Elsevier

Journal of Biomechanics

Volume 31, Issue 11, November 1998, Pages 977-984
Journal of Biomechanics

Skin movement artefact assessment and compensation in the estimation of knee-joint kinematics

https://doi.org/10.1016/S0021-9290(98)00083-9Get rights and content

Abstract

In three dimensional (3-D) human movement analysis using close-range photogrammetry, surface marker clusters deform and rigidly move relative to the underlying bone. This introduces an important artefact (skin movement artefact) which propagates to bone position and orientation and joint kinematics estimates. This occurs to the extent that those joint attitude components that undergo small variations result in totally unreliable values. This paper presents an experimental and analytical procedure, to be included in a subject-specific movement analysis protocol, which allows for the assessment of skin movement artefacts and, based on this knowledge, for their compensation. The effectiveness of this procedure was verified with reference to knee-joint kinematics and to the artefacts caused by the hip movements on markers located on the thigh surface. Quantitative validation was achieved through experimental paradigms whereby prior reliable information on the target joint kinematics was available. When position and orientation of bones were determined during the execution of a motor task, using a least-squares optimal estimator, but the rigid artefactual marker cluster movement was not dealt with, then knee joint translations and rotations were affected by root mean square errors (r.m.s.) up to 14 mm and 6°, respectively. When the rigid artefactual movement was also compensated for, then r.m.s errors were reduced to less than 4 mm and 3°, respectively. In addition, errors originally strongly correlated with hip rotations, after compensation, lost this correlation.

Introduction

In non-invasive three dimensional (3-D) motion analysis using stereophotogrammetry, three or more markers are located, directly or through some kind of fixture, on the skin surface of the body segment of interest. Marker trajectories are recorded while the subject performs a given motor task and relevant laboratory coordinates are estimated in each sampled instant of time. From them, the position vector and orientation matrix (pose) of a frame associated with the underlying bone (BTF) are estimated. Thereafter, by adding information concerning the vectors w of the bony ALs of interest, the movement of which could not be observed directly, the instantaneous laboratory positions of these points are assessed using vector analysis. Vectors w are usually assumed to be time invariant and determined through an ad hoc experiment referred to as ‘anatomical landmark calibration’ (Cappozzo, 1984; Cappozzo et al., 1995a; Cappello et al., 1997). The trajectories of an adequate number of bony ALs allow for the determination of the pose of a BAF versus time. Using the poses of two adjacent BAFs, six variables are derived which thoroughly and effectively describe the relevant joint kinematics.

Unfortunately, reconstructed marker points cannot be considered stationary with respect to the underlying bone. Relative movements belong to two classes (Cappozzo et al., 1996):

Because of these movements, a reconstructed marker cluster can be thought to be deformable (because of zero mean uncorrelated errors in each direction) and endowed with a rigid motion (associated with correlated errors) with respect to the bone.

High-frequency photogrammetric errors may be dealt with by smoothing. Residual photogrammetric errors and skin movement artefact components, which cause cluster deformation, are dealt with using rigid body pose optimal estimators (Spoor and Veldpaus, 1980; Veldpaus et al., 1988; Söderkvist and Wedin, 1993; Challis, 1994; Challis, 1995; Cheze et al., 1995; Cappozzo et al., 1997). However, the problem of the rigid movement of the cluster with respect to the bone, mainly associated with skin movement artefacts, remains unsolved. This movement, if not dealt with, propagates to the AL laboratory trajectories and causes remarkable negative effects on the accuracy of the involved joint kinematic estimates. This occurs to the extent that those joint movement components that, during the analysed exercise, undergo moderate variations result in totally unreliable values, as will be evidenced below. It can certainly be claimed that this represents one of the most important unsolved problems in in vivo joint kinematics analysis.

The objective of this paper is to contribute to the quantitative formulation and to the solution of the above-mentioned problem by proposing a skin movement artefact compensation method. Examples of application of the method are provided with reference to the estimation of knee-joint kinematics and skin movement artefacts caused by hip-joint rotations. The method was quantitatively validated by using experimental paradigms whereby prior reliable information on the target joint kinematics was available.

Section snippets

Conceptual background of the artefact assessment and compensation method

The artefact-related movement of the estimated BTF with respect to the underlying bone is looked upon as a movement of the ALs relative to the BTF, i.e. as a variation of the ALs calibration parameter w. Thus, as opposed to the classical approach, w is considered to be variant and a method is devised to estimate it for each sampled 3-D pose of the body segments involved during movement.

Let a kinetic chain be made of three bony segments A, B, and C. Each of these is described by a BTF and

Instrumentation and subjects

A passive marker optoelectronic stereophotogrammetric system with four cameras was used (sampling frequency 100 f s−1; measurement volume 2×1.6×0.8 m). A spot check on the system (Cappozzo et al., 1994) gave an average precision of 1.0, 0.5, 1.4 mm, and accuracy of 2.9, 1.6, 3.5 mm on the X, Y (vertical), and Z (towards the cameras) laboratory coordinates, respectively.

Four markers (5 mm diameter hemispherical balls) were located on the pelvis through a rigid plate strapped to it. Five and four

Results and discussion

Fig. 3 depicts an example of the femoral ALs artefactual movements during an AAM as a function of the hip flexion-extension angle only, although they were caused by the simultaneously occurring ab-adduction and internal–external rotation. For all subjects examined, the positions of the LE and ME, relative to the shank BTF, and the FH relative to the EBTF, were found to undergo artefactual displacements from their position during orthostatic posture of 17, 14 and 13 mm, respectively. These

Conclusions

A method for assessing the skin movement artefacts, caused by a joint movement on the surface marker cluster located on an adjacent body segment, has been presented. Knowledge of these artefacts permits for their compensation while estimating the 3-D kinematics of the joint at the other end of the same body segment. In order to achieve this result, the subject is asked to perform an ad hoc movement that, to some extent, makes the entire experimental and data reduction procedure more cumbersome.

Acknowledgements

The experiments on patients referred to in this paper were carried out in the Movement Analysis Laboratory of the Orthopaedic Rizzoli Institute, Bologna. The kind collaboration of Drs Maria Grazia Benedetti, Fabio Catani and Alberto Leardini is gratefully acknowledged. This study was partially financed by MURST.

References (17)

There are more references available in the full text version of this article.

Cited by (156)

  • Gait analysis methodology for the measurement of biomechanical parameters in total knee arthroplasties. A literature review

    2018, Journal of Orthopaedics
    Citation Excerpt :

    Following the appropriate standardized kinematic protocol all researchers proceeded to the static trial calibration procedure.21,28,33,39 The collected data ensure the reproducibility of the measurement procedure and minimize errors reported in the literature,43 regarding video capture of external skin markers. A static trial at the anatomical position was captured to ensure that all segments could be correctly reconstructed, before collection of the dynamic trials’ data that were used at the statistical analysis.

View all citing articles on Scopus
View full text