Joint Moments And Powers Variability Estimated By Markerless Motion Capture System May Affect Clinical Interpretations

BACKGROUND: Artificial intelligence advances have popularized markerless motion capture (ML). When a growing body of literature compared and validated the ML system in gait analysis, gait patterns that affect clinical decision-making require attention. PURPOSE: The study compared joint moment and power variability during running between marker-based (MB) and ML systems. We hypothesized that the ML system led to greater variability than the MB one, and the difference would be more pronounced during the swing phase. METHODS: Sixteen healthy runners participated, running at 3.58 m/s on a treadmill for 2 minutes. Running motions were captured by the same amounts of infrared and high-resolution video cameras, which consist of the MB and ML systems, respectively. Ground reaction forces were collected by an instrumented treadmill. We analyzed the last ten strides’ three-dimensional hip, knee, and ankle joint moments and powers using inverse dynamics. The variability (VAR) was quantified as the summation of distances from the mean value, and then separated into stance and swing phases. A two-way (system×phase) MANOVA with repeated measures analysis was conducted for each plane with an alpha level of .05. RESULTS: Significant system and phase interactions were detected in three planes (see Figure). These showed that differences of VARs in the swing phase were greater than that of stance phase in both systems with main effects that VARs of ML were greater than those of MB. CONCLUSION: Clinically, joint kinetics VARs can be used to evaluate dynamic movement coordination, injury risk, gait transition, etc. The greater VARs estimated by ML during running could be interpreted as greater injury risk due to mechanical inefficiencies or potential neuromuscular imbalances. Using the ML system, clinicians are likelier to overestimate injury risk, musculoskeletal dysfunction, rehabilitation needs, compensatory mechanisms, and the challenges in monitoring treatment progress.