Título: Pré-Processamento do Espectro Raman para Detecção de Placas Ateromatosas através de Redes Neurais
Autores: Paula Jr., Alderico R. de; Matta, Claudia E. da; Teixeira, Érika R. M.; Silveira Jr., Landulfo; Galhanone, Paulo R.; Galvão, Roberto K. H.
Resumo: This article presents the results of a study to reduce the number of variables necessary to represent the Raman spectrum acquired from the artery tissue when irradiated by a low power infrared laser. The analyzed methods were: peak detection, genetic algorithm, wavelets and principal components. The reduced vectors generated by these methods were applied to an artificial neural network to classify the target artery tissue into a non-pathological tissue, atherosclerostic lesion or calcified tissue. For each method, the neural network was trained by a set of 15 reduced spectra of samples of the three types of tissues and validated by another set of 10 sample reduced spectra. Among the analyzed methods, the spectrum compression by wavelets and the variable number reduction by principal components presented the best results and they were selected for the next phase of the project.
Código DOI: 10.21528/CBRN2001-059
Artigo em pdf: 5cbrn_059.pdf
Arquivo BibTex: 5cbrn_059.bib