. Damage to the valves in
these perforator veins may result in abnormal blood flow from the deep to the superficial venous
systems, resulting in varices.
The superficial veins of the upper limb include the cephalic vein which perforates the clavipectoral
fascia to join the axillary vein and the basilic vein which pierces the deep fascia halfway up the medial
aspect of the arm to join the paired brachial veins to form the axillary vein. The median vein drains the
subcutaneous tissue of the anterior wrist and forearm, dividing at the elbow to form the median
cephalic and median basilic veins.
Occasionally the left superior vena cava may persist as an embryological remnant and when present
drains into the coronary sinus. The normal right sided vena cava may be either present or absent in
these cases as may the left brachiocephalic vein.
The azygos venous system consists of the ascending azygos vein, azygos arch and the right superior
intercostal veins, while the hemiazygos venous system is composed of the hemiazygos vein, accessory
hemiazygos vein, and the left superior intercostal veins. Obstruction of the superior vena cava below
the insertion of the azygos vein may be compensated for by retrograde flow through the azygos vein
via its arch.20
The inferior vena cava has a complicated embryological development involving the fusion of three
separate segments. Occasionally one of these segments may fail to develop, most frequently the
hepatic segment. The hepatic veins drain normally into the suprarenal potion of the inferior vena cava
and then into the right atrium, while blood from the lower extremities and the pelvis is shunted into the
azygos and hemiazygos veins to drain into the superior vena cava. The estimated prevalence of inferior
vena cava anomalies in the general population is 0.07% to 8.7%.21
There are usually three hepatic veins: central, right, and left, which drain into the inferior vena cava.
Several accessory inferior hepatic veins, usually from the right hepatic lobes may drain more caudally
into the inferior vena cava.5 These vessels generally are less than 2 mm in diameter.22
The portal vein is formed by the union of the splenic and superior mesenteric veins behind the neck of
the pancreas. Most often the portal vein divides into a left and right branch with further subdivision of
the right portal veins into anterior and posterior segmental branches. Surgically relevant portal venous
anatomic variations include: (i) trifurcation of the portal vein into left, right anterior and right posterior
branches at the porta hepatis; (ii) early origin of the right portal vein from the main portal vein; and (iii)
origin of the left portal vein from the right anterior portal branch. The incidence of these portal vein
variations in the general population is estimated at 6% to 20%.22
Similar to the renal arteries, the renal veins may show a number of normal variations in the anatomy,
1615
branching of the main renal vein, accessory veins, retroaortic veins, and circumaortic veins. The
circumaortic configuration of the left renal vein occurs in 5% to 7% of individuals and a single
retroaortic left renal vein is seen in 2% to 3%. Tributaries draining into the left renal vein include the
lumbar veins (75% of individuals) and the adrenal and gonadal veins (100% of individuals).
© 2010 Elsevier
1 of 1 20-04-2010 00:34
CLINICAL APPLICATIONS (Boxes 31-3, 31-4)

Central Thoracic and Upper Extremity Systemic Veins (Fig. 31-4)
page 836
page 837
Box 31-3 Clinical Applications of MR Venography

Thorax and upper extremity veins
1. Diagnosis of central vein thrombus in patients with previous long-term

central venous catheterization, e.g., chemotherapy, total parenteral
nutrition, hemodialysis
2. Determination of central thoracic vein stenosis/occlusion prior to
placement of AV fistula or AV graft in renal failure patients
3. Detection of downstream venous stenoses in renal hemodialysis patients
with poorly functioning AVF or AVG
Pulmonary veins
1. Partial or total anomalous pulmonary venous return

2. Pre-procedure pulmonary vein mapping prior to radiofrequency ablation
(RFA)
Pelvic and lower limb veins
1. Diagnosis of DVT particularly in patients unsuitable for ultrasound, e.g.,

long leg casts
2. Assessment of deep veins for DVT in high-risk patients or before and after
procedures with a high risk of developing DVT
3. Superficial lower limb venous mapping prior to venous surgery
Portal and abdominal veins
1. Portal hypertension and visualization of portocaval communications

(varices)
2. Portal, splenic and superior mesenteric vein thrombosis
3. Portal and hepatic vein anatomy in prospective live related liver donors
4. Portal vein anastomotic strictures in failing liver transplants
5. Hepatic vein occlusion in Budd Chiari syndrome
6. Renal vein anatomy in prospective live kidney donors, particularly if a
laparoscopic nephrectomy is planned
7. Renal vein anastomotic strictures in failing renal transplants
Box 31-4 MR Venography Findings of Venous Thrombosis

Acute thrombus
1. Expansion of the vessel lumen compared to noninvolved veins

2. Bull's eye sign on axial T1-weighted fat-saturated contrast-enhanced
gradient recalled echo (GRE) sequences
3. Enhancing vein rim with dark nonenhancing center
4. Associated with perivenous soft-tissue inflammation and contrast
enhancement
Chronic thrombus
1 of 12 20-04-2010 00:35
1. Small caliber vein, i.e., narrowed lumen

2. Thickened vein wall
3. Partial recanalization with web formation
4. Complete recanalization with return to normal appearance
5. May have complete vein enhancement on T1-weighted fat-saturated
contrast-enhanced GRE sequences
Bland thrombus
1. Variable signal intensity depending on age

2. May be associated with venous expansion if acute
3. Does not enhance following contrast administration
Tumor thrombus
1. Generally associated with vein expansion

2. Shows enhancement after contrast administration, which may be
heterogeneous
Add to lightbox
Add to lightbox
2 of 12 20-04-2010 00:35
Add to lightbox
page 837
page 838
Figure 31-4 A 47-year-old woman with prior right internal jugular central line with large right pleural
effusion being evaluated from a central thoracic venous occlusion. Axial true FISP (A) and
post-contrast SHARP (B) images of the lower neck demonstrate an occluded right internal jugular vein
(arrowheads). The patent left internal jugular vein (arrows) demonstrates artifactual low central signal
on both sequences. C, Coronal MIP demonstrates occlusion of the lower half of the right internal
jugular vein with multiple collateral vessels (arrowheads) present in the neck. The upper half of the right
internal jugular vein (arrow), superior vena cava and brachiocephalic veins are present.
Add to lightbox
Add to lightbox
3 of 12 20-04-2010 00:35
Add to lightbox
Figure 31-5 Tumor thrombus. A, Sagittal SSFP gradient-echo image demonstrating thrombus in the
inferior vena cava extending into the right atrium in a patient with known hepatoma. B, Horizontal long
axis view of the heart confirms the presence of intra-atrial thrombus. C, Axial HASTE image of a
different patient demonstrating left renal vein and IVC tumor thrombus arising from a left renal cell
carcinoma.
Thrombosis of the upper extremity veins was considered to be an uncommon, innocuous self-limiting
disease, however with the increasing use of long-term indwelling central venous catheters for
hemodialysis, chemotherapy, and hyperalimentation, this entity is now appreciated to be more common
23
and more clinically significant than originally thought. Complications of central chest vein thrombo-
occlusive disease include: the restriction or loss of central venous access, venous gangrene,
post-thrombotic syndrome, superior vena cava syndrome, and pulmonary embolism.
MR venography can be used to assess the superior vena cava, brachiocephalic and subclavian veins in
patients with renal insufficiency, in order to exclude central venous occlusion or stenoses prior to
arteriovenous fistula or graft formation or central line placement. Thrombus can be directly identified as
areas of low signal intensity within the veins which may be partially or totally occlusive. The presence
of collateral pathways of venous return may also be demonstrated with these techniques. Acute
thrombus may be differentiated from chronic thrombus by the presence of enhancement in the
occluded vein wall, presumably due to inflammation of the wall by the acutely formed blood clot (Fig.
31-5).
In hemodialysis patients with poorly functioning AV fistulas and graft, the presence of venous stenosis
can be determined with direct injection of diluted gadolinium into the fistula (direct contrast-enhanced
MR venography technique) or with imaging in the equilibrium phase after injection in the contralateral
antecubital fossa vein (indirect MR venography).24
Pulmonary Veins (Fig. 31-6)

