- Email: [email protected]
Non-invasive assessment of soft-tissue artifact and its effect on knee joint kinematics during functional activity
ARTICLE IN PRESS Journal of Biomechanics 43 (2010) 1292–1301
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Non-invasive assessment of soft-tissue artifact and its effect on knee joint kinematics during functional activity Massoud Akbarshahi a, Anthony G. Schache a, Justin W. Fernandez a, Richard Baker a,b, Scott Banks c, Marcus G. Pandy a,n a
Department of Mechanical Engineering, University of Melbourne, Victoria 3010, Australia Murdoch Childrens Research Institute, Parkville, Victoria 3010, Australia c Department of Mechanical and Aerospace Engineering, University of Florida, Gainesville, FL 32611, USA b
a r t i c l e in f o
a b s t r a c t
Article history: Accepted 27 January 2010
The soft-tissue interface between skin-mounted markers and the underlying bones poses a major limitation to accurate, non-invasive measurement of joint kinematics. The aim of this study was twofold: first, to quantify lower limb soft-tissue artifact in young healthy subjects during functional activity; and second, to determine the effect of soft-tissue artifact on the calculation of knee joint kinematics. Subject-specific bone models generated from magnetic resonance imaging (MRI) were used in conjunction with X-ray images obtained from single-plane fluoroscopy to determine threedimensional knee joint kinematics for four separate tasks: open-chain knee flexion, hip axial rotation, level walking, and a step-up. Knee joint kinematics was derived using the anatomical frames from the MRI-based, 3D bone models together with the data from video motion capture and X-ray fluoroscopy. Soft-tissue artifact was defined as the degree of movement of each marker in the anteroposterior, proximodistal and mediolateral directions of the corresponding anatomical frame. A number of different skin-marker clusters (total of 180) were used to calculate knee joint rotations, and the results were compared against those obtained from fluoroscopy. Although a consistent pattern of soft-tissue artifact was found for each task across all subjects, the magnitudes of soft-tissue artifact were subject-, task- and location-dependent. Soft-tissue artifact for the thigh markers was substantially greater than that for the shank markers. Markers positioned in the vicinity of the knee joint showed considerable movement, with root mean square errors as high as 29.3 mm. The maximum root mean square errors for calculating knee joint rotations occurred for the open-chain knee flexion task and were 24.31, 17.81 and 14.51 for flexion, internal–external rotation and abduction–adduction, respectively. The present results on soft-tissue artifact, based on fluoroscopic measurements in healthy adult subjects, may be helpful in developing location- and direction-specific weighting factors for use in global optimization algorithms aimed at minimizing the effects of soft-tissue artifact on calculations of knee joint rotations. & 2010 Elsevier Ltd. All rights reserved.
Keywords: X-ray fluoroscopy Motion capture Gait analysis Skin-mounted markers
1. Introduction Accurate in vivo measurement of knee joint kinematics is important for the evaluation of different surgical techniques, treatment methods and implant designs and for the development and validation of computer models capable of simulating normal and pathological movement (Pandy, 2001; Fernandez et al., 2008). Three-dimensional (3D) motion analysis using skin markers is the most common method for measuring knee joint kinematics in vivo. The accuracy of this approach is determined mainly by errors associated with the non-rigid movement of the soft-tissue
n
Corresponding author. Tel.: + 61 3 8344 4054; fax: +61 3 9347 7717. E-mail addresses: [email protected], [email protected] (M.G. Pandy). 0021-9290/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jbiomech.2010.01.002
interface between the skin markers and the underlying bone, commonly referred to as soft-tissue artifact (STA). Numerous studies have investigated thigh and shank STA for a variety of different motor tasks, such as walking, running and sitto-stand (Cappozzo et al., 1996; Wretenberg et al., 1996; Fuller et al., 1997; Reinschmidt et al., 1997; Stagni et al., 2005; Benoit et al., 2006; Tsai et al., 2009). All of these studies have found STA to be greater for the thigh than for the shank, with STA errors reaching values as high as 50 mm. It is also important to quantify the propagation of STA to the estimation of knee joint kinematics. Previous studies have most often used intrusive techniques for their analyses, such as bone pins (Fuller et al., 1997; Reinschmidt et al., 1997; Benoit et al., 2006), external fixators (Cappozzo et al., 1996) and percutaneous tracking devices (Holden et al., 1997; Manal et al., 2000) to quantify joint motion in vivo. Unfortunately, these devices can restrict the movement of the subject and alter the normal, unimpeded sliding
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
of the soft tissues relative to the underlying bone. To overcome this problem, non-invasive methods such as magnetic resonance imaging (MRI) (Sangeux et al., 2006) and X-ray fluoroscopy (Stagni et al., 2005; Garling et al., 2007; Sudhoff et al., 2007) have been used to quantify joint motion in vivo. However, these studies have been associated with several limitations, such as: (a) capturing several static poses throughout an arc of motion rather than measuring a continuous dynamic motion (Sangeux et al., 2006; Sudhoff et al., 2007); (b) investigating a single motor task only (Sangeux et al.,
1293
2006; Garling et al., 2007; Sudhoff et al., 2007); and (c) using elderly subjects fitted with knee implants (Stagni et al., 2005; Garling et al., 2007). The use of a non-invasive method to quantify the effect of STA on the estimation of knee joint kinematics has not been conducted on a group of healthy young adults for functional motor tasks such as walking. The aim of this study was twofold: first, to combine subjectspecific, MRI-based, 3D bone models and X-ray fluoroscopy to quantify lower limb STA non-invasively in young healthy subjects during functional activity; and second, to determine the effect of STA on the calculation of knee joint kinematics.
2. Methods 2.1. MRI bone models
TH6
TH3
Four able-bodied adult males (age=3073 yr, weight=7177 kg, height=17872 cm) with no history of lower limb musculoskeletal injury participated in this study. Images of each subject’s left lower limb, from the pelvis to the ankle joint, were acquired from a 3T Siemens MRI device using a T2 fat-suppressed sequence (TE=12 ms, RT=23 ms, NEX=1, coronal plane images, slice thickness=1 mm, no gaps). The outlines of the exterior cortical bones were segmented using 3D Doctor (Able Software Corp.). Surface models of the femur, tibia and fibula were created by fitting polygon surfaces to the cloud of segmented points using Geomagic (Geomagic Inc.). Differences between the wrapped surfaces and the segmented points were less than 0.0570.3 mm (average error71 SD) for all the bone models. Bone-embedded anatomical frames for the femur and lower leg (tibia plus fibula) were defined as described in Fernandez et al. (2008) and Eckhoff et al. (2005).
TH5 TH4
TH2
PAT
TH1
SH1
SH2
SH3
2.2. Human motion experiments
Fig. 1. Schematic diagram showing the locations of skin markers mounted on each subject’s left leg. Symbols appearing in the diagram are: TH1–TH6: thigh markers; SH1–SH3: shank markers; and PAT: patellar marker.
Ten reflective markers were placed on each subject’s left leg (Fig. 1). Markers were positioned on the anterior and lateral aspects of the mid and distal third of the thigh (TH2–TH6), the mid anterior and lateral aspects of the shank (SH1–SH3), the lateral femoral epicondyle (TH1) and the patella (PAT). Kinematic data were collected simultaneously from an X-ray fluoroscopy unit (Pulsera, Philips Medical
YF XF ZF
YF XF
RF/FA
ZF
RL/F YL XL
ZFA XFA
ZL
PAT
YTA
YFA ZTA
XTA
RL/FA = RL/F×RF/FA RL/TA = RL/F×RF/TA
Fig. 2. Three retro-reflective/radio-opaque markers were used to derive the transformation matrix (RL/F) between the laboratory frame (XL, YL, ZL) and the fluoroscopy frame (XF, YF, ZF). Following the registration of the 3D bone models to the 2D X-ray images, the transformation matrices (RF/FA and RF/TA) between the anatomical frames of the femur and tibia and the fluoroscopy frame were calculated. (L: laboratory frame; F: fluoroscopy frame; FA: femur anatomical frame; TA: lower leg (tibia/fibula) anatomical frame, PAT: patellar marker used for synchronization).
