SHORT COMMUNICATION

Accuracy of MRI technique in measuring tendon

cross-sectional area
C. Coupp e
1,2
, R. B. Svensson
1
, V. Sdring-Elbrnd
3
, P. Hansen
4
, M. Kjr
1
and S. P. Magnusson
1,2
1
Faculty of Health Sciences, Institute of Sports Medicine, Bispebjerg Hospital and Center for Healthy Aging, University of Copenhagen, Copenhagen NV,
Denmark,
2
Department of Physical Therapy, Bispebjerg Hospital, University of Copenhagen, Copenhagen NV, Denmark,
3
Department of Basic Animal and
Veterinary Sciences/Anatomy & Cell Biology, University of Copenhagen, Copenhagen NV, Denmark, and
4
Department of Radiology, Bispebjerg Hospital,
University of Copenhagen, Copenhagen NV, Denmark
Summary
Correspondence
Christian Couppe, Institute of Sports Medicine
Copenhagen, Bispebjerg Hospital, Building 8, 1st.
oor. Bispebjerg Bakke 23 2400 Copenhagen NV,
Denmark
E-mail: ccouppe@gmail.com
Accepted for publication
Received 01 May 2013;
accepted 29 August 2013
Key words
accuracy; cross-sectional area; mould; MRI;
tendon
Magnetic resonance imaging (MRI) has commonly been applied to determine ten-
don cross-sectional area (CSA) and length either to measure structural changes or
to normalize mechanical measurements to stress and strain. The ability to repro-
duce CSA measurements on MRI images has been reported, but the accuracy in
relation to actual tendon dimensions has never been investigated. The purpose of
this study was to compare tendon CSA measured by MRI with that measured in
vitro with the mould casting technique. The knee of a horse was MRI-scanned with
1.5 and 3 tesla, and two examiners measured the patellar tendon CSA. Thereafter,
the patellar tendon of the horse was completely dissected and embedded in an
alginate cast. The CSA of the embedded tendon was measured directly by optical
imaging of the cast impression. 1.5 tesla grey tendon CSA and 3 tesla grey tendon
CSA were 16.5% and 13.2% lower than the mould tendon CSA, respectively. Also,
3 tesla tendon CSA, based on the red-green border on the National Institute of
Health (NIH) colour scale, was lower than the mould tendon CSA by 2.8%. The
typical error between examiners was below 2% for all the measured CSA. The typi-
cal error between examiners was below 2% for all the measured CSA. These data
show that measuring tendon CSA on the grey-scale MRI images is associated with
an underestimation, but by optimizing the measurement using a 3 tesla MRI and
the appropriate NIH colour scale, this underestimation could be reduced to 2.8%
compared with the direct measurements on the mould.
Introduction
Force generated by muscle is transferred to bones via
tendons to produce movement (Elliott, 1965; Alexander
& Bennet-Clark, 1977; Butler et al., 1978; Biewener &
Baudinette, 1995; Magnusson et al., 2008). Tendons have tra-
ditionally been considered relatively inert structures, but
several recent reports demonstrate that human tendons respond
directly to physical activity by increased metabolism (Hannu-
kainen et al., 2005; Kalliokoski et al., 2005; Bojsen-Moller et al.,
2006) and increased collagen synthesis (Langberg et al., 1999,
2001; Christensen et al., 2011). Furthermore, strength training
and habitual loading of tendons appear to be associated with
an increase in tendon size (Arampatzis et al., 2007; Kongsgaard
et al., 2007; Couppe et al., 2008), conrming that the
aforementioned responses to elevated loading result in a net
increase in tendon tissue. Studies that have reported increased
tendon cross-sectional area (CSA) used magnetic resonance
imaging (MRI) (Arampatzis et al., 2007; Kongsgaard et al.,
2007; Couppe et al., 2008; Seynnes et al., 2009). Because of its
high resolution and seemingly good contrast between different
tissues compared with other available imaging modalities
(Roberts et al., 1999; Reeves et al., 2003; de Boer et al., 2007),
it is the preferable assessment tool to detect modest changes in
tendon CSA. The MRI technique is able to distinguish between
different structures, but how well the contrast in the image
corresponds with the actual borders of the tendon remains
unknown, and as a result, the MRI-based measurement may
either under- or overestimate tendon CSA. Previous attempts to
determine tendon CSA in vitro have resulted in damage to the
Clin Physiol Funct Imaging (2014) 34, pp237241 doi: 10.1111/cpf.12086
237 2013 Scandinavian Society of Clinical Physiology and Nuclear Medicine. Published by John Wiley & Sons Ltd 34, 3, 237241
structure, inaccurate values for non-uniform shapes, labour-
intensive techniques or the requirement of rather expensive
equipment (Walker et al., 1964; Lee & Woo, 1988; Race &
Amis, 1996). However, recently, one study reported accurate
measurements of the cross-sectional area of tendon, in vitro,
using a simple non-destructive mould casting technique com-
monly utilized in dental medical practice (Goodship & Birch,
2005). The purpose of this pilot study was to compare tendon
CSA measured by MRI with that measured in vitro using the
mould casting technique.