Anomalous pulmonary venous connections result from altered embryologic development of the
pulmonary venous system with persistent communications with the systemic veins, e.g., right or left
sided superior vena cava, inferior vena cava. These may be associated with sinus venosus atrial septal
defects. Contrast-enhanced MR venography can be used to determine the presence and
communication of partial anomalous pulmonary venous connections, obviating the use of ionizing
radiation and iodinated contrast medium.
4 of 12 20-04-2010 00:35
Add to lightbox
Add to lightbox
page 838
page 839
Figure 31-6 A 60-year-old man with rheumatic heart disease and atrial fibrillation being evaluated
prior to radiofrequency ablation. A, Contrast-enhanced high-resolution pulmonary vein MRV images
demonstrating excellent opacification of the pulmonary veins and left atrium. The left atrial appendage
is well seen and does not contain thrombus. The right middle lobe pulmonary vein has a separate
ostium to the left atrium. Pre-procedure knowledge of the size and number of pulmonary vein ostia is of
great benefit for the interventionalist performing radiofrequency ablation for atrial fibrillation. B, Single
coronal source image demonstrating the left atrial appendage (arrow) and the right upper lobe
pulmonary vein entering the left atrium (arrowhead).
5 of 12 20-04-2010 00:35
Add to lightbox
Add to lightbox
Add to lightbox
Figure 31-7 Lower limb MRV. A, Composite image made from three MIPs of the lower limb veins
6 of 12 20-04-2010 00:35
obtained from direct contrast-enhanced MRV with bilateral pedal injections demonstrating normal
anatomy with no evidence of venous thrombosis. B, Images of the chest and abdomen in the
equilibrium (mixed arterial-venous phase) using an experimental blood pool agent with excellent
opacification of the venous structures. A pure venous phase could be obtained by subtraction of the
arterial phase data set. C, Coronal post-contrast SHARP sequences demonstrating extensive
thrombus within the pelvic veins continuing cephaladly within the inferior vena cava to its intrahepatic
segment.
A more recent application for MR pulmonary venography is the pre-procedure mapping of the
pulmonary veins prior to radiofrequency ablation for atrial fibrillation. The number and size of the
pulmonary vein ostia with the left atrium can be determined well in advance of the procedure, thus
allowing determination of the appropriate sized catheters. At our institution, the use of pulmonary
venous mapping has significantly decreased the length of the radiofrequency ablation procedure as
well as the radiation dose.25
Lower Extremity Veins (Figs. 31-7 and 31-8)

Both direct and indirect MR venography techniques have been used to evaluate the deep and
superficial veins of the lower limbs.
MR venography can be used to detect deep venous thrombosis (DVT), particularly in patients
unsuitable for accurate assessment by ultrasound, e.g., those in leg casts, and those with anatomic
variations that may not be well appreciated by ultrasound, e.g., duplicated femoral veins. Pelvic and
iliac veins are frequently poorly visualized by ultrasound due to overlying bowel gas and by conventional
venography due to dilution of contrast. This limitation does not occur with contrast-enhanced MR
venography, which may be considered to be the imaging modality of choice for the determination of the
proximal extension of thrombus within the iliocaval venous system and the detection of gonadal vein
thrombus.
MR venography can also be used to assess the status of the deep veins of the lower extremity in
patients with elevated risk for DVT both before and after surgical procedures.
Mapping of the superficial veins can be performed prior to surgery, e.g., vein stripping for coronary
artery bypass surgery or varicose vein surgery with high-quality images of the superficial, muscular and
perforator veins obtained using direct MR venography techniques.
Lower limb vein thrombosis appears as filling defects, which in the acute phase may be outlined by a
thin margin of contrast. Following acute thrombosis, there is clot retraction and either thrombolysis or
recanalization of the affected vein. The venous valves may be injured, resulting in affected deep veins
becoming incompetent. The post-thrombotic syndrome is associated with swelling and pain of the
involved leg which may progress to induration and ulceration. Contrast-enhanced MR venography
findings of this later stage include irregular partially recanalized deep vein lumens with multiple deep or
superficially located collateral vessels.
In our institutional experience, the use of axial and coronal non-contrast-enhanced true FISP and
contrast-enhanced T1-weighted fat-saturated gradient-echo sequences covering the pelvis and thighs
is usually sufficient to diagnose pelvic vein DVT and to determine the proximal extent of IVC thrombus.
page 839
page 840
7 of 12 20-04-2010 00:35
Add to lightbox
Add to lightbox
Add to lightbox
Figure 31-8 A 57-year-old man with endstage renal disease and prior lower limb DVT being evaluated
prior to potential renal transplant for pelvic vein patency. A, Axial contrast-enhanced SHARP image
shows a central filling defect in the right superficial femoral vein (arrow), consistent with a thrombus. B,
Coronal equilibrium (mixed arterial-venous) phase MIP shows no evidence of venous enhancement in
the abdomen or pelvis, highly suggestive of thrombosis. (Occluded right iliac vein (arrow), occluded left
iliac vein (arrowhead).) However as multiple metallic stents were present in the abdomen, artifactual
signal drop-out could not be completely excluded. C, Conventional venogram with bilateral groin
injections confirms that both external iliac veins are occluded at the level of the inguinal ligaments with
filling of superficial abdominal wall collateral veins.
8 of 12 20-04-2010 00:35
Portal and Abdominal Veins (Figs. 31-9 and 31-10)

Portal hypertension is a complication of hepatic failure and cirrhosis. Spontaneous and iatrogenic
portosystemic shunts can be evaluated using indirect MR venography, non-contrast-enhanced SSFP
(true FISP), and contrast-enhanced fat-saturated T1-weighted 2D gradient-echo sequences. If
6
required, the direction of portal venous flow can be determined using TOF or PC techniques.
Portal vein, splenic vein and superior mesenteric vein thrombosis is a clinically important condition seen
in a number of medical disorders, including cirrhosis, pancreatic and hepatic carcinoma, and
hypercoagulable states. Non-contrast SSFP sequences can elegantly demonstrate the thrombus which
may be confirmed on post-contrast T1-weighted gradient-echo sequences. The portal vein can be
imaged by indirect MR venography using subtraction techniques to eliminate arterial signal from the
equilibrium phase image. As there is often reduced signal in the portal vein due to capillary extraction
and dilution effects, use of higher contrast volumes and thicker partition thicknesses may be of use. 4
The portal vein should always be evaluated on the source and MPR images in addition to the
generated MIPs.16 The acutely occluded portal vein is distended with low signal intensity clot and
demonstrates no flow on PC images. Perivascular enhancement may be seen due to the inflammatory
response associated with acute thrombosis. Chronic occlusion with cavernous transformation
demonstrates a classic appearance of small caliber enhancing vessels in the porta hepatis in the portal
venous and equilibrium phases.
Budd Chiari syndrome is caused by hepatic vein thrombosis. The hepatic veins are absent on both the
equilibrium contrast-enhanced 3D MRV images as well as the true FISP and post-contrast
T1-weighted gradient-echo sequences. The peripheral liver may show a higher T2 signal intensity and
more heterogenous contrast enhancement pattern compared to the central liver which eventually
hypertrophies.
page 840
page 841
Add to lightbox
Add to lightbox
9 of 12 20-04-2010 00:35
Add to lightbox
Add to lightbox
Add to lightbox
Add to lightbox
Figure 31-9 Portal venous MRV. Subtracted MIP from contrast-enhanced 3D portal venous MRV
showing normal anatomy (A) and portal vein occlusion and splenorenal shunt with varices (B). C, Time
of flight (TOF) of normal bright hepatopedal flow in the proximal portal vein (thick arrow) with absent
signal in the portion of the portal vein (thin arrow) distal to the black diagonally orientated saturation
10 of 12 20-04-2010 00:35
band. Axial true FISP (D) and axial TOF (E) images demonstrating thrombus in the portal vein
(arrows). F, A 29-year-old potential live related liver donor. Axial true FISP image at the level of the
porta hepatis demonstrates a large inferior accessory hepatic vein joining the inferior vena cava at a
level caudal to the right hepatic vein (arrow).
Add to lightbox
Add to lightbox
11 of 12 20-04-2010 00:35
Add to lightbox
page 841
page 842
Figure 31-10 A 48-year-old female with known portal vein invasion by pancreatic neuroendocrine
tumor. A, Coronal true FISP image demonstrating heterogeneous signal intensity thrombus within the
expanded portal (arrows) and splenic veins (arrowhead). B, Precontrast SHARP image demonstrates
the thrombus to have low signal intensity. C, The thrombus enhances following the administration of
contrast (large arrow, arrowhead). Note the multiple enhancing periportal collateral vessels (thin
arrows).
5,22
MR venography may be used to determine suitability of potential live related liver donors. Portal
vein variants such as portal trifurcation and the origin of the left portal vein from the right anterior portal
vein increase the risk of potential complications and may be grounds for exclusion. The presence of
accessory inferior hepatic veins greater than 5 mm in diameter can also be detected preoperatively,
which impacts on the surgical procedure. In a recent study, true FISP images were found to be as
good if not superior to equilibrium phase 3D contrast-enhanced MRV in the evaluation of the portal vein
and superior for the evaluation of the hepatic veins in prospective live related liver donors. 5
While portal vein and inferior vena cava anastomosis complications are uncommon causes of
post-transplant failure, occurring in 0.5% to 3% and 1% respectively, early detection and treatment
with interventional procedures is crucial to prevent graft failure.26 Contrast-enhanced 3D MRV along
with non-contrast true FISP sequences may be an alternative means to rapidly and noninvasively
evaluate the portal vein and IVC anastomoses.
Contrast-enhanced 3D MR venography also has an increasing role in the preoperative assessment of

