Artificial Intelligence in Physical Therapy for Neurological Rehabilitation: A Systematic Mapping Study

Abstract

Neurological diseases represent a major global health burden, with stroke alone ranking as the second leading cause of mortality and disability. Physical rehabilitation is essential for minimizing impairments and improving quality of life for neurological patients. However, traditional rehabilitation faces significant challenges including high costs, and limited access to specialized staff. Information Technology (IT) systems, particularly those incorporating Artificial Intelligence (AI), have emerged as promising solutions to address these rehabilitation challenges. We present a Systematic Mapping Study (SMS) that analyses studies addressing the challenges of physical rehabilitation for neurological diseases through AI applications. There have been similar SMSs analysing AI on physical rehabilitation, but none were focused on neurological diseases, which require special attention due to their socioeconomic impact. 53 primary studies from the literature were included and analysed in our study. The results indicate that AI has been used to effectively support the rehabilitation of neurological diseases. Machine Learning (ML) techniques, and in particular Convolutional Neural Networks (CNNs), are the most frequently employed approaches. We also identify that most studies lack disease-specific adaptations, representing a major opportunity for improvement. Additionally, we applied the knowledge acquired in this study to our own line of research on the topic, which uses fuzzy logic to adjust rehabilitation routines automatically.