Eur Radiol (2011) 21:14241429

DOI 10.1007/s00330-011-2062-1

COMPUTED TOMOGRAPHY

Metal artifact reduction by dual energy computed

tomography using monoenergetic extrapolation
Fabian Bamberg & Alexander Dierks &
Konstantin Nikolaou & Maximilian F. Reiser &
Christoph R. Becker & Thorsten R. C. Johnson

Received: 29 September 2010 / Revised: 19 November 2010 / Accepted: 1 December 2010 / Published online: 20 January 2011
# European Society of Radiology 2011

Abstract
Objective The aim of the study was to assess the
performance and diagnostic value of a dual energy CT
approach to reduce metal artefacts in subjects with metallic
implants.
Methods 31 patients were examined in the area of their
metallic implants using a dual energy CT protocol
(filtered 140 kVp and 100 kVp spectrum, tube current
relation: 3:1). Specific post-processing was applied to
generate energies of standard 120 and 140 kVp spectra
as well as a filtered 140 kVp spectrum with mean photon
energies of 64, 69 and 88 keV, respectively, and an
optimized hard spectrum of 95150 keV. Image quality
and diagnostic value were subjectively and objectively
determined.
Results Image quality was rated superior to the standard
image in 29/31 high energy reconstructions; the diagnostic
value was rated superior in 27 patients. Image quality and
diagnostic value scores improved significantly from 3.5 to
2.1 and from 3.6 to 1.9, respectively. In several exams
decisive diagnostic features were only discernible in the
high energy reconstructions. The density of the artefacts
decreased from 882 to 341 HU.
F. Bamberg : A. Dierks : K. Nikolaou : M. F. Reiser :
C. R. Becker : T. R. C. Johnson
Department of Clinical Radiology, Ludwig-Maximilians
University,
Klinikum Grosshadern,
Munich, Germany
T. R. C. Johnson (*)
University of Munich,
Grosshadern Campus, Marchioninistrasse 15,
81377 Munich, Germany
e-mail: thorsten.johnson@med.uni-muenchen.de

Conclusions Dual Energy CT with specific postprocessing

can reduce metal artefacts and may significantly enhance
diagnostic value in the evaluation of metallic implants.
Keywords Metal artifacts . Dual energy CT . Computed
tomography . Quanta extrapolation

Introduction
Since the beginning of computed tomography (CT), metal
artefacts have been representing a significant limitation in
diagnostic image evaluation [8]. In the presence of metallic
joint prostheses or osteosynthetic material, it would be
desirable to evaluate the metal implant itself, the interface
between implant and bone and the surrounding tissue.
Important diagnostic criteria include exclusion of fractures
or loosening and verification of sufficient coverage of the
implant [13] and ruling out of hematoma or inflammation in
the adjacent soft tissue. However, due to the occurrence of
metal artefacts, the evaluation of these features remains
challenging with many cases rendered uninterpretable, even
with hard convolution kernels and widened CT density
ranges [8].
The metal artefacts actually comprise two main different
components. The one component is photon starvation due
to full absorption of the x-ray quanta, causing zerotransmission projections. The other component is beam
hardening due to absorption of low energy quanta [1].
While there is no remedy for these artefacts, in clinical
practice, higher energy quanta reduce the extent of artefacts
to some degree and would constitute a potential approach to
improve diagnostic CT image quality in metallic implants.
Specific algorithms such as linear interpolation of reprojected metal traces and multi-dimensional adaptive filtering

Eur Radiol (2011) 21:14241429

of the raw data can also improve image quality but may
also yield other artefacts [12].
Dual Energy CT as a relatively new approach has
brought about several advances in clinical CT interpretation, largely by improving the specificity of diagnostic
information. This is accomplished mainly by displaying the
presence, amount or distribution of specific substances
based on their specific absorption spectrum [6]. For
instance, these methods make it possible to differentiate
kidney stones, to characterize renal lesions, to map lung
perfusion, to depict arteries without superimposing bones or
to differentiate plaque components [2, 3, 911]. Moreover,
images can be optimized in certain aspects based on the
spectral information, i.e. the difference between both
datasets. For instance, it is possible to accentuate the
iodine-related contrast and decrease the noise in the image
[5]. Also, it feasible to extrapolate the beam hardening to
generate images as though they had been acquired with
monoenergetic high energy quanta. This latter approach
may substantially reduce metal artefacts in the acquired
data sets.
Thus, the purpose of this study was to initially evaluate
the efficacy of this technique in removing metal artefacts
based on quantitative CT density measurements and
subjective grading of artefacts and adequacy for diagnostic
evaluation.

