Validation Of Webcam Based Markerless Motion Capture Measurement Of Knee Flexion Angle During Jump Landings

Traditional multi-camera motion capture systems are often inaccessible due to high costs and operational complexity, limiting their use outside specialized labs. Single webcam-based motion capture offers a portable cost-effective alternative, allowing widespread use. PURPOSE: To validate an accessible, open-source webcam-based motion capture system for assessment of peak knee flexion angle (KFA) during jump landings. METHODS: 23 female Division I athletes (age: 20.4 ± 1.2 years, height: 173.2 ± 9.6 cm, mass; 74.0 ± 11.6 kg) performed 5 drop vertical jump (DVJ) landings. Kinematics were recorded concurrently with a single, frontal plane, webcam and a reference eight-camera markerless motion capture system. Both systems use deep learning algorithms to identify key anatomical landmarks for inverse kinematic calculation of KFA. The webcam-based system used automated in-house Python code, which processed webcam video by first scaling participant segments, accounting for global position, then applying vector analysis to estimate KFA for each frame. Peak KFA of each leg was extracted and averaged across trials, for each system. To adjust for systematic error, we fit a linear correction equation and performed a five-fold cross validation. To ensure generalizability, the held out corrected KFA measurements, from each split, were used for validity testing. Before and after correction, KFA was assessed using intraclass correlation coefficients (ICC) for absolute agreement (2,k) and consistency (3,k), and the standard error of the measure (SEM). RESULTS: Webcam-based uncorrected KFA demonstrated moderate agreement and high consistency (ICC2,k = 0.58-0.59, p > 0.05; ICC3,k = 0.91-0.93, p < 0.05), with SEM above five degrees (5.74 - 6.16°) for each leg. After correcting for this systematic error, corrected KFA demonstrated excellent agreement and consistency for each leg (ICC2,k = 0.97, p < 0.05; ICC3,k = 0.96-0.97, p < 0.05), with SEM below five degrees (2.86 - 3.44°). CONCLUSION: Single webcam-based motion capture, paired with vector analysis and a correction equation, provides accurate and reliable knee flexion angle measurements during jump landings. This automated system offers a practical, accessible alternative to traditional lab-based motion capture. Supported by: ChatGPT was used to enhance readability and language of this abstract.