Analysis of Raman spectroscopy data with algorithms based on paraconsistent logic for characterization of skin cancer lesions

Analysis of the Raman data to obtain results in discrimination models is usually done with multivariate statistics based on principal component analysis (PCA). In this work, we present a technique based on a non-classical logic called paraconsistent logic (PL). The aim of this work is to use computational procedures capable of generating efficient expert systems to discriminate cutaneous tissue samples obtained by Raman spectroscopy. First, a set of algorithms originating from PL is presented, and then its application in discrimination analyses is described; the discrimination analysis was conducted using a database of skin tissue samples obtained ex vivo by Raman spectroscopy of spectrum range of 400–1800 cm−1 wavelengths. Data processing, pattern creation, and comparisons were performed using the set of paraconsistent algorithms (SPA-PAL2v). The total number of samples was divided into four histopathological groups, with 115 spectra of basal cell carcinoma (BCC), 21 spectra of squamous cell carcinoma (SCC), 57 spectra of actinic keratosis (AK), and 30 normal skin (NO) spectra. An arrangement type was created for this study, and the samples were randomly selected and analyzed, and the NO group was compared with the group of non-melanoma cancer lesions (BCC + SCC) and the AK tumor lesion. Two analyses were performed. The first (SPA-PAL2v) Mode 1 (no cross-validation) achieved 76% of hits, and the second (SPA-PAL2v) Mode 2 (with cross-validation) achieved 75.78% of hits. These results were compared with discrimination using PCA statistical methods (PCA/DA) and presented superior percentages of hits, which proves the robustness of the SPA-PAL2v, confirming its potential for Raman spectrum data analysis.