Materials and methods

Patient population
All subjects with metallic implants who had been referred
for CT were included in the study. Apart from the indication
for CT, the main inclusion criterion was the presence of a
metallic implant in the examination area. The only
exclusion criterion would have been an age under 18 years,
but all eligible patients referred during the recruitment
period were adults. The exams had been requested with
clinical indication; no CT examinations were acquired
merely for study purposes. Written informed consent was
obtained prior to the examination.
CT image acquisition
All examinations were performed using a Dual Source CT
(Somatom Definition Flash, Siemens Medical, Forchheim,
Germany). As voltages, 140 and 100 kVp were chosen. To
further harden the 140 kVp spectrum, an 0.1 mm tin (Sn)
filter was applied. The relation between the tube currents
was set to approximately 3:1 in favour of the high potential
tube. The purpose was to get as much high energy quanta
as possible and to still get a sufficient low energy image for

1425

beam hardening correction. The CT acquisition parameters

were adjusted such that the dose of the Dual Energy CT
protocol equalled that of the routine single energy examination for the respective body region. In the trunk of the body,
i.e. thorax, abdomen and thighs, the protocol parameters were
100 kVp at 90 mAseff/rot and Sn140 kVp at 270 mAseff/rot
with tube current modulation; collimation was 320.6 mm
and pitch 0.6. For the extremities, tube currents were set to
60 and 180 mAseff/rot, respectively, with 400.6 mm
collimation and a pitch of 0.7. Tube current modulation
was kept on to compensate for the variable attenuation of
different projections due to the metallic implant and to
minimize the dose at the same time.
CT image reconstruction and analysis
For reconstruction, the sharpest available iterative kernel
Q40 and a relatively high slice thickness of 1.5 mm at
1.0 mm interval were applied to optimally exploit the
projection data. Images were reviewed on a 3D workstation (MMWP; Somaris Version CT2008G, Siemens
HealthCare, Forchheim, Germany) using the syngo Dual
Energy software (version VE32B with the monoenergetic application). This algorithm decomposes the CT
numbers of the image into a water-like and an iodinelike component, the former showing a linear relation of
densities in the high and low energy image, the latter an
additional photo effect, i.e. a higher density in the low
energy image. The extrapolation to a certain photon
energy is accomplished scaling this latter part of the
density linearly.
Images were reviewed by two independent observers
(with one and three years of experience in body CT) at
representative levels showing the thickest areas of the
metallic implant in axial plane with standard bone
window settings (width 2400 HU, center 400 HU).
Datasets were generated with extrapolated energies of
64, 69, 88, 105 keV and a manually adjusted optimal
energy (cf. Fig. 1). These energies had been chosen to
match the mean energies of standard 120 kVp (64 keV)
and 140 kVp (69 keV) spectra and an 140 kVp spectrum
with additional 0.1 mm thick tin filter (88 keV). The
105 kVp had been chosen empirically, i.e. based on the
observation that the 105 keV setting mostly provided good
results in five patients we had scanned before initiating the
prospective trial.
To quantify beam hardening artefacts, the density in the
most pronounced, i.e. the most hypodense streak was
measured by both observers adjacent to the metallic implant
at the same site in the reconstructions at the different
energies. These streaks were as wide as the metallic implant
and broad enough to place a region of interest in them. A
reference density measurement was obtained in similar

1426

Eur Radiol (2011) 21:14241429

Fig. 1 Two sets of images showing screws in the spine and in the
tibia reconstructed at 64, 69, 88, 105 and an optimal keV setting (left
to right). Note that the spinal canal the thin layer of bone covering the

left screw are only discernible in the two reconstructions at the highest
energy. Similarly, the screw in the tibia is optimally depicted in the
rightmost image

tissue, i.e. mostly muscle or fat, in an area outside the

artefacts. The regions of interest in the hypodense streak and
the reference area were chosen such that based on anatomy
they had to contain the same type of tissue. Additionally, the
two observers graded both the artefacts and the diagnostic
interpretability on a five point rating scale. Regarding
artefacts, a score of 0 indicates the absence of artefacts, 1
indicates the minor streaks only at the thickest portions of
the metallic implant, 2 represents minor streaks, 3 indicates
pronounced streaks, 4 represents massive artefacts. With
regard to diagnostic evaluation, 0 indicates a fully diagnostic exam, 1 an exam in which minor artefacts in areas of
thick metal that do not affect diagnostic evaluation, 2 minor
streaks without impact on the evaluation of the implant and
the adjacent tissue, 3 a restricted diagnostic interpretation
and 4 an insufficient diagnostic interpretability.

