Partial Least-Squares Discriminant Analysis on Autofluorescence Spectra of Oral Carcinogenesis

Publisher Website
Google Scholar
Abstract
A partial least-squares (PLS) discriminant analysis on the autofluorescence spectra of oral squamous cell carcinoma based on the cross-validation technique was conducted to discriminate among oral tissues at different cancer development stages. These tissues were obtained from hamsters of DMBA-induced buccal pouch carcinogenesis. The study on the fluorescence spectra of the cancer tissues revealed that 320 nm might be the optimal excitation wavelength, and it was selected for the discriminating analysis. The PLS discriminant plot based on cross-validation showed that tissues of oral carcinogenesis belonging to four clinically important cancer development stages—normal tissues, hyperplasia, dysplasia and early cancers, and frankly invasive cancers—could be classified by using the first two PLS factors that emerged from the fluorescence spectra at 320 nm excitation. The PLS factor loading plots of the first PLS factor of 320 and 360 nm excitations showed that the first PLS factor was correlated to the fluorescent structure changes. This study indicates that further development of the PLS discriminant analysis on the autofluorescence spectra may be useful for developing a simple and efficient discriminating algorithm for the identification of different stages of human oral carcinogenesis.