ARTICLE IN PRESS
(7 2.4) (7 1.9) (7 2.5) (7 6.1) (7 10.8) (7 14) (75.4) (7 1.6) (7 2.1) (7 2.5) (7 3.2) (71.9) 5.3 6.4 8.2 10.4 14.6 19.4 10.7 6.7 5.3 6.5 9.0 6.9 (73.6) (75.9) (74.2) (76.5) (74.2) (72.8) (7 4.7) (76.7) (70.5) (73.4) (71.0) (7 0.2) 5.5 9.3 11.3 17.5 15.1 16.8 12.6 30.3 2.6 2.6 2.9 2.7 (7 4.4) (7 2.2) (7 2.1) (7 1.5) (7 1.6) (7 2.3) (7 3.0) (7 2.8) (7 2.3) (7 2.0) (7 2.2) (7 0.2) 10.7 5.3 8.1 2.7 3.5 5.1 5.9 7.2 4.1 4.3 4.5 4.3 (75.6) (76.1) (75.1) (76.5) (76.3) (74.4) (70.9) (77.0) (79.8) (77.4) (79.6) (73.3) 8.8 9.2 8.9 10.9 10.6 10.0 9.7 11.6 7.0 6.0 12.1 8.4 (7 2.5) (7 3.4) (7 3.1) (7 3.1) (7 2.6) (7 3.0) (7 3.1) (7 1.3) (7 1.1) (7 1.1) (7 0.7) (7 0.2) 4.1 4.9 5.4 10.9 10.2 9.9 7.6 17.1 3.2 3.5 3.2 3.3 (7 2.1) (7 1.6) (7 1.9) (7 1.4) (7 2.1) (7 2.6) (7 1.9) (7 2.4) (7 1.7) (7 3.2) (7 1.2) (7 1.2) 9.9 5.6 5.1 4.6 5.7 6.5 6.2 8.7 4.6 6.4 7.0 6.0 (70.7) (71.2) (72.7) (74.6) (77.2) (710.2) (7 5.0) (72.7) (70.9) (71.9) (72.5) (7 0.6) 5.9 6.6 8.3 9.4 13.4 19.0 10.4 7.1 4.1 3.9 5.0 4.3 (7 0.4) (7 1.5) (7 2.1) (7 1.0) (7 0.9) (7 0.7) (7 0.8) (7 1.3) (7 0.3) (7 0.6) (7 0.7) (7 0.1) TH1 TH2 TH3 TH4 TH5 TH6 Thigh PAT SH1 SH2 SH3 Shank
17.0 4.1 7.5 3.4 3.7 5.0 6.8 18.0 4.1 6.0 6.3 5.5
(7 5.7) (7 1.0) (7 1.5) (7 6.4) (7 2.4) (7 0.5) (7 6.3) (7 5.8) (7 0.6) (7 2.2) (7 0.7) (7 1.6)
8.6 7.1 7.5 8.8 8.6 10.7 8.6 8.8 6.0 8.5 11.4 8.6
(7 4.4) (7 2.2) (7 2.1) (7 2.9) (7 3.6) (7 4.5) (7 1.3) (7 3.7) (7 3.2) (7 4.3) (7 1.6) (7 2.7)
3.3 9.0 12.6 5.3 7.6 10.6 8.1 3.6 2.4 3.7 4.1 3.4
( 70.4) ( 71.2) ( 76.1) ( 70.3) ( 72.4) ( 74.8) (73.4) ( 70.5) ( 70.4) ( 71.4) ( 70.9) (70.9) 9.5 3.0 3.2 19.5 13.2 9.2 9.6 40.4 1.6 4.4 1.6 2.5 ( 71.8) ( 72.8) ( 73.0) ( 71.6) ( 71.7) ( 71.7) (75.2) ( 77.0) ( 72.7) ( 72.5) ( 71.6) (71.2)
1.2 3.0 3.5 2.4 2.3 2.0 2.4 2.9 1.3 1.4 1.2 1.3
Av Av Av Av Av Av SD
Av
SD
Av
SD Av
P/D
SD
M/L
SD
A/P
SD
P/D
SD
M/L
SD
Av
SD
Av
SD
M/L P/D A/P
Step-up
SD
Soft-tissue artifact was presented as the root mean square (RMS) error of the position of a given skin marker in its relevant anatomical frame. For each task, the RMS error for a given marker was averaged across subjects to compare the magnitude of STA across tasks. The total amount of STA for the thigh and shank markers was calculated by averaging the RMS values of the TH1–TH6 and SH1–SH3 markers, respectively. Knee joint rotation errors were expressed as the RMS difference between the results obtained from the skin markers and X-ray fluoroscopy.
Av
2.6. Data analysis
A/P
To determine the accuracy of the kinematic measurements derived from X-ray fluoroscopy, additional experiments were performed using dry cadaver bones. Radioopaque markers attached to various anatomical landmarks on the bones were used to define anatomical frames for the femur and tibia/fibula. Computerized tomography (CT) images were recorded at 1 mm intervals, and these images were then segmented to generate 3D bone models. The cadaver bones were fixed in an arbitrary pose using bone cement, with the flexion angle of the knee set at 361. A second series of CT images was then obtained and the images were segmented to generate 3D bone models for the assembled configuration. Differences between the 3D models of the femur, tibia and fibula generated from the two CT datasets were negligible (i.e., the standard deviation between bone model surfaces was less than 0.05 mm). The knee joint rotations for the assembled cadaver model were calculated by describing the pose of the tibia/fibula anatomical frame with respect to the femur anatomical frame, and the result was taken as the true pose of the knee. The assembled cadaver model was then placed in the imaging volume of the fluoroscopy unit, and images were acquired at a 35 cm distance from the image plane. Using the 3D bone models generated from the first set of CT images, knee joint rotations were calculated by iteratively matching the bone models to the CT images, as described above. The knee joint rotations calculated from fluoroscopy were then compared to those obtained for the true pose. Maximum errors obtained for the fluoroscopy measurements across all distances were 1.5 mm for in-plane translations, 3 mm for translations normal to the image plane, and 0.61 for rotations in all planes. These results are consistent with findings reported by Fregly et al. (2005).
M/L
2.5. Evaluation of fluoroscopic measurement error
P/D
To determine the effect of STA on measured knee joint rotations, clusters comprising of different combinations of three skin markers were used to construct technical frames (Cappozzo et al., 1995). Rotation matrices defining the relationship between each technical frame and the relevant MRI-based anatomical frame were established for the neutral standing position, and these matrices were then used to transform each technical frame to the anatomical frame at each instant in time. This procedure was repeated for all the non-collinear triangular combinations derived from the thigh markers (TH1–TH6) and the PAT marker using Gram–Schmidt orthogonalization, resulting in a total of 180 different thigh marker clusters.
A/P
2.4. Calculation of knee joint rotations using skin markers
Level walking
The X-ray images were corrected for image distortion, and the projective parameters of the X-ray system were determined using a calibration routine (Banks and Hodge, 1996). The 3D MRI-based model of the knee joint was registered to the two dimensional X-ray images using custom software that iteratively matched the model pose to the X-ray images. The initial positions of the bones were manually determined, and an edge-based metric, non-linear, least-squares minimization (Fregly et al., 2005) was used to calculate the optimal positions of the bones. Knee joint kinematics were computed using the Joint Coordinate System described by Grood and Suntay (1983). Data for walking were recorded only from mid-swing to mid-stance due to the limited imaging field of view and occlusion from the contralateral leg. The positions of the skin markers were transformed from the laboratory frame to the relevant bone-embedded anatomical frame (Fig. 2A). The PAT marker was described with respect to the femur anatomical frame.
Hip axial rotation
2.3. Calculation of knee joint rotations using X-ray fluoroscopy
Open-chain knee flexion
Systems) operating in pulsed mode at 30 Hz and a video motion capture system (VICON 512, Oxford Metrics) with nine high-resolution M1 cameras sampling at 120 Hz. Subjects were positioned inside the calibration volume of both systems and asked to perform four tasks: open-chain knee flexion; hip axial rotation with the knee extended and the foot resting in the centre of a swivel disc; walking on a treadmill; and a step-up. Three radio-opaque reflective markers, visible in both systems, were placed on a plane parallel to the image plane of the X-ray unit. These markers were used to synchronize data between the two systems and to perform relevant reference frame transformations (Fig. 2). The distances between the radio-opaque reflective PAT marker and the three image-intensifier markers were matched in both the laboratory and image-plane frames to synchronize time between the video motion capture system and the X-ray fluoroscopy unit.
SD
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301 Table 1 Average (7 1SD) RMS error due to STA for each skin marker and for each of the four tasks investigated in this study. The overall degree of error due to STA for the thigh and shank segments (bold rows) was calculated as the average (71SD) of the average RMS error for the thigh (TH1–TH6) and shank (SH1–SH3) markers, respectively. A/P: anteroposterior; P/D: proximodistal; M/L: mediolateral.
1294
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
3. Results The average STA of all subjects was larger for the thigh markers than for the shank markers across all tasks (Table 1). The largest
1295
amount of STA for the thigh occurred in the proximodistal direction during the step-up task (12.674.7 mm), whereas that for the shank occurred in the mediolateral direction during the open-chain knee flexion task (8.672.7 mm).
Open-chain knee flexion 0
350
-500
300 TH6
-100 SH1 Proximodistal (mm)
Proximodistal (mm)
250
200 TH5
TH3 150
100
TH2
TH4
-150 SH2 -200 SH3 -250
-300
50
PAT TH1 -350
0
-50 -40
0
-20
20 40 60 80 Anteroposterior (mm)
100
120
-400 -80
140
-60
-40
0 20 40 -20 Anteroposterior (mm)
60
80
100
60
80
100
Hip axial rotation 0
350
-500
300 TH6
-100 SH1 Proximodistal (mm)
Proximodistal (mm)
250
200 TH5
TH3 150
100
TH2
TH4
-150 SH2 -200 SH3 -250
-300
50
PAT TH1 -350
0
-50 -40
-20
0
20 40 60 80 Anteroposterior (mm)
100
120
140
-400 -80
-60
-40
0 20 40 -20 Anteroposterior (mm)
Fig.3. Trajectories of the skin markers measured for a representative subject (subject 3) for each of the four tasks investigated in this study. Data are presented in the relevant anatomical frame. Note that the vertical and horizontal axes have different scales. Data for the remaining three subjects are given in the Supplementary material. Sign conventions are as follows: anteroposterior axis, anterior (+ )/posterior ( ) and proximodistal axis, proximal (+ )/distal ( ).
