Justine Bourgeot

Project Summary:

Stroke is a leading cause of death and long-term disability worldwide. Among stroke survivors, up to 75% experience upper limb impairments, that hinder their daily activities. Many adopt compensatory strategies, such as using the unaffected limb, which exacerbates motor deficits and limits functional recovery. Effective rehabilitation programs are essential to restore their quality of life. However, standard care often provides insufficient therapy time and relies on unsupervised home exercises, which lack feedback, accountability, and progress tracking, and leads to suboptimal recovery outcomes. This highlights a critical need for telemedicine solutions incorporating feedback-driven techniques to enhance chronic stroke rehabilitation.

To address this need, the Motion Analysis Laboratory seeks to develop a video-based tracking tool for home use. This tool would provide real-time corrective feedback to the patients during upper-limb rehabilitation exercises and estimate clinical scores, enabling clinicians to monitor patient progress and adjust therapy plans effectively. My project focuses on developing and validating the core components of such a system. This includes setting up a pose estimation pipeline for precise upper limb tracking and designing machine learning tools to estimate clinical scores and classify compensatory movements from videos of stroke survivors performing rehabilitation exercises at home.

This proof-of-concept work aims to pave the way for high-quality, home-based rehabilitation solutions. Such solutions would make free time for clinicians to focus on acute-phase recovery, where neuroplasticity is most active and rehabilitation critical, while ensuring chronic stroke survivors benefit from effective and consistent therapy at home.