An Overview of Modern Classification Methods of Histopathological Breast Cancer Scans

The analysis of histopathological scans is a very important test in the diagnosis of breast cancer. In this paper we present an overview of modern methods of classifying histopathology scans on the example of BreakHis dataset. To select the best performing we compared eighteen different architectures of Vision Transformers (ViT). The results were also compared with the best performing deep convolutional neural networks. Our results show often non-obvious ViT architectures that are best at classifying scans in the BreakHis database.