Original research Does tester experience influence the reliability with which 3D gait kinematics are collected in healthy adults? Ryan J. Leigh a, * , Michael B. Pohl b , Reed Ferber a, c a Faculty of Kinesiology, Running Injury Clinic, University of Calgary, 2500 University Dr. NW, Calgary, AB T2N 1N4, Canada b Department of Kinesiology and Health Promotion, University of Kentucky, Lexington, KY, USA c Faculty of Nursing, University of Calgary, Calgary, Alberta, Canada article info Article history: Received 12 September 2012 Received in revised form 19 April 2013 Accepted 30 April 2013 Keywords: Gait analysis Kinematics Reliability Walking abstract Objectives: To determine whether tester experience influences the reliability of three-dimensional gait collections. Design: Reliability study. Participants: Ten healthy subjects visited a university gait laboratory on two separate days and under- went a walking gait analysis. During each visit, kinematic data were collected by a biomechanist with 8 years of 3D gait analysis experience (EXP) and a physical therapist with no previous 3D gait analysis experience (NOV). Main outcome measures: Joint kinematic angles were calculated using either a functional or predictive joint identification method. Within-tester and between-tester measures of reliability were determined by calculating the root mean square error (RMS) and coefficient of multiple correlations (CMC). Results: Within-tester RMS and CMC values were not significantly different (P > 0.05) between the EXP and NOV testers using either a functional or predictive joint approach. Within-tester CMC values exceeded 0.90 for both testers across all kinematic variables. Between-tester CMC reliability values were greater than 0.85 for all variables measured. Conclusions: Following basic training, a physiotherapy clinician with no previous 3D gait experience is as reliable as an experienced gait biomechanist with respect to marker placement accuracy. In addition, reliability comparisons between an experienced and novice tester appear independent of the joint identification method chosen. Ó 2013 Elsevier Ltd. All rights reserved. 1. Introduction Physiotherapists, athletic therapists, and sport biomechanists regularly work with patients and athletes who present with lower extremity ailments. Establishing an effective treatment plan to help manage these injuries often requires that the clinician identify the underlying altered movement patterns and/or aberrant movement biomechanics (McGinley, Baker, Wolfe, & Morris, 2009). Three- dimensional gait analysis (3DGA) is an effective means to mea- sure such movement dysfunctions in a quantitative manner and thus may be an effective tool for the practicing clinician to add to their clinical repertoire. Traditionally, 3DGA has been used in rehabilitation research set- tings to gather quantitative information on the mechanics of the musculoskeletal system during dynamic activities such as walking, running, and functional tasks (Cappozzo, Della Croce, Leardini, & Chiari, 2005; Pohl, Lloyd, & Ferber, 2010). This information is then often used to facilitate a better understanding of several clinical conditions including osteoarthritis, running related injuries, and other neuromuscular conditions such as cerebral palsy and stroke (Barton, Levinger, Menz, & Webster, 2009; Opheim, McGinley, Olsson, Stanghelle, & Jahnsen, 2012). Recently, 3DGA has been adopted by physiotherapists working in musculoskeletal private practice clinics across North America (http://www.3dgaitanalysis.com). It is esti- mated that approximately 28 clinics across North America are now using 3DGA within their private practice to help facilitate clinical decision making. The usefulness with which 3DGA can be used as a measurement tool, either in the research or clinical setting, depends largely in part on the reliability of the measurements obtained from the motion capture system itself (McGinley et al., 2009). To date, several studies have examined the topic of 3DGA reliability and whether 3D motion data can be collected reliably within and * Corresponding author. Tel.: þ1 403 210 7091; fax: þ1 403 220 0546. E-mail addresses: [email protected](R.J. Leigh), [email protected](M.B. Pohl), [email protected](R. Ferber). Contents lists available at SciVerse ScienceDirect Physical Therapy in Sport journal homepage: www.elsevier.com/ptsp 1466-853X/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ptsp.2013.04.003 Physical Therapy in Sport 15 (2014) 112e116
Transcript
at SciVerse ScienceDirect
Physical Therapy in Sport 15 (2014) 112e116
Contents lists available
Physical Therapy in Sport
journal homepage: www.elsevier .com/ptsp
Original research
Does tester experience influence the reliability with which 3D gaitkinematics are collected in healthy adults?