quality score (subjective) as well as absolute and relative HU

measurements (to compensate for different absolute HU values
given the variety of body parts) and address the clustered data
structure (several measurements per subjects), repeat measure
analysis was performed. As such we fitted linear models in
SAS Proc Mixed (SAS Institute, Cary, North Carolina) by
assuming a standard variance components pattern of covariance. All statistical analyses were performed using the SAS
system for Windows, version 9.2 (SAS Institute, Cary, North
Carolina), and statistical significance was defined as p<0.05.

Statistical analysis
Descriptive statistics are provided as mean standard
deviation for continuous and n (%) for categorical variables.
To determine differences of the artefact and diagnostic image
Fig. 2 Improvement of subjectively defined diagnostic image
quality (a) and objectively
measured HU in the most
significant artefact zone
adjacent to the metal implant
(b) at different keV settings
from 64 keV to 105 keV as
well as the individually
selected optimal keV
setting (optimal)

Results
Overall, 31 consecutive subjects were included in the
analysis. They were predominantly elderly women (mean
age: 64.819, 54% female). No adverse event occurred
during the CT examination. Regarding the implants, there
were 22 in the trunk of the body (6 spine, 12 hip, 4 femur)
and 9 in the extremities (2 humerus, 2 radius/ulna, 5 ankle),
examined with the respective protocol. Average CT dose
index (CTDI) was 11.0 mGy in the extremities and

Eur Radiol (2011) 21:14241429

1427

Table 1 Observed diagnostic image quality and extent of artifacts across the different keV settings. Delta indicates the difference between 64 keV
and optimal setting, p the probability of equivalence
Variable

64 keV

69 keV

88 keV

105 keV

Optimal

P-value

Diagnostic Value
Extent of Artifacts
Artifact Density (HU)
Reference Soft Tissue Density (HU)

3.60.7
3.70.7
900.1222
48.044

3.50.7
3.60.7
875.8244
54.946.5

2.80.8
2.90.9
495.4251
56.822

2.20.9
2.01.1
374.8287
52.924.1

2.10.9
1.91.1
341309
45.820.2

1.60.8
1.80.9
558269
2.234

<0.001
<0.001
<0.001
0.67

15.4 mGy in the trunk. On average, the DLP was 398.5

174 mGy*cm.
The mean density measured in the most pronounced
streak was strongly negative and the reference HU
measurement consistent with adjacent soft tissue (average
HU: 597.6357 and 12.916.2, respectively). On average,
the optimal energy setting was found at 119.513.5 keV
(Range: 95 to 150 keV). With increasing photon energy, the
absolute HU in the most pronounced streak increased
significantly from 900222 to 341309 HU (p<0.001;
cf. Fig. 2). In relation to the reference measurements in the
adjacent soft tissue, the relative density also decreased

significantly from 19.4 20 to 8.4 9.8 (p=0.01). As

expected there was no significant difference between these
reference density values in surrounding soft tissue (p=0.67).
Across all energy settings, the scores for diagnostic
value and the extent of artefacts amounted to 2.91.0 and
2.81.0; respectively. The extent of artefacts decreased
with increasing keV (p<0.001) from 3.70.7 to 1.91.1
with an absolute decrease of 1.8 units (relative decrease:
48.6%; cf. Fig. 2). Image quality was rated superior to the
standard image in 29/31 high energy reconstructions. In
pairwise comparison, while there were significant differences
between each group (all p<0.05), the optimal keV setting

Fig. 3 Example of a 72-year old female patient with a lateral femoral

neck fracture who underwent gamma nail osteosynthesis. Images are
displayed at the optimal setting of 109 keV as coronal reformat (a) and
as volume rendered image highlighting the metal in blue (b). Overall,
there are only minimal artefacts allowing to sufficiently evaluate the
prosthesis, and to delineate the fracture in the adjacent bone

Fig. 4 68-year old male patient with a locked intramedullary nail who
underwent dual energy CT for assessment of pseudarthrosis. Images
show a coronal reconstruction (a) and a volume rendered image (b) at
an extrapolated photon energy of 105 keV. There are minimal residual
artefacts at the largest diameter of the metal in axial plane due to
screws

1428

Eur Radiol (2011) 21:14241429

exams had no relevant impairment of diagnostic value and

42% (n=13) had minor streaks without impact on the
evaluation of the implant and the adjacent tissue, resulting
in 68% of exams with fully diagnostic image quality. Only
6% (n=2) were classified as non-diagnostic.

