基于支持向量机的皮肤自体荧光光谱分类方法在糖尿病筛查中的应用

Classification of Human Skin Autofluorescence Spectrum Based on Support Vector Machines for Diagnosis of Diabetes

将63例Ⅱ型糖尿病患者以及140例正常人皮肤的自体荧光光谱分为训练集和测试集两类,针对常用的四种核函数,运用交叉验证、网格寻优法计算最优分类参数,然后结合训练集建模并对测试集分类,结果显示使用径向基核函数时分类效果相对最佳。在此基础上,构建了一种基于线性核函数与径向基核函数的混合核函数,该核函数对人体皮肤自体荧光光谱的分类效果较之于径向基核函数更优,其分类正确率为82.61%,敏感性为69.57%,特异性为95.65%。研究结果表明支持向量机可用于人体皮肤自体荧光光谱的分类,有助于提高糖尿病筛查的正确率。

Skin autofluorescence spectrum of 63 type II diabetics and 140 normal subjects were divided into training set and testing set. According to the four commonly used kernel functions in SVM, cross validation and grid-searching were used to calculate the best parameters for classification. Mode was set up using training set and then verified by testing set. The test result indicated that the best choice for classification is radical basis function. A kind of mixed kernel func- tion based on liner kernel function and radical basis function was built and the result of classification was better than using radical basis function. Its accuracy, sensitivity, and specificity were 82.61% , 69.57% and 95.65% respectively. The result proved that SVM is suitable for the classification of human skin autofluorescence spectrum and conduces to improve the accuracy of diagnosis of diabetes.