Musculoskeletal Imaging Original Research

Morvan et al.
Effect of Femur Position on CT and Stereoradiography Mea-
surements of Femoral Torsion

Musculoskeletal Imaging
Original Research
Downloaded from www.ajronline.org by 114.124.241.227 on 11/09/17 from IP address 114.124.241.227. Copyright ARRS. For personal use only; all rights reserved

Femoral Torsion: Impact of

Femur Position on CT and
Stereoradiography Measurements
Grard Morvan1 OBJECTIVE. The purpose of this study is to evaluate stereoradiographic measurements
Henri Guerini of femoral torsion with different femoral positions, in comparison with CT measurements,
Grgory Carr with use of the current standard axial-slice technique. We hypothesize that CT measurements
Valrie Vuillemin vary with femoral spatial positioning because of the resulting projection onto the CT plane,
whereas stereoradiographic measurements, which are derived from a 3D reconstruction of the
Morvan G, Guerini H, Carr G, Vuillemin V femur, remain constant.
MATERIALS AND METHODS. Both in vitro and in vivo studies were conducted. CT
and stereoradiographic examinations were performed using 30 dry femurs in the following
six femoral positions: neutral position (with the femoral mechanical axis aligned with the lon-
gitudinal axis of the CT scanner or stereoradiography system), 10 of abduction, 10 of ad-
duction, 5 of flexion, 10 of flexion, and 5 of extension. The impact of femoral position on
torsion measurement was assessed using paired t tests. In addition, 18 patients (mean [SD]
age, 42.3 19.9 years) who underwent both CT and stereoradiography examinations were ret-
rospectively assessed. The correlation between femoral positioning and torsion measurement
was determined using the Pearson correlation coefficient.
RESULTS. Flexion and extension statistically significantly affected CT measurement of
femoral torsion (p< 0.01) but not stereoradiography measurement (p> 0.21). A strong cor-
relation existed between hip flexion and the difference between femoral torsion measured by
CT and stereoradiography (r= 0.80).
CONCLUSION. The accuracy of femoral torsion determined by axial CT depends on
the position of the femur. Hip flexion significantly reduced the femoral torsion angle mea-
sured by CT. Conversely, the accuracy of stereoradiography was independent of femur posi-
tioning. Thus, stereoradiography is preferable to CT for accurate measurement of femoral tor-
sion, while it also substantially reduces the radiation dose.

bnormal femoral torsion corre- theory, 3D models can be reconstructed from

Keywords: accuracy, CT, femoral torsion, patient

positioning, stereoradiography
DOI:10.2214/AJR.16.16638
Received April 25, 2016; accepted after revision
December 23, 2016.
1
All authors: Centre dImagerie Mdicale Lonard de
Vinci, 43 rue Cortambert, 75116 Paris, France. Address
correspondence to G. Morvan (gerard.morvan@yahoo.fr).
WEB
This is a web exclusive article.
AJR 2017; 209:W93W99
0361803X/17/2092W93
American Roentgen Ray Society
A lates with several disease pro-
cesses, including hip and knee
osteoarthritis, slipped capital
femoral epiphysis, trochlear dysplasia with
resulting patellar instability, and toe-in or
toe-out gait patterns [1, 2]. Limb disorders
associated with femoral torsion may also be
posttraumatic or may be caused by other
pathologic conditions, such as cerebral palsy.
To deal with these rotational disorders of the
lower limb, an accurate and reproducible
way of measuring femoral torsion is essential
to guide diagnosis as well as treatment [3, 4].
Accurate measurement of the 3D rotation-
al anatomy of the femur with the use of a 2D
projection is a challenge. Therefore, numer-
ous CT or MRI methods for measuring tor-
sion have been described [58]. Although, in
CT and MRI, the current standard practice
is to superimpose CT slices. In clinical prac-
tice, however, the findings of physical ex-
amination have been inconsistent with axi-
al-based CT and MRI assessment of femoral
torsion, and large variations in measurement
have been reported [9, 10]. Some studies have
reported that unadjusted CT-assessed femo-
ral torsion may be affected by femur posi-
tioning and that identification of the femoral
neck axis is unreliable [11, 12]. Thus, there
still is no consensus on the best imaging mo-
dality or technique.
Femoral torsion, also known as femoral
version, is described in the literature as the
angle between the tangent line at the pos-
terior aspect of the condyles and the line
that passes through the centers of the femo-

AJR:209, August 2017 W93

 Morvan et al.
Downloaded from www.ajronline.org by 114.124.241.227 on 11/09/17 from IP address 114.124.241.227. Copyright ARRS. For personal use only; all rights reserved

