摘要
目的 探讨MRI影像组学分析对轻度腕管综合征(CTS)的诊断价值.方法 回顾性选取行腕部MRI检查的70例轻度CTS患者及86例健康志愿者,将MRI脂肪抑制质子密度加权成像(PDWI)导入3D Slicer软件,由2名放射科医师分别进行感兴趣区(ROI)勾画,并提取830个影像组学参数,包括一阶特征、形状特征、纹理特征、小波变换特征.经观察者组内相关系数(ICC)及相关性分析、多因素逻辑回归进行影像组学参数筛选.建立逻辑回归、支持向量机、朴素贝叶斯、决策树、随机森林5种诊断模型,采用受试者工作特征(ROC)曲线分析模型的诊断效能.结果 共筛选出7个影像组学特征纳入诊断模型,逻辑回归模型的曲线下面积(AUC)最佳,训练组AUC 0.91[95%置信区间(CI)0.86～0.96],敏感度88.63%,特异度89.00%;测试组AUC 0.92(95%CI 0.85～0.97),敏感度90.48%,特异度84.62%.结论 MRI影像组学分析可用于诊断轻度CTS,逻辑回归模型具有更优的诊断效能.
Abstract
Objective To explore the diagnostic value of MRI radiomics analysis in mild carpal tunnel syndrome(CTS).Methods Seventy patients with mild CTS and 86 healthy volunteers who underwent wrist MRI examination were retrospectively selected.MRI fat-suppressed proton density weighted imaging(PDWI)were imported into 3D Slicer software,and the region of interest(ROI)delineation was performed by two radiologists independently.The 830 radiomics parameters were extracted,including first-order fea-tures,shape features,texture features,and wavelet-transform features.Radiomics parameter selection was performed through observer intraclass correlation coefficient(ICC),correlation analysis,and multivariate logistic regression.Five diagnostic models were estab-lished,including logistic regression,support vector machine,naive Bayes,decision tree,and random forest.Receiver operating charac-teristic(ROC)curve was used to analyze the diagnostic efficiency of the models.Results Seven radiomics features were selected for inclusion in the diagnostic models.The logistic regression model demonstrated the best performance,with an area under the curve(AUC)of 0.91[95%confidence interval(CI)0.86-0.96],a sensitivity of 88.63%,and a specificity of 89.00%in the training group.In the test group,the AUC was 0.92(95%CI 0.85-0.97),with a sensitivity of 90.48%and a specificity of 84.62%.Conclusion MRI radiomics analysis can be used to diagnose mild CTS,and the logistic regression model demonstrates superior diagnostic per-formance.
NETL
NSTL
万方数据