ARTICLE IN PRESS 1296
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Walking 350
0 -50
300 TH6
-100 Proximodistal (mm)
Proximodistal (mm)
250
200 TH5
TH3 150
100
TH2
50
TH4
SH1
-150 SH2 -200 SH3 -250
-300
PAT TH1
-350
0
-50 -40
0
-20
20 40 60 80 Anteroposterior (mm)
100
120
-400 -80
140
-60
-40
0 20 40 -20 Anteroposterior (mm)
60
80
100
Step-up 0
350
-50
300 TH6
-100
250 Proximodistal (mm)
Proximodistal (mm)
SH1 200 TH5
TH3 150 100
TH2
TH4
-150 SH2 -200 SH3 -250
-300
50
PAT TH1 -350
0
-50 -40
-20
0
20 40 60 80 Anteroposterior (mm)
100
120
140
-400 -80
-60
-40
0 20 40 -20 Anteroposterior (mm)
60
80
100
Fig.3. (Continued)
For the tasks that involved a large amount of knee flexion (open-chain knee flexion, step-up and walking), skin markers on the mid-anterior aspect of the thigh (TH4, TH5 and TH6) moved distally relative to the underlying bone (Fig. 3) (see also
Supplementary material). Amongst the anterior thigh markers, the TH4 marker displayed the largest amount of STA in the proximodistal direction for these tasks, with an average RMS error ranging from 10.973.1 mm (walking) to 19.576.4 mm
ARTICLE IN PRESS
TH5–TH1–TH3 PAT–TH4–TH3 TH1–TH2–TH4 PAT–TH6–TH2 TH1–TH3–TH6 TH3–TH5–TH6 TH1–TH2–TH6 TH5–TH4–TH2 TH1–TH2–TH6 TH5–TH3–TH1 TH1–PAT–TH5 TH2–TH5–TH1 Most accurate thigh marker cluster
2.6 20.2 3.1 33.0 1.9 3.6 3.62 76.1 4.46 58.2 2.7 5.4 2 23.3 2.4 37.9 1.8 32.3 4.9 43.6 4.1 28.8
9.5 82.5 11.8 136.1 9.9 17.5 7.38 149 7.46 110.8 9 17.9 3.8 36.1
4.9 43.0 7.2 76.0
ABD-ADD INT-EXT ROT FLEX-EXT
4.3 7.7 4.45 93.5 5.91 82.6 4.5 8.9
3.7 3.4
(open-chain knee flexion) (Table 1). The TH1 marker had the largest average RMS error in the anteroposterior direction for these tasks (Table 1). As the knee flexed from an extended position, the TH1 marker shifted posteriorly and proximally with respect to the underlying bone (Fig. 3). During walking, markers on the thigh underwent a sudden shift in the proximal and anterior directions immediately after heel strike (Fig. 4). For the open-chain knee flexion, step-up and walking tasks, the average RMS error for the PAT marker in the proximodistal direction was larger than that for any of the thigh markers (Table 1 and Fig. 3), whilst the average RMS error for all the shank markers was smallest in the proximodistal direction (Table 1 and Fig. 3). For the hip axial rotation task, STA was smallest in the proximodistal direction for all markers (Table 1). Average RMS errors for the anterior thigh markers in the anteroposterior and mediolateral directions increased for markers placed more proximally on the thigh (Table 1). The TH6 marker displayed the largest amount of STA for the hip axial rotation task, with an average RMS error of 19.0710.2 mm in the mediolateral direction. Soft-tissue artifact for the shank markers was small, with average RMS errors of less than 8 mm in all directions (Table 1).
Min RMS RMS/ROM
Fig.4. Partial derivatives of the TH1 marker error trajectories calculated with respect to the knee flexion angle for walking. Data are presented for each subject in the anteroposterior (top) and proximodistal (bottom) directions. Symbols appearing in the diagram are: STA, soft-tissue artefact; Q: flexion angle; A/P: anteroposterior; and P/D: proximodistal.
10.2 181.3
0.25 Time (s)
3.3 71.9
0.15
14.5 159.0
0.05
17.8 131.1
0
24.3 21.5
-4
Max RMS RMS/ROM
-3
2.7 29.3
-2
6.1 102.3
-1
2.4 47.3
0
7.2 69.0
P/D
1
6.4 46.7
2
dQ
dSTA
(mm/deg)
3
8.3 7.3
0.45
Subject 1 Subject 2 Subject 3 Subject 4
1297
Average RMS RMS/ROM
0.35
HS 4
ABD-ADD
0.45
INT-EXT ROT
0.35
FLEX-EXT
0.25 Time (s)
ABD-ADD
0.15
INT-EXT ROT
0.05
FLEX-EXT
0
ABD-ADD
-20
INT-EXT ROT
-15
FLEX-EXT
-10
Step-up
-5
Level walking
dQ
dSTA
A/P
0
Hip axial rotation
(mm/deg)
5
Open-chain knee flexion
Subject 1 Subject 2 Subject 3 Subject 4
Error
HS 10
Table 2 Average maximum (max) and minimum (min) RMS errors for knee joint rotations (in degrees). Also shown are RMS errors expressed as a percentage of the range of motion (ROM) for the particular task (RMS/ROM). Results were obtained using all combinations of skin-marker clusters evaluated in this study (180 in total).
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
ARTICLE IN PRESS 1298
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Subject 1 Swing
Stance
Swing
30
Stance
Swing
30
Stance
20 0
20 Int/Ext (deg)
20 40
Abd/Add (deg)
Flex/Ext (deg)
60 10 0 -10 -20
-20 0.2
0.4
0 -10 -20
-30 0
10
-30 0
0.2
0.4
0
0.2
0.4
0
0.2
0.4
0
0.2
0.4
Subject 2 30
30
20
20
10
10
20 0
Int/Ext (deg)
40
Abd/Add (deg)
Flex/Ext (deg)
60
0 -10 -20
-20 0.2
0.4
-10 -20
-30 0
0
-30 0
0.2
0.4
Subject 3 30
30
20
20
10
10
40 20 0
Int/Ext (deg)
Abd/Add (deg)
Flex/Ext (deg)
60
0 -10 -20
-20 0.2
0.4
-10 -20
-30 0
0
-30 0
0.2
0.4
Subject 4 30
30
20
20
20 0
Int/Ext (deg)
40
Abd/Add (deg)
Flex/Ext (deg)
60 10 0 -10 -20
-20 0.2 Time (s)
0.4
0 -10 -20
-30 0
10
-30 0
0.2 Time (s)
0.4
0
0.2 Time (s)
0.4
Fig. 5. Knee joint rotations for walking measured using X-ray fluoroscopy (solid black lines). The gray shaded regions represent the range of the measurements obtained from all skin-marker clusters (total of 180). The RMS errors between the skin-marker measurements and the fluoroscopy measurements are represented by the thick black dashed lines. The vertical dashed line signifies the instant of heel strike. Sign conventions are: flexion (+ )/extension ( ); adduction (+ )/abduction ( ); and internal rotation (+ )/external rotation ( ).
Soft-tissue artifact had a large effect on calculated knee joint rotations, with average RMS errors ranging from 2.41 to 8.31 (Table 2; Figs. 5 and 6). Internal–external rotation displayed the largest average RMS error for three of the four tasks (Table 2). Further, abduction–adduction and internal–external rotation
displayed average RMS errors that were substantially greater than their respective total range of motion (ROM) for three of the four tasks (Table 2). The effect of STA on calculated knee joint rotations varied across subjects, tasks, and plane of motion (Fig. 5). During
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
1299
Open-chain knee flexion 100
30
30
20
20
60 40 20
Int/Ext (deg)
Abd/Add (deg)
Flex/Ext (deg)
80 10 0
-20
-20 0
-20 0
0.4
0.8
0
1.2
0 -10
-10
0
10
0.4
0.8
1.2
0.4
0.8
IntHR
ExtHR
0.5
1.0
1.2
Hip axial rotation IntHR
ExtHR
IntHR
30
0 -5
ExtHR 30
20
Int/Ext (deg)
5
Abd/Add (deg)
Flex/Ext (deg)
10
10 0
0.5
1.0
1.5
0
-20
-20 0
10
-10
-10
-10
20
0
0.5
1.0
1.5
0
1.5
Walking Swing
30
40
Abd/Add (deg)
Flex/Ext (deg)
Stance
20 0
Stance
20 10 0 -10
-20 0.1
0.3
0.5
Stance
20 10 0 -10
-20 0
Swing
30 Int/Ext (deg)
Swing
60
-20 0
0.1
0.3
0.5
0
0.1
0.3
0.5
1.0 1.5 Time (s)
2
Step-up
20 0
30
30
20
20 Int/Ext (deg)
40
Abd/Add (deg)
Flex/Ext (deg)
60
10 0 -10
0
0.5
1.0 Time (s)
1.5
2
0 -10
-20
-20
10
-20 0
0.5
1.0 1.5 Time (s)
2
0
0.5
Fig.6. Knee joint kinematics measured for a representative subject (subject 1) using fluoroscopy (solid black line), the most accurate skin-marker cluster (dotted black line), the least accurate skin-marker cluster (dashed black line), and the best overall skin-marker cluster (gray solid line) for the four tasks investigated. The vertical dashed line shown for walking signifies the instant of heel strike and for hip axial rotation signifies the instance of zero hip axial rotation. Symbols appearing in the diagram are: Ext HR: external hip rotation and Int HR: internal hip rotation. Sign conventions are: flexion (+ )/extension ( ); adduction (+ )/abduction ( ); and internal rotation (+ )/ external rotation ( ).
walking, for example, the RMS error for knee flexion-extension was larger for subjects 2 and 3 than for subjects 1 and 4. In contrast, the RMS error for knee abduction–adduction was larger
for subjects 3 and 4 than for subjects 1 and 2. Also, the most accurate skin-marker cluster was different for each task and for each plane of motion (Table 2; Fig. 6).
ARTICLE IN PRESS 1300
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Heel Strike, t = 0 sec
t = 0.66 sec
t = 1.0 sec
Fig. 7. Snapshots of X-ray images taken during treadmill walking at a speed of 0.78 m/s. Images recorded for subject 1 at the instant of heel strike (t =0 s); at 0.66 s after heel strike; and at 1.0 s after heel strike.
