Detecção de câncer de mama em imagens histopatológicas utilizando CNN com transferência de conhecimento e transferência de cor

Abstract

Breast cancer is a disease characterized by the unrestrained proliferation of cells that can lead to the appearance of tumors near the breast region. In recent years, breast cancer has been considered the second type of cancer with the most occurrences of cases worldwide and nationally, highlighting the importance of early and accurate diagnosis. However, the diagnosis of breast cancer has faced several challenges for health professionals and specialists, given the difficulty of analyzing biopsy samples, for example. With this in mind and considering the computational advance of computer vision techniques for image recognition, many researchers and specialists consider that the use of CAD systems (Computer-Aided Diagnosis) to aid in the diagnosis of breast cancer by through the use of image processing techniques and machine learning can positively contribute to the diagnosis of breast cancer with specialists. Thus, this work implements three approaches of CNN’s of VGG-16 architecture using the techniques of knowledge transfer and color transfer, aiming to propose and evaluate a solution for the detection of breast cancer in histopathological images of cancer using CNN with knowledge transfer and color transfer.