A computer aided diagnosis system for skin cancer detection

Abstract

Melanoma is the deadliest form of skin cancer, accounting for about 75% of deaths related to this type of disease. Fortunately, melanoma early detection can increase the survival rate of victims considering that melanoma skin cancer is often visible to patients and physicians. However, recommended self-examinations or physician-directed exams are not significantly reducing melanoma deadly cases due to the absence of knowledge of the patients or the lack of access to well-trained physicians. Based on that, this paper proposes a computer aided diagnosis system that detects melanoma skin cancer using dermatoscopy images, image processing techniques, and machine learning algorithms. Our main goal is to create a cheap, relatively accurate, and easy-to-use system available as an initial procedure to detect melanomas. The evaluation of the designed system using 748 dermatoscopy images shows sensitivities around 98%, when a simple feature-extraction stage is applied and a classifier based on support vector machines is utilized.