Investigations of skin cancer classification systems in dermoscopy images using machine learning methods

Abstract

Cancer is one of the biggest threats to human beings and is the second leading cause of death in the world. According to the statistical data from WHO(World Health Organization), cancer caused about 7.6 million people death worldwide in 2015, and it is predicted that the number of deaths caused by cancer will increase and the number will possibly reach 13.1 million in 2030. Based on related research, cancer will become the leading cause of death in next 20 years. Of all the known cancers, in world, skin cancer is the most prevalent form of cancer. It is found that each year, more new cases of skin cancer are diagnosed than all the cases of breast cancers, prostate cancers, lung cancers, and colon cancers diagnosed. Due to the manual and slower approach of traditional diagnosis methods, early detection and diagnosis of the disease gets adversely delayed. Also, the accuracy of the state of the art is not up to the mark and so not clinically acceptable. Thus, the proposed work presents an automated approach, which is able to classify dermoscopy image into normal and abnormal. In the first work, a novel CAD system for diagnosis of skin cancer using geometric and Texture features (GLCM, TCM) with Support vector machine is developed. The proposed system consists of preprocessing, ROI segmentation, feature extraction and classification. The overall classification accuracy of SVM is 96.5%, RF classifier is 95 % and Decision Tree is 93%. In the second work, Hybrid SVM based Melanoma Classification System for Dermoscopy images using Multi Structure Descriptor(MSD) is developed newline