Desenvolvimento de um algoritmo baseado em lógica fuzzy para segmentação de lesões em imagens de mamografia digital

Abstract

Breast cancer has been a growing problem for women around the world. According to the World Health Organization (WHO), it is the most common type of cancer among women, with increasing participation, making it one of the most fatal types of cancer worldwide. In Brazil, it is the leading cause of cancer death among women, with 59.000 new cases of cancer in 2018, with an incidence of about 59,70 cases per 100,000 women. Several methods of prevention have been developed, but one of the most effective methods for the detection of lesions is the diagnosis through digital mammography. However, the interpretation of mammography can be a difficult task even for a specialist, since the analysis is affected by several factors, such as image quality, radiologist experience and type of lesion. 12% to 30% of breast cancer cases are not detected because of bad mammography interpretation. The main objective of this work is the study and development of a tumor segmentation technique in mammography images using Fuzzy logic. It aims to insert the Fuzzy approach in the algorithm Random Walker, in order to propose a new solution for lesion segmentation. It aims to insert the Fuzzy approach in the algorithm Random Walker, in order to propose a new solution for segmentation of tumors. Finally, this work compare the results with state of the art techniques. The database has 322 mammography images obtained from 161 patients. However, only 57 of the images contain masses. Results showed that the proposed approach of the Random Walker with Fuzzy logic, used for mass segmentation, obtained better results when compared with the classic Walker Random algorithm, besides decreasing the user effort in the algorithm initialization step.