DISCRETE WAVELET TRANSFORM FOR DENOISING RAMAN SPECTRA OF HUMAN SKIN TISSUES USED IN A DISCRIMINANT DIAGNOSTIC ALGORITHM

In this work, we applied the discrete wavelet transform (DWT) method as a denoising tool for dispersive Raman spectra of skin samples, and we compared the results obtained with the low-order polynomial fitting in a discriminating model based on principal components analysis (PCA). We used a set of 50 Raman spectra of skin tissue fragments diagnosed as normal (N) (25 spectra) and basocellular cell carcinoma (BCC) (25 spectra). A denoising procedure using DWT and its inverse was employed, and the resulting spectra were compared to denoising using low-order polynomial fitting and adjacent averaging smoothing. The tissue spectral profile showed changes in the intensity of bands below 1400 cm^?1 for DWT compared to the denoising by polynomial and smoothing. By applying PCA and Mahalanobis distance in both groups processed, we verified that the filtering method does not alter significantly the discrimination of N and BCC tissues. However, the DWT denoising presented an interesting result, which showed the main components after decomposition of the Raman signal used in the reconstruction.