Identification and characterization of colorectal cancer using Raman spectroscopy and feature selection techniques

• DOI: 10.1364/oe.22.025895
• Publication Date: 2014-10-14T22:34:31Z
• Authors: Shaoxin Li; Gong Chen; Yanjiao Zhang; Zhouyi Guo; Zhiming Liu; Junfa Xu; Xueqiang Li; Lin Lin
• Keywords: Male; Support Vector Machine; Adenocarcinoma; Middle Aged; Spectrum Analysis, Raman; 01 natural sciences; 3. Good health; Imaging, Three-Dimensional; ROC Curve; 0103 physical sciences; Humans; Female; Colorectal Neoplasms; Algorithms

This study aims to detect colorectal cancer with near-infrared Raman spectroscopy and feature selection techniques. A total of 306 spectra tissues normal are acquired from 44 patients. Five diagnostically important bands in the regions 815-830, 935-945, 1131-1141, 1447-1457 1665-1675 cm(-1) related proteins, nucleic acids lipids identified ant colony optimization (ACO) support vector machine (SVM). The diagnostic models built provide a accuracy 93.2% for identifying spectroscopy. demonstrates that associated ACO-SVM algorithms has great potential characterize diagnose cancer.