AI in rehabilitation medicine: navigating the motion capture galaxy
A recent perspective paper by Theresa McGuirk, MS, research engineer, and Professor Carolynn Patten, PhD, PT, FAPTA, appears in a special issue of Restorative Neurology and Neuroscience titled “Breakthroughs in Stroke Rehabilitation: Bridging Engineering, Neuroscience, and Motor Control.”
The paper, “The Clinician’s Guide to Computer Vision in the Motion Capture Galaxy,” explores computer vision motion capture (CVMC), also known as markerless motion capture, and its potential applications in clinical practice, including stroke rehabilitation. The authors use references from The Hitchhiker’s Guide to the Galaxy to frame the discussion.
The authors describe how artificial intelligence (AI) supports CVMC and explain its current state as an emerging technology. While CVMC tools are increasingly available, their capabilities vary widely. The paper outlines key considerations for clinicians interested in using these tools. Clinicians are encouraged explore CVMC in clinical practice with a clear understanding of their capabilities and limitations.
CVMC uses standard digital cameras and computer vision algorithms to track human movement without requiring sensors attached to the body. This contrasts with current marker-based motion capture systems that require placing reflective markers on specific places on the body, a well-established but time-consuming process that can introduce measurement errors. Markerless systems make the process easier for patients and more accessible to clinics.
At the same time, the authors note important limitations. CVMC systems differ in measurement accuracy depending on camera setup, environment, and the algorithms used. Multi-camera systems improve measurement precisions, while single camera systems can be easier to use but are more limited in what they can capture.
Despite these challenges, the authors describe CVMC as a promising and evolving tool. In the long-term, CVMC may help improve stroke rehabilitation by enabling data-informed clinical practice using motion analysis.