Two Paths to Markerless Movement Analysis in Clinical Populations: A Critical Review of MediaPipe and OpenCap for Individuals with Cerebral Palsy

Abstract: Background: Two free, open tools are promising for the development of clinical markerless motion capture. MediaPipe estimates body and hand landmarks from one ordinary camera in real time, and OpenCap reconstructs three-dimensional kinematics and estimates kinetics from two or more smartphones. Research question: How do MediaPipe and OpenCap compare in design and reported accuracy for upper- and lower-body kinematics, balance control, gait, and functional task performance; how well is either validated in children, adolescents, and adults with physical disabilities? Methods: Critical narrative review of studies from PubMed and IEEE Xplore databases. Results: OpenCap has been more extensively validated for sagittal-plane lower-limb kinematics, gait, standing balance, and estimated kinetics, with typical errors of approximately 4 to 7 degrees. However, its performance is weaker in the transverse plane, at the pelvis, and for the upper limbs, and it does not include a hand model. Certain MediaPipe Solutions can track fingers and perform adequately for planar shoulder and trunk angles. However, it exhibits significant errors in out-of-plane measurements and in forearm, ankle, and finger angles. For individuals with cerebral palsy, the evidence is limited to one OpenCap balance feasibility study, one OpenCap gait study conducted without a reference system, and two MediaPipe studies. Significance: Choosing a system depends on the body region examined, the nature of the task or activity performed, and data management and processing requirements. For individuals with CP, further validation is necessary. For clinical use, standardized reporting of camera placement and tracking successful motor patterns and failures is also needed.