Classificação de tumores utilizando imagens hiperspectrais e redes neurais convolucionais

Abstract

The use of hyperspectral imaging (HSI) together with models of classification that use spatial information, is still a little approach. explored. This work proposes a deep learning model, which uses so much of spatial and spectral information, present in the HSI. the long range of the analyzed spectra are in the short wave infrared region (Short Wave InfraRed, SWIR). Samples of human epithelial tissue were analyzed divided between: Melanoma, Dysplastic Nevus and Healthy Skin. The obtained results, show that human epithelial tissue is sensitive to SWIR, and that it is possible to use deep learning models, along with HSI-SWIR images, in differentiation of healthy tissues and tissues containing Melanoma. This differentiation allows the use of an automated tool, as a complementary method to early diagnosis of tumor tissue.