Effect of the calibration procedure of an optoelectronic system on the joint kinematics

Abstract

Optoelectronic systems are largely employed for human movement analysis, where marker trajectories are used to estimate the articular joint kinematics. From a literature analysis it emerged that the error associated to the joint kinematics can be reduced performing the data collection in the center of the system calibration volume. According to human movement analysis literature, the foot-ankle complex appears to be the anatomical joint most affected by instrument inaccuracy, as it moves in the lower bound of the calibration volume during the gait cycle. A multi-segment marker-based model of the lower limb - including the pelvis, thigh, tibia, hindfoot, forefoot and hallux - was investigated in this paper. One healthy subject was asked to walk on the central and on two boundary areas of the capture volume calibrated for the experiments. The calibration procedure was focused on the exploitation of the effects on the joint angles of: (i) calibration volumes (i.e. the global one and two of its sub-volumes) and (ii) number of frames acquired for the calibration procedure (refinement frames). In order to quantify the precision of estimating the joint kinematics when changing the calibration procedure, the RMSE among different refinement frames using both the global volume and the two sub-volumes was computed as an index of the joint angles variation estimated on the sagittal plane. Two two-way ANOVAs were performed to evaluate whether the calibration volumes or the walking areas affect the kinematics. The statistical analysis highlighted a good robustness of the reconstruction algorithm implemented by the optoelectronic system manufacturer. Few variables showed significant differences for the RMSEs, with p-values lower than 0.05. No clear dependence on the body segments here analyzed emerged from the analysis. The coefficient of Multiple Correlations was computed in order to enlighten the similarities among the joint angles time patterns. We conclude that reconstructed trajectories can be affected by the same magnitude errors, regardless to the calibrated volume or the considered walking area. This finding allows conducting the gait analysis without paying attention when calibrating the system and without having to impose excessive restrictions to the tested subjects, allowing to keep their movement as natural as possible.