Marker-based motion capture is the gold standard for 3-dimensional gait analysis but is expensive and largely confined to specialized laboratories. Smartphone-based markerless tools such as OpenCap could scale gait assessment, yet their validity and repeatability during treadmill walking and running remain unclear. We evaluated agreement and within-session repeatability between OpenCap and motion capture during treadmill walking (4 km·h-1) and running (8 and 14 km·h-1) in 10 healthy adults, recorded simultaneously with both systems. We computed spatiotemporal variables, sagittal hip/knee/ankle angle waveforms and range of motion, and center-of-mass displacement waveforms and range of motion. Agreement was assessed with Bland-Altman, intraclass correlation coefficients, Pearson r, and root mean square error; repeatability with trial-to-trial SD, coefficient of variation, and waveform variability index (GaitSD). Spatiotemporal metrics showed bias close to 0 and very strong associations (r ≥ .96), with small minimum detectable changes (eg, ≤0.01 m for stride length and ≤0.03 s for temporal variables). Sagittal range of motion showed good-to-excellent reliability (bias: ≈-4.5° to 2.9°, intraclass correlation coefficients = .76-.93). Waveform root mean square error was ≈1° to 4° for joint angles and ≈0.1 to 0.5 cm for center of mass, and statistical parametric mapping indicated only localized differences. Repeatability and GaitSD were comparable between systems across variables and speeds. Under controlled treadmill conditions, OpenCap showed comparable performance and repeatability metrics to motion capture.
Borba et al. (Thu,) studied this question.