Ryan J. Leigh a,*, Michael B. Pohl b, Reed Ferber a,c
a Faculty of Kinesiology, Running Injury Clinic, University of Calgary, 2500 University Dr. NW, Calgary, AB T2N 1N4, CanadabDepartment of Kinesiology and Health Promotion, University of Kentucky, Lexington, KY, USAc Faculty of Nursing, University of Calgary, Calgary, Alberta, Canada
a r t i c l e i n f o
Article history:Received 12 September 2012Received in revised form19 April 2013Accepted 30 April 2013
1466-853X/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.ptsp.2013.04.003
a b s t r a c t
Objectives: To determine whether tester experience influences the reliability of three-dimensional gaitcollections.Design: Reliability study.Participants: Ten healthy subjects visited a university gait laboratory on two separate days and under-went a walking gait analysis. During each visit, kinematic data were collected by a biomechanist with 8years of 3D gait analysis experience (EXP) and a physical therapist with no previous 3D gait analysisexperience (NOV).Main outcome measures: Joint kinematic angles were calculated using either a functional or predictivejoint identification method. Within-tester and between-tester measures of reliability were determinedby calculating the root mean square error (RMS) and coefficient of multiple correlations (CMC).Results: Within-tester RMS and CMC values were not significantly different (P > 0.05) between the EXPand NOV testers using either a functional or predictive joint approach. Within-tester CMC valuesexceeded 0.90 for both testers across all kinematic variables. Between-tester CMC reliability values weregreater than 0.85 for all variables measured.Conclusions: Following basic training, a physiotherapy clinician with no previous 3D gait experience is asreliable as an experienced gait biomechanist with respect to marker placement accuracy. In addition,reliability comparisons between an experienced and novice tester appear independent of the jointidentification method chosen.
� 2013 Elsevier Ltd. All rights reserved.
1. Introduction
Physiotherapists, athletic therapists, and sport biomechanistsregularly work with patients and athletes who present with lowerextremity ailments. Establishing an effective treatment plan to helpmanage these injuries often requires that the clinician identify theunderlying altered movement patterns and/or aberrant movementbiomechanics (McGinley, Baker, Wolfe, & Morris, 2009). Three-dimensional gait analysis (3DGA) is an effective means to mea-sure such movement dysfunctions in a quantitative manner andthus may be an effective tool for the practicing clinician to add totheir clinical repertoire.
Traditionally, 3DGA has been used in rehabilitation research set-tings to gather quantitative information on the mechanics of the
musculoskeletal system during dynamic activities such as walking,running, and functional tasks (Cappozzo, Della Croce, Leardini, &Chiari, 2005; Pohl, Lloyd, & Ferber, 2010). This information is thenoften used to facilitate a better understanding of several clinicalconditions including osteoarthritis, running related injuries, andother neuromuscular conditions such as cerebral palsy and stroke(Barton, Levinger,Menz, &Webster, 2009;Opheim,McGinley, Olsson,Stanghelle, & Jahnsen, 2012). Recently, 3DGA has been adopted byphysiotherapists working in musculoskeletal private practice clinicsacross North America (http://www.3dgaitanalysis.com). It is esti-mated that approximately 28 clinics across North America are nowusing 3DGA within their private practice to help facilitate clinicaldecision making.