Discussion

Fig. 5 Volume rendered image of a hip prosthesis at the optimal

energy of 110 keV. Note the facture of the dorsal screw

did not provide significantly less artefacts as compared with

the 105 keV setting (p=0.68). Using the optimal energy
setting, 36% had none or minor artefacts in areas of thick
metal which did not affect diagnostic evaluation (n=11) and
only 6% (n=2) were classified as non-diagnostic.
The diagnostic value score was significantly different
among the keV settings (p<0.001) with significantly higher
diagnostic value in the 105 keV and optimal energy group
(from 3.60.7 to 2.10.9 for 64 keV and optimal energy;
respectively; Table 1, Fig. 1), consistent with a decrease of
1.6 units (relative decrease: 44.4%). Overall, the diagnostic
value was rated superior in the high energy reconstruction
in 27 of 31 patients, and equal in the remaining cases. In
five (16%) cases, decisive diagnostic features were only
discernible in the high energy reconstructions. Examples
include a fracture of the bone close to the implant (Fig. 3), a
pseudoarthrosis adjacent to the implant (Fig. 4) a fracture of
a screw of the metallic implant (Fig. 5), or loosening of a
screw (Fig. 6). In pairwise comparison, while the diagnostic
value score decreased by each group significantly (all p<
0.05), there was no significant difference between the
105 keV and the optimal energy (2.20.9 vs. 2.10.9, p=
0.63 for 105 keV and the optimal energy; respectively).
Using the optimal energy setting, 26% (n=8, score 1 and 2)

In this study, we provide initial evidence that metal artefacts

can be substantially reduced using high energy extrapolation based on Dual Energy CT. The results indicate a
significantly improved diagnostic assessment of implants,
of the surrounding tissue as well as the interface between
implant and bone. We also show that 105 keV may serve as
a robust standard energy setting to provide an optimal
visualization of metallic implants.
The effects of photon starvation and beam hardening
have limited the assessability of metal implants and its
surrounding tissue since the introduction of CT imaging
[13]. The result is that the metallic implant itself is depicted
as maximum density value without internal structure, the
interface between implant and surrounding tissue is
distorted or at least not clearly depicted, and the surrounding soft tissue is partially not evaluable due to pronounced
hypodense streaks. In clinical practice, high tube voltage
settings have been used to reduce metal artefacts because
high energy quanta traverse metal better, i.e. with less
absorption or filtering. However, this approach is not
always sufficiently effective and requires a dedicated high
voltage acquisition protocol with associated disadvantages
in soft tissue differentiation. Factors influencing the degree of
artefacts are the metallic alloy, orientation and thickness of

Fig. 6 Multiplanar reformat at 115 keV in the plane of pedicle screws

after dorsal spondylodesis. Note the loosening of both screws. The
remaining screws at other levels did not show these hypodense fringes
around the threads

