Classification of rat FTIR colon cancer data using wavelets and BPNN

Abstract

A feature extracting method based on wavelets for horizontal attenuated total reflectance Fourier transform infrared spectroscopy (HATR-FTIR) and the cancer classification using artificial neural network trained with back-propagation algorithm is presented. The FTIR data collected from 36 normal Sprague-dawley (SD) rats, 60 1,2-DMH-induced SD rats, and 44 second generation rats of those induced rats were first preprocessed. Then, 12 feature variants were extracted using continuous wavelet analysis. Based on BPNN classification, all spectra were classified into two categories: normal and abnormal ones. The accuracy values of identifying normal, dysplastic, early carcinoma, and advanced carcinoma were 100%, 94%, 97.5%, and 100%, respectively. This result indicated that FTIR with continuous wavelet transform (CWT) and the back-propagation neural network (BPNN) could effectively and easily diagnose colon cancer in its early stages. © 2009 SIOC, CAS, Shanghai, & WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.