4. Discussion The aim of this study was firstly, to quantify lower limb STA in young healthy subjects during functional activity; and secondly, to determine the effect of STA on the calculation of knee joint kinematics. Subject-specific, MRI-based bone models were generated and used, in conjunction with X-ray images obtained from video fluoroscopy, to accurately measure 3D knee joint rotations during four functional tasks. Soft-tissue artifact for the thigh was found to be larger than that for the shank (Table 1). While there was some consistency in the patterns of STA measured across subjects for a given task (Fig. 3 and Supplementary material), the magnitudes of the errors were found to be both subject- and location-dependent. The magnitudes of the errors in calculating knee joint rotations were similar for the different planes of motion (Table 2). However, abduction–adduction and internal– external rotation calculated via skin-marker clusters demonstrated misleading patterns (Figs. 5 and 6). The degree of error associated with a given skin-marker cluster differed considerably across different tasks, subjects and planes of motion. The results from this study may aid in the development of location- and direction-specific weighting factors for use in global optimization algorithms aimed at minimizing the effects of STA on calculations of knee joint rotations. When comparing results from previous research investigating errors due to STA, differences in methodology, subject characteristics, tasks performed, anatomical frame definitions, skin-marker locations, and data analysis procedures must all be taken into account. Despite this, our results are consistent with those given by others (Cappozzo et al., 1996; Sati et al., 1996; Fuller et al., 1997; Stagni et al., 2005; Tsai et al., 2009). For example, using X-ray imaging, Sati et al. (1996) and Tsai et al. (2009) found markers positioned on the medial and lateral femoral condyles to move posteriorly and proximally during knee flexion, with STA errors as high as 44 mm in the anteroposterior direction. These errors are similar in magnitude and pattern to those obtained for the TH1 marker during the open-chain knee flexion, walking and step-up tasks in the present study. Also, Tsai et al. (2009) reported distal movement of the anterior thigh markers during knee flexion, which is consistent with our results. In a similar study, Stagni et al. (2005) used single-plane X-ray fluoroscopy on patients fitted with knee implants and reported STA errors as high as 31 and 20 mm for thigh and shank markers, respectively, in the mediolateral direction. Smaller errors for the thigh and shank markers in the mediolateral direction were observed in the present study (Table 1). These differences may be due to inconsistencies in measuring mediolateral translations using single-plane X-ray fluoroscopy and differences in the soft-tissue properties of the subjects; specifically, subjects in the present study were young adults, while those in the Stagni et al.
(2005) study were older patients who had undergone knee replacement surgery. The effect of STA on knee joint kinematics during walking has previously been investigated by Reinschmidt et al. (1997) and Benoit et al. (2006). These studies reported errors ranging in magnitude from 41 to 131, with errors for non-sagittal-plane rotations exceeding the total ROM. Stagni et al. (2005) evaluated the effect of STA on knee joint rotations during four motor tasks and found that knee flexion angle estimates were least affected by STA, whereas RMS/ ROM errors of up to 192% and 117% were found for knee abduction– adduction and internal–external rotation, respectively. In the present study, higher absolute RMS errors were found (Table 2), which is most likely due to the larger knee ROM investigated. The relative errors were similar to the results reported by Reinschmidt et al. (1997) and Benoit et al. (2006). A number of limitations were associated with the present study. First, the procedure used to validate the fluoroscopy measurements did not replicate the exact experimental conditions. The cadaver bone models constructed for the validation process were derived from CT images, whereas those for the subjects were derived from MRI. Moro-oka et al. (2007) evaluated the accuracy of MRI- and CT-based bone models and found that the mean differences between model surfaces were no greater than 0.08 and 0.14 mm for the femur and tibia, respectively. Given the magnitudes of the STA being investigated, the small error associated with MRI-based bone models is unlikely to have influenced the main findings in the present study. Second, the quality of the fluoroscopic images, and hence the accuracy of the pose estimation process, may have been influenced by dynamic motion and the presence of soft tissue. To enhance image quality, subjects were instructed to perform each task at a speed slower than their self-selected speed in order to reduce image blur. Further, a relatively high X-ray voltage (55 KV) was used to reduce the effect of soft tissue on bone-edge clarity, resulting in a higher contrast between soft and hard tissues (see Fig. 7 and compare with Fig. 1 in Tashman (2008)). Third, the reduced speed of movement used to enhance X-ray image quality may have resulted in an underestimation of the effect of inertia on the magnitude of STA. These problems may be overcome by using high-speed, bi-planar, X-ray fluoroscopy (Tashman and Anderst, 2002, 2003). Finally, the time and computational cost of deriving subject-specific models limited the sample used in this study to four subjects. Nonetheless, compared to other X-ray fluoroscopy-based studies in the literature (Stagni et al., 2005; Garling et al., 2007; Sudhoff et al., 2007), the present study provides novel data for the effect of STA on the estimation of knee joint kinematics during a variety of dynamic functional tasks for the natural (intact) knee.
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Conflict of Interest The authors do not have any financial or personal relationships with other people or organizations that could inappropriately influence this manuscript.
Acknowledgments This work was supported by the Australian Research Council under Discovery Project Grants DP0772838 and DP0878705 and by a VESKI Innovation Fellowship to M.G. Pandy.
Appendix A. Supporting materials Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jbiomech.2010.01.002.
References Banks, S.A., Hodge, W.A., 1996. Accurate measurement of three-dimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Transactions on Biomedical Engineering 43, 638–649. Benoit, D.L., Ramsey, D.K., Lamontagne, M., Xu, L., Wretenberg, P., Renstrom, P., 2006. Effect of skin movement artifact on knee kinematics during gait and cutting motions measured in vivo. Gait & Posture 24, 152–164. Cappozzo, A., Catani, F., Croce, U.D., Leardini, A., 1995. Position and orientation in space of bones during movement: anatomical frame definition and determination. Clinical Biomechanics 10, 171–178. Cappozzo, A., Catani, F., Leardini, A., Benedetti, M., Croce, U.D., 1996. Position and orientation in space of bones during movement: experimental artefacts. Clinical Biomechanics 11, 90–100. Eckhoff, D.G., Bach, J.M., Spitzer, V.M., Reinig, K.D., Bagur, M.M., Baldini, T.H., Flannery, N.M.P., 2005. Three-dimensional mechanics, kinematics, and morphology of the knee viewed in virtual reality. Journal of Bone and Joint Surgery (American) 87, 71–80. Fernandez, J.W., Akbarshahi, M., Kim, H.J., Pandy, M.G., 2008. Integrating modelling, motion capture and X-ray fluoroscopy to investigate patellofemoral function during dynamic activity. Computer Methods in Biomechanics and Biomedical Engineering 11, 41–53. Fregly, B.J., Rahman, H.A., Banks, S.A., 2005. Theoretical accuracy of model-based shape matching for measuring natural knee kinematics with single-plane fluoroscopy. Journal of Biomechanical Engineering 127, 692–699.
1301
Fuller, J., Liu, L.-J., Murphy, M.C., Mann, R.W., 1997. A comparison of lowerextremity skeletal kinematics measured using skin- and pin-mounted markers. Human Movement Science 16, 219–242. Garling, E.H., Kaptein, B.L., Mertens, B., Barendregt, W., Veeger, H.E.J., Nelissen, R.G.H.H., Valstar, E.R., 2007. Soft-tissue artefact assessment during step-up using fluoroscopy and skin-mounted markers. Journal of Biomechanics 40, S18–S24. Grood, E., Suntay, W., 1983. A joint coordinate system for the clinical description motions: application to the knee. Journal of Biomechanical Engineering 105, 136–144. Holden, J.P., Orsini, J.A., Siegel, K.L., Kepple, T.M., Gerber, L.H., Stanhope, S.J., 1997. Surface movement errors in shank kinematics and knee kinetics during gait. Gait & Posture 5, 217–227. Manal, K., McClay, I., Stanhope, S., Richards, J., Galinat, B., 2000. Comparison of surface mounted markers and attachment methods in estimating tibial rotations during walking: an in vivo study. Gait & Posture 11, 38–45. Moro-oka, T.-a, Hamai, Satoshi, Miura, Hiromasa, Shimoto, Takeshi, Higaki, Hidehiko, Fregly, Benjamin J., Iwamoto, Yukihide, Banks, Scott A., 2007. Can magnetic resonance imaging-derived bone models be used for accurate motion measurement with single-plane three-dimensional shape registration? Journal of Orthopaedic Research 25, 867–872 Pandy, M.G., 2001. Computer modeling and simulation of human movement. Annual Review of Biomedical Engineering 3, 245–273. Reinschmidt, C., van den Bogert, A.J., Nigg, B.M., Lundberg, A., Murphy, N., 1997. Effect of skin movement on the analysis of skeletal knee joint motion during running. Journal of Biomechanics 30, 729–732. Sangeux, M., Marin, F., Charleux, F., Durselen, L., Ho Ba, Tho, M.C., 2006. Quantification of the 3D relative movement of external marker sets vs. bones based on magnetic resonance imaging. Clinical Biomechanics 21, 984–991. Sati, M., de Guise, J.A., Larouche, S., Drouin, G., 1996. Quantitative assessment of skin-bone movement at the knee. The Knee 3, 121–138. Stagni, R., Fantozzi, S., Cappello, A., Leardini, A., 2005. Quantification of soft tissue artefact in motion analysis by combining 3D fluoroscopy and stereophotogrammetry: a study on two subjects. Clinical Biomechanics 20, 320–329. Sudhoff, I., Van Driessche, S., Laporte, S., de Guise, J.A., Skalli, W., 2007. Comparing three attachment systems used to determine knee kinematics during gait. Gait Posture 25, 533–543. Tashman, S., 2008. Comments on ‘‘validation of a non-invasive fluoroscopic imaging technique for the measurement of dynamic knee joint motion’’. Journal of Biomechanics 41, 3290–3291. Tashman, S., Anderst, W., 2002. Skin motion artifacts at the knee during impact movements. In: Proceedings of the 7th Annual Meeting of Gait and Clinical Movement Analysis Society, Chattanooga. Tashman, S., Anderst, W., 2003. In-vivo measurement of dynamic joint motion using high speed biplane radiography and CT: application to canine ACL deficiency. Journal of Biomechanical Engineering 125, 238–245. Tsai, T.Y., Lu, T.W., Kuo, M.Y., Hsu, H.C., 2009. Quantification of three-dimensional movement of skin markers relative to the underlying bones during functional activities. Biomedical Engineering: Applications, Basis and Communications 21, 223–232. Wretenberg, P., Nemeth, G., Lamontagne, M., Lundin, B., 1996. Passive knee muscle moment arms measured in vivo with MRI. Clinical Biomechanics 11, 439–446.
