From Lab to Studio: Implementing Markerless AI for Scalable ACL Prevention in Female Dancers

Background: Female dancers experience non-contact anterior cruciate ligament (ACL) injuries at rates comparable to high-risk contact sports, yet laboratory-based marker systems have remained inaccessible for routine screening. Objectives: To compare the accuracy, feasibility, and ACL-risk detection performance of AI-enhanced markerless versus marker-based motion analysis in female dancers. Methods: Following a prospectively registered protocol, we searched PubMed, Scopus, Web of Science, SPORTDiscus, CINAHL, IEEE Xplore, and dance-specific databases from 2015 to November 2025. Eligible studies performed direct head-to-head comparisons during dance-specific tasks (e.g., grand jeté, turnout plié, pointe relevé) in female dancers aged 10–30 years. Primary outcome: root-mean-square error (RMSE) for knee valgus angle. Risk of bias was assessed with ROBINS-I; evidence certainty with GRADE. Results: Twelve studies (n = 456 female dancers, mean age 18.2 years) were included. Markerless systems achieved a pooled RMSE of 2.9° (95% CI 2.1–3.7°, I2 = 48%, k = 8) for knee valgus during landings and turnout tasks, with a pooled sensitivity of 84% (95% CI 76–90%) for high-risk profiles. Setup time was reduced by 80–95% and cost by >99% compared with marker-based systems. Certainty of evidence was moderate for accuracy and low for sensitivity. Conclusion: AI-enhanced markerless motion analysis provides clinically acceptable accuracy and unprecedented feasibility for ACL-risk screening in female dancers. Integration into studio-based prevention programmes is now justified and urgently needed. Level of evidence: Level II (systematic review of Level I–II studies).