The usefulness with which 3DGA can be used as a measurementtool, either in the research or clinical setting, depends largely inpart on the reliability of the measurements obtained from themotion capture system itself (McGinley et al., 2009). To date,several studies have examined the topic of 3DGA reliability andwhether 3D motion data can be collected reliably within and
R.J. Leigh et al. / Physical Therapy in Sport 15 (2014) 112e116 113
between days during walking and running (Besier, Sturnieks,Alderson, & Lloyd, 2003; Pohl et al., 2010; Wilken, Rodriguez,Brawner, & Darter, 2012). These studies collectively found thatgait kinematic data could be acquired with good to excellent reli-ability across both walking and running conditions in healthy adultsubjects. In the reliability studies published by Besier et al. (2003)and Pohl et al. (2010), the authors also sought to determinewhether gait kinematic reliability was different when using either apredictive (MAN) or functional (FUN) approach to determine jointcentres and joint axes of rotation. The predictive approach calcu-lates joint centres and anatomical coordinate systems based solelyon the placement of anatomical markers (Bell, Brand, & Pederson,1989). The functional approach, in which the joint centre/axesand anatomical coordinate systems are calculated using functionalmovement tasks (Schwartz & Rozumalski, 2005), is thus lessdependent on the placement of anatomical markers. Given thatmarker placement inaccuracy is thought to be one of the leadingcauses of error in 3DGA (Della Croce, Cappozzo, & Kerrigan, 1999),an advantage of the functional method over the predictive method,was thought to lie in its decreased reliance on the reliable place-ment of anatomical markers (Leardini et al., 1999). Despite thesuggested advantage for the functional method, Pohl et al. (2010)found no improvement in within-tester (between-day) orbetween-tester reliability when comparing the FUN and MANtechniques. Similarly, Besier et al. (2003) found that only frontalplane knee kinematic reliability was improved using the FUNmethod compared to the MAN method. While good to excellentgait kinematic reliability was demonstrated in all of the abovestudies, and achieving excellent reliability does not appear todepend on the joint centre technique used, both authorsacknowledged that they used experienced testers and that testerswith limited 3DGA experience may not demonstrate the samereliability as those with 3DGA experience.
Given the increased use of 3DGA by clinicians within the clinicalenvironment, and the suggestion that tester experience may play arole in determining 3DGA reliability, the purpose of the presentstudy is to determine whether a clinician with no 3DGA experiencecan collect 3DGA data as reliably as a biomechanist experienced in3DGA collections. A secondary purpose is to determinewhether thechosen joint centre methodology (FUN or MAN) influences thisreliability comparison. It was hypothesized that: 1) the testerexperienced in 3DGA would exhibit improved within-tester reli-ability compared to the clinician when a MAN technique was usedgiven previous experience with gait analysis specific anatomicalland-marking; 2) within-tester (between-day) reliability would besimilar between the 3DGA tester and the clinician when using theFUN approach.
2. Methods
A convenience sample of ten (six females, four males) healthyindividuals (age ¼ 22.5 � 2.8 years; mass ¼ 65.2 � 13.7 kg;height¼ 1.73� 0.11m; BMI¼ 18.9� 0.6) volunteered to participatein the study. The inclusion criteria required that subjects werecurrently free from lower extremity injury and familiar withwalking on a treadmill. Subjects were excluded if they had a historyof major lower extremity surgery, or if they had experienced lowerextremity musculoskeletal pain in the six weeks prior to the study.All subjects provided written informed consent for the study thatreceived approval from the Institutional Review Board.
Prior to commencement of the study data collections, a four parttraining session was conducted by the experienced biomechanisttester (8 years’ of 3D gait analysis experience). The purpose of thetraining session was to train the novice tester (a physical therapistwith 4 years clinical experience but no 3D gait analysis experience)
on marker location placement used during motion analysis. Thefirst two training sessions involved the experienced tester (EXP)describing and placing markers on a pilot subject while the clini-cian untrained in gait analysis (NOV) observed. Each of the first twotraining sessions lasted 45 min. The rationale for having the NOVtester observe the first two training sessions was to orient the NOVtester to the general protocol of motion capture marker set-up.Session three involved the NOV tester placing markers under theEXP testers’ direct supervision while allowing the NOV tester toseek advice when needed. In the final training session, the NOVtester placed the markers independently of the EXP testers’ feed-back until after all markers had been placed, at which time the EXPtester provided feedback. The final two training sessions took60 min given that the NOV tester was entirely responsible forsystem set-up and subject marker placement. The gait analysissystem was calibrated before each testing session by a lab techni-cian working within the clinic. It was established a priori wherecameras should best be placed and their positions remained unal-tered between testing sessions.