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Non-invasive assessment of soft-tissue artifact and its effect on knee joint kinematics during functional activity Massoud Akbarshahi a, Anthony G. Schache a, Justin W. Fernandez a, Richard Baker a,b, Scott Banks c, Marcus G. Pandy a,n a
Department of Mechanical Engineering, University of Melbourne, Victoria 3010, Australia Murdoch Childrens Research Institute, Parkville, Victoria 3010, Australia c Department of Mechanical and Aerospace Engineering, University of Florida, Gainesville, FL 32611, USA b
a r t i c l e in f o
a b s t r a c t
Article history: Accepted 27 January 2010
The soft-tissue interface between skin-mounted markers and the underlying bones poses a major limitation to accurate, non-invasive measurement of joint kinematics. The aim of this study was twofold: first, to quantify lower limb soft-tissue artifact in young healthy subjects during functional activity; and second, to determine the effect of soft-tissue artifact on the calculation of knee joint kinematics. Subject-specific bone models generated from magnetic resonance imaging (MRI) were used in conjunction with X-ray images obtained from single-plane fluoroscopy to determine threedimensional knee joint kinematics for four separate tasks: open-chain knee flexion, hip axial rotation, level walking, and a step-up. Knee joint kinematics was derived using the anatomical frames from the MRI-based, 3D bone models together with the data from video motion capture and X-ray fluoroscopy. Soft-tissue artifact was defined as the degree of movement of each marker in the anteroposterior, proximodistal and mediolateral directions of the corresponding anatomical frame. A number of different skin-marker clusters (total of 180) were used to calculate knee joint rotations, and the results were compared against those obtained from fluoroscopy. Although a consistent pattern of soft-tissue artifact was found for each task across all subjects, the magnitudes of soft-tissue artifact were subject-, task- and location-dependent. Soft-tissue artifact for the thigh markers was substantially greater than that for the shank markers. Markers positioned in the vicinity of the knee joint showed considerable movement, with root mean square errors as high as 29.3 mm. The maximum root mean square errors for calculating knee joint rotations occurred for the open-chain knee flexion task and were 24.31, 17.81 and 14.51 for flexion, internal–external rotation and abduction–adduction, respectively. The present results on soft-tissue artifact, based on fluoroscopic measurements in healthy adult subjects, may be helpful in developing location- and direction-specific weighting factors for use in global optimization algorithms aimed at minimizing the effects of soft-tissue artifact on calculations of knee joint rotations. & 2010 Elsevier Ltd. All rights reserved.
Keywords: X-ray fluoroscopy Motion capture Gait analysis Skin-mounted markers
1. Introduction Accurate in vivo measurement of knee joint kinematics is important for the evaluation of different surgical techniques, treatment methods and implant designs and for the development and validation of computer models capable of simulating normal and pathological movement (Pandy, 2001; Fernandez et al., 2008). Three-dimensional (3D) motion analysis using skin markers is the most common method for measuring knee joint kinematics in vivo. The accuracy of this approach is determined mainly by errors associated with the non-rigid movement of the soft-tissue
n
Corresponding author. Tel.: + 61 3 8344 4054; fax: +61 3 9347 7717. E-mail addresses: [email protected], [email protected] (M.G. Pandy). 0021-9290/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jbiomech.2010.01.002
interface between the skin markers and the underlying bone, commonly referred to as soft-tissue artifact (STA). Numerous studies have investigated thigh and shank STA for a variety of different motor tasks, such as walking, running and sitto-stand (Cappozzo et al., 1996; Wretenberg et al., 1996; Fuller et al., 1997; Reinschmidt et al., 1997; Stagni et al., 2005; Benoit et al., 2006; Tsai et al., 2009). All of these studies have found STA to be greater for the thigh than for the shank, with STA errors reaching values as high as 50 mm. It is also important to quantify the propagation of STA to the estimation of knee joint kinematics. Previous studies have most often used intrusive techniques for their analyses, such as bone pins (Fuller et al., 1997; Reinschmidt et al., 1997; Benoit et al., 2006), external fixators (Cappozzo et al., 1996) and percutaneous tracking devices (Holden et al., 1997; Manal et al., 2000) to quantify joint motion in vivo. Unfortunately, these devices can restrict the movement of the subject and alter the normal, unimpeded sliding
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
of the soft tissues relative to the underlying bone. To overcome this problem, non-invasive methods such as magnetic resonance imaging (MRI) (Sangeux et al., 2006) and X-ray fluoroscopy (Stagni et al., 2005; Garling et al., 2007; Sudhoff et al., 2007) have been used to quantify joint motion in vivo. However, these studies have been associated with several limitations, such as: (a) capturing several static poses throughout an arc of motion rather than measuring a continuous dynamic motion (Sangeux et al., 2006; Sudhoff et al., 2007); (b) investigating a single motor task only (Sangeux et al.,
1293
2006; Garling et al., 2007; Sudhoff et al., 2007); and (c) using elderly subjects fitted with knee implants (Stagni et al., 2005; Garling et al., 2007). The use of a non-invasive method to quantify the effect of STA on the estimation of knee joint kinematics has not been conducted on a group of healthy young adults for functional motor tasks such as walking. The aim of this study was twofold: first, to combine subjectspecific, MRI-based, 3D bone models and X-ray fluoroscopy to quantify lower limb STA non-invasively in young healthy subjects during functional activity; and second, to determine the effect of STA on the calculation of knee joint kinematics.
2. Methods 2.1. MRI bone models
TH6
TH3
Four able-bodied adult males (age=3073 yr, weight=7177 kg, height=17872 cm) with no history of lower limb musculoskeletal injury participated in this study. Images of each subject’s left lower limb, from the pelvis to the ankle joint, were acquired from a 3T Siemens MRI device using a T2 fat-suppressed sequence (TE=12 ms, RT=23 ms, NEX=1, coronal plane images, slice thickness=1 mm, no gaps). The outlines of the exterior cortical bones were segmented using 3D Doctor (Able Software Corp.). Surface models of the femur, tibia and fibula were created by fitting polygon surfaces to the cloud of segmented points using Geomagic (Geomagic Inc.). Differences between the wrapped surfaces and the segmented points were less than 0.0570.3 mm (average error71 SD) for all the bone models. Bone-embedded anatomical frames for the femur and lower leg (tibia plus fibula) were defined as described in Fernandez et al. (2008) and Eckhoff et al. (2005).
TH5 TH4
TH2
PAT
TH1
SH1
SH2
SH3
2.2. Human motion experiments
Fig. 1. Schematic diagram showing the locations of skin markers mounted on each subject’s left leg. Symbols appearing in the diagram are: TH1–TH6: thigh markers; SH1–SH3: shank markers; and PAT: patellar marker.
Ten reflective markers were placed on each subject’s left leg (Fig. 1). Markers were positioned on the anterior and lateral aspects of the mid and distal third of the thigh (TH2–TH6), the mid anterior and lateral aspects of the shank (SH1–SH3), the lateral femoral epicondyle (TH1) and the patella (PAT). Kinematic data were collected simultaneously from an X-ray fluoroscopy unit (Pulsera, Philips Medical
YF XF ZF
YF XF
RF/FA
ZF
RL/F YL XL
ZFA XFA
ZL
PAT
YTA
YFA ZTA
XTA
RL/FA = RL/F×RF/FA RL/TA = RL/F×RF/TA
Fig. 2. Three retro-reflective/radio-opaque markers were used to derive the transformation matrix (RL/F) between the laboratory frame (XL, YL, ZL) and the fluoroscopy frame (XF, YF, ZF). Following the registration of the 3D bone models to the 2D X-ray images, the transformation matrices (RF/FA and RF/TA) between the anatomical frames of the femur and tibia and the fluoroscopy frame were calculated. (L: laboratory frame; F: fluoroscopy frame; FA: femur anatomical frame; TA: lower leg (tibia/fibula) anatomical frame, PAT: patellar marker used for synchronization).
