Artificial intelligence applications in sport-related concussion: an updated scoping review

OBJECTIVES: Sport-related concussion is a complex mild traumatic brain injury for which diagnosis, monitoring, and prognosis remain largely dependent on subjective clinical assessment. Artificial intelligence has emerged as a potential tool to enhance objectivity by integrating large, multimodal datasets across the concussion care pathway. DESIGN: Scoping review. METHODS: A systematic literature search was conducted across six databases (MEDLINE, EMBASE, SPORTDiscus, Scopus, Web of Science, and Cochrane Central) from inception to December 2025. Eligible studies were classified into four domains: Detection & Diagnosis, Monitoring & Surveillance, Prognosis & Recovery, and Prevention & Risk Modeling. RESULTS: Fifty-five studies met the inclusion criteria. Detection & Diagnosis was the most represented domain, primarily leveraging electroencephalography, speech, motor, and multimodal clinical data. Monitoring & Surveillance studies focused on wearable sensors, mouthguards, and video-based impact detection to quantify exposure and reduce false-positive events. Prognosis & Recovery models examined recovery trajectories, persistent symptoms, and reinjury risk, while Prevention & Risk Modeling studies predominantly relied on biomechanical and finite element-derived data to estimate injury risk. Despite promising performance, studies were highly heterogeneous and frequently limited by small or imbalanced samples, inconsistent outcome definitions, limited external validation, and poor model interpretability. CONCLUSIONS: Artificial intelligence shows growing potential to support sport-related concussion management across multiple clinical domains. However, current evidence supports its use primarily as a decision-support tool. Future research should prioritize large, multicenter studies, transparent labeling strategies, explainable artificial intelligence frameworks, and rigorous external validation to enable safe implementation.