Quantification of the errors associated with marker occlusion in stereophotogrammetric systems and implications on gait analysis

Optoelectronic stereophotogrammetric systems (OSSs) represent the standard for gait analysis. Despite widespread, their reported accuracy in nominal working conditions shows a variability of several orders of magnitude, ranging from few microns to several millimetres. No clear explanation for this variability has been provided yet. We hypothesized that this reflects an error affecting OSS outcomes when some of the tracked markers are totally or partially occluded. The aim of this paper is to quantify this error in static and dynamic conditions, also distinguishing between total and partial marker occlusion. A Vicon system featuring 8 cameras is employed in this study. Two camera distributions, one designed to maximize OSS accuracy and another one representative of a typical gait setup, are investigated. For both the setups, static and dynamic tests are performed, evaluating the different impact of partial and total marker occlusions. Marker occlusions significantly affected the system performances. The maximum measure variation reached 1.86 mm and 7.20 mm in static and dynamic conditions, respectively, both obtained in the case of partial occlusion. This systematic source of error is likely to affect gait measures: markers placed on the patient body are often visible only by half of the cameras, with swinging arms and legs providing moving occlusions. The maximum error observed in this study can potentially affect the kinematics outcomes of conventional gait models, particularly on frontal and coronal plane, and consequently the peak muscle forces estimated with musculoskeletal models.