Artificial intelligence models for sports-related knee injuries showed high performance in early studies, with some achieving strong discriminatory ability, but most lack external validation and remain investigational for routine clinical use.
The findings highlight both the rapid progress and current limitations of artificial intelligence in sports medicine, particularly in prediction, imaging, and recovery modeling.
Machine learning models analyzing biomechanical, physiological, and demographic data showed variable performance in predicting injury risk among athletes. In one study of 791 female athletes, a support vector machine model achieved a mean area under the curve of 0.63 for anterior cruciate ligament (ACL) injury prediction, indicating modest predictive performance despite comprehensive inputs.
By contrast, tree-based models demonstrated stronger performance in specific conditions. Random forest models predicting medial tibial stress syndrome achieved an area under the curve of 0.98 and classification accuracy of 0.96 in a military cohort. External validation in a separate cohort showed an area under the curve of 0.95 and accuracy of 0.94.
Deep learning models applied to magnetic resonance imaging achieved diagnostic accuracy ranging from 55% to nearly 100% for detecting ACL and meniscal tears. These systems can automate tear detection, localization, and grading and have demonstrated performance comparable to expert radiologists in some settings when used as adjunctive tools. Multimodal approaches that combine imaging with clinical data have shown improved predictive performance in specific applications, particularly for postoperative outcomes.
Machine learning models also predicted postoperative outcomes and recovery trajectories. In one study of 680 patients, models predicting graft failure after ACL reconstruction achieved area under the curve values ranging from 0.71 to 0.85, with knee hyperextension identified as a key predictor.
For return to sport, a random forest model in 102 athletes achieved an area under the curve of 0.952 using early postoperative functional measures such as hop testing, balance scores, and strength metrics. Younger age, greater strength, and lower body mass index were associated with improved outcomes.
Across domains, most models were derived from small, single-center or highly specific cohorts, limiting generalizability. Data sets were often unbalanced, with underrepresentation of female and non-elite athletes and frequent reliance on region-specific populations. Many models also lacked external validation. Much of the current evidence is based on retrospective studies with limited prospective validation.
The researchers also highlighted interpretability challenges, as many models function as “black box” systems without transparent reasoning, as well as technical and regulatory barriers to integration into clinical workflows.
In a narrative review of studies identified through PubMed, Embase, MEDLINE, and Web of Science, researchers evaluated applications of artificial intelligence and machine learning across injury prediction, diagnosis, prognosis, and rehabilitation in sports-related knee injuries.
“AI has the potential to transform the management of sports-related knee injuries through more predictive, personalised, and precise care,” wrote lead researcher Saran Singh Gill of the Department of Surgery and Cancer at Imperial College London in the United Kingdom, and colleagues.
The research team also included Nasir Kharma and Chinmay Madhukar Gupte.
Chinmay Madhukar Gupte reported serving as President-Elect of the British Association for Surgery of the Knee, and the researchers reported no known competing financial or personal interests related to the review.
Source: The Knee
