基于支持向量机的类风湿关节炎早期诊断模型构建及其效果评价

Effect evaluation of rheumatoid arthritis early diagnosis model based on support vector machine

目的构建基于支持向量机(SVM)的类风湿关节炎(RA)早期诊断模型,并评价其预测效果。方法以240例RA患者和180例其他风湿免疫病患者作为研究对象,测定其血清抗环瓜氨酸多肽(CCP)抗体和类风湿因子(RF),采用SVM构建早期诊断模型,并采用五次交叉验证法评价其效果。结果 SVM仿真诊断正确率为85.48%,高于RF(70.71%)和抗CCP抗体(84.05%)。五次交叉验证结果显示,SVM仿真模型诊断RA的灵敏度(Sen)为88.33%、特异度(Spe)为81.67%,MCC值为0.702 65,说明模型性能较好。RF诊断RA的Sen为74.17%,Spe为66.11%,抗CCP抗体诊断RA的Sen为78.75%、Spe为91.11%。三者Sen、Spe比较,P均<0.01。结论成功构建基于SVM的RA早期诊断模型,其对RA的预测效果较好。

Objective To establish an early diagnosis of rheumatoid arthritis based on support bector machine and evaluate its predictive effect. Methods Included 240 rheumatoid arthritis patients and 180 other rheumatic autoimmune disease patients,anti-CCP and RF was measured. Establish an early diagnosis of rheumatoid arthritis based on support vector machine,and evaluate its predictive effect by using 5-fold cross validation. Results The correct diagnostic rate of 5-fold cross validation was 85. 48%,diagnostic sensitivity was 88. 33% and diagnostic specificity was 81. 67%,prediction diagnostic accuracy was better than RF and anti-CCP( all P < 0. 01). MMC was 0. 702 65. Conclusion The study suggests that rheumatoid arthritis early diagnosis model based on support vector machine have a prediction diagnostic accuracy.