Subjects visited the lab on two occasions separated by a mini-mum of two days. During each visit, subjects underwent twoseparate 3D gait analyses; one of these gait analyses was conductedby the EXP tester while the other test was conducted by the NOVtester. Each of these 3D gait analyses were separated by 30min. TheEXP and NOV testers alternated who tested first during each sub-jects visit to the lab on day one. For each gait analysis, 9 mmspherical markers were placed on the pelvis and right lower ex-tremity. Anatomical markers and technical marker clusters wereplaced in the manner described by Pohl et al. (2010). All subjectswore standard laboratory shoes (Nike Air Pegasus, Nike Inc.) foreach of the testing sessions.
Eight Vicon cameras (Vicon, Oxford, UK) were used to collectmarker co-ordinate data at 120 Hz. Prior to the motion trials, astanding calibration trial was performed while subjects stood in astandardized position with their feet positioned 0.3 m apart andpointing straight ahead. Following the standing trial, subjects per-formed separate functional movements of the hip and knee whichwere subsequently used to determine functional joint centres ofrotation (Pohl et al., 2010). Subjects thenwalked on the treadmill at1.1 m/s and kinematic data for five complete gait cycles of the rightlimbwere collected following a brief accommodation period (Fellin,Rose, Royer, & Davis, 2009). The use of a treadmill was requiredgiven the space restrictions of the laboratory. Care was taken toensure that any residual marker placement markings were notvisible to the tester performing the second gait analysis.
Visual 3D software (C-Motion Inc., Germantown, USA) was usedfor filtering co-ordinate data, identifying functional joint centresand axes of rotation, and performing joint angle calculations. Alldata processing was performed by a separate individual who wasblinded to each tester. Three-dimensional marker co-ordinate datawere filtered at 10 Hz using a fourth order Butterworth filter. Twocustom models (MAN and FUN) were created based on manualmarker placement only (MAN) and functional joint methodology(FUN), respectively. Both technical (TCS) and anatomical co-ordinate systems (ACS) were defined for the pelvis, thigh, shank,and foot for both the MAN and FUNmodels. For a definition of bothan anatomical (ACS) and technical (TCS) frame, please refer toCappozzo, Catani, Croce, and Leardini (1995). In the MAN model,the hip joint centre was determined as per the techniquesdescribed by Bell, Pedersen, and Brand (1990). The FUNmodel usedfunctional techniques to determine both the hip and knee jointcenter/axis of rotation as described by Schwartz and Rozumalski(2005). A comprehensive description of the individual segmentalACS’s for each model together with the calculation of the three-dimensional joint angles at the hip, knee, and ankle has been
Fig. 2. Within-tester ensemble mean (SD) RMS errors for all subjects using the MAN(top graph) and FUN (bottom graph) approach. EXP; (Experienced tester), NOV;(Novice tester).
R.J. Leigh et al. / Physical Therapy in Sport 15 (2014) 112e116114
reported elsewhere (Pohl et al., 2010). Kinematic data were thenanalysed for the gait cycle and normalized to 101 data points. Timenormalization was necessary to enable the coefficient of multiplecorrelation (CMC) analysis described below.
Two statistical measures of reliability were used to comparedifferent aspects of the kinematic curves and thus calculate thewithin-tester (between-day) and between-tester reliability. TheCMC was used to compare the overall shape of the kinematic curveand was calculated as the average value subtracted from each curve(Growney, Meglan, Johnson, Cahalan, & An, 1997). A second mea-sure of reliability, root mean square (RMS) error, was used to esti-mate the kinematic offset between curves in the FUN and MANconditions separately between sessions. Within-tester (between-day) reliability was calculated for each tester, using both a FUN andMAN approach, by comparing the two visits made by each subjecton separate days. The experienced tester’s within-tester reliabilityindices were then statistically compared to the indices of the novicetester in the MAN and FUN conditions separately using Wilcoxonsigned rank tests (Fig. 1). Non-parametric Wilcoxon signed ranktests were chosen given the small sample size and given thatnormality could not be assumed (Portney & Watkins, 2009).Between-tester reliability was calculated by using the datacollected by both testers on day 2 only. Between-tester reliabilityusing the MAN approach was then statistically compared to theFUN approach using Wilcoxon signed rank tests (Fig. 1). Statisticalsignificance was set at P < 0.05 and statistical analyses were per-formed using SPSS 19 (SPSS Inc., Chicago, USA).
