基于Matlab的受试者操作特征曲线分析

Matlab Based Receiver Operating Characteristic Curve Analysis

目的介绍新开发的受试者操作特征(Receiver Operating Characteristic,ROC)曲线分析软件。方法根据非参数法ROC曲线分析的基本原理,利用Matlab编写可视化的非参数法ROC曲线分析软件,对文献中的数据分别用本软件、SPSS及Analyse-it软件进行验证。最后利用本软件对两种技术检测的血清钠水平诊断洛矶山斑疹热(RMSF)的诊断作用进行评价和比较。结果本软件提供了非参数法ROC曲线分析的基本功能,能够直接输入或导入诊断试验数据,计算ROC曲线下面积并进行比较,并能保存ROC曲线操作点。与其他两个软件及文献对验证数据的ROC曲线分析结果完全相同,对应用实例进行了恰当的分析。结论本软件具有较完整的非参数法ROC曲线分析功能,在实际工作中具有一定的实用价值。

Objective To introduce a new software for receiver operating characteristic curve analysis.Methods Non-parametric ROC curve analysis software was developed visually in Matlab language.To validate the software,two data sets for ROC curve analysis were processed via SPSS,Analyse-it and the developed software,respectively.Finally,the software was applied to estimate and compare the diagnostic performance of two techniques of measuring serum sodium levels in diagnosing Rocky Mountain Spotted Fever.Results The software provided the basic functions of non-parametric ROC curve analysis,including entering directly or exporting diagnostic test data,calculating and comparing the area under the ROC curve,and saving the operating points on the ROC curve.For the two validating data sets,the software gave the same ROC curve analysis results to those provided by SPSS,Analyse-it and the literatures.Conclusions The practical application showed that the software could perform the ROC curve analysis appropriately.