FPGA based diagnostic system for Malignant Melanoma dermatoscopy image recognition

Abstract

The use of Deep Learning (DL) and Convolution Neural Network (CNN) combined with Digital Image Processing (DIP) techniques could improve the diagnostics and make it more accurate. This project proposes an approach of Malignant Melanoma detection through the use of NN and DIP, training the neural network with a large number of dermatoscopic images previously verified by biopsy (histology). Moreover, an optimization of the processing time of DIP algorithms and CNN weights through FPGA (Field-Programmable Gate Array) has been discussed. It is necessary for a lower power consumption and portability, as well as to create an embedded system that will be able to assist in making the medical diagnosis of melanoma on its early stages.