Automatic Pipeline for Kinematic Tracking of Clinically Relevant Upper-Limb Motor Tasks in Humans and Macaque

Reliable quantification of finger and hand kinematic metrics is critical for evaluating recovery and rehabilitation outcomes in neurological diseases such as stroke. Traditional methods of kinematic tracking, such as Vicon, are effective but prohibitively expensive ($50-75k), creating a high entry barrier for academic research and rendering them impractical for non-human primate (NHP) studies. This work presents an automatic pipeline for markerless tracking that integrates DeepLabCut for 2D pose estimation and Anipose for 3D reconstruction. The pipeline was validated against Vicon (ground truth) using data collected from healthy human participants (n =3) performing the box and block test (BBT) a clinically used motor task. The pipeline achieved 90% accuracy compared to Vicon's joint angle computation for key joints, including the shoulder, elbow, and wrist. Eliminating the need for physical markers, the pipeline enabled accurate finger tracking in also a macaque (n=1), which is particularly challenging with traditional marker-based systems. Beyond technical improvements, this pipeline enables the analysis of finger dexterity during clinically relevant hand tasks, such as the Brinkman Board Test, in both humans and NHPs. By facilitating cross-species comparisons of motor performance, it provides a scalable tool to monitor recovery and evaluate the effects of neurotherapies in animal models of neurological diseases affecting upper-limb function, specifically finger dexterity, thereby enhancing translatability in preclinical research.Clinical Relevance- This scalable approach provides a robust framework for clinical and pre-clinical motor control studies and lays the groundwork for future research into neuromodulation applications in motor function rehabilitation.