ARTICLE IN PRESS
(7 2.4) (7 1.9) (7 2.5) (7 6.1) (7 10.8) (7 14) (75.4) (7 1.6) (7 2.1) (7 2.5) (7 3.2) (71.9) 5.3 6.4 8.2 10.4 14.6 19.4 10.7 6.7 5.3 6.5 9.0 6.9 (73.6) (75.9) (74.2) (76.5) (74.2) (72.8) (7 4.7) (76.7) (70.5) (73.4) (71.0) (7 0.2) 5.5 9.3 11.3 17.5 15.1 16.8 12.6 30.3 2.6 2.6 2.9 2.7 (7 4.4) (7 2.2) (7 2.1) (7 1.5) (7 1.6) (7 2.3) (7 3.0) (7 2.8) (7 2.3) (7 2.0) (7 2.2) (7 0.2) 10.7 5.3 8.1 2.7 3.5 5.1 5.9 7.2 4.1 4.3 4.5 4.3 (75.6) (76.1) (75.1) (76.5) (76.3) (74.4) (70.9) (77.0) (79.8) (77.4) (79.6) (73.3) 8.8 9.2 8.9 10.9 10.6 10.0 9.7 11.6 7.0 6.0 12.1 8.4 (7 2.5) (7 3.4) (7 3.1) (7 3.1) (7 2.6) (7 3.0) (7 3.1) (7 1.3) (7 1.1) (7 1.1) (7 0.7) (7 0.2) 4.1 4.9 5.4 10.9 10.2 9.9 7.6 17.1 3.2 3.5 3.2 3.3 (7 2.1) (7 1.6) (7 1.9) (7 1.4) (7 2.1) (7 2.6) (7 1.9) (7 2.4) (7 1.7) (7 3.2) (7 1.2) (7 1.2) 9.9 5.6 5.1 4.6 5.7 6.5 6.2 8.7 4.6 6.4 7.0 6.0 (70.7) (71.2) (72.7) (74.6) (77.2) (710.2) (7 5.0) (72.7) (70.9) (71.9) (72.5) (7 0.6) 5.9 6.6 8.3 9.4 13.4 19.0 10.4 7.1 4.1 3.9 5.0 4.3 (7 0.4) (7 1.5) (7 2.1) (7 1.0) (7 0.9) (7 0.7) (7 0.8) (7 1.3) (7 0.3) (7 0.6) (7 0.7) (7 0.1) TH1 TH2 TH3 TH4 TH5 TH6 Thigh PAT SH1 SH2 SH3 Shank
17.0 4.1 7.5 3.4 3.7 5.0 6.8 18.0 4.1 6.0 6.3 5.5
(7 5.7) (7 1.0) (7 1.5) (7 6.4) (7 2.4) (7 0.5) (7 6.3) (7 5.8) (7 0.6) (7 2.2) (7 0.7) (7 1.6)
8.6 7.1 7.5 8.8 8.6 10.7 8.6 8.8 6.0 8.5 11.4 8.6
(7 4.4) (7 2.2) (7 2.1) (7 2.9) (7 3.6) (7 4.5) (7 1.3) (7 3.7) (7 3.2) (7 4.3) (7 1.6) (7 2.7)
3.3 9.0 12.6 5.3 7.6 10.6 8.1 3.6 2.4 3.7 4.1 3.4
( 70.4) ( 71.2) ( 76.1) ( 70.3) ( 72.4) ( 74.8) (73.4) ( 70.5) ( 70.4) ( 71.4) ( 70.9) (70.9) 9.5 3.0 3.2 19.5 13.2 9.2 9.6 40.4 1.6 4.4 1.6 2.5 ( 71.8) ( 72.8) ( 73.0) ( 71.6) ( 71.7) ( 71.7) (75.2) ( 77.0) ( 72.7) ( 72.5) ( 71.6) (71.2)
1.2 3.0 3.5 2.4 2.3 2.0 2.4 2.9 1.3 1.4 1.2 1.3
Av Av Av Av Av Av SD
Av
SD
Av
SD Av
P/D
SD
M/L
SD
A/P
SD
P/D
SD
M/L
SD
Av
SD
Av
SD
M/L P/D A/P
Step-up
SD
Soft-tissue artifact was presented as the root mean square (RMS) error of the position of a given skin marker in its relevant anatomical frame. For each task, the RMS error for a given marker was averaged across subjects to compare the magnitude of STA across tasks. The total amount of STA for the thigh and shank markers was calculated by averaging the RMS values of the TH1–TH6 and SH1–SH3 markers, respectively. Knee joint rotation errors were expressed as the RMS difference between the results obtained from the skin markers and X-ray fluoroscopy.
Av
2.6. Data analysis
A/P
To determine the accuracy of the kinematic measurements derived from X-ray fluoroscopy, additional experiments were performed using dry cadaver bones. Radioopaque markers attached to various anatomical landmarks on the bones were used to define anatomical frames for the femur and tibia/fibula. Computerized tomography (CT) images were recorded at 1 mm intervals, and these images were then segmented to generate 3D bone models. The cadaver bones were fixed in an arbitrary pose using bone cement, with the flexion angle of the knee set at 361. A second series of CT images was then obtained and the images were segmented to generate 3D bone models for the assembled configuration. Differences between the 3D models of the femur, tibia and fibula generated from the two CT datasets were negligible (i.e., the standard deviation between bone model surfaces was less than 0.05 mm). The knee joint rotations for the assembled cadaver model were calculated by describing the pose of the tibia/fibula anatomical frame with respect to the femur anatomical frame, and the result was taken as the true pose of the knee. The assembled cadaver model was then placed in the imaging volume of the fluoroscopy unit, and images were acquired at a 35 cm distance from the image plane. Using the 3D bone models generated from the first set of CT images, knee joint rotations were calculated by iteratively matching the bone models to the CT images, as described above. The knee joint rotations calculated from fluoroscopy were then compared to those obtained for the true pose. Maximum errors obtained for the fluoroscopy measurements across all distances were 1.5 mm for in-plane translations, 3 mm for translations normal to the image plane, and 0.61 for rotations in all planes. These results are consistent with findings reported by Fregly et al. (2005).
M/L
2.5. Evaluation of fluoroscopic measurement error
P/D
To determine the effect of STA on measured knee joint rotations, clusters comprising of different combinations of three skin markers were used to construct technical frames (Cappozzo et al., 1995). Rotation matrices defining the relationship between each technical frame and the relevant MRI-based anatomical frame were established for the neutral standing position, and these matrices were then used to transform each technical frame to the anatomical frame at each instant in time. This procedure was repeated for all the non-collinear triangular combinations derived from the thigh markers (TH1–TH6) and the PAT marker using Gram–Schmidt orthogonalization, resulting in a total of 180 different thigh marker clusters.
A/P
2.4. Calculation of knee joint rotations using skin markers
Level walking
The X-ray images were corrected for image distortion, and the projective parameters of the X-ray system were determined using a calibration routine (Banks and Hodge, 1996). The 3D MRI-based model of the knee joint was registered to the two dimensional X-ray images using custom software that iteratively matched the model pose to the X-ray images. The initial positions of the bones were manually determined, and an edge-based metric, non-linear, least-squares minimization (Fregly et al., 2005) was used to calculate the optimal positions of the bones. Knee joint kinematics were computed using the Joint Coordinate System described by Grood and Suntay (1983). Data for walking were recorded only from mid-swing to mid-stance due to the limited imaging field of view and occlusion from the contralateral leg. The positions of the skin markers were transformed from the laboratory frame to the relevant bone-embedded anatomical frame (Fig. 2A). The PAT marker was described with respect to the femur anatomical frame.
Hip axial rotation
2.3. Calculation of knee joint rotations using X-ray fluoroscopy
Open-chain knee flexion
Systems) operating in pulsed mode at 30 Hz and a video motion capture system (VICON 512, Oxford Metrics) with nine high-resolution M1 cameras sampling at 120 Hz. Subjects were positioned inside the calibration volume of both systems and asked to perform four tasks: open-chain knee flexion; hip axial rotation with the knee extended and the foot resting in the centre of a swivel disc; walking on a treadmill; and a step-up. Three radio-opaque reflective markers, visible in both systems, were placed on a plane parallel to the image plane of the X-ray unit. These markers were used to synchronize data between the two systems and to perform relevant reference frame transformations (Fig. 2). The distances between the radio-opaque reflective PAT marker and the three image-intensifier markers were matched in both the laboratory and image-plane frames to synchronize time between the video motion capture system and the X-ray fluoroscopy unit.
SD
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301 Table 1 Average (7 1SD) RMS error due to STA for each skin marker and for each of the four tasks investigated in this study. The overall degree of error due to STA for the thigh and shank segments (bold rows) was calculated as the average (71SD) of the average RMS error for the thigh (TH1–TH6) and shank (SH1–SH3) markers, respectively. A/P: anteroposterior; P/D: proximodistal; M/L: mediolateral.
1294
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
3. Results The average STA of all subjects was larger for the thigh markers than for the shank markers across all tasks (Table 1). The largest
1295
amount of STA for the thigh occurred in the proximodistal direction during the step-up task (12.674.7 mm), whereas that for the shank occurred in the mediolateral direction during the open-chain knee flexion task (8.672.7 mm).
Open-chain knee flexion 0
350
-500
300 TH6
-100 SH1 Proximodistal (mm)
Proximodistal (mm)
250
200 TH5
TH3 150
100
TH2
TH4
-150 SH2 -200 SH3 -250
-300
50
PAT TH1 -350
0
-50 -40
0
-20
20 40 60 80 Anteroposterior (mm)
100
120
-400 -80
140
-60
-40
0 20 40 -20 Anteroposterior (mm)
60
80
100
60
80
100
Hip axial rotation 0
350
-500
300 TH6
-100 SH1 Proximodistal (mm)
Proximodistal (mm)
250
200 TH5
TH3 150
100
TH2
TH4
-150 SH2 -200 SH3 -250
-300
50
PAT TH1 -350
0
-50 -40
-20
0
20 40 60 80 Anteroposterior (mm)
100
120
140
-400 -80
-60
-40
0 20 40 -20 Anteroposterior (mm)
Fig.3. Trajectories of the skin markers measured for a representative subject (subject 3) for each of the four tasks investigated in this study. Data are presented in the relevant anatomical frame. Note that the vertical and horizontal axes have different scales. Data for the remaining three subjects are given in the Supplementary material. Sign conventions are as follows: anteroposterior axis, anterior (+ )/posterior ( ) and proximodistal axis, proximal (+ )/distal ( ).