potential live kidney donors. While in the past renal venous anatomy was not a major factor in the
surgical planning for organ donation, the increasing use of laparoscopic techniques for renal graft
harvesting is changing this. The presence of large retroaortic or circumaortic renal veins, particularly on
the left, may preclude a laparoscopic approach and necessitate a conventional right nephrectomy. 15
Following renal transplantation, contrast-enhanced 3D MR venography can be used to detect strictures

at the anastomosis of the transplant renal vein and native iliac vein without the use of nephrotoxic
16
iodinated contrast medium or ionizing radiation.
© 2010 Elsevier
12 of 12 20-04-2010 00:35
CONCLUSION
Currently, MR venography is not a mainstream clinical image modality despite the recent development
of new techniques including direct and indirect contrast-enhanced 3D MR venography. The clinical
indications for MR venography are growing. Use of non-contrast-enhanced steady-state free
precession gradient-echo and contrast-enhanced T1-weighted fat-saturated 2D spoiled gradient-echo
sequences may be sufficient in most cases to diagnose the presence or absence of central thoracic,
abdominal or pelvic thrombosis or occlusion.
REFERENCES
1. Ruehm S, Zimmy K, Debatin J: Direct contrast-enhanced 3D MR venography. Eur Radiol 11:102-112, 2001.
2. Neimatallah M, Ho V, Dong Q, et al: Gadolinium-enhanced 3D magnetic resonance angiography of the thoracic vessels. J
3. Kanne J, Lalani T: Role of computed tomography and magnetic resonance imaging for deep venous thrombosis and
pulmonary embolism. Circulation 109(suppl I): I-15-I-21, 2004.
4. Prince M, Grist T, Debatin J (eds): 3D contrast MR angiography, 3rd edn. Berlin: Springer-Verlag, 2003.
5. Carr J, Nemcek A, Abecassis M, et al: Preoperative evaluation of the entire vasculature in living liver donors with use of
contrast-enhanced MR angiography and true fast imaging with steady-state precession. J Vasc Interv Radiol 14:441-449,
6. Pereles FS, Baskaran V: Abdominal magnetic resonace angiography: principles and practical applications. Top Magn
7. Ho V, Corse W, Hood M, Rowedder A: MRA of the thoracic vessels. Semin Ultrasound CT MRI 24:192-216, 2003.
8. Foo T, Ho V, Marcos H, et al: MR angiography using steady-state free precession. Magn Reson Med 48:699-706, 2002.
9. Bosmans H, Wilms G, Dymarkowski S, Marchal G: Basic principles of MRA. Eur J Radiol 38:2-9, 2001. Medline
Similar articles
10. Green D, Parker D: CTA and MRA: visualization without catheterization. Semin Ultrasound CT MRI 24:185-191, 2003.
11. Prince M, Sostman D: MR venography: unsung and underutilized. Radiology 226:630-632, 2003. Medline
Similar articles
12. Ruehm S, Wiesner W, Debatin J: Pelvic and lower extremity veins: contrast-enhanced three-dimensional MR venography
with a dedicated vascular coil-initial experience. Radiology 215:421-427, 2000. Medline Similar articles
13. Swan J, Carroll T, Kennell T, et al: Time resolved three-dimensional contrast enhanced angiography of the peripheral
vessels. Radiology 225:43-52, 2002. Medline Similar articles
14. Lebowitz J, Rofsky N, Krinsky G, Weinreb J: Gadolinium-enhanced body MR venography with subtraction technique. Am J
Roentgenol 169:755-758, 1997.
15. Hussain S, Kock M, Izermans J, et al: MR imaging: a "one-stop shop" modality for preoperative evaluation of potential living
kidney donors. RadioGraphics 23:505-520, 2003. Medline Similar articles
16. Glockner J: Three-dimensional gadolinium-enhanced MR angiography: applications for abdominal imaging.
RadioGraphics 21:357-370, 2001. Medline Similar articles
17. Oxtoby J, Widjaja E, Gibson K, Uzoka K: 3D gadolinium-enhanced MRI venography: evaluation of central chest veins and
impact on patient management. Clin Radiol 56:887-894, 2001.
18. Fraser D, Moody A, Davidson I, et al: Deep venous thrombosis: diagnosis by using venous enhanced subtracted peak
arterial MR venography versus conventional venography. Radiology 226:812-820, 2003. Medline Similar articles
19. Sharafuddin M, Stolpen A, Dang Y, et al: Comparison of MS-325- and gadodiamide-enhanced MR venography of iliocaval
veins. J Vasc Interv Radiol 13:1021-1027, 2002. Medline Similar articles
20. Chasen M, Charnsangavej C: Venous chest anatomy: clinical implications. Eur J Radiol 27:2-14, 1998. Medline
Similar articles
21. Obernosterer A, Aschauer M, Schnedel W, Lipp R: Anomalies of the inferior vena cava in patients with iliac vein
thrombosis. Ann Intern Med 136:37-41, 2002. Medline Similar articles
22. Lee V, Morgan G, Teperman L, et al: MR imaging as the sole preoperative modality for right hepatectomy: a prospective
study of living adult-to-adult liver donor candidates. Am J Roentgenol 176:1475-1482, 2001.
23. Kroencke T, Taupitz M, Arnold R, et al: Three-dimensional gadolinium-enhanced magnetic resonance venography in
suspected thrombo-occlusive disease of the central chest veins. Chest 120:1570-1576, 2001. Medline Similar articles
24. Smits J, Bis C, Elgersma O, et al: Hemodialysis access imaging: comparison of flow-interrupted contrast-enhanced MR
angiography and digital subtraction angiography. Radiology 225:829-834, 2002. Medline Similar articles
25. Collins J, Pereles FS, Song G, et al: Pre-ablative pulmonary venous mapping by high resolution MR angiography facilitates
electrophysiologic pulmonary vein ablation and reduces fluoroscopy time in patients, Presented at the North American Society
1 of 2 20-04-2010 00:36
for Cardiac Imaging 30th Annual Meeting and Acientific Session/Third International Workshop on Coronary MR and CT
Angiography, Dallas, Texas, September 2002.
26. Kim B, Kim T, Jung D, et al: Vascular complications after liver related transplantation: evaluation with gadolinium-enhanced
three-dimensional MR angiography. Am J Roentgenol 181:467-474, 2003.
SUGGESTED READING
Baarslag H, van Beek E, Reekers J: Magnetic resonance venography in consecutive patients with suspected deep vein
thrombosis of the upper extremity: initial experience. Acta Radiol 45:38-43, 2004. Medline Similar articles
Ernst O, Asnar V, Sergent G, et al: Comparing contrast-enhanced breath-hold MR angiography and conventional angiography
in the evaluation of mesenteric circulation. Am J Roentgenol 174:433-439, 2000.
Gallix B, Achard-Lichere C, Dauzat M, et al: Flow-independent magnetic resonance venography of the calf. J Magn Reson
Halpern E, Mitchell D, Wechsler R, et al: Preoperative evaluation of living renal donors: comparison of CT angiography and MR
Korosec F, Frayne R, Grist T, Mistretta C: Time resolved contrast-enhanced 3D MR angiography. Magn Reson Med
36:345-351, 1996.
Ruehm S, Wiesner W, Debatin J: Pelvic and lower extremity veins: contrast-enhanced three-dimensional MR venography with
a dedicated vascular coil-initial experience. Radiology 215:412-427, 2000.
Shinde T, Lee V, Rofsky N, et al: Three-dimensional gadolinium-enhanced MR venographic evaluation of patency of central
veins in the thorax: initial experience. Radiology 213:555-560, 1999. Medline Similar articles
© 2010 Elsevier
2 of 2 20-04-2010 00:36
ARDIAC MAGING ECHNIQUES

