ARTIFICIAL INTELLIGENCE IN PERSONALIZED PHYSICAL REHABILITATION: ADVANCES, CLINICAL TRANSLATION, AND FUTURE DIRECTIONS

Abstract

Artificial intelligence (AI) is rapidly transforming physical rehabilitation by enabling more personalized, adaptive, data-driven, and accessible approaches to patient assessment and therapeutic intervention. This systematic review critically synthesizes advances in AI-enabled personalized physical rehabilitation, with particular emphasis on technological development, clinical translation, and future implementation. Following PRISMA 2020 principles, evidence was examined across machine learning, deep learning, computer vision, wearable and biosensor technologies, rehabilitation robotics and exoskeletons, virtual and augmented reality, digital platforms, and telerehabilitation. Current applications demonstrate substantial potential for automated functional and movement assessment, patient-specific outcome prediction, individualized exercise prescription, adaptive robotic assistance, continuous monitoring, and real-time therapeutic feedback. Clinical translation is most developed in neurological and stroke rehabilitation, while growing applications are evident across musculoskeletal and orthopedic, geriatric, and cardiopulmonary rehabilitation. Collectively, these technologies support a transition from episodic and standardized therapy toward closed-loop rehabilitation in which patient performance is continuously measured, interpreted, predicted, and used to adapt treatment. However, clinical effectiveness remains heterogeneous, and strong algorithmic performance does not necessarily translate into meaningful improvements in mobility, independence, participation, quality of life, or treatment adherence. Major barriers include small and unrepresentative datasets, algorithmic bias, limited external validation, insufficient explainability, privacy and regulatory concerns, poor interoperability, workflow integration challenges, and unequal digital access. Emerging multimodal AI and patient-specific digital twins may further strengthen precision rehabilitation by integrating diverse clinical, biomechanical, physiological, and behavioral information. Future progress requires multicenter prospective validation, standardized clinically meaningful outcomes, fairness assessment, cost-effectiveness evaluation, and human-centered implementation. AI should ultimately augment rather than replace rehabilitation professionals, combining computational intelligence with clinical judgement to deliver safe, equitable, trustworthy, and genuinely personalized rehabilitation.