Discriminative Pattern Mining for Breast CancerHistopathology Image Classification via FullyConvolutional Autoencoder

Accurate diagnosis of breast cancer in histopathology images is challenging due to theheterogeneity of cancer cell growth as well as a variety of benign breast tissue proliferative lesions. In thispaper, we propose a practical and self-interpretable invasive cancer diagnosis solution. With minimumannotation information, the proposed method mines contrast patterns between normal and malignant imagesin a weak-supervised manner and generate a probability map of abnormalities to verify its reasoning.Particularly, a fully convolutional autoencoder is used to learn the dominant structural patterns among normalimage patches. Patches that do not share the characteristics of this normal population are detected andanalyzed by one-class support vector machine and one-layer neural network. We apply the proposed methodto a public breast cancer image set. Our results, in consultation with a senior pathologist, demonstrate that theproposed method outperforms existing methods. The obtained probability map could benefit the pathologypractice by providing visualized verification data and potentially leads to a better understanding of data-driven diagnosis solutions.