Andrew C. Larson
Debiao Li
Orlando P. Simonetti
Niels Oesingmann
INTRODUCTION
While over 20 million cardiovascular imaging examinations are performed annually using X-ray
angiography, echocardiography, or nuclear medicine, the rapidly evolving techniques of cardiovascular
MRI (CMRI) offer many advantages which may ultimately position MRI as the modality of choice for
cardiovascular imaging. While X-ray catheter angiography can provide high-resolution images in
realtime and is therefore particularly useful for angioplasty and stenting, these lumenographic
techniques generally lack sufficient soft-tissue contrast for diagnostic interrogation external to the
vasculature and require the use of invasive catheterization strategies, radiation exposure, and
nephrotoxic contrast agents. Cardiac ultrasound, or echocardiography, offers realtime imaging
capabilities with economical and convenient bedside instrumentation. However, the ultrasound signal-
to-noise ratio (SNR) and contrast-to-noise ratio (CNR) are generally inferior to that of MRI and
accurate quantification of functional parameters can be difficult with ultrasound, a dilemma that is
compounded by a significant dependence on operator selection of imaging windows for appropriate
cardiac views. Nuclear medicine imaging techniques are relatively noninvasive and reproducible with
high prognostic value for both cardiac function and myocardial perfusion. While these techniques have
been readily adapted by cardiologists and benefit from over 20 years of clinical validation, nuclear
medicine techniques require radioactive isotopes and provide relatively poor spatial resolution and low
diagnostic specificity. CT techniques can provide high spatial resolution, which is particularly useful for
coronary artery angiography, and allows visualization of coronary artery calcification. However, these
techniques require exposure to X-ray radiation, provide relatively poor temporal resolution, and lack
large-scale clinical validation.
The overall capabilities of MR imaging techniques have been closely related to advances in scanner
hardware technology. Recent advances in magnet, gradient, and radiofrequency-coil designs have
improved imaging efficiency (i.e., reduced acquisition time while increasing SNR and CNR) and allowed
the utilization of pulse sequences and acquisition strategies for CMRI inconceivable only a few
decades previously. MRI offers a host of contrast mechanisms which can be readily adapted to
delineate tissues and physiologic phenomenon of particular diagnostic importance. This flexibility has
led to the promise of a "one-stop shop", comprehensive CMRI examination using only a single modality
for diagnostic cardiovascular imaging.1 This chapter will first introduce the MR pulse sequences most
commonly utilized for cardiovascular imaging. Later sections will describe CMRI techniques for
anatomic localization, interrogation of cardiovascular morphology, cardiac function and flow, myocardial
perfusion, myocardial viability, and coronary MR angiography.
page 843
page 844
1 of 2 20-04-2010 00:37
Add to lightbox
Add to lightbox
Figure 32-1 A, Spin-echo and B, turbo spin-echo pulse sequence diagrams. GSS, slice-select
gradient; GRO, read-out gradient; GPE, phase-encode gradient; ADC, radiofrequency receive.
© 2010 Elsevier
2 of 2 20-04-2010 00:37
PULSE SEQUENCES
Spin-Echo
Changes in tissue T1 and T2 relaxation times and proton density have been shown in a variety of
pathologic conditions, including myocardial infarction, 2-5 tumors,6,7 infiltrative and deposition
6,8,9 6,10 11
disease, cardiac transplant rejection, and atherosclerotic plaques. The traditional spin-echo
sequence provides flexible T1, T2, and proton-density weighting depending on the choice of time to
echo (TE) and repetition time (TR) and can therefore be very useful for cardiovascular imaging
applications. The spin-echo pulse sequence consists of a 90-degree radiofrequency excitation pulse
followed by a 180-degree refocusing pulse prior to readout of imaging data (Fig. 32-1A). Image
contrast can be freely changed from moderate T1 to strong T2 weighting depending on the choice of
TE and TR. Spin-echo sequences are not susceptible to the T2* magnetic field inhomogeneity and
susceptibility effects particularly problematic in many cardiovascular imaging orientations. However,
spin-echo sequences are sensitive to the effects of flow and motion, generally restricting spin-echo
data acquisition to diastole. The turbo spin-echo (TSE) sequence accelerates spin-echo data
acquisition by the application of multiple refocusing pulses during each TR, thereby permitting the
sampling of multiple phase-encoding lines within each TR (Fig. 32-1B). Further efficiency improvements
are possible with the HASTE strategy, a combination of half-Fourier sampling with single-shot TSE. 12
Gradient-Echo
The traditional gradient-recalled echo (GRE) pulse sequence consists of a radiofrequency excitation
pulse of flip angle α followed by the application of opposing polarity magnetic field gradients to bring
the spins into coherence for the formation of the imaging echo (Fig. 32-2A). In addition to proton
density, T1, and T2 weighting, GRE signals are T2* weighted because the dephasing of spins resulting
from magnetic field inhomogeneities is not refocused with gradient reversal. T2* decay requires the
use of a sufficiently shorter TE for GRE sequences, usually not longer than 8 to 10 ms in
cardiovascular imaging applications. Rather than extend the TR to allow full magnetization relaxation
between radiofrequency excitations, most GRE approaches involve spoiling the remaining transverse
magnetization each TR with radiofrequency phase-cycling strategies and crusher gradients, techniques
described as fast low angle shot (FLASH).13 While these strategies allow more rapid image acquisition
than traditional spin-echo sequences, the TR and flip angle α are ultimately limited by the T1 relaxation
times of the tissue. The TR and flip angle must be optimized in order to maximize signal intensities from
tissues of interest while maintaining sufficiently short TR for efficient image acquisition. Using too short
a TR and/or too large a flip angle does not allow sufficient regrowth of the longitudinal magnetization
prior to subsequent radiofrequency excitations. Flowing blood can replenish an imaging section with
fully recovered spins, producing images with signal intensity proportional to through plane flow velocity.
For FLASH pulse sequences, contrast between blood pool and neighboring myocardial or vascular
tissues is highly dependent on such inflow effects. The efficiency of GRE pulse sequences can be
enhanced by the utilization of an echo-train readout to sample multiple echoes with each
radiofrequency excitation (Fig. 32-2B).14-16 Depending on tissue T2*, the longer TR inherent in such an
approach may also offer increased SNR and CNR because of greater longitudinal magnetization
recovery and larger influx of fully recovered spins between excitations.
page 844
page 845
1 of 6 20-04-2010 00:39
Add to lightbox
Add to lightbox
Figure 32-2 A, Gradient-recalled echo (GRE) and B, GRE echo-train pulse sequence diagrams. GSS,
slice-select gradient; GRO, read-out gradient; GPE, phase-encode gradient; ADC, radiofrequency
receive.
The extravascular contrast agent gadolinium diethylenetriamine pentetate (Gd-DTPA) significantly

