Deep Learning for Cancer Diagnosis

This chapter presents a methodology for diagnosis of pigmented skin lesions using convolutional neural networks. The architecture is based on convolu-tional networks and it evaluated new CNN models as well re-trained modification pre-existing were used. experi-mental results showed that pre-trained big datasets gen-eral purpose image classification when in order to identify le-sion types offer more accurate compared network trained explicitly from the dermatoscopic images. best performance was achieved by re-training modified version ResNet-50 with accuracy equal 93.89%. Analysis lesion pathology type also performed melanoma basal cell carcinoma being 79.13% 82.88%, respectively.