Gait & Posture 45 (2016) 211–216

Contents lists available at ScienceDirect

Gait & Posture

journal homepage: www.elsevier.com/locate/gaitpost

Defining the medial-lateral axis of an anatomical femur coordinate

system using freehand 3D ultrasound imaging
Elyse Passmore a,b,c,*, Morgan Sangeux a,b,c
a
Hugh Williamson Gait Analysis Laboratory, The Royal Children’s Hospital, Melbourne, Australia
b
Murdoch Childrens Research Institute, Melbourne, Australia
c
Department of Mechanical Engineering, The University of Melbourne, Australia

A R T I C L E I N F O A B S T R A C T

Article history: Hip rotation from gait analysis informs clinical decisions regarding correction of femoral torsional
Received 25 September 2015 deformities. However, it is among the least repeatable due to discrepancies in determining the medial-
Received in revised form 4 February 2016 lateral axis of the femur. Conventional or functional calibration methods may be used to define the axis
Accepted 5 February 2016
but there is no benchmark to evaluate these methods. Freehand 3D ultrasound, the coupling of
ultrasound with 3D motion capture, may provide such a benchmark.
Keywords: We measured the accuracy in vitro and repeatability in vivo of determining the femur condylar axis
Freehand 3D ultrasound
from freehand 3D ultrasound. The condylar axis provided the reference medial-lateral axis of the femur
Anatomical femur coordinate system
and was used to evaluate one conventional method and three functional calibration methods, applied to
Condylar axis
Gait analysis three calibration movements. Ten healthy subjects (20 limbs) underwent 3D gait analysis and freehand
Functional calibration 3D ultrasound. The functional calibration methods were a transformation technique, a geometrical
method and a method that minimises variance of knee varus-valgus kinematics (DynaKAD). The
conventional method used markers over the femoral epicondyles.
The condylar axis determined by 3D ultrasound showed good accuracy in vitro, 1.68 (SD: 0.38) and
good repeatability in vivo, 0.28 (RSMD: 2.38). The DynaKAD method applied to the walking calibration
movement determined the medial-lateral axis closest to the ultrasound reference. The average angular
difference in the transverse plane was 3.18 (SD: 6.18).
Freehand 3D ultrasound offers an accurate, non-invasive and relatively fast method to locate the
medial-lateral axis of the femur for gait analysis.
ß 2016 Elsevier B.V. All rights reserved.

1. Introduction
Gait analysis provides quantitative information of a subjects
gait pattern in the form of joint angles, moments and powers. In the
clinical setting, it informs decision making regarding surgical
correction of gait deformities. For example, the hip rotation profile
during gait is a major determinant in recommending femoral
derotation osteotomy and in predicting functional outcomes
[1]. However, hip rotation kinematics has been found to be among
the least repeatable data from gait analysis [2–4].
The lack of repeatability of the hip rotation profile can primarily
be attributed to a lack of repeatability in determining the frontal
* Corresponding author at: The Royal Children’s Hospital Melbourne 50 Fle-
mington Road Parkville Victoria 3052 Australia. Tel.: +61 3 9345 5354.
E-mail addresses: elyse.passmore@gmail.com, elyse.passmore@rch.org.au
(E. Passmore).
plane of the femur anatomical coordinate system. The frontal plane
of the femur is defined by the cross-product of its longitudinal axis
(hip joint centre to knee joint centre) and its medial-lateral axis. It
is the difficulty in determining the medial-lateral axis of the femur
that affects hip rotation kinematics [4]. Conventional gait models,
widely implemented in commercial software packages (e.g. Plug in
Gait, Vicon motion systems, Oxford UK), define the medial-lateral
axis of the femur by the transepicondylar axis. This axis is
identified by placement of markers over the medial and lateral
epicondyles of the femur, or the positioning of a Knee Alignment
Device (KAD) which clamps on the medial and lateral epicondyles
[5].