shortens the T1 relaxation time of those tissues containing sufficient concentrations of the contrast
agent. GRE pulse sequences are particularly useful for Gd-DTPA contrast-enhanced MRI because of
the strong T1 weighting and rapid acquisition times when using large flip angles and short TR. For
these reasons GRE pulse sequences are commonly utilized for contrast-enhanced MR
angiography,17,18 first-pass myocardial perfusion,19 and infarct imaging.20
TrueFISP
True fast imaging with steady-state precession (TrueFISP) is a steady-state free-precession (SSFP)
pulse sequence that involves complete refocusing of the magnetization over each TR (balanced
21,22
gradients) and 180-degree alternation in radiofrequency excitation phase (Fig. 32-3A). These
sequences may also be called balanced fast field echo (FFE), fast imaging employing steady-state
acquisition (FIESTA), or simply balanced SSFP. TrueFISP sequences have recently become ubiquitous
for many state-of-the-art CMRI techniques because of significant improvements in SNR and blood-
to-myocardium contrast when compared to FLASH sequences.23 These advantages are readily
2 of 6 20-04-2010 00:39
apparent when comparing an image acquired with the TrueFISP sequence to an image acquired with
the FLASH sequence while utilizing optimal imaging parameters for each technique (Fig. 32-3B).
Add to lightbox
Add to lightbox
page 845
page 846
Figure 32-3 A, TrueFISP pulse sequence diagram and B, comparison cardiac long-axis images
demonstrating the visibly improved contrast-to-noise ratio provided by TrueFISP over FLASH
techniques when using optimized sequence parameters for each acquisition strategy. GSS, slice-
select gradient; GRO, read-out gradient; GPE, phase-encode gradient; ADC, radiofrequency receive.
3 of 6 20-04-2010 00:39
Add to lightbox
Add to lightbox
Add to lightbox
Figure 32-4 A, TrueFISP off-resonance banding artifact (arrow) resulting from a region of greater than
180-degree spin dephasing over each TR. B and C, TrueFISP images exhibiting blood flow ghosting
artifacts (arrow) originating from the descending aorta present during systole (B) but not diastole (C).
4 of 6 20-04-2010 00:39
Unlike FLASH sequences, rather than spoiling the remaining transverse magnetization each TR,
TrueFISP sequences refocus the magnetization such that it can readily contribute to the imaging echo
in subsequent sequence repetitions. For cineangiography, TrueFISP sequences allow utilization of
much larger flip angles and shorter TR, which increases signal intensity and decreases acquisition
time. In contrast to FLASH, the TR allowed by TrueFISP sequences is generally only limited by
gradient hardware rather than T1 and through-plane flow rates. With |α| greater than 50 degrees and
TR much less than T2, the TrueFISP signal is primarily dependent on the T2/T1 ratio which is high for
blood relative to myocardium. A detailed description of TrueFISP signal characteristics is beyond the
scope of this chapter but several important characteristics are presented to facilitate the avoidance of
SSFP-related image artifacts for CMRI.
TrueFISP sequences require the development and subsequent maintenance of coherent transverse
magnetization throughout image acquisition. A lack of steady-state coherence caused by main field
nonuniformity, eddy currents, or inhomogeneous magnetic susceptibility can result in significant
off-resonance artifacts manifest as dark stripes or ghosting of blood flow (Fig. 32-4).
A crucial requirement for artifact-free TrueFISP imaging is to limit off-resonant spin dephasing within a
voxel to less than 180 degrees over each TR. The off-resonance artifacts associated with TrueFISP
sequences are most readily described by amplitude and phase response profiles as functions of the
phase evolution over each TR (Fig. 32-5). While the TrueFISP signal response profile is specific to flip
angle, TR, and tissue relaxation properties, the profile shown in Figure 32-5 is quite typical for the
imaging parameters and tissues encountered during CMRI (70-degree flip angle; 3.0/1.5 ms TR/TE;
250/1300 ms T2/T1). In Figure 32-5, notice the abrupt local amplitude minima at the ±180-degree
dephasing positions along the profile. These positions correspond to the dark bands depicted in Figure
32-4A. Between roughly ±100 degrees the magnetization is completely refocused and of nearly
uniform amplitude. The main magnetic field should be shimmed to position spins within the imaging
volume-of-interest (VOI) between these boundaries. Utilization of the shortest possible TR minimizes
spin dephasing, thereby decreasing the sensitivity to off-resonance banding artifacts. A shorter TR is
realized by using strong gradients and removing any unnecessary quiescent dead-times in the
implemented TrueFISP pulsing sequences.
page 846
page 847
Add to lightbox
5 of 6 20-04-2010 00:39
Add to lightbox
Figure 32-5 True FISP amplitude and phase-response profiles as functions of the phase evolution
over each sequence repetition time for α = 70 degrees; TR/TE = 3.0/1.5 ms; T2/T1 = 250/1300 ms.
SSFP blood-flow ghosting artifacts (see Fig. 32-4B), are the result of rapidly flowing spins maintaining
sufficient transverse magnetization to form coherent echoes after exiting the imaging slice. These
artifacts are most commonly encountered while acquiring images during systolic phases of the cardiac
cycle in orientations including portions of the aortic trunk.
Achieving a completely uniform magnetic field shim can be difficult in cardiovascular imaging
applications because of heart motion, blood flow, chemical shift effects, and susceptibility variations at
air-tissue interfaces.24 Most scanners utilize an iterative approach to determine shim and synthesizer
(excitation) center frequency (ωcf) settings. Imperfections in the magnetic field shim leads to a
spectrum of resonant water frequencies across the VOI. The ωcf determines the position of the water
spectrum on the response profile in Figure 32-5. While the water spectral bandwidth may represent
only a small range of dephasing angles relative to the size of the optimal ±100-degree dephasing
region, improper choice of ωcf may position portions of the VOI external to the ±100-degree region.
These complications may ultimately be eliminated with improved shimming and frequency adjustment
regimens. However, recently developed frequency scouting techniques can also be very helpful in
mitigating these complications, particularly for high spatial resolution CMRI applications requiring the
use of longer TR. These scouting strategies involve acquiring a series of images using identical
sequence-timing parameters but a range of ωcf settings. The ωcf setting corresponding to the image
demonstrating minimal artifact is chosen for subsequent scans. This approach can also be very helpful
for optimizing ωcf for frequency-selective fat-saturation pulses.24
A further complication related to shimming and frequency settings for TrueFISP CMRI involves the
intravoxel partial volume effects occurring at the fat-water interfaces. Fat and water spins within the
same voxel may be located at positions of opposite polarity along the TrueFISP phase-response curve
leading to signal cancellation. While this effect can provide improved edge delineation between
pericardium and myocardium for cineangiography and morphologic studies, signal cancellation can
potentially be problematic when using low-resolution single-shot acquisition strategies with relatively
large voxel sizes causing excessive signal voids along tissue edges bordered by fatty structures.
These intravoxel partial volume effects can also result in erroneous low-resolution B1-field maps for
parallel imaging strategies. However, typically such effects can be eliminated (at least within a VOI) by
simply adjusting ωcf to bring water and fat into phase within the VOI (Fig. 32-6).
For TrueFISP sequences, signal amplitude and phase will be modulated during the initial excitation
repetitions until achieving a steady-state value. Depending on tissue relaxation and flip angle, the
approach to steady state may require several 100 repetitions and therefore most TrueFISP acquisition
strategies mitigate these effects with dummy pulse preparation schemes involving the application of a
number of radiofrequency excitations prior to imaging data acquisition. Several of these strategies are
described in greater detail in the context of particular CMRI imaging applications in later sections.
© 2010 Elsevier
6 of 6 20-04-2010 00:39
SYNCHRONIZATION STRATEGIES
page 847
page 848
Add to lightbox
Add to lightbox
Figure 32-6 A, TrueFISP intravoxel fat-water cancellation effects are evident, particularly at the
pericardium-epicardium interface. B, These intravoxel signal cancellation effects can be mitigated by
adjusting the synthesizer frequency.
While emerging ultrafast imaging techniques may allow single-shot image acquisition for some
applications, most CMRI strategies require the segmentation of the acquisition of imaging data over
multiple heart beats in order to reduce cardiac motion artifacts. These strategies assume that the
motion of the heart does not change over multiple cycles and therefore data acquired during the same
short interval within the cardiac cycle over a number of heart beats can be combined without producing
motion artifacts. Cardiac cycle synchronization is necessary in order to utilize segmented acquisition
techniques, while both cardiac and respiratory synchronization are necessary if imaging while free
breathing. In addition to segmented acquisition strategies, cardiac synchronization is also necessary
for serial single-shot imaging over multiple heart beats.
Cardiac Cycle Synchronization
1 of 4 20-04-2010 00:39
Add to lightbox
Figure 32-7 ECG R-wave triggered synchronization of MRI data acquisition during diastole.
Cardiac cycle synchronization, or gating, is generally achieved by recording the ECG during the scan
and triggering MR data acquisition based on the detected position of the ECG R-wave (Fig. 32-7).
Delay times can be added to this trigger position in order to enable data acquisition during diastole.
Rapidly switching magnetic field gradients, radiofrequency pulsing, and magnetohydrodynamic effects
can induce significant artifacts in the acquired ECG signal.25-28 Such artifacts were particularly
problematic in early CMRI but recent advancements in ECG hardware and software have mitigated
most of these complications. Some recent advances include the use of fiberoptic and wireless lead
28
systems, artifact and noise suppression algorithms, and vectorcardiogram-based gating systems.
However, while such techniques have made significant improvements in ECG gating, signals acquired
within the scanner bore continue to lack diagnostic quality. Diagnostic-quality ECG recording must be
performed external to the MR scanner bore.
page 848
page 849
Peripheral gating is the most commonly utilized alternative to ECG gating for CMRI. For peripheral
gating, a signal representative of the cardiac cycle is sampled using a pulse oximeter or theoretically
any MR-compatible device capable of generating a waveform representative of pulsation in the
peripheral vasculature. Systolic peaks in the pulse waveform are detected in order to trigger MR data
acquisition. Peripheral gating strategies avoid many of the problems associated with ECG recording
within the MR scanner bore. However, these techniques are generally considered inferior to
ECG-gating techniques because: 1. the relatively smooth morphology of the systolic peaks (as
opposed to the sharp ECG R-wave) can result in less reproducible trigger positions relative to the
cardiac cycle over multiple heart beats; and 2. unlike R-wave trigger positions which precede
myocardial contraction, peripheral gating trigger times follow systolic myocardial contraction. The latter
effect can be incompatible for CMRI strategies requiring triggered magnetization preparation during
early diastole to be followed by data acquisition during late diastole.
Respiratory Cycle Synchronization

