Virtual Physical Therapist System Using Full-Chain Patient Data — Focusing on Rehabilitation of Orthopedic Lower Limb Postoperative

Abstract

<title>Abstract</title> Personalized weight-bearing rehabilitation is crucial for orthopedic lower limb postoperative patients with different health backgrounds. However, physical therapists are in significant shortage, and the effectiveness of one-size-fits-all rehabilitation plans varies greatly. Here, using full-chain data from hospitalization to rehabilitation of over 1,200 patients across 25 medical institutions, a comprehensive AI-driven rehabilitation system was proposed as a Virtual Physical Therapist (VPT). The system achieves automatic formulation of overall rehabilitation plans at discharge (Multimodal Rehabilitation Schedule Network) and real-time dynamic interventions during the rehabilitation process (InsFormer). Multi-center validation showed that over 95% of patients achieved reasonable weight-bearing training in the appropriate rehabilitation period. Mean absolute error (MAE) of recovery time prediction is 0.8 week. Specifically, it showed superior rehabilitation outcomes for some patients with underlying diseases or the elderly, surpassing the diagnostic level of senior physicians. This generative AI approach adapts to changes in patients’ physical function and specific rehabilitation feedback, shortening recovery time, improving rehabilitation outcomes, and opening up new pathways for intelligent rehabilitation paradigms.

Ähnliche Arbeiten