Alternatively, several functional calibration methods have been
proposed. These use the motion of the femur and the tibia during a
calibration movement to determine the axis that best describes the
flexion–extension movement of the knee joint. The knee axis is
then used, explicitly [4,6] or implicitly [7], as a proxy to define the

http://dx.doi.org/10.1016/j.gaitpost.2016.02.006
0966-6362/ß 2016 Elsevier B.V. All rights reserved.
212 E. Passmore, M. Sangeux / Gait & Posture 45 (2016) 211–216
medial-lateral axis of the femur. These methods have the
advantage of removing the need for precise marker placement
over anatomic landmarks but may be sensitive to the type of
calibration movement and soft tissue artefact between the skin-
mounted markers and the underlying bones. The majority of
functional calibration methods model the movement of the knee
by a fixed single axis of rotation. This is seen in axis transformation
techniques (ATT or SARA) [8] as well as in geometrical methods
[6,8–10]. An alternative to the single axis of rotation approach are
those methods which determine an axis which minimises the
variance of the frontal plane (varus\valgus) knee kinematics [4,11–
13]. These methods assume the variance is a result of cross-talk
from the sagittal plane movement.
Validation of the methods to define the medial-lateral axis of
the femur are limited due to the absence of a benchmark able to
provide the respective position of the femur anatomical coordinate
system to the skin-mounted markers. Currently, validation has
been limited to in silico data [8,10], mechanical or surrogate models
[14], or in vivo data with indirect outcome measures such as inter-/
intra-assessor repeatability of the hip and knee kinematics or the
absence of cross-talk [4,6,7].
Coordinate systems that are anatomically sound (coordinate
systems that match the bone/segment planes) and repeatable are
required to ensure accurate gait analysis results. We hypothe-
sized that a freehand 3D ultrasound method may be used to
define an anatomical coordinate system of the femur that is more
accurate and repeatable than conventional and functional
calibration methods. Freehand 3D ultrasound combines ultra-
sound imaging with 3D motion capture [15] and has been
utilised in gait analysis to determine the location of the hip joint
centres [16–18]. Freehand 3D ultrasound allows for the position
of the bones to be determined in relation to the skin-mounted
markers. We propose to use the most posterior aspects of the
medial and lateral condyles to define the medial-lateral axis of
the femur. This corresponds to the table top axis described by
Murphy and Simon [19] for measuring femoral torsion, also
called femoral anteversion, and to the condylar axis presented by
Eckhoff et al. [20].
The objectives of this study were (1) to determine in vitro
the accuracy of the freehand 3D ultrasound system to locate
the medial and lateral condyles, (2) to determine in vivo the
repeatability of defining an anatomical coordinate system of the
femur using freehand 3D ultrasound imaging and (3) to use
the anatomical coordinate system of the femur from freehand 3D
ultrasound as a benchmark to evaluate conventional and
functional methods commonly used in gait analysis.
2. Methods
Five reflective markers were rigidly attached to a 59 mm Echo
Blaster 128-1Z ultrasound probe (Telemed, Lithuania). The
coordinate system of the probe was built from the marker
positions and the pose of the probe tracked using least squares
fitting [21]. A calibration procedure using the Cambridge stylus
method was performed to determine the position of the
ultrasound image in the coordinate system of the probe
[18]. A ten-camera video-motion-capture system (Vicon Motion
Systems) recorded marker positions at 100 Hz and a radio
frequency button triggered the motion-capture and ultrasound
systems simultaneously.
Ten healthy adults (5 males, 5 females), mean age 29  9 years
and BMI of 24.4  3.1 kg/m2 with no history of gait pathology, joint
disease, injury or neurological problems were recruited to evaluate
the freehand 3D ultrasound protocol in vivo. The participants gave
written informed consent and ethics approval for the study was
granted from the local institutions ethics committee.
Each subject underwent 3D gait analysis and freehand 3D
ultrasound imaging of the posterior aspect of the femoral condyles.