Most CMRI images can be acquired within a 5 to 15 second breath-hold with respiratory suspension
remaining the most common form of respiratory cycle synchronization. Achieving a consistent
breath-hold position between acquisitions is important in order to accurately adjust the orientations of
impending acquisitions based on images acquired during previous breath-holds. Patients tend to
achieve better consistency when breath-holding at expiration and practicing breath-holding procedures
prior to scanning can be very helpful. However, achieving sufficient spatial resolution and anatomic
coverage can require using an acquisition time exceeding that compatible with breath-holding,
particularly when examining severally ill, pediatric, or uncooperative patients. Successful
synchronization of image acquisition with respiration while free breathing can decouple acquisition time
and related imaging parameters from the limits of breath-hold durations. Image acquisition is typically
synchronized to intervals of expiration because of the greater reproducibility of tissue position at
expiration over multiple respiratory cycles. However, each respiratory cycle may not reproduce the
same shift in heart position as previous cycles and therefore optimal synchronization strategies should
take into account respiratory cycle phase as well as absolute changes in tissue position between
respiratory cycles.
2 of 4 20-04-2010 00:39
Early MR scanner hardware and pulse sequences required relatively long acquisition times and
therefore many body imaging strategies required the utilization of respiratory-gating techniques to limit
motion artifacts. The first respiratory-gating strategies involved sampling signals representative of
abdominal expansion and relaxation accompanying each respiratory cycle. Based on these signals,
data acquisition would be suspended during periods of inspiration29 or alternatively the order in which
30
phase-encoding lines were acquired would be modulated based on respiratory cycle position. These
synchronization signals can be derived with a variety of methods but most commonly using a pneumatic
pressure belt fastened across the abdomen. However, this type of respiratory gating has not gained
widespread acceptance for CMRI principally because of the potential inconsistencies in heart position
over multiple respiratory cycles which are not well represented in these types of synchronization
signals.
More recently, the navigator echo (NAV) techniques31 were developed primarily to allow longer
acquisition times and therefore improve coverage and spatial resolution for three-dimensional (3D)
coronary MR angiography.32-34 A NAV echo is acquired by exciting the spins within a spatially delimited
column, typically at the dome of the right hemidiaphragm (Fig. 32-8A). Following excitation of the tissue
column, either a gradient or spin-echo is recorded during the application of a frequency-encoding
gradient along the axis aligned with the excitation column. Excitation of the column is achieved with
either a 2D spatially-selective radiofrequency pulse,35 or the application of two slice-selective pulses
with the first 90-degree pulse exciting a sagitally-oriented slice and the second 180-degree pulse
exciting a slice roughly orthogonal to the first such that only those spins within the column are
refocused.34,36 Each acquired NAV echo produces a 1D image descriptive of the position of the liver-
to-lung interface at the superior edge of the diaphragm. A series of NAV echo measurements over
multiple respiratory cycles is shown in Figure 32-8B. A NAV echo is typically sampled immediately prior
to the acquisition of a single segment of phase-encoding lines and typically requires less than 50 ms.
Proper orientation of the NAV echo column is necessary in order to optimize respiratory
synchronization and to avoid corruption of the spins within the imaging volume of interest. The later
complication is particularly relevant when utilizing the orthogonal-slice approach to NAV echo column
excitation.
Add to lightbox
3 of 4 20-04-2010 00:39
Add to lightbox
page 849
page 850
Figure 32-8 A, Position of NAV echo excitation column (white bar) and B, series of NAV echo
measurements over multiple respiratory cycles.
For free-breathing NAV echo respiratory synchronized image acquisitions, the information regarding
diaphragm edge position can be utilized to: 1. accept or reject imaging data; 2. reposition the imaging
section based on positional shifts, dubbed "slice-following"; or 3. motion correct the acquired imaging
data based on these positional shifts. The latter two approaches rely on established indices describing
changes in heart position relative to changes in diaphragm air-to-liver interface position. 37-39 Prior to
the acquisition of imaging data, a 10 to 15 second NAV echo scout scan is typically necessary to
establish appropriate thresholds and acceptance windows corresponding to expiration. In addition to
allowing free-breathing acquisitions, NAV echo techniques can also be useful for prospective correction
of breath-hold drifts.
© 2010 Elsevier
4 of 4 20-04-2010 00:39
ANATOMIC LOCALIZATION
The first step in a CMRI examination involves the anatomic localization of those imaging planes
necessary for the orientation of subsequent scans. Localization is generally achieved by acquiring a
series of images with the orientation of each successive image adjusted based on the observed
anatomic structures in the previously acquired images. Several strategies have been proposed for
cardiovascular localization. The two most common strategies involve either serial ECG-triggered
single-shot acquisition of low-resolution images with scanning halted after the acquisition of each image
in order to adjust the imaging plane for successive scans; or repeated single-shot acquisition of
low-resolution images with realtime "steering" of the imaging plane with the localization scan only
halted upon reaching the proper orientation. The latter strategy requires realtime slice-plane controls
which are not available on all clinical scanners. For the former and most commonly utilized strategy,
the imaging plane for successive scans is typically oriented perpendicular to the immediately previous
imaging plane to ultimately migrate to an imaging plane in the desired orientation. To avoid motion
artifacts due to systolic contraction and early diastolic relaxation of the heart, these measurements are
typically made during late diastole. Typically, single-shot TrueFISP or FLASH sequences are used for
localization procedures. Alternatively, single-shot HASTE techniques can be useful for localizing
vascular structures.
The images in Figure 32-9 demonstrate the utilization of this type of localization strategy to achieve a
conventional cardiac short-axis orientation. From the simple transverse orientation (Fig. 32-9A), a
perpendicular slice is acquired bisecting the left ventricle to produce the roughly two-chamber
orientation in Figure 32-9B. The left ventricle is then obliquely bisected along the long axis to produce
the roughly four-chamber orientation in Figure 32-9C. Finally, short-axis images are acquired by
orienting the slice plane perpendicular to both the four-chamber slice plane in Fig. 32-9C and the
long-axis of the heart. This same general approach can be utilized to reach two-, three-, and
four-chamber long-axis orientations as well as orientations crossing valve planes and bisecting the
great vessels. The ability to reproducibly adapt imaging orientations in 3D space independent of
complicated morphology, pathology, and anatomic position is an important characteristic of MRI
impossible with many competing cardiovascular imaging modalities.
© 2010 Elsevier
1 of 1 20-04-2010 00:40
CARDIOVASCULAR MORPHOLOGY
Static anatomic imaging of the cardiovascular system has been performed using spin-echo-based MRI
techniques for a number of years. In fact, some of the earliest MRI results obtained in human subjects
using a whole-body superconducting magnet included gated spin-echo images of the heart.40 With the
introduction of multiecho spin-echo techniques, known as turbo spin-echo (TSE) or fast spin-echo
(FSE), the possibility of acquiring high-resolution, black-blood images of the heart and blood vessels
41 42,43
within a breath-hold, or even a single heart beat, became a reality. These rapid multiecho
techniques have generally replaced standard spin-echo sequences for anatomic cardiovascular
imaging. In this section, a technical description of black-blood imaging techniques will be provided and
more recent advances in black-blood sequences and applications will be discussed.
Fast Black-Blood Cardiac Morphologic Imaging

Spin-echo imaging of the heart and vascular system relies on the washout of blood signal to generate
contrast between the blood and vessel or heart wall.44 Blood flowing rapidly through the imaging plane
does not experience both the 90-degree excitation pulse and the 180-degree refocusing pulse and
therefore exhibits little or no signal. Spin-echo imaging has proven to be relatively successful in
generating black-blood images of the heart and blood vessels; however, it does suffer from several
drawbacks. First, blood must flow completely through the slice in the time between the 90- and
180-degree pulses. This time is typically on the order of 5 to 10 ms and may not be long enough,
particularly if the primary direction of flow is not perpendicular to the imaging plane. Second, this same
washout effect can cause signal dropout and motion artifact within the heart itself if data are acquired
during phases of rapid cardiac motion. Third, at an acquisition rate of only one line per cardiac cycle,
ECG-gated spin-echo sequences must run for several minutes and respiratory motion artifacts can be
problematic.
To address these problems, TSE sequences designed specifically for cardiovascular imaging were
introduced.41 One of the primary benefits of TSE over standard spin-echo is that the acquisition time is
sufficiently reduced to allow image acquisition during a short breath-hold. This is accomplished by
applying multiple refocusing 180-degree pulses after each excitation. Typically 8 to 32 echoes are
acquired following each excitation, reducing the acquisition time of a 128 × 256 image down from 16 to
4 heart beats. While reducing the acquisition time to a single breath-hold successfully eliminates
respiratory motion artifacts, the washout effect can still be insufficient to eliminate blood signal and
provide high contrast between blood and vessel or heart wall. A double-inversion blood nulling
45
preparation has been used successfully to suppress blood signal in both TurboFLASH and TSE pulse
sequences.
page 850
page 851
Add to lightbox
1 of 7 20-04-2010 00:40
Add to lightbox
Add to lightbox
Add to lightbox
Figure 32-9 A-C, Series of single-shot TrueFISP localization images sequentially acquired to achieve
D, the cardiac short-axis orientation. Dashed line in C indicates the perpendicular slice position
selected to achieve the orientation in D.
2 of 7 20-04-2010 00:40
The combination of a spatially nonselective 180-degree pulse followed immediately by a section-