Materials and methods
Specimen procedure
A horse cadaver knee was obtained from a 15-year-old horse
donor within 30 min postmortem. The horse knee was placed
in a refrigerator at 4C for 24 h, and then, it was stored at
20C. The knee was allowed to thaw at 4C for 24 h before
preparation.
MRI procedure
The horse patellar tendon CSA was determined by MRI 1.5
tesla (General Electric, Sigma Horizon LX 1.5 Tesla,
T1-weighted Spin Echo (SE), Millwaukee, WI, USA) and MRI
3 tesla (Siemens Verio, 3 Tesla, T1-weighted SE, Erlangen,
Germany). Axial scans were performed perpendicular to
the tendon using the following parameters for 1.5 tesla
(Kongsgaard et al., 2007; Couppe et al., 2008): relaxation time
(TR)/echo time (TE) 400/14 ms, eld of view (FOV) 20,
matrix 256 9 256, slice thickness 5.0 mm and spacing
0 mm. For 3 tesla, the following parameters were used: TR/
TE 400/14 ms, FOV 20, matrix 512 9 512, slice thickness
1.0 mm and spacing 0 mm. Tendon CSA was measured along
its length as previously described (Kongsgaard et al., 2007;
Couppe et al., 2008) at three positions, the central slice used
for aiming the MRI and two slices 1 cm to either side of the
central slice (Fig. 1a,d). A pill containing sh oil was attached
in front of the patellar tendon prior to MRI examination in
order to measure MRI and mould examinations in the same
location. The patellar tendon CSA was manually outlined using
the software program Osiris 4.19 (http://www.sim.hcuge.ch/
osiris/). A phantom (plastic tube) containing 1.0% CuSO
4
,
which was also scanned in the MRI, was placed in FOV to
normalize the tendon signal intensity (Carroll et al., 2008).
The colour intensity of each image was adjusted using the
National Institute of Health (NIH) colour scale mode of the
software by setting the intensity of the phantom as the maxi-
mum value. Tendon CSA was measured using the grey scale
for 1.5 tesla and 3 tesla. To optimize the measuring protocol,
the NIH colour scale was also used to measure tendon CSA
for 3 tesla. Here, the red-green border was used as the border
for the tendon (Fig. 1f). The details of the measurement,
including the reliability of the method in our laboratory (Kon-
gsgaard et al., 2007, 2009; Couppe et al., 2008) and elsewhere
(Carroll et al., 2008; Seynnes et al., 2009), has been reported
previously. Two different examiners measured tendon CSA.
(a) (d)
(b) (e)
(c) (f)
Figure 1 Tendon cross-sectional area (CSA)
was measured using the grey scale for 1.5
tesla and 3 tesla (b+e). To optimize the
measuring protocol, the National Institute of
Health (NIH) colour scale was also used to
measure tendon CSA for 3 tesla (f). Here, the
red-green border was used as the border for
the tendon (f).
2013 Scandinavian Society of Clinical Physiology and Nuclear Medicine. Published by John Wiley & Sons Ltd 34, 3, 237241
Accuracy of MRI in measuring tendon size, C. Coupp e et al. 238
Mould procedure
The patella and tibial tubercle were removed from the knee
with a saw. The horse patellar tendon was then embedded in
a mould made of an alginate dental impression material (Blue-
print Cremi, DENTPLY De trey, Germany) as described by
Goodship & Birch (2005) (Fig. 2). Using a glass tube of well-
dened CSA, we found that the mould shrinks by 3.6%. This
was accounted for in the reported moulds CSA.
