Explainable Artificial Intelligence in Response to the Failures of Musculoskeletal Disorder Rehabilitation

Osteoarticular pathologies, and particularly low back pain and ankle sprains, due to their number and recurrence, constitute a public health issue. Practitioners do not have enough data describing the impact of treatments on the evolution of pathologies, which is necessary to develop a program that can dynamically adapt to changing patient conditions. We have therefore designed an application based on a series of medical consensus rules capable of generating dynamic exercise sessions adapted to the patient’s pathology and its evolution. In the traditional care pathway, it is difficult in retrospect to understand the failures in management (which amount to 40% in ankle sprains). Our approach, which adapts to the patient’s state of health over time, allows us to better understand how exercises are generated and then to analyze the pathways in order to monitor their effectiveness. The application, resulting from this work, is available as a WebApp.