The gait analysis protocol consisted of a static standing calibration,
functional calibration movements, and walking at self-selected
speed. Passive reflective markers were attached to the subject
according to the Plug-in-Gait marker set [22] (Vicon Motion
Systems) with additional markers on the thigh and shank, Fig. 2(a).
An experienced tester (EP) performed marker placement for each
subject.
We defined the medial-lateral axis of the femur as the condylar
axis [23,24]. The condylar axis models the posterior aspect of the
medial and lateral condyles as spherical or cylindrical. The medial-
lateral axis of the femur is defined by the line connecting the
centres of the medial and lateral spheres/cylinders that best fit the
posterior aspects of the condyles [25,26]. In our freehand 3D
protocol, we used the most posterior aspect of the condyles which
were identified with the subject standing, Fig. 2(b). The probe was
positioned over the knee popliteal fossa in a medial-lateral
orientation, the image depth was set at 60 mm with focus between
20 and 30 mm. The probe was moved up and down to identify the
most posterior aspect of each condyle. The freehand 3D ultrasound
images were loaded into the software Stradwin [15] and two
landmarks were positioned manually to locate the most posterior
aspect of the condyles, Fig. 2(c). For some subjects it was
impossible to view both condyles simultaneously and separate
images and landmarks were obtained for the medial and lateral
condyles.
The conventional method utilised markers over the medial and
lateral epicondyles to define the medial-lateral axis of the femur,
Fig. 2(a). Three functional calibration methods were tested, each
applied to three different calibration movements. The calibration
methods were: the ATT [8], which determines the knee axis that
moves the least during the calibration movement, the geometric
method of Chang and Pollard [10], which assumes marker
trajectories form concentric circles around the knee flexion axis
and the DynaKAD method [4,11], which minimises variance in the
frontal plane (varus\valgus) knee kinematics. The calibration
movements were three repetitions of active knee flexion–
extension (open kinematic chain), bilateral squats (closed kine-
matic chain) and walking strides at a self-selected speed. Markers
on the thigh and the shank (Fig. 2(a)) were tracked using least
squares fitting [27] for the functional calibration methods.
The reference anatomical coordinate system of the femur was
determined as follows. The primary axis was the longitudinal axis
of the femur (Z- axis) and defined by the vector from knee joint
centre to hip joint centre. The positions of the hip and knee joint
centres were determined from the static calibration using Plug-in-
Gait (Vicon motion systems). The anterior-posterior axis (X- axis,
perpendicular to the frontal plane of the femur) was determined by
the cross-product of the longitudinal axis and the medial-lateral
axis of the femur as identified by freehand 3D ultrasound imaging.
The Y- axis of the femur was defined as the cross-product between
the Z- and X- axes, which corresponds to the medial-lateral axis
projected onto the transverse plane of the femur.
Similar coordinate systems of the femur were constructed for
the conventional and functional methods. For all methods, the
longitudinal axis of the femur coordinate system was the same
because the locations of the hip and knee joint centres were held
constant. However, each conventional or functional method led to
a different estimate of the medial-lateral axis of the femur and
therefore led to a different frontal plane of the femur. The
conventional and functional calibration methods were compared
to the freehand 3D ultrasound method as the reference. The
angular difference between the frontal planes of the femur was
calculated for each method\calibration movement. This angular
difference reflects the offset in hip rotation that would be observed
 E. Passmore, M. Sangeux / Gait & Posture 45 (2016) 211–216
between the methods. Results were calculated for both knees of
each subject (20 samples). The hip and knee joint kinematics for
each subject during a representative gait cycle were calculated
using the femur coordinate system from each method/calibration
movement combination. The pelvic and tibia coordinate system of
each subject remained constant for all methods.