A B C
Fig. 1Measurement of femoral torsion performed using CT in dry femur.
A, Superimposed axial CT slices of femoral head and femoral neck show femoral neck axis (green line). Circle denotes femoral head.
B, Axial CT slice through most posterior part of femoral condyles shows bicondylar femoral axis (yellow line).
C, Femoral torsion (i.e., angle between axes [green line and yellow line]) measured using CT scan protocol. Circle denotes femoral head.

ral head and femoral neck [13, 14] (Fig. 1).
This is meant to be projected onto the trans-
verse plane of the patient (i.e., perpendicular
to the mechanical axis of the femur) but, in
standard practice, is projected onto the axial
plane on CT or MRI. In cases of pain, swell-
ing, or flexion contracture in which the fe-
mur is not aligned with the CT or MRI axis,
the anatomic and CT or MRI planes are not
the same [11].
The main difficulty with femoral torsion
measurement is determining the axis of the
femoral neck. Several methods rely on axi-
al slices to determine the orientation of the
femoral neck; however, the choice of slices
can strongly influence the measurement. In-
accuracies are mostly due to the difficulty of
drawing a precise line along the middle of
the femoral neck on a CT image, even though
proper measurement is necessary, for exam-
ple, to perform osteotomies to correct post-
traumatic rotational deformities [12, 15].
Some questions arise in this context. Is
the CT measurement of femoral torsion reli-
able? Does positioning of the femur have an
influence on the measurement of its torsion?
Could EOS stereoradiographic imaging
(EOS Imaging) be an alternative for assess-
ing this parameter accurately and indepen-
dent of lower limb positioning?
Several studies have explored related
questions [1621]; however, to our knowl-
edge, no study has addressed femur posi-
tioning. Although previous studies appeared
to address a similar question, important dif-
ferences existed between them. Two of the
studies [16, 17] artificially created a range
of femoral torsion values by cutting and ro-
tating the femur mid diaphysis; they found
that stereoradiographic measurements had
strong reliability, compared with 3D recon-
structions from CT (i.e., ground truth), but
they did not evaluate different flexion angles
or axial CT. Four other studies examined
measurements of femoral torsion in adults
[18, 19] and children [20, 21], finding com-
parable results between stereoradiography
and CT on average, but they likewise did not
explore the effects of hip flexion.
The purpose of the present study, there-
fore, is to evaluate the accuracy of both
CT and stereoradiography measurements
of femoral torsion with different femur po-
sitions. We hypothesize that axial CT mea-
surements of torsion will be related to spatial
positioning of the femur, whereas stereora-
diographic measurements of torsion will be
unchanged regardless of the spatial position-
ing of the femur.
Materials and Methods
Patients were scanned using two different mo-
dalities, as performed in daily practice at Centre
dImagerie Mdicale Lonard de Vinci, with the
use of a 16-MDCT scanner (BrightSpeed CT99,
GE Healthcare) with a detector size of 20 mm and
a low-dose stereoradiographic device that com-
bines a particle detector with an innovative linear
scanning technique (EOS). When performed with
the patient in the standing position, stereoradiogra-
phy allows for simultaneous acquisition of whole-
body, frontal, and lateral images, with the use of
slot-scanning technology, followed by 3D model-
ing of the lower limb and deformities in the weight-
bearing position [22].
Two studies were performed. An in vitro study
was performed to test our hypothesis under con-
trolled conditions, and an in vivo study was con-
ducted to investigate the impact of femur position-
ing in a standard patient population. In each case,
images from stereoradiography and CT were ac-
quired under the same conditions.
In Vitro Evaluation
Thirty dry femurs were selected and imaged
using both CT and stereoradiography. Each femur
was studied in six positions: neutral position, 10
of abduction, 10 of adduction, 5 of flexion, 10 of
flexion, and 5 of extension. The six positions were
chosen because they represent the positioning of
patients in our practice (e.g., to evaluate hip flex-
ion secondary to hip or knee osteoarthritis). The
neutral position should provide the true femoral
torsion value.
Neutral positioning in CT examinations was
established by using the laser line to align the
femoral mechanical axis, which was defined as
the line passing through the center of the femoral
head and the intercondylar notch, with the longi-
tudinal axis of the CT scanner platform. Neutral
positioning in stereoradiography was established
by ensuring (with use of a vertical line on the pre-
view image) that the femoral mechanical axis was
strictly vertical in both the frontal and sagittal
planes (i.e., that it was aligned with the direction
of the movement of the x-ray sources). To produce
the inclinations, wooden wedges were used with
a radiolucent foam placed between the femur and
the wood (Fig. 2).
For the CT scans, working from frontal and sag-
ittal scout views, two acquisition zones were de-
fined: the proximal femur and the distal femur. For
each acquisition zone, 2.5-mm-thick slices with a
pitch of 0.531:1 were acquired in compliance with
the protocol set forth by the manufacturer of the
CT scanner. The acquisition parameters were the
same as those used in our routine clinical prac-
tice: tube voltage, 120 kV; tube current, 200 mA;
matrix, 512 512; and a bone convolution kernel.
The femurs were fixed in the different positions,
attached to the wedges. For stereoradiographic ac-
quisitions, femurs were placed in the EOS system
in the six different positions previously described,
with the wooden wedges used to produce the in-

