基于MRI影像组学和临床特征的列线图在膝关节半月板军事训练伤风险评估中的应用价值OA

Objective To develop a combined nomogram integrating MRI radiomics and clinical features for assessing the risk of military training-related injuries to knee meniscus.Methods The clinical baseline data and knee joint MRI data were retrospectively collected from 261 frontline military personnel who underwent MRI examinations at a certain hospital for knee discomfort following military training between January 2022 and December 2024.Scans were performed using Siemens Skyra 3.0T MRI and GE Discovery 3.0T MRI.The participants were divided into a tear group and a non-tear group based on the presence or absence of meniscal tears in the knee,and allocated into a training set and a testing set with a 7∶3 stratified random sampling method.Clinical predictors were identified through univariate and multivariate analyses,and a clinical feature model was established using a logistic regression algorithm;Pyradiomics software was used to extract MRI radiomics features,and radiomics scores were calculated using the Mann-Whitney U test,Spearman's correlation analysis and the least absolute shrinkage and selection operator(LASSO)regression.With these radiomics scores as predictors,a radiomics feature model was constructed via logistic regression;finally,radiomics and clinical features were integrated,and a clinical-radiomics nomogram was developed as a combined model using logistic regression.The predictive performance of the combined model was evaluated with the ROC curve,calibration curve and decision curve analysis(DCA),and Delong test was employed to compare differences in AUC among different models.Statistical analyses were carried out using SPSS 26.0 and Python(v3.7).Results The combined model had the AUC value in the test set being 0.881,demonstrating superior predictive performance compared with the clinical feature model(AUC=0.737 in the test set)and the radiomics feature model(AUC=0.853 in the test set).Delong test revealed that the difference in AUC between the clinical feature model and the combined model in the test set was statistically significant(P<0.05),while the differences in AUC between the radiomics feature model and each of the other two models were not statistically significant(P>0.05).DCA indicated that the combined model had significant clinical net benefit within a threshold probability range of 0.2 to 0.8.Calibration curves confirmed high agreement between the predicted probabilities and observed outcomes(Hosmer-Lemeshow test,P>0.05).Conclusion The nomogram generated based on MRI radiomics and clinical features can serve as a visual tool for early risk assessment of military training-related meniscus injuries in the knee,thereby providing a scientific basis for developing personalized military training protection strategies.[Chinese Medical Equipment Journal,2026,47(5):62-69]