Mould tendon CSA
The mould was cut in the transverse plane at the three posi-
tions corresponding to the MRI sections, and the tendon was
removed. A small piece of graph paper was placed on top of
the mould sections to aid in focusing and to function as an
internal standard to account for small differences in focus
(Fig. 2). Digital images were taken using microscope
(SMZ1000, Nikon, Tokyo, Japan) to measure the tendon CSA
by image analysis (Scion image, NIH). Two examiners mea-
sured the CSA.
Statistical analysis
Typical error percentage was used to analyse the reliability of
measures between 2 examiners. The typical error percentage is
calculated as SD
diff
=

2
p

Meanoverall 100 and provides a
measure of the relative measurement error (Hopkins, 2000).
Results are reported as means (SD).
Results
The values of the measurements of the moulding technique
and MRI are shown in Table 1. 1.5 tesla grey tendon CSA and
3 tesla grey tendon CSA were 16.5% and 13.2% lower than
the mould tendon CSA, respectively. Also, 3 tesla tendon CSA,
based on the redgreen border on the NIH colour scale, was
lower than the mould tendon CSA by 2.8%. The typical error
between examiners was below 2% for all the measured CSA
(Table 2).
Discussion
To the best of our knowledge, this is the rst pilot study that
has examined how well MRI-measured CSA corresponds to the
actual tendon CSA. We found an underestimation of 2.8% of
the tendon CSA using 3 tesla red-green MRI compared with the
direct measurements from the mould, which indicates that with
this optimized CSA measurement procedure, the contrast of
MRI represents the tissue uniquely. By comparison, measure-
ments on the grey-scale images resulted in a signicantly greater
underestimation of tendon CSA by 16.5% (1.5 tesla grey MRI)
and 13.2% (3 tesla grey MRI). The reason for this underestima-
tion is likely that the high-intensity signal of the surrounding
tissue overlaps the low-intensity signal of the tendon.
While tendon CSA may seem relatively easy to measure,
previous studies have shown that it may be a challenge to
accurately determine tendon dimension due to dehydration,
compression or damage of tendon (Lee & Woo, 1988; Race &
Amis, 1996). Furthermore, inaccurate values for non-uniform
shapes, labour-intensive techniques and requirement of expen-
sive equipment are also difculties that have been associated
with measuring the tendon in vitro (Walker et al., 1964; Lee &
Woo, 1988; Race & Amis, 1996). In the present study, we
used a simple non-destructive mould casting technique
that have reported accurate measurements of tendon size
(Goodship & Birch, 2005). However, using a glass tube with
Figure 2 The tendon was dissected out and embedded in a mould-
ing material. The mould was then cut into sections equivalent to MRI
scans and cross-sectional areas compared.
Table 1 The values of the mould and MRI measurements.
CSA mm
2
Difference from
mould (mm
2
)
Mould 88.4 (1.1)
1.5 tesla Grey 73.6 (0.7) 14.6 [16.5%]
3 tesla Grey 76.5 (0.8) 11.7 [13.2%]
3 tesla Red-green 85.8 (0.6) 2.5 [2.8%]
Values are mean (SD) between two examiners for mould and for MRI.
Table 2 The values of the mould and MRI measurements between
the two examiners.
Examiner 1
(mm
2
)
Examiner 2
(mm
2
)
Inter
examiner
typical error
(mm
2
)
Inter
examiner
typical
error %
Mould 88.9 88.0 0.39 0.44
1.5 tesla
Grey
75.2 72.0 1.30 1.76
3 tesla Grey 75.9 77.2 0.76 1.00
3 tesla
Red-green
84.1 87.5 1.35 1.58
2013 Scandinavian Society of Clinical Physiology and Nuclear Medicine. Published by John Wiley & Sons Ltd 34, 3, 237241
Accuracy of MRI in measuring tendon size, C. Couppe et al. 239
a well-dened CSA, we found that the mould consistently
shrunk by 3.6%, which was accounted for in the reported ten-
don CSA values based on mould measurement. This shrinking
of the mould has not been reported before. Although the
mould casting technique may not necessarily be a gold stan-
dard to use as reference because shrinkage was detected, it
may still be a useful technique to validate other tendon struc-
tures in future studies.
Therefore, only the mould was used for determining the ten-
don CSA by the optical imaging, because the corresponding
tendon pieces were affected by aring and dehydration
(Goodship & Birch, 2005). Therefore, we assumed that the
mould would best represent the tendon CSA compared with
MRI scanning. As shown in Fig. 1f, the contrast of the sur-
rounding tissue is greater on 3 tesla MRI compared with 1.5
tesla (Fig. 1c) (the transition from blue to red is narrower).
