It is not just soft tissue artifact (STA) and it is not just the marker set that determines 3DMA validity and reliability of joint angle data. If you interested in measuring STA or correcting errors in joint angle data then there are larger sources of error that need to be addressed first before STA can be considered. 3DMA involves a series of processes, each should be based on a sound understanding of the sources of errors involved and with checks in place to assess and reduce these errors at each step in the process.

The major influences on validity and reliability of joint angle data during gait can be groups as:
1. 3D system (cameras, calibration, volume and reconstruction of 3D marker coordinates)
2. Post processing (tracking, identification, smoothing and gap filling)
3. Defining segment axes location and orientation (functional, regression, optimization)
4. Analytical methods (least squares, multi segment, joint DoF)
5. Skin movement artefact (marker placement, RFD).

Also see link to 3DMA error flow chart:
3DMA Flow Chart.jpg

I have presented data previously comparing the 3DMA reliability literature of traditional minimalist marker based (PiG, HelenHayes, Kit Vaughan, T3Gait), KAD, Optimization methods and rigid fixation devices with and without functional joint centres. The unpublished gait reliability data from the University of Otago is included and is essentially using the guidelines mentioned below.

A summary can be found at:
3DMA Cluster Design.pdf
Summary 3D Reliability.pdf
Results Table Systematic Review brief.xls

I have also presented data previously on normal gait and the widely varying, inconsistent and often unrealistic joint angle data that has been presented in the literature, including marker based, bone pins and radiographic.

Normal gait, axes misalignment and nonlinear error.xls

All studies suffer greatly from axes misalignment errors. The examples demonstrate the influence and correction of axes misalignment and non-linear errors on joint angle data. The examples also show the close agreement in gait joint angle data (that seemed unrelated) that can be obtained by post hoc correction of axes misalignment. Where systematic errors in thigh axes alignment about the longitudinal axes of -12 to +22 degrees were observed across the studies. The examples also shows the relatively small and repeatable influence of soft tissue artefact (STA) makes on gait joint kinematics in marker based studies.
This is the first time that I am aware of that normal gait joint angles have been demonstrated.

From the 3DMA cluster design file, here are some basic guidelines used when placing markers:

  • At least four markers per segment [1,4,5], however six to ten markers has been recommended [6]. These may be real or virtual joint centers, defined relative to adjacent segments.
  • Markers are well distribution along at least two axes [7].
  • Markers are non-symmetrical about two axes [5].
  • Markers avoid areas of high skin movement artefact (proximal 1/3 thigh, Greater trochanter [8, 9], calf musculature [9], anterior forearm, and other large muscle bellies.
  • Do not place rigid fixation devices (RFD) containing fixed markers over large muscle bellies, such as anterior-lateral thigh. You are interested in the movement of the segment underneath and the RFD will only exacerbate skin movement artefact.
  • Placing markers at varying locations so individual markers contain varying motion due to skin movement artefact within the cluster [6]. This may involve placing markers on the anterior, lateral and posterior aspects, placing markers lateral to joints as well as on the body of the segment, and including virtual joint centers.

And some methodological guides:

