Classificação de Lesões de Pele com IA Utilizando uma Abordagem em Cascata

Abstract

Skin cancer is a public health issue with high incidence in countries with intense sun exposure, such as Brazil. In remote regions far from major urban centers, the shortage of dermatologists hinders access to specialized diagnosis and increases the risk of late detection. This work evaluates a cascade-based Machine Learning approach for skin lesion screening, using classification algorithms applied to dermatoscopic images. The proposed method aims to support primary healthcare professionals, such as general practitioners, in identifying potentially malignant lesions, thus contributing to faster referral of suspicious cases and to the early detection of skin cancer in areas with limited access to specialists.