Melanoma Classification Based on Three Different Very Deep Neural Networks

The present study proposes a methodology for classifying dermoscopic skin lesions images with melanoma and non-melanoma issues. Earlier diagnosis of melanoma gives the patient an estimate of a 5-year survival rate of 95%. Our main objective is to develop a computational tool that assists medical specialists in the task of recognizing melanoma in its initial stages. For this reason, our system is intended to generalize a prediction model, applying image processing techniques and very deep neural networks.