Answers are grounded in this paper and cite their sources.
Artificial intelligence (AI)-enabled markerless motion capture (MMC) estimates gait-related measures from video and may make repeated lower-limb assessment more accessible. Its validity depends on the task, output, population, acquisition protocol, software version, reference method, and intended use. This narrative review uses a fit-for-purpose validation framework and a descriptive task-by-output evidence overview to distinguish measurement evidence from clinical utility. Selected spatiotemporal measures and sagittal-plane kinematics during standardized gait have the most consistent support. A published meta-analysis reported a pooled inter-rater reliability intraclass correlation coefficient (ICC) of 0.97 (95% confidence interval (CI) 0.93–0.99) for walking speed. In the original OpenCap validation with 10 healthy participants, joint-angle mean absolute error averaged 4.5° across four tasks. These results describe specific studies and do not establish interchangeability or clinical benefit. Frontal/transverse kinematics, fast or occluded tasks, and kinetic or tissue-loading estimates remain more dependent on acquisition and modeling assumptions. Clinician-assisted monitoring is a more defensible near-term application than standalone diagnosis, referral, or return-to-sport clearance. Supervised feedback means clinician-shown visual or numerical information; feasibility and measurement studies do not establish that it improves rehabilitation outcomes. Higher-consequence use requires external validation and prospective evaluation of decision impact.
No takes yet. Share an insight, caveat, or question.
Han et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: