Análise e extração de características de câncer de pele usando processamento digital de imagens

Abstract

In Brazil, 33% of all medical diagnoses are related to skin cancer. There are many types of skin cancer among which stand out the melanoma and non-melanoma types. Non-melanoma skin cancer is the most common type of skin cancer. Melanoma skin cancer, although less frequent, is by far more aggressive. There are about 8,400 cases of melanoma skin cancer in Brazil per year. Among the many techniques for diagnosing skin lesions, we highlight dermatoscopy. The analysis of dermatoscopic images is particularly challenging because lesions vary in color, shape, and size. In this context, this work applies digital image processing techniques to develop an algorithm capable of segmenting and extracting features of melanoma and non-melanoma skin cancer, which serves as a tool to aid in skin lesion diagnosis using digital image processing techniques. The algorithm developed was based on the ABCD rule and brought relevant results with the extraction of features such as asymmetry, border, color intensities, diameter and area of the lesion. The results were promising, showing the effectiveness of the techniques applied in the analysis and helping the specialist in the diagnosis of skin cancer.