selective 180-degree pulse is applied when the R-wave trigger is detected (Fig. 32-10). The result of
this combination is that all spins outside of the selected slice are inverted, while the magnetization
within the slice experiences both of the 180-degree pulses and therefore undergoes zero net nutation.
The inversion time extends over systole, allowing the noninverted blood within the slice plane to be
replaced by inverted blood (Fig. 32-11). The 90-degree excitation pulse of the TSE readout is then
applied in diastole after the heart has returned to approximately the same shape and position that it
was in at the R-wave when the blood nulling preparation was applied. Blood, which has a relatively
long T1 relaxation time, approaches its null point in 500 to 600 ms, depending on the time allowed for
recovery between inversion pulses. The shorter the time between successive inversion pulses, the
shorter the inversion time required to null the blood signal. This fact enables the technique to work over
a wide range of heart rates. At shorter heart rates, the time available within the cardiac cycle for
inversion recovery delay is shorter, but fortuitously the repetition time and inversion time to null blood
are also shorter. Acquiring the image data during mid-to-late diastole also avoids rapid cardiac motion
which can occur during systolic contraction and the early filling phase of diastole.
Correct synchronization of the acquisition to the cardiac cycle is critical to the effectiveness of the
black-blood preparation. The TSE acquisition is sensitive to motion, and image acquisition must take
place during a quiescent phase of the cardiac cycle. If the inversion time is too long, the blood signal
will begin to recover. These potential pitfalls are illustrated in Figure 32-12.
page 851
page 852
Add to lightbox
Figure 32-10 Timing diagram for black-blood turbo spin-echo. The double inversion pulse pair is
applied at the R-wave trigger, effectively inverting all spins outside of the imaged slice. During
inversion time (TI1), inverted blood replaces blood within the slice plane. The 90-degree excitation
pulse is applied near the time that inverted blood recovers to its null point. A third inversion pulse can
be optionally applied to generate STIR contrast for fat suppression and greater sensitivity to tissue
fluid.
3 of 7 20-04-2010 00:40
Add to lightbox
Figure 32-11 Proper timing of preparatory pulses and acquisition window within the cardiac cycle
allows for replacement of blood within the slice plane with inverted blood during systole. The imaged
slice plane experiences both nonselective and section-selective inversion, resulting in no net change in
magnetization. Image acquisition takes place during diastole when the heart has returned to
approximately the same position as when the preparatory pulses were applied at the following R-wave
trigger. (From Simonetti OP, Finn JP, White RD, Laub G, Henry DA: "Black blood" T2-weighted
41
inversion-recovery MR imaging of the heart, Radiology 199:49-57, 1996.)
Add to lightbox
page 852
page 853
Figure 32-12 Short-axis black-blood images obtained at different times following the double-inversion
pulse. A, Image is the result of timing the acquisition during systolic motion. The inversion time was set
too short to allow the prepared slice to return to its original position. Myocardial signal is lost due to the
combination of motion and inversion nulling. B, Image demonstrates proper inversion nulling of blood
and acquisition timing to diastole. C, Image shows the recovery of blood signal which can occur if the
inversion time is too long.
4 of 7 20-04-2010 00:40
Add to lightbox
Figure 32-13 Black-blood STIR image. Note the uniform suppression of fat in the chest wall, abdomen,
and surrounding the heart. Note also the high signal intensity in the inferolateral wall due to acute
myocardial infarction. (Courtesy of J. Bogaert and S. Dymarkowski, Department of Radiology,
Gasthuisberg University Hospital, Leuven Belgium.)
Several variations of the basic black-blood TSE scheme diagrammed in Figure 32-10 are currently in
widespread clinical use. T1-weighted imaging can be performed by using a short echo train (7-15
echoes) and centric reordering to keep the effective TE short and triggering every cardiac cycle to
keep the effective TR as short as possible. T2-weighted imaging is achieved by triggering every two or
three cardiac cycles and using a relatively long echo train (15-35 echoes) and long effective echo time.
STIR imaging for fat suppression and enhanced sensitivity to tissue fluid is attained through the use of
a third inversion pulse (see Fig. 32-10). An example of a STIR image is shown in Figure 32-13.
HASTE42,43 is a single-shot variation of TSE for black-blood imaging in a single heartbeat. HASTE
employs partial-Fourier data acquisition and very short echo spacing to reduce the scan time for an
entire image to less than 300 ms. HASTE images may appear to be slightly blurred relative to TSE due
to both the longer acquisition window and T2 decay, which may suppress the signal from higher spatial
frequencies. Example single-shot HASTE images are shown in Figure 32-14.
Black-blood TSE is currently in widespread use in clinical cardiac MRI practice. Its primary application
is for morphologic imaging of the myocardium, valves,46 and thoracic blood vessels. T2-weighted and
STIR variants of the technique are sensitive to tissue edema and have been used for evaluation and
47
characterization of cardiac and mediastinal masses as well as acute myocardial infarction. More
recently, high-resolution variations of black-blood TSE have been applied to imaging of vessel walls
and atherosclerotic plaque in the aorta48 and coronary arteries.49
5 of 7 20-04-2010 00:40
Add to lightbox
page 853
page 854
Figure 32-14 Axial, single-shot, black-blood HASTE images. Each complete image was acquired in
about 300 ms, a sufficiently short acquisition duration to avoid most respiratory motion artifacts.
Add to lightbox
6 of 7 20-04-2010 00:40
Add to lightbox
Add to lightbox
Figure 32-15 A, Prospectively-gated segmented cine acquisition; B, retrospectively-gated segmented
cine acquisition; and C, hypothetical k-space diagram corresponding to the phase-encoding lines
sampled in A and B.
Black-blood TSE has generally replaced standard spin-echo for morphologic imaging of the heart and
mediastinum. Effective blood signal suppression, rapid scan times, and the ability to generate contrast
sensitive to pathology have all contributed to the widespread acceptance of this technique. Promising
new results indicate that this technique will probably play an important role in the characterization of
atherosclerotic plaques in the peripheral and coronary vessels.
© 2010 Elsevier
7 of 7 20-04-2010 00:40
CARDIAC FUNCTION AND FLOW

MR cineangiography ("cine") involves acquiring a series of images depicting tissue position throughout
the cardiac cycle, which allows interrogation of cardiovascular morphology and motion as well as
assessment of cardiac function.13,50 MR cine techniques allow measurement of myocardial mass,
ejection fraction, and wall thickness. Cinematic viewing of these image series allows detection of hypo-
or hyper-kinetic wall motion abnormalities. MR flow imaging techniques allow the measurement of
blood flow velocity for functional measurements and the detection of hemodynamic abnormalities.
Segmented Cine
MR cineangiography is generally performed with segmented acquisition strategies during a breath-hold
using either FLASH or TrueFISP pulse sequences. The set of phase-encoding lines necessary to
achieve a desired spatial resolution is divided into a number of segments, Ns. For prospectively-gated
strategies (Fig. 32-15A), detection of the R-wave triggers repeated acquisition of a k-space segment
until Np total phases are sampled during an acquisition window shorter than the shortest expected R-R
interval. Each successive R-wave triggers acquisition of Np phases for another k-space segment.
Imaging data are synchronized by simply combining those segments acquired at the same cardiac
phase to produce an image. For retrospectively-gated strategies (Fig. 32-15B),51 a k-space segment
is repeatedly acquired during an acquisition window of a predefined duration spanning at least the
longest expected R-R interval before incrementing to the next segment during the next acquisition
window. The time between the acquisition of a phase-encoding line and the immediately previous
R-wave trigger is stored with that line. Data are synchronized retrospectively by using weighted
interpolation schemes to generate the necessary phase-encoding lines for reconstruction of images at
each cardiac phase.
page 854
page 855
Add to lightbox
Figure 32-16 Factor-of-two echo sharing for improving temporal resolution for prospectively-gated
segmented cine MRI.
The total acquisition time for prospective gating techniques depends on heart rate and Ns, whereas the
total acquisition time for retrospectively-gated techniques depends on the acquisition window length for
each segment and Ns. The temporal resolution of the cine imaging frames is dependent on the number
of views acquired for each segment (views/segment, VPS) and the sequence repetition time.
Retrospectively-gated techniques require a slightly longer overall acquisition time than prospectively-
triggered techniques, depending on the degree of temporal oversampling of each cardiac cycle.
1 of 11 20-04-2010 00:41
However, retrospectively-gated techniques are not as susceptible to heart-rate fluctuations and can
provide images throughout the cardiac cycle, which is important for assessing systolic and/or diastolic
disease processes. Prospectively-triggered techniques struggle to provide late-diastolic images.
Ideally, chosen cine acquisition parameters should allow reconstruction of 15 to 20 cardiac phases
depending on a patient's heart rate.
Temporal resolution for cineangiography can be increased by using "sliding-window" echo-sharing