3. Results
3.1. Within-tester reliability
Within-tester reliability of the NOV tester was not significantlydifferent (P > 0.05) from the within-tester reliability of the EXPtester when each tester’s RMS error (Fig. 2) and CMC reliabilitycoefficients were compared (Table 1). This similarity in reliabilitybetween the NOV and EXP testers was seen in both the FUN andMAN conditions. Within-tester CMC values exceeded 0.90 for bothtesters in all kinematic variables measured and within-tester RMS
Fig. 1. Schematic of the testing protocol for the within (left figure) and between-tester (right figure) comparisons. (4 ¼ comparison by Wilcoxon Signed Rank Test).
Table 1Within-tester (between-day) mean (SD) CMC values for the experienced (EXP) andnovice (NOV) testers. MAN (manual); FUN (functional).
R.J. Leigh et al. / Physical Therapy in Sport 15 (2014) 112e116 115
error was less than 5� for both testers. The largest difference in RMSerror values between the EXP and NOV testers was 1.3� (ankle in-ternal/external rotation) using the MAN technique and 1.3� (ankleinternal/external rotation) in the FUN technique.
3.2. Between-tester reliability
Between-tester RMS error was significantly lower (P < 0.05) inthe FUN condition compared to the MAN method for knee flexion/extension (difference 2.5�) and hip abduction/adduction (differ-ence 0.6�) (Fig. 3). Between-tester CMC values exceeded 0.87 acrossall kinematic variables measured using either a FUN or MANapproach (Table 2). Between-tester CMC values were significantlydifferent (P < 0.05) for knee internal/external rotation (FUN; 0.93,MAN 0.88), hip flexion/extension (MAN; 0.97, FUN; 0.97) and kneeflexion/extension (MAN; 0.97, FUN; 0.96) (Table 2).
4. Discussion
The purpose of the present study was to determine whether aclinician, with no previous 3DGA experience, could collect gait ki-nematic data as reliably as a tester experienced in 3DGA collections.A further purposewas to determine if the joint centremethodologychosen influenced this reliability. It was hypothesized that thetester experienced in 3D gait analysis and the clinician untrained in3D gait analysis would demonstrate similar within-tester reliabilityusing a functional joint center technique (FUN) but dissimilar
Fig. 3. Between-tester ensemble mean (SD) RMS errors for all subjects in the FUN andMAN approach. MAN (manual); FUN (functional). *P < 0.05.
reliabilities when utilizing the more marker dependent predictivemethod (MAN).
Our results suggest that a physiotherapy clinician previouslyuntrained in 3D gait analysis marker placement can demonstratethe same within-tester reliability as an experienced tester whenmarker placement accuracy is the variable of interest. This is, to ourknowledge, the first study to compare the influence of testerexperience on 3D gait kinematic reliability. The excellent within-tester reliability values obtained by the inexperienced and experi-enced 3D gait analysis testers in the present study are in accordancewith those observed in previous studies in which the kinematicreliability values were good to excellent (Besier et al., 2003; Pohlet al., 2010; Wilken et al., 2012).
The finding that the experienced and inexperienced testersdemonstrated similar within-tester reliability appears irrespectiveof whether a functional or predictive joint centre methodology isused. This finding was surprising given that joint center and axis ofrotation calculations in the predictive approach (MAN) is moreheavily dependent on marker placement accuracy as comparedwith the functional approach (Leardini et al., 1999). With markerplacement inaccuracy thought to be the largest contributor to 3Dgait analysis error (Della Croce et al., 1999), we hypothesized thatthe experienced tester’s additional experience palpating andlocating 3DGA specific landmarks would confer reliability benefitswhen using a predictive approach (MAN). However, while theclinician did not have previous 3DGA marker placement experi-ence, it appears that the clinician’s previous clinical experience (4years) was sufficient to ensure the same reliability as an experi-enced 3DGA tester. The finding that the FUN technique performedno better with respect to reliability than the MAN technique is alsoin accordancewith Pohl et al. (2010) who found no difference whenusing experienced testers.