The results were compared to the variability of the hip rotation
kinematics from our subjects a posteriori. The standard deviation of
hip rotation kinematics was 7.78 in our dataset. We considered that
results farther than 15.48 (2 standard deviations) from the
reference method may affect clinical decision making. We therefore
reported the proportion of the results within 15.48 of the reference
method for each method/calibration movement combination. The
results were analysed through a general linear model ANOVA with the
following two fixed effects: calibration method (ATT, Geometric and
DynaKAD) and calibration movement (flexion–extension, bilateral
squats and walking). Tukey’s simultaneous tests and grouping
analysis was performed post hoc to determine the difference between
methods at a = 0.05.
The repeatability of the freehand 3D ultrasound method was
estimated in vivo. Two separate acquisitions of ultrasound data and
identification of the posterior aspect of the condyles were
performed for each knee and the angular difference in the
transverse plane between the two measurements calculated.
We estimated the accuracy of the freehand 3D ultrasound
protocol in vitro. The femur was positioned on a level surface with
the shaft of the femur horizontal to the surface and an inkpad
under the condyles. As a result, the ink marked the most posterior
aspects of the condyles corresponding to the anatomical axis (table
top axis) described by Murphy and Simon [19], Fig. 1(a). The femur
was then firmly secured to a rigid container. Three markers, placed
on the outside, defined the container’s coordinate system. We
obtained the reference coordinates of the medial and lateral
condyles with respect to the container’s coordinate system by
pointing to the centre of the ink areas with the Cambridge stylus.
The container was then filled with water and the freehand 3D
Fig. 1. (a) Dry femur bone sample with inked areas used to identify the most
posterior aspects of the medial and lateral femoral condyles. (b) Examiner
performing freehand 3D ultrasound of the dry femur bone sample.
213
ultrasound protocol was performed three times with the examiner
blinded to the location of the femur within the container, Fig. 1(b).
Two markers attached to the femur defined its shaft axis. The
angular difference between the condylar axes defined from the
Cambridge stylus and from freehand 3D ultrasound was calculated
in the plane perpendicular to the shaft axis of the femur.
All statistical analyses were performed in Minitab (Minitab Inc.,
State College, USA).
3. Results
The angular difference between the axes obtained in vitro from
the freehand 3D ultrasound protocol and the Cambridge stylus was
1.68 (SD: 0.38). In vivo, the repeatability of the freehand 3D
ultrasound protocol was good with average difference between
repeated measures of 0.28 (RMSD: 2.38). Results for all methods
and calibration movements are presented in Fig. 3. The general
linear model ANOVA results showed a significant effect (p = 0.037)
for interaction of the functional calibration method and calibration
movement. The DynaKAD method consistently produced the
results closest to the ultrasound reference, walking calibration
movement average difference: 3.18 (SD: 6.18). All data using
DynaKAD with the walking calibration movement were within the
range of hip rotation variability of our data set. The conventional,
ATT and geometric methods all resulted in axes that were
predominantly external compared to the ultrasound reference.
Details of the range of motion achieved for each of the
functional calibration movements can be found in Table 1. Similar
amounts of knee varus and knee rotation were observed for all
movements. The walking movement produced a reduced range of
knee flexion 708 (SD: 48) compared to the flexion–extension 1088
(SD: 308) and bilateral squats 1068 (SD: 158) movement.
Kinematic curves for the knee (sagittal and frontal planes) and
the hip (transverse plane) during a representative gait cycle for
each limb are shown in Fig. 4. The graphs show the group average
plus and minus one standard deviation. The ultrasound reference
method had on average, 5.78 (SD: 7.78) of internal hip rotation
throughout the gait cycle and minimal variance in knee varus/
valgus. The DynaKAD method consistently produced group
averages closest to the ultrasound reference method. The ATT
and geometric methods were consistently further away. In
particular, using the bilateral squats calibration movement, the
knee kinematics showed a marked valgus dip during knee flexion
consistent with cross-talk at the knee for both methods.
