Study on spectral parameters and the support vector machine in surface enhanced Raman spectroscopy of serum for the detection of colon cancer

Abstract

Surface enhanced Raman spectroscopy (SERS) has been recognized as an effective tool for the analysis of tissue samples and biofluids. In this work, a total of 27 spectral parameters were chosen and compared using SERS. Four parameters with the highest prediction ability were selected for further support vector machine (SVM) analysis. As a comparison, principal component analysis (PCA) was used on the same dataset for feature extraction. SVM was used with the above two data reduction methods separately to differentiate colon cancer and the control groups. Serum taken from 52 colon cancer patients and 60 healthy volunteers were collected and tested by SERS. The accuracy for Parameter-SVM was 95.0%, the sensitivity was 96.2%, and the specificity was 95.5%, which was much higher than the results using only one parameter, while for PCA-SVM, the results are 93.3%, 92.3%, and 92.9%, respectively. These results demonstrate that the SERS analysis method can be used to identify serum differences between colon cancer patients and normal people.