W94 AJR:209, August 2017

 Effect of Femur Position on CT and Stereoradiography Measurements of Femoral Torsion

clinations. A neutral position was obtained when
the femoral mechanical axis was vertical, aligned
with the axis of movement of the x-ray sources.
Biplanar acquisitions were performed using the
standard low-dose protocol provided by the manu-
facturer for lower limb indications (for anteropos-
Downloaded from www.ajronline.org by 114.124.241.227 on 11/09/17 from IP address 114.124.241.227. Copyright ARRS. For personal use only; all rights reserved
er limbs for which knee arthroplasty (two femurs) or
hip arthroplasty (four femurs) had been performed
were excluded from the analysis. A total of 30 fe-
murs were then used to assess femoral torsion.
For CT acquisitions, 1.25-mm-thick slices were
acquired. Working from frontal and sagittal scout
enced senior radiologist, to confirm that the posi-
tioning of anatomic landmarks was respected as
defined by CT and stereoradiography protocols.
For the in vivo study, the objective was to com-
pare the torsion for each patient, whereby torsion
is a parameter inherent to the patient that should

terior acquisitions, a tube voltage of 85 kV and a
tube current of 200 mA were used; for lateral ac-
quisitions, a tube voltage of 110 kV and a tube cur-
rent of 320 mA were used).
CT and stereoradiography measurements were
directly comparable because the same anatomic
landmarks were used for calculations of femoral
torsion (i.e., the femoral neck axis and the poste-
rior bicondylar axis). In the neutral position, CT
and stereoradiography measurements are expect-
ed to be equivalent because the calculations are
performed in the same plane.
Clinical Study
This study was approved by the institutional re-
view board at Centre dImagerie Mdicale Lo-
nard de Vinci. Informed consent from patients
was not required because patient care was not
modified by the study design.
From June 2012 to March 2014, after being re-
ferred by orthopedic surgeons for lower limb tor-
sional measurements, 18 patients (13 female pa-
tients and five male patients; mean age, 42.3 19.9
(SD) years; range, 1377 years) underwent both CT
and stereoradiographic imaging with a mean inter-
val of 10 days occurring between the two examina-
tions. The patients were identified through a search
of the database at our institution and were retro-
spectively included in the present study. All images
were anonymized before undergoing analysis. Low-
views, two acquisition zones (hip and knee) were
defined, in accordance with the protocol at our
institution. The distance between the ankles was
set the same as the distance between the femoral
heads, to keep the lower limbs aligned with the
z-axis (i.e., the longitudinal axis) of the CT scan-
ner. Acquisition parameters were the same as
those used in the in vitro study.
For stereoradiographic acquisitions, patients
stood upright within the stereoradiography sys-
tem in a weight-bearing position. They were asked
to shift their feet slightly (approximately 5 cm),
to allow radiographers to distinguish the femoral
condyles in the sagittal view and to perform 3D
modeling of the lower limbs accurately, as per the
recommendations of the manufacturer of the ste-
reoradiography system. Biplanar (frontal and sag-
ittal) acquisition of images of the lower limbs, from
the top of the pelvis to the feet, was performed for
each patient, with use of the same acquisition pro-
tocol that was used for the in vitro study.
CT and Stereoradiography Measurements
All measurements were performed indepen-
dently, without reference to earlier values, and in
a random order. Two experienced radiographers
(with 4 and 5 years of experience) performed
measurements using both modalities and for both
studies (of dry bones and patients). All measure-
ments were verified and validated by an experi-
not be affected by the flexion angle. Therefore, al-
though the flexion angle was necessarily different
between the two cases, this was exactly the goal of
the study. Given that the in vitro study showed the
relative independence of the stereoradiography
torsion values from the flexion angle, these were
assumed to be invariant. By contrast, the in vivo
data were analyzed according to the flexion angle
within the CT image, because this was the focus of
our investigation. The specific measurement pro-
tocols used for each modality are described in the
following subsection.
CT Measurement
Measurement of femoral torsion was conduct-
ed in accordance with the method described by
Reikeras et al. [13]: the neck axis was defined
from two superimposed slices as the line pass-
ing from the center of the femoral head (i.e., the
slice where the femoral head is the largest) to the
midpoint of the base of the neck (where the ante-
rior and posterior cortical walls are parallel). The
femoral bicondylar axis was determined to be the
line tangent to the femoral condyles, on a slice ac-
quired at the most convex part of the femoral con-
dyles, which is often distinguished by the Roman
arch shape of the intercondylar notch. The angle
of femoral torsion was measured between the neck
axis and the bicondylar axis (Fig. 1). All mea-
surements were performed independently, with-
out reference to earlier values, and in a random
order. Torsion was considered positive for ante-
version (neck axis oriented forward, with regard
to the bicondylar axis) and negative for retrover-