techniques.52,53 Rather than simply reconstructing images at the cardiac cycle positions of the acquired
phases, intermediate images can be reconstructed by "sharing" subsets of the phase-encoding lines
acquired during adjacent phases to reconstruct intermediate images and thereby provide a
"reconstructed" temporal resolution greater than that defined by VPS. An example of factor-of-two
echo sharing for prospectively-triggered cine MRI is shown in Figure 32-16. The echo-sharing window
must be sufficiently short to minimize motion artifacts. An excessive echo-sharing window duration can
result in segmentation ghost artifacts and/or object blurring. Central k-space lines may be updated
more frequently to reduce the difference between acquired and reconstructed temporal resolution for
54-56
the lower spatial frequency image components. The following set of imaging parameters are
typical of those commonly utilized for TrueFISP segmented cine imaging with state-of-the art 1.5-T
clinical scanners for an eight-heart-beat breath-held acquisition: 6 mm slice thickness; 300 × 250 mm2
field of view; 3.0/1.5 ms TR/TE; 60-degree flip angle; 975 Hz/pixel bandwidth; 15 lines/segment; 256 ×
2
120 matrix; and 1.17 × 2.08 mm inplane voxel size. Imaging parameters for a segmented cine FLASH
acquisition with similar temporal and spatial resolution would be 8.0/4.0 ms TR/TE; 20-degree flip
angle; 195 Hz/pixel bandwidth; and 6 lines/segment. Recall that the FLASH sequence requires a longer
TR to allow inflow for sufficient blood-to-myocardium contrast. Therefore, to provide the same spatial
resolution with the FLASH imaging techniques, an acquisition time of 20 heart beats would be required.
Even with identical resolution, the FLASH images would typically remain inferior in both CNR and SNR
to comparable TrueFISP images. Furthermore, the FLASH sequence flip angle often must be adjusted
depending on imaging orientation in order to accommodate orientation-dependent inflow velocities.
Therefore, short-axis views with primarily through-plane flow require moderate flip angles of typically
20 to 25 degrees, whereas long-axis views with inplane flow require lower flip angles of 15 to 20
degrees to reduce saturation. TrueFISP segmented cine techniques are generally considered superior
to FLASH techniques. However, FLASH techniques continue to be useful for visualizing blood flow jets
and valve orifices not always well depicted by TrueFISP.
To avoid signal oscillations during the approach to steady state, most TrueFISP cine strategies involve
application of a single cardiac cycle of dummy pulses, prior to sampling of imaging data. Imaging data
are not acquired during application of the dummy pulses. Application of these dummy pulses is
generally preceded by application of a single α/2 radiofrequency pulse which significantly reduces the
number of dummy pulses necessary for the magnetization to reach steady state. 57 To maintain steady
state throughout a prospectively-gated TrueFISP cine acquisition, dummy pulses must also be
continuously applied during the time interval between the acquisition of the final phase of a segment
and the next R-wave triggering acquisition of the following segment. This approach is commonly
described as a constant radiofrequency acquisition. FLASH approaches generally do not utilize dummy
pulse preparation schemes and constant radiofrequency excitation and therefore the first images of
prospectively-triggered FLASH cine image series may appear much brighter than the rest of the image
series. Dynamic cinematic viewing of these image series demonstrate the well-known "lightning"
artifact.
page 855
page 856
For the assessment of myocardial function, cine image series are typically acquired in four-, three-,
and two-chamber, and short-axis orientations. Short-axis images are typically acquired as a stack of 8
to 10 parallel slices evenly spaced to cover the left ventricle from base to apex. Examples of these
four imaging orientations are shown in Figure 32-17 with an example of a short-axis stack (single
diastolic phase images from each cine series of the stack) shown in Figure 32-18.
Realtime Cine
2 of 11 20-04-2010 00:41
Realtime cine imaging techniques can be utilized when either breath-holding or ECG gating are not
possible. For these techniques, k-space data for the reconstruction of each cardiac image are
acquired within temporally distinct acquisition windows rather than multiple acquisition windows from
multiple cardiac cycles. Realtime techniques inherently eliminate the issues associated with cardiac
gating but inevitably increase the difficulty of achieving a temporal resolution, spatial resolution, SNR,
and overall image quality sufficient for clinical utility. ECG-triggered realtime acquisition techniques can
also be utilized as a single-shot cine imaging strategy to acquire cine image series at multiple slice
positions covering the entire heart in a single breath-hold.
Add to lightbox
Add to lightbox
3 of 11 20-04-2010 00:41
Add to lightbox
Add to lightbox

Clinical Magnetic Resonance Imaging PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Clinical Magnetic Resonance Imaging PDF

Hochgeladen von

Copyright:

Verfügbare Formate

Clinical Magnetic Resonance Imaging: 3-Volume Set

by Robert R. Edelman, John Hesselink, and Michael Zlatkin

I. PHYSICS, INSTRUMENTATION, AND ADVANCED TECHNIQUES

27. Magnetic Resonance Angiography: Basic Principles

39. Brain: Indications, Technique, and Atlas

HYSICS NSTRUMENTATION AND DVANCED

ISTORY OF AGNETIC ESONANCE

OVERVIEW OF THE HISTORY OF MAGNETIC RESONANCE

by Zavoisky at the Kazan State University.

The Nobel Prize for Physics 1952

where B includes the various magnetic fields applied.

First Magnetic Resonance Images

NMR diffraction in solids

First Commercial Magnetic Resonance Scanners

A GUIDED TOUR OF THE FOURIER TRANSFORM

The discussion in this section is centered around the following questions.

What is the Fourier transform?

A more in-depth coverage can be found in many texbooks.24,25,29

Magnetic Resonance Imaging

DEVELOPMENT OF MAGNETIC RESONANCE IMAGING

Table 1-1. Development of MR Gradient Amplitude and Rise Time

Table 1-2. Echo-Planar Imaging Terminology

Note: terminology is the same for all the manufacturers.

Table 1-3. Spin-Echo Sequences

Single shot FSE

Table 1-4. Gradient-Echo Sequences

Box 1-1 Historical overview: the development of MR. As a

Box 1-2 Nobel Prizes Directly Related to MR. (From ref. 44

Box 1-3 Nobel Prizes in Other Fields, Awarded to Individuals

How is the nuclear magnetic resonance signal generated?

ORIGIN OF THE NUCLEAR MAGNETIC RESONANCE SIGNAL

In a nucleus, the magnetic moments of proton-proton or neutron-neutron pairs tend to align in

Behavior of Nuclear Spins in a Magnetic Field

With regard to the classical-physics perspective portrayed in Figure 2-2, it is

Response of the Magnetization to a Radiofrequency Pulse

Generation of the Magnetic Resonance Signal

BASIC CHARACTERISTICS OF THE MAGNETIC RESONANCE SIGNAL

Longitudinal relaxation is characterized by an exponential regrowth of Mz with a time constant that is

Dephasing and T2*

SPATIAL LOCALIZATION OF THE MR SIGNAL TO FORM AN IMAGE

The characteristics of an applied magnetic-field gradient are commonly represented by a timing

The Fourier Transform and an Introduction to k-Space

information in spatial-frequency space, or k-space.

Generalizing this expression to three dimensions gives:

The Pulse-Sequence Timing Diagram

Multislice, Sequential-Slice, and Three-Dimensional Imaging

BASIC FORMS OF IMAGE CONTRAST

A proton-density-weighted image is obtained by minimizing the contributions from T1 and T2 relaxation,

Conceptually, T2*-weighted imaging is identical to T2-weighted imaging; contributions from T1

RACTICAL ONSIDERATIONS AND MAGE PTIMIZATION

carefully regardless of size or location.

MR Safety for Pregnant Patients and Technical Staff

PREPARING A PATIENT FOR MRI

Education of the Patient

Physical Preparation of the Patient

If a protocol includes IV contrast medium administration, a technologist, nurse, or physician should

case the patient cannot tolerate the full study.

If these techniques fail, the patient may need oral or IV sedation.

IV sedation may be arranged for a severely claustrophobic or anxious patient. If IV sedation is

SELECTING AN IMAGING PROTOCOL

bandwidths, such as certain artifacts, that will be considered later on.

Selecting a Pulse Sequence

Incoherent MPGR FLASH T1-FAST, SHORT GRE STAG