Oral cancers are the serious health problem in developing as well as developed world, and more so in India and other south Asian countries. Survival rate of these cancers, despite advances in treatment modalities are one of the poorest which is attributed to lack of reliable screening and early detection methods. The hamster buccal pouch (HBP)carcinogenesis model closely mimics human oral cancers. Optical spectroscopy methods are sensitive enough to detect
subtle biochemical changes and thus hold great potential in early detection of cancers. However, efficacy of these techniques in classifying of sequential evolution of tumors has never been tested. Therefore, in this study, we have explored the feasibility of Raman spectroscopic classification of different stages of cancers in hamster model. Strong vibrational modes of lipids (1440, 1654, and 1746 cm-1) are seen in control tissue spectra, whereas strong protein bands are seen in spectra of DMBA treated tissues. These differences were exploited to classify control and treated tissues
using Linear Discriminant Analysis (LDA), Principle Component Analysis (PCA)-Limit test, Factorial Discriminant Analysis (FDA), Quadratic Discriminant Analysis (QDA), PLS-DA and non- linear decision tree methods. All these techniques have shown good classification between spectra of different stages of tumor evolution and results were further successfully verified by leave-one-out and single blinded methods. Thus findings of this study, first of its kind,demonstrate the feasibility of Raman spectroscopic detection of early changes in tumor evolution.