4. Discussion
This study proposed a freehand 3D ultrasound method to define
an anatomical coordinate system from the position of the medial-
lateral axis of the femur. Ultrasound offers an alternative to
measure bony anatomy without exposure to ionising radiation.
Previously, freehand 3D ultrasound has been used in gait analysis
to locate the hip joint centre in various populations [16–
18]. However, this is the first freehand 3D ultrasound protocol
to determine the medial-lateral axis of the femur for gait analysis.
The freehand 3D ultrasound method presented no major practical
difficulties for the assessor or the subjects and required ten
minutes to acquire both knees. Clear ultrasound images of the
condyles were obtained while the subject was standing (extended
knees, Fig. 2).
The accuracy of the freehand 3D ultrasound system was
estimated in vitro against data obtained with a precision-machined
stylus [18]. The accuracy of the ultrasound protocol was 1.68 (SD:
0.38). The accuracy could not be determined in vivo because there is
no concurrent ‘gold standard’ at this point in time. The freehand 3D
ultrasound protocol also demonstrated good repeatability in vivo
214 E. Passmore, M. Sangeux / Gait & Posture 45 (2016) 211–216

Fig. 2. (a) Marker set definition; The four markers on the thigh and the three markers on the shank (blue) were used to track segments during functional calibration
movements; Medial and lateral epicondyle markers were used to define the transepicondylar axis of the conventional method (red). (b) Subject during 3D freehand
ultrasound imaging of the posterior aspect of the condyles. The femoral coordinate system was defined as follows. The primary axis (blue) was the longitudinal axis of the
femur (Z- axis) defined by the vector from knee joint centre to hip joint centre. The hip and knee joint centres are marked by &. The anterior–posterior axis (red) (X- axis,
perpendicular to the frontal plane of the femur) was determined by the cross-product of the longitudinal axis and the medial–lateral axis of the femur (dashed green) as
identified by freehand 3D ultrasound imaging. The Y- axis of the femur (green) was defined as the cross-product between the Z- and X- axes, which is the medial-lateral axis
projected onto the transverse plane of the femur. (c) Ultrasound image of the posterior aspect of the condyles, the crosses identify the landmarks placed to represent the most
posterior aspect of the condyles. The medial–lateral axis of the femur is represented by the dashed green line.

with average difference between repeated measures of 0.28
(RMSD: 2.38).
We used the most posterior aspects of the medial and lateral
condyles to define the medial-lateral axis of the femur. This
corresponds to the table top axis described by Murphy and Simon
[19] for measuring femoral torsion, and the condylar axis
presented by Eckhoff et al. [20]. There is a growing body of
literature supporting the condylar axis as a bone-fixed surrogate
for the sagittal plane flexion–extension axis of the knee [25,28–
30]. For example, in Hancock et al. [24], the authors advocate the
use of the condylar axis for the positioning of the knee prosthesis,
which has a single-fixed axis of rotation. Recently Yin et al. [31]
utilised a bi-plane X-ray system to compare the orientation of the
knee flexion–extension axis, and the condylar axis the authors
found minimal angular difference between the condylar axis and
the knee flexion–extension axis, 18 (38 with the transepicondylar)
in the transverse plane. In addition, Johal et al. [30] found that tibial
rotation was facultative rather than obligatory over the 208–908 arc
of knee flexion. This supports an overall understanding of the
movement at the knee being decoupled in terms of sagittal plane
knee flexion–extension and internal–external tibial rotation.
The use of an anatomical definition of the femur coordinate
system is also essential to support clinical interpretation of hip
rotation kinematics during gait and to inform clinical decision on
femoral de-rotation osteotomy. We used the condylar axis
determined from freehand 3D ultrasound as a reference for the
medial-lateral axis of the femur. We used this reference to evaluate
several conventional and functional methods commonly employed
in gait analysis. Results showed that the DynaKAD method was
consistently the closest to the ultrasound reference. Nearly all
results from the DynaKAD method were within the variability
(2SD) of hip rotation of our dataset for the flexion–extension (95%),
squats (100%) and walking (100%) calibration movements,
Fig. 3. Therefore, data for which only walking is available may be
calibrated using the DynaKAD functional axis. In populations with
walking disabilities the range of knee flexion during walking may be
reduced compared to the population we investigated in this study. For
these populations we anticipate the need for additional, and probably
assisted, knee flexion–extension movement.
