Automatic Detection and Classification Of Melanoma Skin Cancer through Deep Learning Techniques

Abstract

In terms of mortality, skin cancer is one among the deadliest forms of cancer. A consistent automated method for skin lesion recognition is required for initial identification. This paper proposes an automated method for classifying skin cancers. The goal is to create a model that uses Deep Learning algorithms to diagnose skin cancer and classify it into several types. Various computer-aided solutions for the correct identification of melanoma cancer have been offered. A reliable CAD system, however, for exact melanoma identification is extremely difficult to develop. Classic learning machines or deep learning approaches are used in existing systems. We suggest the use of an intelligent Region of Interest (ROI) system based on transference learning to recognize and distinguish between melanoma and other cancers.   The ROIs are obtained using an enhanced  k-mean method. The suggested ROI-based transference learning strategy shows good performance than the previous classification systems that utilize entire images.