Validity of using a smartphone-based markerless motion capture system for quantitative analysis of human dynamic movements

Quantitative motion analysis is often restricted in clinical and sports settings due to high cost and technical requirements of 3D optical systems (MoCap). This study assessed the validity of a smartphone-based markerless system (OpenCap) against MoCap across eight dynamic tasks. Participants (N = 41) performed movements including walking, running, jumping, and cutting, with simultaneous dual-system capture. Lower-extremity joint kinematic agreement was evaluated using discrete metrics (Root-Mean-Square-Error [RMSE], normalised RMSE [NRMSE], Pearson’s r, and Bland-Altman analyses) and continuous Statistical Parametric Mapping (SPM). OpenCap demonstrated strong sagittal-plane agreement (RMSE = 7.0°–13.4°, NRMSE = 7.4%–24.5%, r = 0.70–0.99) and similar SPM waveforms, despite systematically overestimating flexion/dorsiflexion by 5°–15°. Conversely, out-of-plane kinematics exhibited high normalised errors (RMSE = 3.5°–16.6°, NRMSE = 29.0%–136.3%), highly variable correlations (r = –0.09–0.80), and divergent SPM waveforms. Additionally, OpenCap performed generally better during spatially constrained jumping movements than expansive translating tasks. This highlights algorithmic challenges regarding temporal synchronisation and distal tracking when individuals translate rapidly across the cameras’ depth of field. While OpenCap currently lacks the multi-planar precision necessary for absolute out-of-plane evaluation, it successfully captures overarching sagittal trajectories. Ultimately, users must carefully weigh the system’s accessibility advantages against its kinematic limitations before implementation in field or clinical environments.