Clinical Validity of Markerless Motion Capture for Biomechanical Decisions
Year: 2026
Type: article
Abstract: Background: Markerless motion capture can extend quantitative assessment beyond specialist laboratories. Its clinical value depends on whether a specified output supports a defined biomechanical decision. Objective: To examine the evidence linking video-derived measurements to biomechanical interpretation and propose a practical framework for evaluating clinical use. Methods: A critical narrative synthesis combined targeted scholarly searches, reference verification, accessible primary studies, and methodological guidance. Searches were updated on 10 November 2026. Evidence was organised by measurement construct, reference, population, task, and intended interpretation. Selection was purposive; no exhaustive search, pooled accuracy estimate, or formal certainty grade was undertaken. Synthesis: Validation results depend on joint, rotation plane, task, recording conditions, and reference model. Clinical studies show that agreement in average sagittal waveforms can coexist with missed individual abnormalities or gait-classification disagreement. Multicentre group differentiation establishes a different property from individual measurement agreement. Kinetic estimation introduces additional dependence on external forces, contact assumptions, and musculoskeletal model structure. Correlation, repeatability, detectable change, clinically important change, and improved decisions require separate evidence. Conclusion: Markerless motion capture should be evaluated as linked measurements and interpretations. Credible use requires explicit anatomical definitions, reference-aware agreement analysis, representative external evaluation, transparent treatment of failed recordings, and version-controlled longitudinal assessment. The proposed framework identifies evidence requirements without prescribing universal angular-error thresholds or presenting video-derived outputs as direct tissue-load measurements.
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7225822584