Fig. 3Femoral torsion calculated by dedicated

A B software (sterEOS, EOS Imaging) as angle between
Fig. 2Use of wooden wedges to produce different inclinations for in vitro study with CT and stereoradiography. green and yellow axes (femoral neck axis and
A and B, Photographs show use of wooden wedge to produce 5 of flexion (A) and 10 of flexion (B). For both CT tangent to posterior femoral condyles) projected
and stereoradiography, radiolucent foam block (seen in profile) was used between femur and wood to improve onto orthogonal plane to femoral mechanical axis.
image contrast. Vertical line was drawn on foam to simplify correct positioning. (Adapted with permission from EOS Imaging)

AJR:209, August 2017 W95

 Morvan et al.

and the five other positions, in both CT and stereo-

25
radiography. Statistical significance was set at p<
0.01 because of the multiple comparisons. Interop-
20 erator agreement was assessed using the intraclass
correlation coefficient (ICC) for stereoradiography
and CT measurements in vitro and in vivo [24]. For
*
Downloaded from www.ajronline.org by 114.124.241.227 on 11/09/17 from IP address 114.124.241.227. Copyright ARRS. For personal use only; all rights reserved

15
Femoral Torsion ()

the in vivo study, the Pearson correlation coefficient

(r) between CT and stereoradiography measure-
10
* ments was calculated to determine the impact of
hip flexion and extension on the difference between

5
* stereoradiography and CT measurements.

Results
0 In Vitro Evaluation
As expected, in the neutral position, no
statistically significant difference existed in
5
Neutral 5 of 10 of 5 of 10 of 10 of mean femoral torsion between stereoradiog-
Position Flexion Flexion Extension Abduction Adduction raphy and CT (for stereoradiography, 12.1
6.4; for CT, 11.5 8.3; p= 0.30).
Fig. 4Bar graph of mean (SD) femoral torsion values measured using stereoradiography (blue bars) and CT The mean ICC for interoperator agree-
(red bars) for each of six positions, with values for neutral position (green bars) shown for reference. Error bars ment of femoral torsion was 0.95 or greater
denote SDs. Blue and red horizontal lines denote mean femoral torsion from six positions measured using CT for stereoradiography, compared with 0.90
and stereoradiography, respectively. Asterisks denote statistically significant differences (in all three cases,
p< 0.01) based on paired t tests comparing neutral position with each of five other positions.
or greater for CT (Table 1). Given this high
level of agreement, the results for both read-
sion (neck axis oriented backward, with regard to software then automatically calculated femoral ers have been combined for greater clarity.
the bicondylar axis). Hip flexion was assessed in torsion between the femoral neck axis and the For the stereoradiography measurements
the sagittal scout view relative to the longitudinal posterior bicondylar axis, projected in the trans- in the other five positions, no statistically
axis, by drawing a line between the center of the verse plane of the femur (defined as orthogonal to significant differences were noted between
femoral head and the femoral notch. Femur ab- the femoral mechanical axis) (Fig. 3). Flexion and the neutral position (i.e., the femoral me-
duction or adduction was assessed in the frontal extension were not directly measured on stereora- chanical axis aligned with the vertical axis
scout view, by measuring the angle between the diography because there was negligible effect be- of the stereoradiography system) and any of
femoral mechanical axis and the longitudinal axis. cause of the 3D reconstruction. the five positions (p= 0.210.99) (Fig. 4).
The measurements were performed on the PACS For the CT measurements, no statistical-
(Centricity, GE Healthcare) used at our institution. Data Analysis ly significant difference was noted for ab-
Statistical software (MedCalc, version 12, duction and adduction, compared with the
Stereoradiography Measurement MedCalc Software) was used for statistical analy- neutral position (i.e., the femoral mechanical
On the basis of the biplanar acquisition, an oper- sis. A descriptive statistical analysis was first per- axis aligned with the longitudinal axis of CT
ator manually identified some anatomic landmarks formed for the complete datasets. For the in vitro scan) (p= 0.93 and 0.04, respectively). How-
(e.g., the femoral head, femoral condyles, and fem- evaluation, a repeated-measures ANOVA was per- ever, a statistically significant difference was
oral shaft at its distal third) used to determine the formed across the different femur positions; given noted between the flexion and extension po-
bony envelope of the femur in 3D, through the use significance, post hoc t tests for paired samples were sitions and the neutral position (p< 0.01), re-
of dedicated software (SterEOS, EOS imaging), as used to determine whether there was a statistically vealing a dependence on femoral spatial po-
described later in this subsection. Working from significant difference between the neutral position sition (Fig. 4).
this 3D model, the software automatically calcu-
TABLE 1: Interoperator Agreement of Femoral Torsion Measurement on Dry
lated a set of clinical parameters of the lower limb,
Bones in Each Position
including femoral torsion [23].
The operator first manually identified on CT Stereoradiography
both radiographs the femoral head, the condyle,
Rotation Angle ICC 95% CI ICC 95% CI
and the diaphysis of the femur at its distal third.
Second, an initial solution of the 3D bony enve- Neutral position 0.94 0.880.97 0.97 0.930.98
lope was proposed by the software on the basis 5 of flexion 0.92 0.840.96 0.95 0.910.98
of statistical shape modeling of expected femo- 10 of flexion 0.90 0.810.95 0.95 0.900.98
ral shapes, by focusing on the key landmarks and
5 of extension 0.92 0.840.96 0.96 0.930.98
resulting clinical parameters. This first 3D bony
envelope was retroprojected on both radiographs, 10 of abduction 0.90 0.800.95 0.96 0.920.98
and a parametric model of the femur was created. 10 of adduction 0.92 0.840.96 0.97 0.940.99
Finally, a fine adjustment of anatomic landmarks NoteMinimum intraclass correlation coefficient (ICC) for CT measurement of femoral torsion was 0.90,
and contours was performed by the operator. The whereas that for stereoradiographic measurement was 0.95.

