Artificial intelligence in physical rehabilitation

Abstract

The rapid development of modern artificial intelligence systems (generative models, which are most often considered AI), particularly powerful basic and multimodal models, is opening up new prospects for personalization and increased effectiveness of physical rehabilitation. These technologies enable the analysis and integration of diverse patient data: from objective indicators of movement and muscle activity (biomechanics, EMG) to imaging data (MRI, ultrasound) and subjective reports. Modern AI models are capable of identifying complex spatiotemporal patterns in the process of motor function restoration. This is made possible by training on large volumes of raw clinical and instrumental data. Such AI solutions have the potential to transform all key areas of rehabilitation, including assessment and monitoring, automated movement and gait analysis using video, interpretation of wearable sensor data for objective progress tracking, creation of adaptive, "smart" therapy plans based on predictive models that predict individual responses to various types of exercise, risk modeling, and prediction of long-term rehabilitation outcomes after stroke, spinal cord injury, or orthopedic surgery. The purpose of this article is to evaluate the applicability of available models for developing rehabilitation programs after stroke.