The conventional method uses the transepicondylar axis and
was predominantly external compared to the freehand 3D
ultrasound condylar axis. Eighty percent of the results from the

Table 1
Range of movement for the knee joint during functional calibration movements.
Fig. 3. Boxplot of the angular difference from the 3D freehand ultrasound reference
for 20 limbs. Proportion of data in % within 15.48 (2 SD of the variability of hip Calibration movement Knee flexion (8) Knee varus (8) Knee rotation (8)
rotation in our subjects) of the ultrasound axis is displayed at the top of the graph.
Flexion–extension 108  30 11  5 18  9
Tukey’s grouping results (a = 0.05) are specified at the bottom of the graph, each letter
Bilateral squats 106  15 11  4 25  8
specifies a statistically separate group from (A) closest to the 3D freehand ultrasound
Walking 70  4 13  4 24  3
reference to (D) furthest. * represents outlying data points. Signifies the mean.
 E. Passmore, M. Sangeux / Gait & Posture 45 (2016) 211–216 215

Fig. 4. Kinematic curves for knee flexion, knee varus and internal hip rotation during one gait cycle. Data presented as group average plus and minus one standard deviation for
each method. The values presented alongside the internal hip rotation graphs are the value of the group average at the last time point. By construction, the hip rotation offset
between the groups remains constant throughout the gait cycle.

conventional method were within the hip rotation variability from
our data set. The transepicondylar and condylar axes may not be
parallel anatomically. Studies, using a range of medical imaging
techniques, found the two axes to have an angular difference
between 28 to 78 [20,32]. It is therefore likely that the difference we
measured was due to inaccuracy in marker placement rather than
genuine anatomical difference between the transepicondylar and
condylar axes. Marker placement was performed by a gait analyst
with 5 years’ experience but who rarely performs clinical marker
placement. We cannot exclude that improved results may have
been obtained with an assessor who is more frequently involved in
clinical marker placement. However, even when performed by
experienced gait analysis personnel, the accuracy and repeatability
of marker placement may still be an issue [2–4]. A limitation of our
study is the use of a small sample of 10 healthy adults (20 limbs) to
calculate the standard deviation of the hip rotation kinematics.
216 E. Passmore, M. Sangeux / Gait & Posture 45 (2016) 211–216
Both ATT and geometric functional calibration methods model
the knee joint as a single axis hinge joint fixed in the femur
coordinate system. Our results indicate that this leads to an axis
external to the condylar axis, ATT 6.18 (SD: 6.88) to 14.78 (SD: 8.18)
and Geometric 5.78 (SD: 11.88) to 15.98 (SD: 8.68) depending on the
calibration movement. With 60-80% of the results within the
variability of hip rotation in our dataset, the ATT and geometric
methods did not provide any improvement over the conventional
method. In particular, the geometric method presented results
with larger standard deviations and several outliers. This is in
contrast with previous studies on the hip joint centre where
geometric methods outperformed transformation techniques and
predictive methods [16,17,33].
Functional calibration methods perform well under rigid body
conditions, such as a mechanical hinge joint [8,14]. The variability
of results for the functional calibration methods in our study may
be attributed to two causes. The functional model (e.g. single fixed
hinge joint) may not accurately represent the behaviour of the
knee joint in vivo. Soft tissue artefact may result in noisy data
adversely affecting the optimisation procedure. For example,
considerable soft tissue artefact has been measured on the thigh
segment [34] which would affect determination of the knee
functional axis.