W96 AJR:209, August 2017

 Effect of Femur Position on CT and Stereoradiography Measurements of Femoral Torsion

8 neutral position, leading to the similar mean

results between modalities (Fig. 6).
6
In the clinical study, the mean ICC for in-
teroperator agreement of femoral torsion was
Femoral Torsion vs Neutral Position ()

0.95 (95% CI, 0.900.98) for stereoradiogra-

4
phy and 0.85 (95% CI, 0.700.92) for CT.
Downloaded from www.ajronline.org by 114.124.241.227 on 11/09/17 from IP address 114.124.241.227. Copyright ARRS. For personal use only; all rights reserved

2
Discussion
The present study investigated the effect
0 of femoral position on the measurement of
femoral torsion, finding that CT measure-
2 ments depended significantly on femur spa-
tial position (in flexion and extension), where-
4 as stereoradiography measurements did not.
Both modalities showed good reliability
6 (ICC for CT,> 0.90; ICC for stereoradiogra-
phy,>0.95), regardless of femoral positioning.
8 CT currently is considered the best meth-
5 of 10 of 5 of 10 of 10 of od with which to determine torsion or rota-
Flexion Flexion Extension Abduction Adduction
tional malalignment of the femur. Although
some authors have questioned its use, CT
Fig. 5Difference in femoral torsion measurement from neutral position. Bar graph shows mean (SD) femoral measurements are usually considered ac-
torsion, as measured with stereoradiography (blue bars) and CT (red bars). Error bars denote SDs.
curate in the literature and in daily prac-
ticethat is, they are considered the refer-
Differences between measurements for findings from the in vitro study. Maximum ence standard. A high accuracy of femoral
the neutral position and any of the other five flexion and extension were 6 and 8, respec- torsion measurements is necessary to assess
positions were, on average, less than 0.3 tively, among the patients studied, with max- and correct posttraumatic or idiopathic rota-
with the stereoradiography system, where- imum torsional differences between stereo- tional deformities.
as with CT the difference between measure- radiography and CT ranging from 9 to The present study reveals, however, that
ments for the neutral position and 10 of 12.5 (i.e., similar to a comparison between the accuracy of femoral torsion determined
flexion was a mean of 5.2 1.7 and that CT performed with a neutral femoral posi- by CT is questionable. Our findings show
between measurements for the neutral posi- tion and CT performed with a flexed or ex- that the inaccuracy of this modality cannot
tion and 5 of extension was 3.3 2.4 (Fig. tended femoral position), with a mean dif- be explained only by the difficulty of identi-
5). The mean difference from the neutral po- ference of only 0.9 from the value for the fication of the femoral neck axis, as previous-
sition with hip abduction and adduction was
less than 0.9 (Fig. 5). SDs for CT were ap-
15
proximately twice as high as those for stereo-
radiography (2 vs 1).
Difference in Femoral Torsion Measured

