基于血清SERS的肝癌分类模型的机器学习算法优化研究

Machine learning algorithm optimization of liver cancer classification model based on serum SERS

目的通过比较主成分分析(PCA)、线性判别分析(LDA)、支持向量机(SVM)和正交偏最小二乘判别分析(OPLS-DA),优化获得基于血清表面增强拉曼光谱术(SERS)的肝癌分类模型的最佳算法,为研发新型肝癌诊断技术奠定基础。方法采集检测52例肝癌患者、64例肝硬化患者和53例健康志愿者的血清增强拉曼信号,分析不同SERS光谱特征,分别建立PCA-LDA、PCA-SVM、SVM-线性核函数、SVM-高斯径向基函数(SVM-RBF)和OPLS-DA算法的肝癌分类模型,并使用敏感性、特异性、准确度和受试者操作特征曲线评估5种算法对肝癌的诊断效能。结果在三分类中,基于OPLS-DA算法对肝癌的分类准确度均高于其他4种算法的。在二分类肝癌组和非肝癌组中,基于SVM-RBF算法准确度高于其他4种算法的。二分类模型的受试者操作特征曲线显示,SVM-RBF算法对肝癌的预测能力均高于其他4种算法的。结论SVM-RBF算法结合血清SERS建立的肝癌分类模型是一种快速、准确地检测肝癌的新技术。

Objective To optimize the best algorithm of liver cancer via a comparison between the principal component analysis(PCA),linear discriminant analysis(LDA),support vector machine(SVM)and orthogonal partial least squares discriminant analysis(OPLS-DA).A classification model based on serum surfaceenhanced Raman spectroscopy(SERS)was built,and a foundation laid for developing new liver cancer diagnosis technology.Methods The serum enhanced Raman signals of 52 liver cancer,64 liver cirrhosis and 53 healthy volunteers were detected and collected,and the different SERS spectral characteristics analyzed.Liver cancer classification models based on PCA-LDA,PCA-SVM,SVM-Linear,SVM-RBF and OPLS-DA algorithms were established respectively.Sensitivity,specificity,accuracy and receiver operator characteristic curve were used to evaluate the diagnostic efficacy of the five algorithms for liver cancer.Results Of the three classifications,the accuracy of liver cancer based on OPLS-DA was higher than that of the others.Of the two types of liver cancer groups and non-liver cancer groups,the accuracy of SVM-RBF was higher thanthat of the others.The receiver operator characteristic curve of the binary model showed that the predictive ability of SVM-RBF algorithm for liver cancer was higher than that of the others.Conclusion The classification model of liver cancer established by SVM-RBF algorithm combined with serum SERS technology is a new technology for rapid and accurate detection of liver cancer.