The higher contrast enables an examiner to better trace the out-
line of the tendon and thereby reduce the error in measuring
tendon CSA, in vivo. The typical percentage error of repeated
measures of site-specic tendon CSA of MRI images for one
examiner has previously been reported to be 2.02.5%
(Couppe et al., 2008), and in the present study, the typical error
percentage of measures between two examiners was below 2%.
These data highlight that a well-dened protocol must be used
to ensure low variability, especially if the aim is to detect those
modest changes (56%) that can be expected to accompany
short-term resistance training interventions (Arampatzis et al.,
2007; Kongsgaard et al., 2007; Seynnes et al., 2009).
It should be noted that more inexpensive imaging tech-
niques, such as ultrasound imaging, have not been able to
detect changes in tendon CSA in prospective studies (Roberts
et al., 1999; Reeves et al., 2003; de Boer et al., 2007; Malliaras
et al., 2013). The reason for this is maybe due to lower con-
trast between different tissues, which becomes inherently
associated with measurement inaccuracy. In addition, it is
necessary to standardize the image contrast (here using a
CuSO
4
phantom) to accurately dene the tendon outline,
which is not possible with ultrasound. Taken together, 3 tesla
red-green MRI provides seemingly good contrast between dif-
ferent tissues compared with other available imaging modali-
ties and is therefore the preferable assessment tool to detect
modest changes in tendon CSA.
For this experiment, a tendon of roughly the same size as
that of humans was used and it had to be freshly harvested to
ensure normal hydration and thus contrast in the MRI. For
this purpose, horse tendon is suitable, but because of the low
availability, it precludes a greater sample size. To reduce the
risk of systematic error caused by the reduced contrast, we
used two examiners.
In conclusion, these data show that measuring CSA on the
grey-scale MRI images is associated with an underestimation,
but by optimizing the measurement using a 3 tesla MRI and
the appropriate NIH colour scale, this underestimation could
be reduced to 2.8% compared with the direct measurements
on the mould.
Acknowledgments
We thank Professor Dr.med. Preben Dybdahl-Thomsen,
Department of Basic Animal and Veterinary Sciences/Anatomy
& Cell Biology, University of Copenhagen, Denmark, and MRI
technician, Berit Sorensen, Department of Radiology, Bispeb-
jerg Hospital, University of Copenhagen, Denmark, for their
exemplary technical assistance. The Danish Medical Research
Council and the Danish Rheumatism Association supported
this work.
Conict of interest
The authors have no conict of interests.
References
Alexander RM, Bennet-Clark HC. Storage of
elastic strain energy in muscle and other tis-
sues. Nature, (1977); 265: 114117.
Arampatzis A, Karamanidis K, Albracht K.
Adaptational responses of the human Achil-
les tendon by modulation of the applied
cyclic strain magnitude. J Exp Biol (2007);
210: 27432753.
Biewener A, Baudinette R. In vivo muscle force
and elastic energy storage during steady-
speed hopping of tammar wallabies (Macropus
eugenii). J Exp Biol (1995); 198: 18291841.
de Boer MD, Maganaris CN, Seynnes OR,
Rennie MJ, Narici MV. Time course of mus-
cular, neural and tendinous adaptations to
23 day unilateral lower-limb suspension in
young men. J Physiol (2007); 583:
10791091.
Bojsen-Moller J, Kalliokoski KK, Seppanen M,
Kjaer M, Magnusson SP. Low-intensity ten-
sile loading increases intratendinous glucose
uptake in the Achilles tendon. J Appl Physiol
(2006); 101: 196201.
Butler DL, Grood ES, Noyes FR, Zernicke RF.
Biomechanics of ligaments and tendons.
Exerc Sport Sci Rev (1978); 6: 125181.
Carroll CC, Dickinson JM, Haus JM, Lee GA,
Hollon CJ, Aagaard P, Magnusson SP,
Trappe TA. Inuence of aging on the in
vivo properties of human patellar tendon.
J Appl Physiol (2008); 105: 19071915.
Christensen B, Dandanell S, Kjaer M,
Langberg H. Effect of anti-inammatory
medication on the running-induced rise in
patella tendon collagen synthesis in humans.
J Appl Physiol (2011); 110: 137141.
Couppe C, Kongsgaard M, Aagaard P, Hansen
P, Bojsen-Moller J, Kjaer M, Magnusson SP.