10
by CT vs Stereoradiography ()

Clinical Study
The mean femoral torsion measured 5
was 14.6 9.1 with stereoradiography
and 15.5 8.5 with CT. No statistical- 0
ly significant difference was found between
these two modalities (p= 0.39). 5
Despite the similar mean values noted,
there was a statistically significant negative
10
correlation between the amount of hip flex-
ion and the difference between femoral tor-
sion measured with CT and stereoradiogra- 15
10 8 6 4 2 0 2 4 6 8
phy (r = 0.80, p < 0.01) (Fig. 6). In other Extension () Flexion ()
words, the femoral torsion measured with
CT was significantly less than that measured Fig. 6Scatterplot showing impact of flexion and extension on difference between femoral torsion measured
with stereoradiography at higher flexion val- using CT and stereoradiography. Because stereoradiography was shown to be largely unaffected by femoral
ues. Conversely, at higher extension values, flexion and extension (Fig. 5), difference between CT and stereoradiography shows error in CT measurements
the femoral torsion measured with CT was resulting from out-of-plane movement (i.e., flexion and extension) of femur, ranging from 13 to 9 in femoral
torsion, with statistically significant correlation noted (r= 0.80). Flexion and extension values (blue dots) are
significantly greater than that measured with derived from CT sagittal scout scan as angle between longitudinal axis and line connecting center of femoral
stereoradiography. This corresponds with head with femoral notch. Diagonal line denotes linear correlation.

AJR:209, August 2017 W97

 Morvan et al.

ly reported elsewhere [5]. Although the inter-
operator agreement for CT measurements of
femoral torsion was slightly smaller than that
for stereoradiographic measurements, it still
remained very good for the six different fem-
oral positions (ICC, > 0.90), which means
Downloaded from www.ajronline.org by 114.124.241.227 on 11/09/17 from IP address 114.124.241.227. Copyright ARRS. For personal use only; all rights reserved
Given that a left-right difference in torsion
greater than 10 indicates a possible deform-
ity, this magnitude of error in femoral torsion
measurement could have a substantial impact
on diagnosis and treatment.
The dependence of measurement of femo-
oral torsion are projected into the CT scan
reference frame so that modifying the femur
position changes its projection. By contrast,
femoral torsion in the stereoradiography sys-
tem is measured by projecting the anatom-
ic structures onto the anatomic frame of the

that the femoral neck can be reliably iden- ral torsion on hip flexion and extension can femur, which is independent of its position.
tified regardless of femur positioning. Nev- be explained by simple geometric projection These geometric considerations explain that,
ertheless, by comparing CT measurements in which the forward or backward rotation of on axial CT, a significant statistical differ-
in two different positions (10 of flexion and the femur changes the projected line of the ence was noted in torsion measurement be-
5 of extension), we found a mean difference femoral neck (Fig. 7). In axial CT analysis, tween the flexion and extension positions,
of 8.5 and a maximum difference of 12.6. the anatomic structures used to assess fem- compared with the neutral position. A previ-
ous study found that bias is associated with
the axial CT techniques in simulated femur
flexion and extension positions, confirming
our findings [9]. Also, although CT has the
capability to create 3D reconstructions, this is
not done in standard clinical practice. Stereo-
radiographic measurement of femoral torsion
was not affected by femur positioning (Figs. 4
and 5) because of the 3D reconstructions: no
statistically significant difference was noted
between the multiple spatial positions, when
compared with the neutral position.
A B In the clinical study, when we compared
only the mean femoral torsion between the
CT and stereoradiographic measurements,
the mean difference of 0.9 (range, 9 to
12.5) was not statistically significant (p=
0.34), thereby confirming the results of previ-
ous studies that likewise did not find a statis-
tically significant difference between the two
modalities [1722]. One study, for example,
reported a mean difference of 1 (range, 10
to 8) [21], whereas another reported a mean
difference of 2.1 (range, 11 to 17.3) [20].
However, the investigators in the latter study
did not investigate the variabilities between
CT and stereoradiographic measurements,
C D nor did they describe the effect of lower limb
positioning on femoral torsion measurements,
as was done in the present study. Interopera-
tor agreement regarding femoral torsion mea-
surement in the clinical study was very good
for both modalities (ICC for CT, 0.85; ICC for
stereoradiography, 0.95). To our knowledge,
only one previous study [25] measured lower
reproducibility in stereoradiography, but the
study was conducted using an older version
of the software that did not provide a warn-
ing when the condyles were potentially re-
E
versed, and the operators were inexperienced.
Fig. 7Impact of femur positioning on femoral torsion measurement obtained using CT scan in dry femur.
AE, Diagrams of geometric projections show neutral position (A), 5 of flexion (B), 10 of flexion (C), and 5
All other studies, to our knowledge, have re-
of extension (D) superimposed on CT scans of femur in relation to illustration (E) depicting femur positions ported high ICC values for both modalities
in orange, blue, green and red, respectively. As femur rotates forward during hip flexion (B and C), greater [1722]. Given the weight of evidence from
trochanter (for example) moves forward relative to femoral head, causing projected femoral neck axis to be the present study and previous studies, fem-
progressively more aligned with coronal plane. Hip flexion (B and C) tends to underestimate femoral torsion on
CT, whereas hip extension (D) overestimates this measurement. Yellow lines (AD) denote distance between oral torsion appears to have excellent repro-
anterior and posterior cortical walls, used to define midpoint of base of neck. ducibility for both stereoradiography and CT.

