Feasibility of Patient-Uploaded Videos for Gait Assessment in Multiple Sclerosis

Background: Gait impairment is common among people with multiple sclerosis (PwMS) and is an important marker of disease progression. However, gait assessments typically require in-person evaluations. Objective: To describe the pose-estimation-based method for estimating spatiotemporal gait parameters from a single consumer-grade video, and evaluate the feasibility of home video collection by PwMS. Methods: In a single-center longitudinal digital phenotyping study, ambulatory adults with MS completed a standardized walking task recorded in the frontal plane. Pose estimation (MediaPipe Pose, Ultralytics) and custom scripts were used to estimate gait parameters from videos. Participants were invited to record walking videos at home using personal devices. Adoption and technical feasibility were evaluated across two home video data acquisition phases, with iterative protocol refinements. Results: The in-clinic study included 132 participants; 55 contributed home videos. In Phase I, while home video adoption was low (45% [30/66]), 87% [26/30] uploaded [≥]1 video of sufficient quality for gait analysis. After protocol refinements, 100% [25/25] uploaded [≥]1 high-quality video. Overall, high-quality frontal-plane videos were obtained at similar rates at home (92% [97/105]) and in-clinic (91% [423/467]). Conclusions: Home walking videos can feasibly be collected by PwMS to estimate gait parameters, providing an accessible approach for remote gait monitoring.