目的 本研究旨在通过系统综述和荟萃分析，总结并更新过去十年间磁共振成像（MRI）在诊断前交叉韧带（ACL）损伤方面的表现。并进一步通过构建基于X线和MRI的影像学预测模型，帮助临床医生更客观地通过影像学检查判断ACL功能，并对比两种检查方法的诊断性能，从而为临床医生的治疗决策提供支持。

方法 研究分为两部分，第一部分为系统综述和荟萃分析，检索和筛选近十年有关MRI诊断ACL损伤的文献，提取诊断性能数据，计算汇总敏感度、特异度和曲线下面积（AUC）等诊断准确性指标。根据撕裂类型、样本量、研究设计、MRI场强和发表年份进行亚组分析，并使用QUADAS-2工具评估偏倚风险。第二部分研究为一项双中心回顾性观察研究，纳入了2021年10月至2024年12月期间因膝骨关节炎接受膝关节置换术的患者。研究共纳入一个中心238名患者作为训练队列，另一个中心220名患者作为验证队列。所有患者均接受了X线和MRI检查，并在术中通过观察和拉力测试评估ACL功能。研究从X线和MRI影像中提取了多个影像学特征，通过LASSO回归筛选关键影像学特征，并构建逻辑回归模型。最后，通过列线图将模型可视化，并分别在训练队列和验证队列中使用ROC曲线和校准曲线评估模型的诊断性能和不同人群间的一致性。

结果 第一部分研究共纳入31项研究，其中25项提供了足够的数据用于荟萃分析。在诊断ACL损伤方面，MRI诊断的汇总AUC、敏感度和特异度分别为0.95、0.88和0.90。在亚组分析中，除了部分ACL撕裂的诊断敏感度（50%）、DOR（18）显著降低，NLR（0.53）显著提升外，其他所有结果均显示出良好的诊断性能。第二部分研究的X线模型中，筛选出4个关键影像学特征：内侧胫骨平台磨损最深的位置所在平台百分比（LDW%）、内侧胫骨平台后三分之一处的磨损深度（D2）、胫骨平台后倾（PTS）和矢状面胫骨前移（SATT）。该模型在训练队列中的AUC为0.869，敏感度为92.3%，特异度为61.9%。在MRI模型中，筛选出4个关键影像学特征：LDW%、D2、SATT和后交叉韧带角度（PCLA）。该MRI模型1在训练队列中的AUC为0.883，敏感度为83.5%，特异度为75.5%。结合评估ACL形态的“直观定义法”的MRI模型2在训练队列的AUC为0.939，敏感度为89.0%，特异度为85.7%。验证队列显示出相似的结果。

结论 本研究再次证实MRI是诊断ACL损伤的重要成像工具，具有高度敏感度和特异度。经过对影像学变量的筛选，我们成功构建了基于X线和MRI的影像学预测模型，并通过列线图实现了模型的可视化，为临床医生提供了一种用于判断ACL功能的客观、准确的评分工具，而且在外部验证中取得了一致性的结果。研究结果表明，LDW%、D2、SATT和PTS构建的X线预测模型和LDW%、D2、SATT和PCLA构建的MRI预测模型在预测ACL功能方面均具有较高的诊断价值，尤其是当MRI预测模型与ACL形态判断结合后，诊断性能进一步提升。

【Objectives】 This study aims to summarize and update the performance of MRI in diagnosing ACL injuries over the past decade through a systematic review and meta-analysis. Furthermore, by constructing X-ray and MRI imaging prediction models, this study aims to help clinicians more objectively assess ACL function through imaging examinations and compare the diagnostic performance of the two methods, thereby supporting clinical decision-making. 【Methods】 The study is divided into two parts. The first part is a systematic review and meta-analysis, which involves searching and screening literature on MRI diagnosis of ACL injuries over the past decade, extracting diagnostic performance data and calculating summary sensitivity, specificity and area under the curve (AUC) as diagnostic accuracy metrics. Subgroup analyses were performed based on tear type, sample size, study design, MRI field strength and publication year. The risk of bias was assessed using the QUADAS-2 tool. The second part of the study is a two-center retrospective observational study, which included patients who underwent knee replacement surgery for knee osteoarthritis from October 2021 to December 2024. A total of 238 patients from one center were included as the training cohort, and 220 patients from another center were included as the validation cohort. All patients underwent X-ray and MRI, and ACL function was assessed intraoperatively through observation and tension testing. Multiple X-ray and MRI features were extracted from the images, and key imaging features were selected using LASSO regression to construct logistic regression models. Finally, the models were visualized using nomograms. The diagnostic performance and consistency across different populations were evaluated using ROC curves and calibration curves in both the training and validation cohorts. 【Results】 The first part of the study included 31 studies, of which 25 provided sufficient data for meta-analysis. The summary AUC, sensitivity and specificity of MRI for diagnosing ACL injuries were 0.95, 0.88 and 0.90, respectively. In subgroup analyses, except for the diagnosis of partial ACL tears, where the diagnostic sensitivity (50%) and DOR (18) were significantly reduced, and the NLR (0.53) was significantly increased. All other results showed good diagnostic performance. In the second part of the study, the X-ray model identified four key imaging features: the percentage of the medial tibial plateau with the deepest wear location (LDW%), the wear depth at the posterior third of the medial tibial plateau (D2), tibial plateau slope (PTS) and sagittal anterior tibial translation (SATT). The AUC of this model in the training cohort was 0.869, with a sensitivity of 92.3% and specificity of 61.9%. In the MRI model, four key imaging features were selected: LDW%, D2, SATT and posterior cruciate ligament angle (PCLA). The AUC of this MRI model 1 in the training cohort was 0.883, with a sensitivity of 83.5% and specificity of 75.5%. The MRI model 2, which combined the "visual definition method" for assessing ACL morphology, had an AUC of 0.939 in the training cohort, with a sensitivity of 89.0% and specificity of 85.7%. The validation cohort showed similar results. 【Conclusions】 This study reaffirms that MRI is an important imaging tool for diagnosing ACL injuries, with high sensitivity and specificity. After screening imaging variables, we successfully constructed X-ray and MRI imaging prediction models. And we visualized the models using nomograms, providing clinicians with an objective and accurate scoring tool for assessing ACL function, which achieved consistent results in external validation. The study results indicate that the X-ray prediction model constructed with LDW%, D2, SATT and PTS and the MRI prediction model constructed with LDW%, D2, SATT and PCLA both have high diagnostic value in predicting ACL function, especially when the MRI prediction model is combined with ACL morphology assessment, the diagnostic performance is further enhanced.