骨形态是前交叉韧带损伤的重要危险因素
简介
一项基于5000例MRI的多中心研究，利用深度学习模型（ACL-P）分析3D骨形态点云，评估其与前交叉韧带（ACL）损伤风险的相关性。模型在外部队列（n=7797）和职业运动员队列（n=269）中区分损伤的AUC分别达0.887和0.907，且性能不受损伤急慢性期（0-365天）或继发性关节脱位影响。关键发现：3D骨形态是ACL损伤的独立解剖风险因素，而非单…
英文摘要
Anterior cruciate ligament (ACL) injury is a prevalent sports-related trauma with significant impact on both professional athletes and young individuals. While often viewed as accidental events, high recurrence rates suggest the presence of underlying structural risk factors. This study investigated the association between 3D bone morphology and ACL injury susceptibility. Leveraging a multicenter dataset of 5000 MRIs, we developed ACL-injury Patterning (ACL-P), a deep learning model that utilizes 3D bone morphology point clouds for injury classification. Given the short-term stability of bone geometry following trauma, this classification serves as a proxy task to quantify the correlation between pre-existing morphological traits and ACL injury through discriminative performance. The model's discriminative performance was validated across an external civilian test set (n = 7797) and a professional athlete test set (n = 269), yielding an Area Under the Curve (AUC) of 0.887 (95% CI: 0.879-0.895) and 0.907 (95% CI: 0.872-0.941), respectively. Robustness checks confirmed that the model's performance remained consistent across both acute and chronic injury phases (0-365 days) and was independent of secondary joint dislocation. These findings establish 3D bone morphology as a significant anatomical factor linked to injury susceptibility, enhancing our understanding of the structural correlates of knee trauma.