Similar to the within-tester reliability findings, all between-tester CMC reliability values exceeded 0.870, suggesting excel-lent between-tester reliability when using either a functional ormanual joint methodology approach. Our finding that between-tester reliability was excellent irrespective of the joint centrecalculation technique chosen is in agreement with Pohl et al.(2010) who found that between-tester reliability values, usingeither a functional or manual approach, exceeded 0.90 across mostjoint angles. These findings suggest that reliability is not detri-mentally affected when kinematic data is collected between tes-ters with different levels of experience or when using differentjoint centre calculation methodologies. It should be noted thatstatistically significant between-tester differences in CMC valueswere observed across the knee transverse plane and knee and hipsagittal plane (Table 2). However, given that the CMC valuesexceeded 0.875 in these cases and that the difference in CMCvalues in the transverse and sagittal planes was between 0.002and 0.05, the clinical significance of this difference is questionable.
R.J. Leigh et al. / Physical Therapy in Sport 15 (2014) 112e116116
Similarly, while statistically significant, the clinical significance ofa 0.6 and 2.5� difference in RMS error in hip abduction and kneeflexion respectively between the functional and manual tech-niques is questionable (Fig. 3). A recent study published by Wilkenet al. (2012) found that, in a healthy lean population similar toours, the minimal detectable change in hip frontal plane motionand knee sagittal plane motion is in excess of the 0.6 and 2.5�
difference observed in our study. Given that our RMS error valuesmay be below the minimal detectable change, the importance ofsuch a small difference is questionable.
The results presented herein should be interpreted within thecontext of potential study limitations. The first pertains to theacknowledgement that the testers in the present study were notresponsible for all aspects of the 3DGA collection which includescalibration of the motion analysis system, camera set-up, anatom-ical marker placement, and data modelling. Since these other as-pects, in addition to marker placement, can introduce error into thecollection of repeated gait analyses (Chiari, Della Croce, Leardini, &Cappozzo, 2005), tester experience may play a role in the reli-ability of gait analysis if calibration and analysis is part of thecollection. However, given that anatomical marker placement ac-curacy is considered to be one of the largest contributor of error ingait analysis (Della Croce et al., 1999), we aimed to determine theinfluence of tester experience on 3DGA reliability across this vari-able specifically. The second limitation pertains to the use of onlyone 3D gait analysis experienced tester and one clinician tester.Including more than one experienced and one clinician tester ineach group may have increased the generalisability of the results.Comparing thewithin-tester reliability of an experienced tester to atrue “novice” clinician tester, who has neither 3D gait analysis norclinical experience may have been helpful in further determiningthe influence of tester experience on the reliability of kinematic gaitvariables. However, given that gait analysis is often performed bothin the research and clinical settings bya testerwith prior anatomicalknowledge (e.g. kinesiologists, athletic therapists, physical thera-pists), it seemed more generalizable to select a clinician as opposedto someonewith no anatomical knowledge. A final limitation is thatthe subjects recruited in the present study were all young, leanparticipants. It has been postulated by Besier et al. (2003) thatincreased percent body fat may increase the difficulty with whichanatomical landmarks are identified and thus influence the accu-racy of marker placement. Further study investigating the effect ofadiposity on the within-tester and between-tester reliability of gaitanalysis is therefore needed. Lastly, it will be important to examinewhether the reliability of 3DGA in an untrained clinician is similar tothat of an experienced gait analysis tester when examining subjectswith musculoskeletal impairments or injuries.
5. Conclusion
The results of the present study suggest that a physiotherapyclinician untrained in 3DGA data collection is as reliable as anexperienced 3DGA tester with respect to marker placement accu-racy. In addition, this similarity in reliability is found irrespective ofwhether a functional (FUN) or predictive (MAN) joint methodologyis used to determine joint centres.
Conflict of interestDr. Ferber is a shareholder in 3DGait Analysis Systems Inc. Hewas
involved indeveloping the rationale for this study, posing the research
question, helping interpret the findings, andmaking final edits to themanuscript.Heplayedno role indata collectionordata analysis andassuch had no influence on the results of the study in any manner.