Habitual loading results in tendon hypertro-
phy and increased stiffness of the human
patellar tendon. J Appl Physiol (2008); 105:
805810.
Elliott DH. Structure and function of mamma-
lian tendon. Biol Rev Camb Philos Soc (1965);
40: 392421.
Goodship AE, Birch HL. Cross sectional area
measurement of tendon and ligament in
vitro: a simple, rapid, non-destructive tech-
nique. J Biomech (2005); 38: 605608.
Hannukainen J, Kalliokoski KK, Nuutila P,
Fujimoto T, Kemppainen J, Viljanen T,
Laaksonen MS, Parkkola R, Knuuti J, Kjaer
M. In vivo measurements of glucose uptake
in human Achilles tendon during different
2013 Scandinavian Society of Clinical Physiology and Nuclear Medicine. Published by John Wiley & Sons Ltd 34, 3, 237241
Accuracy of MRI in measuring tendon size, C. Coupp e et al. 240
exercise intensities. Int J Sports Med (2005);
26: 727731.
Hopkins WG. Measures of reliability in sports
medicine and science. Sports Med (2000); 30:
115.
Kalliokoski KK, Langberg H, Ryberg AK,
Scheede-Bergdahl C, Doessing S, Kjaer A,
Boushel R, Kjaer M. The effect of dynamic
knee-extension exercise on patellar tendon
and quadriceps femoris muscle glucose
uptake in humans studied by positron emis-
sion tomography. J Appl Physiol (2005); 99:
11891192.
Kongsgaard M, Reitelseder S, Pedersen TG,
Holm L, Aagaard P, Kjaer M, Magnusson SP.
Region specic patellar tendon hypertrophy
in humans following resistance training. Acta
Physiol (Oxf) (2007); 191: 111121.
Kongsgaard M, Kovanen V, Aagaard P,
Doessing S, Hansen P, Laursen AH, Kaldau
NC, Kjaer M, Magnusson SP. Corticosteroid
injections, eccentric decline squat training
and heavy slow resistance training in
patellar tendinopathy. Scand J Med Sci Sports
(2009); 19: 790802.
Langberg H, Skovgaard D, Petersen LJ, Bulow
J, Kjaer M. Type I collagen synthesis and deg-
radation in peritendinous tissue after exercise
determined by microdialysis in humans.
J Physiol (1999); 521(Pt 1): 299306.
Langberg H, Rosendal L, Kjaer M. Training-
induced changes in peritendinous type I col-
lagen turnover determined by microdialysis
in humans. J Physiol (2001); 534: 297302.
Lee TQ, Woo SL. A new method for determin-
ing cross-sectional shape and area of soft tis-
sues. J Biomech Eng (1988); 110: 110114.
Magnusson SP, Narici MV, Maganaris CN,
Kjaer M. Human tendon behaviour and adap-
tation, in vivo. J Physiol (2008); 586: 7181.
Malliaras P, Kamal B, Nowell A, Farley T,
Dhamu H, Simpson V, Morrissey D,
Langberg H, Maffulli N, Reeves ND. Patellar
tendon adaptation in relation to load-
intensity and contraction type. J Biomech,
(2013); 46: 18931899.
Race A, Amis AA. Cross-sectional area
measurement of soft tissue. A new casting
method. J Biomech (1996); 29: 12071212.
Reeves ND, Maganaris CN, Narici MV.
Effect of strength training on human
patella tendon mechanical properties of
older individuals. J Physiol (2003); 548:
971981.
Roberts CS, King DH, Goldsmith LJ. A statisti-
cal analysis of the accuracy of sonography
of the patellar tendon. Arthroscopy (1999);
15: 388391.
Seynnes OR, Erskine RM, Maganaris CN,
Longo S, Simoneau EM, Grosset JF, Narici
MV. Training-induced changes in structural
and mechanical properties of the patellar
tendon are related to muscle hypertrophy,
but not to strength gains. J Appl Physiol,
(2009); 107: 523530.
Walker LB, Harris EH, Benedict JV. Stress-
strain in human cadaveric plantaris tendon:
a preliminary study. Med Electron Biol Eng,
(1964); 2: 3138.
2013 Scandinavian Society of Clinical Physiology and Nuclear Medicine. Published by John Wiley & Sons Ltd 34, 3, 237241
Accuracy of MRI in measuring tendon size, C. Couppe et al. 241