ARTICLE IN PRESS 1296
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Walking 350
0 -50
300 TH6
-100 Proximodistal (mm)
Proximodistal (mm)
250
200 TH5
TH3 150
100
TH2
50
TH4
SH1
-150 SH2 -200 SH3 -250
-300
PAT TH1
-350
0
-50 -40
0
-20
20 40 60 80 Anteroposterior (mm)
100
120
-400 -80
140
-60
-40
0 20 40 -20 Anteroposterior (mm)
60
80
100
Step-up 0
350
-50
300 TH6
-100
250 Proximodistal (mm)
Proximodistal (mm)
SH1 200 TH5
TH3 150 100
TH2
TH4
-150 SH2 -200 SH3 -250
-300
50
PAT TH1 -350
0
-50 -40
-20
0
20 40 60 80 Anteroposterior (mm)
100
120
140
-400 -80
-60
-40
0 20 40 -20 Anteroposterior (mm)
60
80
100
Fig.3. (Continued)
For the tasks that involved a large amount of knee flexion (open-chain knee flexion, step-up and walking), skin markers on the mid-anterior aspect of the thigh (TH4, TH5 and TH6) moved distally relative to the underlying bone (Fig. 3) (see also
Supplementary material). Amongst the anterior thigh markers, the TH4 marker displayed the largest amount of STA in the proximodistal direction for these tasks, with an average RMS error ranging from 10.973.1 mm (walking) to 19.576.4 mm
ARTICLE IN PRESS
TH5–TH1–TH3 PAT–TH4–TH3 TH1–TH2–TH4 PAT–TH6–TH2 TH1–TH3–TH6 TH3–TH5–TH6 TH1–TH2–TH6 TH5–TH4–TH2 TH1–TH2–TH6 TH5–TH3–TH1 TH1–PAT–TH5 TH2–TH5–TH1 Most accurate thigh marker cluster
2.6 20.2 3.1 33.0 1.9 3.6 3.62 76.1 4.46 58.2 2.7 5.4 2 23.3 2.4 37.9 1.8 32.3 4.9 43.6 4.1 28.8
9.5 82.5 11.8 136.1 9.9 17.5 7.38 149 7.46 110.8 9 17.9 3.8 36.1
4.9 43.0 7.2 76.0
ABD-ADD INT-EXT ROT FLEX-EXT
4.3 7.7 4.45 93.5 5.91 82.6 4.5 8.9
3.7 3.4
(open-chain knee flexion) (Table 1). The TH1 marker had the largest average RMS error in the anteroposterior direction for these tasks (Table 1). As the knee flexed from an extended position, the TH1 marker shifted posteriorly and proximally with respect to the underlying bone (Fig. 3). During walking, markers on the thigh underwent a sudden shift in the proximal and anterior directions immediately after heel strike (Fig. 4). For the open-chain knee flexion, step-up and walking tasks, the average RMS error for the PAT marker in the proximodistal direction was larger than that for any of the thigh markers (Table 1 and Fig. 3), whilst the average RMS error for all the shank markers was smallest in the proximodistal direction (Table 1 and Fig. 3). For the hip axial rotation task, STA was smallest in the proximodistal direction for all markers (Table 1). Average RMS errors for the anterior thigh markers in the anteroposterior and mediolateral directions increased for markers placed more proximally on the thigh (Table 1). The TH6 marker displayed the largest amount of STA for the hip axial rotation task, with an average RMS error of 19.0710.2 mm in the mediolateral direction. Soft-tissue artifact for the shank markers was small, with average RMS errors of less than 8 mm in all directions (Table 1).
Min RMS RMS/ROM
Fig.4. Partial derivatives of the TH1 marker error trajectories calculated with respect to the knee flexion angle for walking. Data are presented for each subject in the anteroposterior (top) and proximodistal (bottom) directions. Symbols appearing in the diagram are: STA, soft-tissue artefact; Q: flexion angle; A/P: anteroposterior; and P/D: proximodistal.
10.2 181.3
0.25 Time (s)
3.3 71.9
0.15
14.5 159.0
0.05
17.8 131.1
0
24.3 21.5
-4
Max RMS RMS/ROM
-3
2.7 29.3
-2
6.1 102.3
-1
2.4 47.3
0
7.2 69.0
P/D
1
6.4 46.7
2
dQ
dSTA
(mm/deg)
3
8.3 7.3
0.45
Subject 1 Subject 2 Subject 3 Subject 4
1297
Average RMS RMS/ROM
0.35
HS 4
ABD-ADD
0.45
INT-EXT ROT
0.35
FLEX-EXT
0.25 Time (s)
ABD-ADD
0.15
INT-EXT ROT
0.05
FLEX-EXT
0
ABD-ADD
-20
INT-EXT ROT
-15
FLEX-EXT
-10
Step-up
-5
Level walking
dQ
dSTA
A/P
0
Hip axial rotation
(mm/deg)
5
Open-chain knee flexion
Subject 1 Subject 2 Subject 3 Subject 4
Error
HS 10
Table 2 Average maximum (max) and minimum (min) RMS errors for knee joint rotations (in degrees). Also shown are RMS errors expressed as a percentage of the range of motion (ROM) for the particular task (RMS/ROM). Results were obtained using all combinations of skin-marker clusters evaluated in this study (180 in total).
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
ARTICLE IN PRESS 1298
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Subject 1 Swing
Stance
Swing
30
Stance
Swing
30
Stance
20 0
20 Int/Ext (deg)
20 40
Abd/Add (deg)
Flex/Ext (deg)
60 10 0 -10 -20
-20 0.2
0.4
0 -10 -20
-30 0
10
-30 0
0.2
0.4
0
0.2
0.4
0
0.2
0.4
0
0.2
0.4
Subject 2 30
30
20
20
10
10
20 0
Int/Ext (deg)
40
Abd/Add (deg)
Flex/Ext (deg)
60
0 -10 -20
-20 0.2
0.4
-10 -20
-30 0
0
-30 0
0.2
0.4
Subject 3 30
30
20
20
10
10
40 20 0
Int/Ext (deg)
Abd/Add (deg)
Flex/Ext (deg)
60
0 -10 -20
-20 0.2
0.4
-10 -20
-30 0
0
-30 0
0.2
0.4
Subject 4 30
30
20
20
20 0
Int/Ext (deg)
40
Abd/Add (deg)
Flex/Ext (deg)
60 10 0 -10 -20
-20 0.2 Time (s)
0.4
0 -10 -20
-30 0
10
-30 0
0.2 Time (s)
0.4
0
0.2 Time (s)
0.4
Fig. 5. Knee joint rotations for walking measured using X-ray fluoroscopy (solid black lines). The gray shaded regions represent the range of the measurements obtained from all skin-marker clusters (total of 180). The RMS errors between the skin-marker measurements and the fluoroscopy measurements are represented by the thick black dashed lines. The vertical dashed line signifies the instant of heel strike. Sign conventions are: flexion (+ )/extension ( ); adduction (+ )/abduction ( ); and internal rotation (+ )/external rotation ( ).
Soft-tissue artifact had a large effect on calculated knee joint rotations, with average RMS errors ranging from 2.41 to 8.31 (Table 2; Figs. 5 and 6). Internal–external rotation displayed the largest average RMS error for three of the four tasks (Table 2). Further, abduction–adduction and internal–external rotation
displayed average RMS errors that were substantially greater than their respective total range of motion (ROM) for three of the four tasks (Table 2). The effect of STA on calculated knee joint rotations varied across subjects, tasks, and plane of motion (Fig. 5). During
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
1299
Open-chain knee flexion 100
30
30
20
20
60 40 20
Int/Ext (deg)
Abd/Add (deg)
Flex/Ext (deg)
80 10 0
-20
-20 0
-20 0
0.4
0.8
0
1.2
0 -10
-10
0
10
0.4
0.8
1.2
0.4
0.8
IntHR
ExtHR
0.5
1.0
1.2
Hip axial rotation IntHR
ExtHR
IntHR
30
0 -5
ExtHR 30
20
Int/Ext (deg)
5
Abd/Add (deg)
Flex/Ext (deg)
10
10 0
0.5
1.0
1.5
0
-20
-20 0
10
-10
-10
-10
20
0
0.5
1.0
1.5
0
1.5
Walking Swing
30
40
Abd/Add (deg)
Flex/Ext (deg)
Stance
20 0
Stance
20 10 0 -10
-20 0.1
0.3
0.5
Stance
20 10 0 -10
-20 0
Swing
30 Int/Ext (deg)
Swing
60
-20 0
0.1
0.3
0.5
0
0.1
0.3
0.5
1.0 1.5 Time (s)
2
Step-up
20 0
30
30
20
20 Int/Ext (deg)
40
Abd/Add (deg)
Flex/Ext (deg)
60
10 0 -10
0
0.5
1.0 Time (s)
1.5
2
0 -10
-20
-20
10
-20 0
0.5
1.0 1.5 Time (s)
2
0
0.5
Fig.6. Knee joint kinematics measured for a representative subject (subject 1) using fluoroscopy (solid black line), the most accurate skin-marker cluster (dotted black line), the least accurate skin-marker cluster (dashed black line), and the best overall skin-marker cluster (gray solid line) for the four tasks investigated. The vertical dashed line shown for walking signifies the instant of heel strike and for hip axial rotation signifies the instance of zero hip axial rotation. Symbols appearing in the diagram are: Ext HR: external hip rotation and Int HR: internal hip rotation. Sign conventions are: flexion (+ )/extension ( ); adduction (+ )/abduction ( ); and internal rotation (+ )/ external rotation ( ).
walking, for example, the RMS error for knee flexion-extension was larger for subjects 2 and 3 than for subjects 1 and 4. In contrast, the RMS error for knee abduction–adduction was larger
for subjects 3 and 4 than for subjects 1 and 2. Also, the most accurate skin-marker cluster was different for each task and for each plane of motion (Table 2; Fig. 6).
ARTICLE IN PRESS 1300
M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Heel Strike, t = 0 sec
t = 0.66 sec
t = 1.0 sec
Fig. 7. Snapshots of X-ray images taken during treadmill walking at a speed of 0.78 m/s. Images recorded for subject 1 at the instant of heel strike (t =0 s); at 0.66 s after heel strike; and at 1.0 s after heel strike.
