Download PDF
ABSTRACT
Background
Artificial intelligence (AI) is increasingly embedded in the clinical and performance ecosystems of sports medicine and rehabilitation, spanning injury-risk prediction, wearable-sensor analytics, medical imaging, rehabilitation robotics, and clinical decision support. Despite rapid growth in the literature, the field lacks a consolidated, methodologically transparent synthesis covering the 2020-2026 period.
Objective
This systematic literature review (SLR) maps the state of the art of AI applications across sports medicine and rehabilitation, appraises methodological quality and translational maturity, and identifies research gaps and future directions.
Methods
Following PRISMA 2020 guidance, seven information sources (PubMed/MEDLINE, Scopus, Web of Science, IEEE Xplore, SpringerLink, ScienceDirect, and Google Scholar as a supplementary source) were searched using structured Boolean strings. Eligible studies were original peer-reviewed research, systematic/scoping reviews, or methodologically important conference papers that used at least one AI/ML/DL/computer-vision/NLP technique, addressed a sports-medicine or rehabilitation outcome, were published in English between 2020 and 2026, and provided sufficient methodological detail for quality evaluation. Records underwent two-stage screening and structured quality appraisal, yielding 72 including studies.
Results
Machine learning models, particularly ensemble methods such as random forests and gradient boosting, dominate injury-risk prediction, while convolutional and recurrent architectures underpin medical imaging interpretation and time-series athlete monitoring. Computer vision and markerless motion capture have matured rapidly as substitutes for laboratory-based biomechanical analysis. Wearable sensors integrated with edge-AI inference enable continuous, real-time monitoring, and rehabilitation robotics increasingly incorporate adaptive, AI-driven control. Explainable AI (XAI) and clinical decision support systems remain comparatively underdeveloped in the sports-medicine context relative to general clinical AI. Persistent limitations include small and non-representative datasets, limited external validation, algorithmic opacity, and unresolved ethical and governance questions.
Conclusion
AI offers substantial, evidence-supported potential to enhance injury prevention, diagnostic precision, and personalised rehabilitation in sports medicine, but clinical translation remains constrained by data, validation, explainability, and real-world implementation gaps. Future research should prioritise standardised multimodal datasets, external and prospective validation, explainable and clinician-centered model design, and harmonised ethical governance frameworks.
Keywords:Artificial Intelligence; Sports Medicine; Rehabilitation; Injury Prediction; Systematic Literature Review