W98 AJR:209, August 2017

 Effect of Femur Position on CT and Stereoradiography Measurements of Femoral Torsion
Hip flexion was the main source of inac-
curacy in CT measurements of femoral tor-
sion. This is important because, in daily
practice, abnormal positions of the femur are
often in flexion, particularly in subjects with
hip or knee osteoarthritis. In this case, CT
Downloaded from www.ajronline.org by 114.124.241.227 on 11/09/17 from IP address 114.124.241.227. Copyright ARRS. For personal use only; all rights reserved
measurements clearly underestimate the val-
ue of the femoral torsion angle, which could
affect clinical decision making.
The main limitation of the present study is
that 3D modeling of the femur using SterEOS
software requires biplanar image acquisi-
tion performed with the patient in a position
with the feet slightly shifted, as per the rec-
ommendation of the manufacturer. This is
recommended to distinguish the anatomic
structures on the lateral view and to ensure
accurate bicondylar axis identification. Al-
though this could lead to a small change in fe-
mur orientation on stereoradiography, we have
shown with studies of dry bones that there is
no impact on stereoradiographic measure-
ments. Another limitation of the present study
is that we did not investigate intraoperator re-
peatability; however, given the excellent inter-
operator repeatability, it can be assumed that
intraoperator repeatability is likewise reliable.
Moreover, a previous study found that femo-
ral anteversion measurement assessed with
stereoradiography showed good intraoperator
and interoperator reproducibility (ICC, 0.91
and 0.82, respectively) [26]. Finally, radiation
dose was not highlighted in our analysis. Ste-
reoradiography has been reported to greatly
reduce the radiation dose to which the patient
is exposed, providing a dose up to 30 times
less than that associated with CT examination
of the lower limb [21, 27].
We conclude that femoral torsion mea-
surements made using stereoradiography are
largely independent of femur positioning. By
contrast, the accuracy of femoral torsion as-
sessed by axial CT does depend on lower
limb positioning because the axes are pro-
jected onto the 2D radiographic plane rather
than perpendicular to the functional femoral
plane. Therefore, a method of measurement
like stereoradiography is warranted to reli-
ably assess femoral torsion in routine clini-
cal practice, mainly for patients who present
with sagittal malalignment of the lower limb.
References
1. Zeng WN, Wang FY, Chen C, et al. Investigation
of association between hip morphology and prev-
alence of osteoarthritis. Sci Rep 2016; 6:23477
AJR:209, August 2017 W99
2. Liebensteiner MC, Ressler J, Seitlinger G,
Djurdjevic T, El Attal R, Ferlic PW. High femoral
anteversion is related to femoral trochlea dyspla-
sia. Arthroscopy 2016; 32:22952299
3. Dickschas J, Harrer J, Reuter B, Schwitulla J,
Strecker W. Torsional osteotomies of the femur.
JOrthop Res 2015; 33:318324
4. Fabry G, MacEwen GD, Shands AR Jr. Torsion of
the femur: a follow-up study in normal and abnor-
mal conditions. JBone Joint Surg Am 1973;
55:17261738
5. Murphy SB, Simon SR, Kijewski PK, Wilkinson
RH, Griscom NT. Femoral anteversion. JBone
Joint Surg Am 1987; 69:11691176
6. Tomczak RJ, Guenther KP, Rieber A, Mergo P,
Ros PR, Brambs HJ. MR imaging measurement of
the femoral antetorsional angle as a new tech-
nique: comparison with CT in children and adults.
AJR 1997; 168:791794
7. Hiseth A, Reikers O, Fnstelien E. Evaluation of
three methods for measurement of femoral neck
anteversion: femoral neck anteversion, definition,
measuring methods and errors. Acta Radiol 1989;
30:6973
8. Sugano N, Noble PC, Kamaric E. A comparison of