4. Discussion The aim of this study was firstly, to quantify lower limb STA in young healthy subjects during functional activity; and secondly, to determine the effect of STA on the calculation of knee joint kinematics. Subject-specific, MRI-based bone models were generated and used, in conjunction with X-ray images obtained from video fluoroscopy, to accurately measure 3D knee joint rotations during four functional tasks. Soft-tissue artifact for the thigh was found to be larger than that for the shank (Table 1). While there was some consistency in the patterns of STA measured across subjects for a given task (Fig. 3 and Supplementary material), the magnitudes of the errors were found to be both subject- and location-dependent. The magnitudes of the errors in calculating knee joint rotations were similar for the different planes of motion (Table 2). However, abduction–adduction and internal– external rotation calculated via skin-marker clusters demonstrated misleading patterns (Figs. 5 and 6). The degree of error associated with a given skin-marker cluster differed considerably across different tasks, subjects and planes of motion. The results from this study may aid in the development of location- and direction-specific weighting factors for use in global optimization algorithms aimed at minimizing the effects of STA on calculations of knee joint rotations. When comparing results from previous research investigating errors due to STA, differences in methodology, subject characteristics, tasks performed, anatomical frame definitions, skin-marker locations, and data analysis procedures must all be taken into account. Despite this, our results are consistent with those given by others (Cappozzo et al., 1996; Sati et al., 1996; Fuller et al., 1997; Stagni et al., 2005; Tsai et al., 2009). For example, using X-ray imaging, Sati et al. (1996) and Tsai et al. (2009) found markers positioned on the medial and lateral femoral condyles to move posteriorly and proximally during knee flexion, with STA errors as high as 44 mm in the anteroposterior direction. These errors are similar in magnitude and pattern to those obtained for the TH1 marker during the open-chain knee flexion, walking and step-up tasks in the present study. Also, Tsai et al. (2009) reported distal movement of the anterior thigh markers during knee flexion, which is consistent with our results. In a similar study, Stagni et al. (2005) used single-plane X-ray fluoroscopy on patients fitted with knee implants and reported STA errors as high as 31 and 20 mm for thigh and shank markers, respectively, in the mediolateral direction. Smaller errors for the thigh and shank markers in the mediolateral direction were observed in the present study (Table 1). These differences may be due to inconsistencies in measuring mediolateral translations using single-plane X-ray fluoroscopy and differences in the soft-tissue properties of the subjects; specifically, subjects in the present study were young adults, while those in the Stagni et al.
(2005) study were older patients who had undergone knee replacement surgery. The effect of STA on knee joint kinematics during walking has previously been investigated by Reinschmidt et al. (1997) and Benoit et al. (2006). These studies reported errors ranging in magnitude from 41 to 131, with errors for non-sagittal-plane rotations exceeding the total ROM. Stagni et al. (2005) evaluated the effect of STA on knee joint rotations during four motor tasks and found that knee flexion angle estimates were least affected by STA, whereas RMS/ ROM errors of up to 192% and 117% were found for knee abduction– adduction and internal–external rotation, respectively. In the present study, higher absolute RMS errors were found (Table 2), which is most likely due to the larger knee ROM investigated. The relative errors were similar to the results reported by Reinschmidt et al. (1997) and Benoit et al. (2006). A number of limitations were associated with the present study. First, the procedure used to validate the fluoroscopy measurements did not replicate the exact experimental conditions. The cadaver bone models constructed for the validation process were derived from CT images, whereas those for the subjects were derived from MRI. Moro-oka et al. (2007) evaluated the accuracy of MRI- and CT-based bone models and found that the mean differences between model surfaces were no greater than 0.08 and 0.14 mm for the femur and tibia, respectively. Given the magnitudes of the STA being investigated, the small error associated with MRI-based bone models is unlikely to have influenced the main findings in the present study. Second, the quality of the fluoroscopic images, and hence the accuracy of the pose estimation process, may have been influenced by dynamic motion and the presence of soft tissue. To enhance image quality, subjects were instructed to perform each task at a speed slower than their self-selected speed in order to reduce image blur. Further, a relatively high X-ray voltage (55 KV) was used to reduce the effect of soft tissue on bone-edge clarity, resulting in a higher contrast between soft and hard tissues (see Fig. 7 and compare with Fig. 1 in Tashman (2008)). Third, the reduced speed of movement used to enhance X-ray image quality may have resulted in an underestimation of the effect of inertia on the magnitude of STA. These problems may be overcome by using high-speed, bi-planar, X-ray fluoroscopy (Tashman and Anderst, 2002, 2003). Finally, the time and computational cost of deriving subject-specific models limited the sample used in this study to four subjects. Nonetheless, compared to other X-ray fluoroscopy-based studies in the literature (Stagni et al., 2005; Garling et al., 2007; Sudhoff et al., 2007), the present study provides novel data for the effect of STA on the estimation of knee joint kinematics during a variety of dynamic functional tasks for the natural (intact) knee.
ARTICLE IN PRESS M. Akbarshahi et al. / Journal of Biomechanics 43 (2010) 1292–1301
Conflict of Interest The authors do not have any financial or personal relationships with other people or organizations that could inappropriately influence this manuscript.
Acknowledgments This work was supported by the Australian Research Council under Discovery Project Grants DP0772838 and DP0878705 and by a VESKI Innovation Fellowship to M.G. Pandy.
Appendix A. Supporting materials Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jbiomech.2010.01.002.
References Banks, S.A., Hodge, W.A., 1996. Accurate measurement of three-dimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Transactions on Biomedical Engineering 43, 638–649. Benoit, D.L., Ramsey, D.K., Lamontagne, M., Xu, L., Wretenberg, P., Renstrom, P., 2006. Effect of skin movement artifact on knee kinematics during gait and cutting motions measured in vivo. Gait & Posture 24, 152–164. Cappozzo, A., Catani, F., Croce, U.D., Leardini, A., 1995. Position and orientation in space of bones during movement: anatomical frame definition and determination. Clinical Biomechanics 10, 171–178. Cappozzo, A., Catani, F., Leardini, A., Benedetti, M., Croce, U.D., 1996. Position and orientation in space of bones during movement: experimental artefacts. Clinical Biomechanics 11, 90–100. Eckhoff, D.G., Bach, J.M., Spitzer, V.M., Reinig, K.D., Bagur, M.M., Baldini, T.H., Flannery, N.M.P., 2005. Three-dimensional mechanics, kinematics, and morphology of the knee viewed in virtual reality. Journal of Bone and Joint Surgery (American) 87, 71–80. Fernandez, J.W., Akbarshahi, M., Kim, H.J., Pandy, M.G., 2008. Integrating modelling, motion capture and X-ray fluoroscopy to investigate patellofemoral function during dynamic activity. Computer Methods in Biomechanics and Biomedical Engineering 11, 41–53. Fregly, B.J., Rahman, H.A., Banks, S.A., 2005. Theoretical accuracy of model-based shape matching for measuring natural knee kinematics with single-plane fluoroscopy. Journal of Biomechanical Engineering 127, 692–699.
1301
Fuller, J., Liu, L.-J., Murphy, M.C., Mann, R.W., 1997. A comparison of lowerextremity skeletal kinematics measured using skin- and pin-mounted markers. Human Movement Science 16, 219–242. Garling, E.H., Kaptein, B.L., Mertens, B., Barendregt, W., Veeger, H.E.J., Nelissen, R.G.H.H., Valstar, E.R., 2007. Soft-tissue artefact assessment during step-up using fluoroscopy and skin-mounted markers. Journal of Biomechanics 40, S18–S24. Grood, E., Suntay, W., 1983. A joint coordinate system for the clinical description motions: application to the knee. Journal of Biomechanical Engineering 105, 136–144. Holden, J.P., Orsini, J.A., Siegel, K.L., Kepple, T.M., Gerber, L.H., Stanhope, S.J., 1997. Surface movement errors in shank kinematics and knee kinetics during gait. Gait & Posture 5, 217–227. Manal, K., McClay, I., Stanhope, S., Richards, J., Galinat, B., 2000. Comparison of surface mounted markers and attachment methods in estimating tibial rotations during walking: an in vivo study. Gait & Posture 11, 38–45. Moro-oka, T.-a, Hamai, Satoshi, Miura, Hiromasa, Shimoto, Takeshi, Higaki, Hidehiko, Fregly, Benjamin J., Iwamoto, Yukihide, Banks, Scott A., 2007. Can magnetic resonance imaging-derived bone models be used for accurate motion measurement with single-plane three-dimensional shape registration? Journal of Orthopaedic Research 25, 867–872 Pandy, M.G., 2001. Computer modeling and simulation of human movement. Annual Review of Biomedical Engineering 3, 245–273. Reinschmidt, C., van den Bogert, A.J., Nigg, B.M., Lundberg, A., Murphy, N., 1997. Effect of skin movement on the analysis of skeletal knee joint motion during running. Journal of Biomechanics 30, 729–732. Sangeux, M., Marin, F., Charleux, F., Durselen, L., Ho Ba, Tho, M.C., 2006. Quantification of the 3D relative movement of external marker sets vs. bones based on magnetic resonance imaging. Clinical Biomechanics 21, 984–991. Sati, M., de Guise, J.A., Larouche, S., Drouin, G., 1996. Quantitative assessment of skin-bone movement at the knee. The Knee 3, 121–138. Stagni, R., Fantozzi, S., Cappello, A., Leardini, A., 2005. Quantification of soft tissue artefact in motion analysis by combining 3D fluoroscopy and stereophotogrammetry: a study on two subjects. Clinical Biomechanics 20, 320–329. Sudhoff, I., Van Driessche, S., Laporte, S., de Guise, J.A., Skalli, W., 2007. Comparing three attachment systems used to determine knee kinematics during gait. Gait Posture 25, 533–543. Tashman, S., 2008. Comments on ‘‘validation of a non-invasive fluoroscopic imaging technique for the measurement of dynamic knee joint motion’’. Journal of Biomechanics 41, 3290–3291. Tashman, S., Anderst, W., 2002. Skin motion artifacts at the knee during impact movements. In: Proceedings of the 7th Annual Meeting of Gait and Clinical Movement Analysis Society, Chattanooga. Tashman, S., Anderst, W., 2003. In-vivo measurement of dynamic joint motion using high speed biplane radiography and CT: application to canine ACL deficiency. Journal of Biomechanical Engineering 125, 238–245. Tsai, T.Y., Lu, T.W., Kuo, M.Y., Hsu, H.C., 2009. Quantification of three-dimensional movement of skin markers relative to the underlying bones during functional activities. Biomedical Engineering: Applications, Basis and Communications 21, 223–232. Wretenberg, P., Nemeth, G., Lamontagne, M., Lundin, B., 1996. Passive knee muscle moment arms measured in vivo with MRI. Clinical Biomechanics 11, 439–446.