The DynaKAD method determines the position of the medial-
lateral axis of the femur which minimize knee varus-valgus. This is
equivalent to modelling the knee as two hinge joints, one for
sagittal plane flexion–extension around the Y- axis of the femur
and one for internal–external rotation around the longitudinal axis
of the tibia. Such a model may be more consistent with the body of
literature described previously [25,28–30] and probably why the
medial-lateral axes found by the DynaKAD method was the closest
to our ultrasound reference.
5. Conclusion
The objective of this study was to propose and evaluate
methods to achieve anatomically sound and repeatable measures
of hip rotation during gait analysis. A method utilising freehand 3D
ultrasound was proposed to locate the medial-lateral axis of the
femur. The method provided repeatable measurements (mean
0.28, RMSD 2.38) in less than 10 minutes additional time for the
subjects. In the absence of freehand 3D ultrasound imaging, the
DynaKAD functional calibration method applied to the walking
calibration movement produced results closest to the ultrasound
reference. Therefore, the processing of retrospective data may be
performed in the absence of additional calibration movements. It
must be noted however, that all subjects in this study had adequate
range of knee flexion. In subjects with reduced range of knee
flexion during walking, an assisted knee flexion–extension
calibration movement may be necessary.
Acknowledgements
The authors would like to acknowledge Professor Richard Baker
for his expertise and comments regarding the manuscript.
Conflict of interest
No financial or personal relationships were conducted with
individuals or organizations that could inappropriately influence
or bias this work.
References
[1] Schwartz MH, Rozumalski A, Novacheck TF. Femoral derotational osteotomy:
surgical indications and outcomes in children with cerebral palsy. Gait Posture
2014;39:778–83.
[2] McGinley JL, Baker R, Wolfe R, Morris ME. The reliability of three-dimensional
kinematic gait measurements: a systematic review. Gait Posture 2009;29:
360–9.
[3] Schwartz MH, Trost JP, Wervey RA. Measurement and management of errors in
quantitative gait data. Gait Posture 2004;20:196–203.
[4] Schache AG, Baker R, Lamoreux LW. Defining the knee joint flexion-extension
axis for purposes of quantitative gait analysis: an evaluation of methods. Gait
Posture 2006;24:100–9.
[5] Davis R, DeLuca P. Clinical gait analysis: current methods and future directions.
In: Harris G, Smith P, editors. Hum. Motion Anal. Curr. Appl. Futur. Dir.. New York:
The Institute of Electrical and Electronic Engineers Press, Inc.; 1996. p. 17–42.
[6] Besier TF, Sturnieks DL, Alderson JA, Lloyd DG. Repeatability of gait data using a
functional hip joint centre and a mean helical knee axis. J Biomech
2003;36:1159–68.
[7] Schwartz MH, Rozumalski A. A new method for estimating joint parameters
from motion data. J Biomech 2005;38:107–16.
[8] Ehrig RM, Taylor WR, Duda GN, Heller MO. A survey of formal methods for
determining functional joint axes. J Biomech 2007;40:2150–7.
[9] Shakarji CM. Least-squares fitting algorithms of the NIST algorithm testing
system. J Res Natl Inst Stand Technol 1998;103:633.
[10] Chang LY, Pollard NS. Robust estimation of dominant axis of rotation. J
Biomech 2007;40:2707–15.
[11] Baker R, Finney L, Orr J. A new approach to determine the hip rotation profile
from clinical gait analysis data. Hum Mov Sci 1999;18:655–67.
[12] Marin F, Mannel H, Claes L, Dürselen L. Correction of axis misalignment in the
analysis of knee rotations. Hum Mov Sci 2003;22:285–96.
[13] Rivest L-P. A correction for axis misalignment in the joint angle curves
representing knee movement in gait analysis. J Biomech 2005;38:1604–11.
[14] MacWilliams BA. A comparison of four functional methods to determine
centers and axes of rotations. Gait Posture 2008;28:673–9.