Eur Radiol (2011) 21:14241429

the implant and CT acquisition parameters, including tube

voltage, tube current and collimation. Studies have shown
that Titanium is the least obtrusive metal [4] and that artefacts
increase with thickess of the implant and angulation to the
long axis of the gantry, while high tube voltage and current
as well as narrow collimation decrease artefacts [1].
Dual energy CT acquisitions, which are increasingly
used in radiological imaging, may provide an attractive
option to mitigate metal artefacts. The dual energy datasets
may be used to extrapolate to much higher photon energies,
which can be adapted according to metal alloy, or to
generate standard 120 kVp images for soft tissue visualization. Our initial results indicate that this approach is
successful and quite effective. It appears to be more
efficient than a mere increase of tube voltage, even if an
additional tin filter is applied to rid the spectrum of low
energy quanta, as we showed with the 88 keV setting.
The artefacts were not completely eliminated but substantially reduced, the image quality improved by 49% and the
diagnostic value was enhanced by approximately 44%.
Overall, we found that 68% of examinations became
diagnostic. The optimal keV setting providing the lowest
artefacts and highest diagnostic value varied depending on
the prosthesis, i.e. probably depending on its diameter and
alloy (Range: 95150 keV). Thus, dual energy CT may allow
for individual photon energy optimisation which can be
performed retrospectively without knowledge of the material
used. The fact that we did not observe significantly better
values in artefact reduction and diagnostic image quality with
the individually optimized energy extrapolation setting can
probably be attributed to the limited sample size. Further
systematic research will be necessary to determine accurate
differences in optimal settings between prosthesis material
compositions, which was not objective of the present study and
not feasible as data on the alloy of the implants was lacking.
Importantly, this technique works without additional
radiation dose. The tube currents are adjusted such that the
dose equals that of the routine protocols and complies with the
respective legal requirements or recommendations. In this
initial trial, we adjusted the tube current relation in favor of the
high energy spectrum in order to reduce both truncation and
beam hardening artefacts, while the low energy spectrum was
acquired with low dose just sufficient to use it for beam
hardening correction. This approach was chosen based on
results of Monte Carlo simulations [7]. However, depending
on the exact alloy and diameter of the metallic implant, on the
body region and its diameter, the optimal tube current relation
may vary, and it will require more specific clinical trials to
optimize the protocols in this respect, presumably resulting in
variable tube current relations for different body regions.
Regarding workflow and automation, it is of interest that
we did not observe significant differences between the
105 keV and the optimized high energy reconstructions.

1429

This implies that high energy reconstructions can be

performed routinely at 105 keV without the need for
optimizing the energy level individually. This approach
can also be applied in combination with purely software
based algorithms for metal artefact reduction [14].

Conclusion
High energy reconstructions of Dual Energy CT datasets can
significantly reduce metal artefacts and improve image quality
and diagnostic value. The evaluation of metallic implants
and adjacent bone or tissue is considerably enhanced.

References
1. Barrett JF, Keat N (2004) Artifacts in CT: recognition and
avoidance. Radiographics 24:16791691
2. Graser A, Becker CR, Staehler M, et al. Single-phase dual-energy
CT allows for characterization of renal masses as benign or
malignant. Invest Radiol 45:399405
3. Graser A, Johnson TR, Bader M et al (2008) Dual energy CT
characterization of urinary calculi: initial in vitro and clinical
experience. Invest Radiol 43:112119
4. Haramati N, Staron RB, Mazel-Sperling K et al (1994) CT scans
through metal scanning technique versus hardware composition.
Comput Med Imaging Graph 18:429434
5. Holmes DR 3rd, Fletcher JG, Apel A et al (2008) Evaluation of
non-linear blending in dual-energy computed tomography. Eur J
Radiol 68:409413
6. Johnson TR, Krauss B, Sedlmair M et al (2007) Material
differentiation by dual energy CT: initial experience. Eur Radiol
17:15101517
7. Krauss B, Schmidt B, Sedlmair M, Flohr T (2010) Evaluation of an
image-based method to calculate monoenergetic images from dual
energy images(ed)^(eds) Abstract Book, European Congress of
Radiology 2010. European Society of Radiology, Vienna, p S201
8. Lee MJ, Kim S, Lee SA et al (2007) Overcoming artifacts from
metallic orthopedic implants at high-field-strength MR imaging
and multi-detector CT. Radiographics 27:791803
9. Thieme SF, Hoegl S, Nikolaou K, et al. Pulmonary ventilation and
perfusion imaging with dual-energy CT. Eur Radiol 20:28822889
10. Uotani K, Watanabe Y, Higashi M et al (2009) Dual-energy CT
head bone and hard plaque removal for quantification of calcified
carotid stenosis: utility and comparison with digital subtraction
angiography. Eur Radiol 19:20602065
11. Watanabe Y, Uotani K, Nakazawa T et al (2009) Dual-energy
direct bone removal CT angiography for evaluation of intracranial
aneurysm or stenosis: comparison with conventional digital
subtraction angiography. Eur Radiol 19:10191024
12. Watzke O, Kalender WA (2004) A pragmatic approach to metal
artifact reduction in CT: merging of metal artifact reduced images.
Eur Radiol 14:849856
13. White LM, Buckwalter KA (2002) Technical considerations: CT
and MR imaging in the postoperative orthopedic patient. Semin
Musculoskelet Radiol 6:517
14. Yu L, Li H, Mueller J et al (2009) Metal artifact reduction from
reformatted projections for hip prostheses in multislice helical
computed tomography: techniques and initial clinical results.
Invest Radiol 44:691696