Machine Learning Techniques for Automated Melanoma Detection

The malignant melanoma is one of the most aggressive forms of skin cancer. Modern Dermatology recognizes early diagnosis as a fundamental role in reducing the mortality rate and to guarantee less invasive treatments for patients. Computer-Aided Diagnosis (CAD) systems are increasingly adopted for the early diagnosis of skin lesions. These systems consist of different phases that must be chosen appropriately based on the characteristics of digital images aiming to obtain a reliable diagnosis. Acquisition, pre-processing, segmentation, feature extraction and selection, and finally classification of dermoscopic images hold challenges to be faced and overcome to improve the automatic diagnosis of dangerous lesions such as melanoma. The classification phase is particularly delicate: over time, a series of automatic learning algorithms have been proposed to better face this issue. In this paper, we refer to the various machine learning approaches that have been proposed and that provide inspiration for the implementation of effective frameworks.