[15] Gee A, Prager R, Treece G, Cash C, Berman L. Processing and visualizing three-
dimensional ultrasound data. Br J Radiol 2004;77:S186–93.
[16] Sangeux M, Peters A, Baker R. Hip joint centre localization: Evaluation on
normal subjects in the context of gait analysis. Gait Posture 2011;34:324–8.
[17] Peters A, Baker R, Morris ME, Sangeux M. A comparison of hip joint centre
localisation techniques with 3-DUS for clinical gait analysis in children with
cerebral palsy. Gait Posture 2012;36:282–6.
[18] Peters A, Baker R, Sangeux M. Validation of 3-D freehand ultrasound for the
determination of the hip joint centre. Gait Posture 2010;31:530–2.
[19] Murphy S, Simon S. Femoral anteversion. J Bone Jt Surg 1987;69:1169–76.
[20] Eckhoff D, Hogan C, DiMatteo L, Robinson M, Bach J. Difference between the
epicondylar and cylindrical axis of the knee. Clin Orthop Relat Res
2007;461:238–44.
[21] Soderkkvist I, Wedin PA. Determining the movements of the skeleton using
well-configured markers. J Biomech 1993;26:1473–7.
[22] Davis R, Ounpuu S, Tyburski D, Gage J. A gait analysis data collection and
reduction technique. Hum Mov Sci 1991;10:575–87.
[23] Eckhoff DG, Bach JM, Spitzer VM, Reinig KD, Bagur MM, Baldini TH, et al. Three-
dimensional mechanics, kinematics, and morphology of the knee viewed in
virtual reality. J Bone Joint Surg Am 2005;87(Suppl 2):71–80.
[24] Hancock CW, Winston MJ, Bach JM, Davidson BS, Eckhoff DG. Cylindrical axis,
not epicondyles, approximates perpendicular to knee axes. Clin Orthop Relat
Res 2013;471:2278–83.
[25] Iwaki H, Pinskerova V, Freeman MA. Tibiofemoral movement 1: the shapes and
relative movements of the femur and tibia in the unloaded cadaver knee. J
Bone Joint Surg Br 2000;82:1189–95.
[26] McPherson a, Kärrholm J, Pinskerova V, Sosna a, Martelli S. Imaging knee
position using MRI RSA/CT and 3D digitisation. J Biomech 2005;38:263–8.
[27] Söderkvist I, Wedin PÅ. On condition numbers and algorithms for determining
a rigid body movement. Bit 1994;34:424–36.
[28] Freeman MAR, Pinskerova V. The movement of the normal tibio-femoral joint.
J Biomech 2005;38:197–208.
[29] Hill PF, Vedi V, Williams a, Iwaki H, Pinskerova V, Freeman MA. Tibiofemoral
movement 2: the loaded and unloaded living knee studied by MRI. J Bone Joint
Surg Br 2000;82:1196–8.
[30] Johal P, Williams A, Wragg P, Hunt D, Gedroyc W. Tibio-femoral movement in
the living knee. A study of weight bearing and non-weight bearing knee
kinematics using interventional MRI. J Biomech 2005;38:269–76.
[31] Yin L, Chen K, Guo L, Cheng L, Wang F, Yang L. Identifying the functional
flexion-extension axis of the knee: an in-vivo kinematics study. PLoS One
2015;10:e0128877.
[32] Asano T, Akagi M, Nakamura T. The functional flexion-extension axis of the
knee corresponds to the surgical epicondylar axis: In vivo analysis using a
biplanar image-matching technique. J Arthroplasty 2005;20:1060–7.
[33] Sangeux M, Pillet H, Skalli W. Which method of hip joint centre localisation
should be used in gait analysis? Gait Posture 2014;40:20–5.
[34] Peters A, Galna B, Sangeux M, Morris M, Baker R. Quantification of soft tissue
artifact in lower limb human motion analysis: a systematic review. Gait
Posture 2010;31:1–8.