Markerless Motion Capture in Routine Clinical Upper Limb Assessments: Validity and Insights Beyond Ordinal Scoring

This study evaluated the feasibility and validity of using AI-based markerless motion capture (MMC) to augment the Action Research Arm Test (ARAT), a standard but subjective clinical tool for assessing upper limb function. By deploying a three-camera setup in a routine clinical environment, the researchers captured 1,174 task performances from 20 patients with various neurological conditions. The study aimed to determine if MMC could provide objective kinematic data that is both accurate across different levels of impairment and sensitive enough to capture recovery nuances that the ordinal ARAT scale misses.

The researchers found that MMC-derived kinematic metrics successfully captured the construct of upper limb function. At the population level, these metrics showed strong discrimination between different clinical performance tiers. Crucially, the longitudinal case studies demonstrated that MMC could decompose recovery into specific domains—such as range of motion, velocity, and compensatory movements—revealing distinct recovery trajectories for patients who otherwise appeared identical on the ARAT. Furthermore, the kinematic metrics remained sensitive to improvement even after patients reached the 'ceiling' of the ARAT, where the clinical score could no longer register progress.

Clinical neurorehabilitation currently relies heavily on ordinal scales that are prone to subjectivity and lack the sensitivity to detect subtle changes in motor quality. By integrating MMC into existing clinical workflows, therapists can obtain objective, continuous data without the need for cumbersome markers or laboratory-grade equipment. This approach supports 'precision neurorehabilitation,' allowing for more granular monitoring of patient recovery and potentially enabling more tailored, data-driven interventions that remain effective even as patients approach functional recovery.