Abstract

Introduction: The study validates a novel system, FlexiGait-P1, a single-camera, computer vision-based gait analysis tool, against the gold-standard 3D motion capture system, Vicon, by evaluating its accuracy in extracting spatiotemporal parameters.

Methods: Data from 31 participants in the GPJATK dataset were analyzed. The FlexiGait-P1 system used video-based pose estimation to calculate key gait parameters, including stance, swing, step, and stride times. These were compared to Vicon outputs using statistical analyses, including mean absolute errors (MAE), intraclass correlation coefficients (ICC), and Bland-Altman plots.

Results: FlexiGait-P1 demonstrated high accuracy, with MAEs for joint kinematics below 5° and strong correlations (ICC > 0.90) for most temporal parameters. Stride time (ICC = 0.97) and cadence (ICC = 0.96) showed excellent agreement, while Bland-Altman analysis indicated minimal bias across parameters (e.g., mean difference ~0.01 seconds for step time). Deviations in stance and swing phases were within acceptable ranges, confirming robust agreement with Vicon's outputs. Although swing phase reliability was moderate (ICC = 0.72), stride time and other metrics consistently exhibited near-perfect agreement.

Conclusions: The study shows that FlexiGait-P1 provides a cost-effective, portable alternative to traditional motion capture systems, achieving robust accuracy for clinical and research applications. Its ease of use and reliability position it as a transformative tool for broader adoption in diverse settings.

Demo Videos