  • Do not smooth 3D data. It will introduce distortions in 3D trajectories around sudden de- accelerations, particularly foot strike, hitting or striking. In the case of foot strike producing large errors between the foot segment and COP from force platforms. Which in turn produce large oscillations in resultant joint moments and forces immediately prior to and after foot strike.
  • Do not gap fill. The best way to reconstruct a 3D path if required is to reconstruct the segment location from all available markers and use the segment location to fill the missing frames.
  • Use a least squares method [4,5,6] even when using the minimum three markers [4]. The most stable and preferred least square method is that of Veldpaus (1988) [5]. Direct methods and inertial (principal axes) methods that are based purely only on the global marker coordinates are poor by comparison and should not be used to define segment axes location [4-5].
  • Use 6 degrees of freedom joints. However, do include virtual joint centers from proximal and/or distal segments in the least square approach.
  • Do not use a multi-segment least squares or optimization approach with constrained joints (3 DoF). It is the least reliable of 3DMA methods as the model cannot accommodate unavoidable errors in 3D marker location, joint centers and skin movement artefact. Joint translations and changes in joint center of rotation will also adversely affect the model.
  • RMS Errors in reconstructed 3D markers as well as 3D segment location should be assess and reported as part of routine 3DMA. Marker and segment tracking errors are commonly step discontinuities which can be readily identified and corrected using RMS errors on a frame by frame basis. Allowing the identification of the marker in error on a 3D segment or 2D camera coordinate in error comprising a 3D marker location. Do not smooth raw 3D marker data as this will prevent routine identification and correction of errors in 3D markers and least squares reconstruction of segment locations.
  • Assess and correct axes misalignment in the subject calibration procedure by collecting and analyzing joint angle data from a controlled and repeatable movement pattern (squat or gait). This can be used to refine segment axes alignment by reducing non-linear error (cross talk) in non-sagital joint rotations and checking validity of rotations before the analysis of the trials of interest.
  • Do not add or subtract offsets directly from joint angle data. This ignores the source of the error (axes misalignment) and interdependency of joint angle data. Instead correct the segment axes miss-alignment within the subject calibration procedure. This will correct the nonlinear errors propagated through the Cardan rotations of both the proximal and distal joints of the segment.
  • Collect and analyze a controlled and repeatable movement pattern (squat or gait) both pre and post trials of each session. This can be used to assess consistency of methods, axes alignment and calculated joint angle data against known or expected joint rotations.

Of interest is a special issue in the Journal of Biomechanics (Vol 62, 2017) on multi body kinematics optimisation methods (MBO or MKO) with soft tissue artefact (STA) compensation.

What is evident in this issue is the lack of awareness of the multiple factors and their importance in influencing axes misalignment between modelled and underlying segment and on the validity and reliability of joint angle. Misconceptions and poor understandings are evident, including; assuming STA is the dominant source of error in axes misalignment; use of a model that introduces additional errors into the reconstruction of segment locations, is known to lack validity and reliability in joint angle data and is highly dependent on joint DoF, markers used and movement pattern; that bone pins give criterion joint angle data, and; a STA compensation approach that uses a predictive function based on derived joint angle data. Despite the extensive efforts presented in this special issue and in the wider literature related to the MBO model with STA compensation, the method has been unsuccessful and has yet to describe normal gait joint angles, has produced no improvement in validity of knee joint kinematics, and is yet to describe the relatively small and repeatable STA present in gait, which has been perceived as large, varied and inconsistent.

A review can be found here.
JBiomech STA Vol62.pdf

There is a lot of information contained in the links, let me know if you have difficulty accessing the files. This may well raise more questions than answers.


[1] Miller, N.R., Shapiro, R., McLauchlan, T.M. (1980), A technique for obtaining spatial kinematic parameters of segment of biomechanical systems from cinematographic data. JBiomechanics, 13, 535-547.
[2] Veldpaus, F.E., Woltring, H.J., Dortmans, L.j.M.G. (1988) A least squares algorithm for the equiform transformation from spatial marker co-ordinates. J. Biomechanics, 21, 45-54.
[3] Challis, J.H. (1995) A procedure for determining rigid body transformation parameters. J. Biomechanics, 28, 733-737.
[4] Cappozzo, A, Cappello, A, Della Croce, U, Pensalfini, F (1997) Surface marker cluster design criteria for 3-D bone movement reconstruction, IEEE Trans. Biom. Eng., 44, 1165-1174.
[5] Carman, AB, Milburn, PD (2006) Determining rigid body transformation parameters from ill-conditioned spatial marker coordinates. J. Biomechanics, 39, 1778-1786.
[6] Miller, N.R., Shapiro, R., and McLaughlin, T.M. (1980) A technique for obtaining spatial kinematic parameters of segments of biomechanical systems from cinematographic data. Journal of Biomechanics, 13(7), 535-548.
[7] Solderkvist, I, and Wedin, PA (1993) Determining the movement of the skeleton using well-configured markers, J. Biomechanics, 26, 1473-1477.
[8] Cappozzo, A, Catani, F, Leardini, A, Benedetti, MG, Della Croce, U (1996) Position and orientation in space of bones during movement: experimental artifacts, Clinical Biomechancis, 11, 90-100.
[9] Stagni, R, Fantozzi, S, Cappello, A, Leardini, A (2005) Quantification of soft tissue artifact in motion analysis by combining 3D fluoroscopy and sterophotogrammetry: a study on two subjects. Clinical Biomechanics, 20, 320-329.