Accuracy, repeatability and sensitivity of IMU based motion capture systems

This study aims to evaluate an inertial measurement unit (IMU)-based motion capture system for lower body gait analysis on treadmills in terms of its accuracy, repeatability and sensitivity. The Xsens MVN BIOMECH is a popular inertial sensor-based motion capture system widely used by the gait community. However, there is insufficient information regarding its validation for use in gait. Accuracy of pelvic and lower body segments was evaluated with respect to a PhaseSpace Motion Capture System with a thirteen-camera setup. RMS errors for joint angles were evaluated at gait speeds of 1 m/s and 0.5 m/s. Repeatability was evaluated using the Coefficient of Multiple Correlation (CMC) on two different days. Sensitivity of the IMU-based motion capture (mo-cap) system was analyzed based on its ability to distinguish between gait for 1 m/s (symmetric) and 0.95 m/s, 0.9 m/s, 0.85 m/s and 0.8 m/s (asymmetric) conditions. Data from 10 healthy, young individuals were collected and analyzed. We found that the IMU-based system demonstrates reasonably high accuracy when measuring joint angles (0.47 to 3.9 degrees). Accuracy was affected by speed with higher accuracy at lower speeds, especially for Ankle dorsi/plantarflexion. Repeatability was established, with high CMC values (0.76 to 0.98), lower than similar previous studies. Gait cycles were treated as coherent entities and the ability to distinguish between small changes (0.05 m/s) was demonstrated. Sensitivity of gait cycles were compared using High Dimensional Analysis of Variance and the Adaptive Neyman test both for groups and individuals. Smaller differences can be detected at the individual level since gait can vary considerably across individuals. A representative case demonstrated significant differences (p<0.001) for Ankle dorsi/plantarflexion and Hip abduction/adduction on a comparison between 1 m/s (symmetric) and 0.95 m/s (asymmetric). These comparisons are strongly affected by calibration repeatability and more research in this domain is recommended. The IMU motion capture system demonstrates high sensitivity with the ability to differentiate small changes (0.05 m/s at 99.75% confidence) in gait.