基于临床及影像组学构建中轴性脊柱关节病支持向量机预测模型

Support vector machine based on clinical risk factors and CT radiomics for diagnosing of axial spondyloarthritis

目的:基于临床及影像组学采用支持向量机(SVM)构建中轴性脊柱关节病(ax Sp A)的预测模型。方法:回顾性收集2012年10月至2019年2月在温州医科大学附属第一医院就诊的568例腰背痛患者,最终诊断axSpA 319例,非axSpA 249例。按7:3将患者随机分为训练组与验证组。于骶髂关节CT上手动勾画三维感兴趣区(VOI)并提取影像组学特征,应用最小冗余最大相关性(mRMR)及最小绝对收缩和选择算子(LASSO)算法进行降维及选择最优影像组学特征;采用单因素和多因素Logistic回归分析寻找诊断axSpA的临床危险因素。最后使用SVM分别构建临床、影像组学及临床-影像组学联合模型,利用受试者工作特征(ROC)曲线及Delong检验评估模型的诊断效能。结果:临床-影像组学联合模型在验证组中具有最佳诊断效能,诊断准确性为0.83,灵敏度和特异度分别为85.2%、79.7%,其ROC曲线下面积(AUC=0.91)高于临床模型(AUC=0.81)及影像组学模型(AUC=0.83),差异均有统计学意义(P<0.05)。结论:基于临床和影像组学构建SVM模型对诊断axSpA具有较高价值。

Objective: To construct a prediction model of axial spondyloarthritis(axSpA) based on CT imaging and clinical features by support vector machine(SVM). Methods: A group of 568 patients diagnosed as ax Sp A(n=319) or non-axSpA(n=249) between October 2012 and February 2019 were included retrospectively,who were randomly divided into training and validation cohorts at a ratio of 7:3. The volume of interest(VOI)was manually sketched on the sacroiliac joint CT and the radiomics were extracted. The mRMR and LASSO algorithms were used to select the optimal parameters. Univariate and multivariate Logistic regression analysis was made to find the clinical risk factors of axSpA. Finally, SVM was used to build clinical, radiomics and clinicalradiomics combined models, and their performance was evaluated by the receiver operating characteristic curve(ROC) and Delong test. Results: The clinical-radiomics model showed the best diagnostic effcacy in the validation cohort and higher area under curve(AUC) values(AUC=0.91) than radiomics model(AUC=0.83) or clinical model(AUC=0.81)(P<0.05), with high accuracy(0.83), sensitivity(85.2%) and specificity(79.7%). Conclusion:Support vector machine based on CT radiomics and clinical risk factors has high value in the diagnosis of axSpA.