Classification of colonic tissues using near-infrared Raman spectroscopy and support vector machines

Abstract

The ability of combining near-infrared (NIR) Raman spectroscopy with support vector machines (SVM) for improving multi-class classification between different histopathological groups in tissues was evaluated in this study. A total of 105 colonic tissue specimens from 59 patients including 41 normal, 18 hyperplastic polyps and 46 adenocarcinomas were used for this purpose. A rapid-acquisition dispersive-type NIR Raman system was utilized for tissue Raman spectroscopic measurements at 785-nm laser excitation. A total of 817 tissue Raman spectra were acquired and subjected to principal components analysis (PCA) for SVM-based multi-class classification, in which 324 Raman spectra were from normal, 184 from polyps and 309 from adenocarcinomatous colonic tissue. Two types of SVM (i.e., C-SVM and v-SVM) with three different kernel functions (linear, polynomial and Gaussian radial basis function (RBF) in combination with PCA were used to develop effective diagnostic algorithms for classification of Raman spectra of different colonic tissues. The performance of various SVM-based algorithms was evaluated and compared using a leave-one-out, cross-validation method. The results showed that in the C-SVM classification, the maximum overall diagnostic accuracy of 99.3, 99.4 and 99.9% can be achieved using the linear, polynomial and RBF kernels, respectively; while in the v-SVM classification, the maximum overall diagnostic accuracy of 98.4, 98.5 and 99.6% can be obtained using the linear, polynomial and RBF kernels, respectively. All the polyps can be identified from normal and adenocarcinomatous tissue using the C-SVM algorithms. The RBF C-SVM algorithm was proven to be the best classifier for providing the highest diagnostic accuracy (99.9%) for multi-class classification. This study demonstrates that NIR Raman spectroscopy in combination with a powerful SVM technique has great potential for providing an effective and accurate diagnostic schema for cancer diagnosis in the colon.