Melanoma skin cancer images classification based on deep neural network and super classifier

Abstract

Nowadays, the industrialization of societies and subsequent major changes in the style of human life and human habits. more pollution of the environment and its destruction, the legalization of greenhouse gas production, and subsequent destruction of the ozone layer. It leads to the spread of various cancers more than before. In the meantime, the spread of skin cancer as one of the most common types of cancer has spread faster and now this disease has been recognized as an epidemic cancer all over the world. Several researches has been conducted in the field related to skin cancer, and in this research, conventional methods of machine vision have been used. In this article, a new method for the automatic diagnosis of melanoma skin cancer using deep neural networks is proposed and to the fusion of results a support vector machine super-classification has been used, which ultimately improves the results as much as possible in comparison with the other conventional methods. Achieving 97.33% accuracy in the automatic detection of melanoma skin cancer images is evidence of the efficient performance of the proposed method in detecting skin cancer cases in a set of human skin images. Using the proposed method in this article is completely applicable in real-world applications and also can be considered as a reliable method due to the high accuracy in simulations.