Objective: To analyze MRI and clinical characteristics of idiopathic inflammatory myopathy (IIM) activity and construct a prediction model. Methods: A retrospective analysis was conducted on 326 patients with IIM from December 2019 to December 2023 at General Hospital of Ningxia Medical University, including 112 males and 214 females, aged(53.7±15.3) years. According to histopathology and electromyography, they were divided into active phase group(n=86) and inactive phase group (n=240). The two groups were randomly divided into the training set and the verification set according to the ratio of 7∶3. The single factor analysis, least absolute shrinkage and selection operator (Lasso), random forest algorithm, and multivariate logistic regression model were used to screen the risk factors of IIM activity and construct a prediction model. Receiver operating characteristic (ROC) curve and calibration curve were used to evaluate the performance of prediction model. Results: There were significant differences in gender, age, T1 value, T2 value, creatine kinase-MB(CKMB), creatine kinase (CK) and lactate dehydrogenase (LDH) between the two groups(all P<0.05). Lasso and random forest algorithm screened 5 variables for analysis, age (λ=-0.009), T2 value (λ=-2.564), CKMB (λ=-0.256), CK (λ=-0.492), LDH (λ=-2.786) respectively. Multivariate logistic regression model showed that age (OR=1.603, 95%CI: 1.030-1.096), T2(OR=352.269, 95%CI: 13.303-9 328.053), CKMB (OR=2.470, 95%CI: 1.497-4.075), CK(OR=4.973, 95%CI: 2.583-9.575), LDH(OR=1 155.247, 95%CI: 152.387-8 757.954) were risk factors for active IIM patients. A prediction model nomograms were drawn with the above risk factors included. The area under the ROC curve (AUC) of the prediction model for the training set MRI combined with clinical indicators was higher than that of the clinical indicator model [0.914 (95%CI: 0.873-0.955) vs 0.901 (95%CI: 0.858-0.945), P<0.001], with sensitivity of 88.3% and 90.7%, and specificity of 81.7% and 75.0%, respectively. The AUC of the prediction model for the validation set MRI combined with clinical indicators was higher than that of the clinical model [0.982 (95%CI: 0.873-0.955) vs 0.934 (95%CI: 0.858-0.945), P<0.001], with sensitivity of 97.2% and 88.5%, and specificity of 100.0% and 92.3%, respectively. The calibration curves plotted in the training set and test set, respectively, fit well with the ideal curve. Conclusion: The nomogram model of MRI combined with clinical indicators can effectively predict the activity of IIM.
目的： 分析特发性炎性肌病（IIM）活动性的MRI和临床特征，构建预测模型。 方法： 回顾性分析2019年12月至2023年12月宁夏医科大学总医院收治的326例IIM患者的MRI及临床资料，其中男112例，女214例，年龄（53.7±15.3）岁，根据组织病理学及肌电图检查结果确认是否为活动期，将患者分为活动期组（n=86）和非活动期组（n=240），两组中分别按照7∶3的比例随机分为训练集和验证集。采用单因素分析、最小绝对收缩和选择算子（Lasso）及随机森林算法、多因素logistic回归模型筛选IIM活动性的危险因素并构建预测模型。采用受试者工作特征（ROC）曲线、校准曲线对预测模型效能进行评价。 结果： 两组患者性别、年龄、T1值、T2值、肌酸激酶同工酶（CKMB）、肌酸激酶（CK）、乳酸脱氢酶（LDH）差异均有统计学意义（均P<0.05）。Lasso及随机森林算法筛选出5个变量可用于分析，分别为年龄（λ=-0.009）、T2值（λ=-2.564）、CKMB（λ=-0.256）、CK（λ=-0.492）、LDH（λ=-2.786）。多因素logistic回归模型结果提示：年龄（OR=1.603，95%CI：1.030~1.096），T2值（OR=352.269，95%CI：13.303~9 328.053），CKMB（OR=2.470，95%CI：1.497~4.075），CK（OR=4.973，95%CI：2.583~9.575），LDH（OR=1 155.247，95%CI：152.387~8 757.954）为IIM患者活动性的危险因素。纳入以上危险因素，绘制预测模型列线图。训练集MRI联合临床指标列线图预测模型ROC曲线下面积（AUC）高于单纯临床指标模型［0.914（95%CI：0.873~0.955）比0.901（95%CI：0.858~0.945），P<0.001］，灵敏度分别为88.3%和90.7%，特异度分别为81.7%和75.0%。验证集数据MRI联合临床指标列线图预测模型ROC曲线AUC高于单纯临床指标模型［0.982（95%CI：0.873~0.955）比0.934（95%CI：0.858~0.945），P<0.001］，灵敏度分别为97.2%和88.5%，特异度分别为100.0%和92.3%。分别在训练集和测试集绘制的校准曲线与理想曲线拟合较好。 结论： MRI联合临床指标的列线图模型可有效预测IIM活动性。.