alternative methods of measuring femoral antever-
sion. JComput Assist Tomogr 1998; 22:610614
9. Jarrett DY, Oliveira AM, Zou KH, Snyder BD,
Kleinman PK. Axial oblique CT to assess femoral
anteversion. AJR 2010; 194:12301233
10. Botser IB, Ozoude GC, Martin DE, Siddiqi AJ,
Kuppuswami S, Domb BG. Femoral anteversion
in the hip: comparison of measurement by com-
puted tomography, magnetic resonance imaging,
and physical examination. Arthroscopy 2012;
28:619627
11. Hermann KL, Egund N. CT measurement of ante-
version in the femoral neck: the influence of femur
positioning. Acta Radiol 1997; 38:527532
12. Jaarsma RL, Bruggeman AW, Pakvis DF,
Verdonschot N, Lemmens JA, van Kampen A.
Computed tomography determined femoral tor-
sion is not accurate. Arch Orthop Trauma Surg
2004; 124:552554
13. Reikers O, Bjerkreim I, Kolbenstvedt A. Ante-
version of the acetabulum and femoral neck in
normals and in patients with osteoarthritis of the
hip. Acta Orthop Scand 1983; 54:1823
14. Hernandez RJ, Tachdjian MO, Poznanski AK,
Dias LS. CT determination of femoral torsion.
AJR 1981; 137:97101
15. Jaarsma RL, van Kampen A. Rotational malalign-
ment after fractures of the femur. JBone Joint
Surg Br 2004; 86:11001104
16. Thpaut M, Brochard S, Leboucher J, et al. Mea-
suring physiological and pathological femoral an-
teversion using a biplanar low-dose x-ray system:
validity, reliability, and discriminative ability in
cerebral palsy. Skeletal Radiol 2016; 45:243250
17. Pomerantz ML, Glaser D, Doan J, Kumar S,
Edmonds EW. Three-dimensional biplanar radiog-
raphy as a new means of accessing femoral version:
a comparative study of EOS three-dimensional ra-
diography versus computed tomography. Skeletal
Radiol 2015; 44:255260
18. Buck FM, Guggenberger R, Koch PP, Pfirrmann
CW. Femoral and tibial torsion measurements
with 3D models based on low-dose biplanar radio-
graphs in comparison with standard CT measure-
ments. AJR 2012; 199:[web]W607W612
19. Folinais D, Thelen P, Delin C, Radier C, Catonne
Y, Lazennec JY. Measuring femoral and rotation-
al alignment: EOS system versus computed to-
mography. Orthop Traumatol Surg Res 2013;
99:509516
20. Rosskopf AB, Ramseier LE, Sutter R, Pfirrmann
CW, Buck FM. Femoral and tibial torsion mea-
surement in children and adolescents: comparison
of 3D models based on low-dose biplanar radiog-
raphy and low-dose CT. AJR 2014; 202:[web]
W285W291
21. Meyrignac O, Moreno R, Baunin C, et al. Low-
dose biplanar radiography can be used in children
and adolescents to accurately assess femoral and
tibial torsion and greatly reduce irradiation. Eur
Radiol 2015; 25:17521760
22. Wybier M, Bossard P. Musculoskeletal imaging in
progress: the EOS imaging system. Joint Bone
Spine 2012; 80:238243
23. Chaibi Y, Cresson T, Aubert B, et al. Fast 3D re-
construction of the lower limb using a parametric
model and statistical inferences and clinical mea-
surements calculation from biplanar x-rays.
Comput Methods Biomech Biomed Engin 2012;
15:457466
24. Shrout PE, Fleiss JL. Intraclass correlations: uses
in assessing rater reliability. Psychol Bull 1979;
86:420428
25. Knafo J, Thelen T, Verdier D, Creppy L, Tournier C,
Fabre T. Reproducibility of low-dose stereography
measurements of femoral torsion after IM nailing of
femoral shaft fractures and in intact femurs. Orthop
Traumatol Surg Res 2016; 102:595599
26. Guenoun B, Zadegan F, Aim F, Hannouche D,
Nizard R. Reliability of a new method for lower-
extremity measurements based on stereoradio-
graphic three-dimensional reconstruction. Orthop
Traumatol Surg Res 2012; 98:506513
27. Delin C, Silvera S, Bassinet C, et al. Ionizing radia-
tion doses during lower limb torsion and anteversion
measurements by EOS stereoradiography and com-
puted tomography. EurJ Radiol 2014; 83:371377