Logistic回归算法结合近红外光谱对子宫内膜癌组织切片分类的研究

子宫内膜癌是一种常见的妇科癌症。实验将Logistic回归作为一种建模方法引入到子宫内膜癌分类诊断模型中。77个样本通过主成分判别分析和支持向量机判别分析进行降维，应用拉丁配分方法选择训练集和测试集并确定Logistic回归模型参数。结果表明，Logistic回归模型不仅能够对样本进行正确的分类，而且能将样本的分类归属趋势与临床诊断结果很好的一致。主成分判别分析结合Logistic回归有望发展为一种近红外光谱检测癌症组织的新方法。

Diagnosis of Endometrial Cancer Based on Logistic Regression and Near Infrared Spectroscopy

Endometrial carcinoma is one of the most common gynecologic cancers. The present paper reports a new application of Logistic regression to building model of endometrial cancer. Near infrared (NIR) spectra was introduced. In our study, the NIR spectra of 77 specimens were pretreated by principal component-linear discriminant analysis (PC-LDA) and support vector machine discriminant analysis (SVM-DA). Latin partition method for selecting training and test sets was used to determine the significant parameters for Logistic regression model. From the predicted results of logistic regression model, both the categories of samples and the trends of samples belonging to other class were clear and concordant with the clinical result. The proposed procedure proved to be suitable to being developed as a noninvasive diagnosis method for cancer tissue.