Ethical ApprovalAll subjects provided written informed consent for the study
that received approval from the Institutional Review Board.
FundingThe present study was funded by Alberta Innovates: Health
Solutions.
Acknowledgements
Wewould like to thank Chandra Lloyd and Talia Weber for theirassistance with the data processing.
References
Barton, C. J., Levinger, P., Menz, H. B., & Webster, K. E. (2009). Kinematic gaitcharacteristics associated with patella-femoral pain syndrome: a systematicreview. Gait & Posture, 30, 405e416.
Bell, A. L., Brand, R. A., & Pederson, D. R. (1989). Prediction of hip-joint centrelocation from external landmarks. Human Movement Science, 8, 3e16.
Bell, A. L., Pedersen, D. R., & Brand, R. A. (1990). A comparison of several hip centerlocation prediction methods. Journal of Biomechanics, 23, 617e621.
Besier, T. F., Sturnieks, D. L., Alderson, J. A., & Lloyd, D. G. (2003). Repeatability of gaitdata using a functional hip joint centre and a mean helical knee axis. Journal ofBiomechanics, 36, 1159e1168.
Cappozzo, A., Catani, F., Croce, U. D., & Leardini, A. (1995). Position and orientationin space of bones during movement: anatomical frame definition and deter-mination. Clinical Biomechanics, 10(4), 171e178.
Cappozzo, A., Della Croce, U., Leardini, A., & Chiari, L. (2005). Human movementanalysis using stereophotogrammetry. Part 1: theoretical background. Gait &Posture, 21, 186e196.
Chiari, L., Della Croce, U., Leardini, A., & Cappozzo, A. (2005). Human movementanalysis using stereophtogrammetry. Part 2: instrumental errors. Gait & Posture,21, 197e211.
Della Croce, U., Cappozzo, A., & Kerrigan, D. C. (1999). Pelvis and lower limbanatomical landmark calibration precision and its propagation to bone geom-etry and joint angles.Medical & Biological Engineering & Computing, 37, 155e161.
Fellin, R. E., Rose, W. C., Royer, T. D., & Davis, I. S. (2009). Comparison of methods forkinematic identification of footstrike and toe-off during overground andtreadmill running. Journal of Science and Medicine, 13, 646e650.
Growney, E., Meglan, D., Johnson, M., Cahalan, T., & An, K. N. (1997). Repeatedmeasures of adult normal walking using a video tracking system. Gait & Posture,6, 147e162.
[http://www.3dgaitanalysis.com]Leardini, A., Cappozzo, A., Catani, F., Toksvig-Larsen, S., Petitto, A., Sforza, V., et al.
(1999). Validation of a functional method for the estimation of hip joint centrelocation. Journal of Biomechanics, 32, 99e103.
McGinley, J. L., Baker, R., Wolfe, R., & Morris, M. E. (2009). The reliability of three-dimensional kinematic gait measurements: a systematic review. Gait &Posture, 29, 360e369.
Opheim, A., McGinley, J. L., Olsson, E., Stanghelle, J. K., & Jahnsen, R. (2013). Walkingdeterioration and gait analysis in adults with spastic bilateral cerebral palsy.Gait & Posture, 37(2), 165e171.
Pohl, M. B., Lloyd, C., & Ferber, R. (2010). Can the reliability of three-dimensionalrunning kinematics be improved using functional joint methodology? Gait &Posture, 32, 559e563.
Portney, L. G., & Watkins, M. P. (2009). Nonparametric tests for group comparisons.In L. G. Portney, & M. P. Watkins (Eds.), Foundations of clinical research: Appli-cations to practice (3rd ed.). (pp. 503e522) New Jersey: Pearson Prentice Hall.
Schwartz, M. H., & Rozumalski, A. (2005). A new method for estimating joint pa-rameters from motion data. Journal of Biomechanics, 38, 107e116.
Wilken, J. M., Rodriguez, K. M., Brawner, M., & Darter, B. J. (2012). Reliability andminimal detectable change values for gait kinematic and kinetics in healthyadults. Gait & Posture, 35(2), 301e307.
