Fast and Interpretable Deep Learning Pipeline for Breast Cancer Recognition

Abstract

Breast cancer is one of the main causes of death across the world in women. Early diagnosis of this type of cancer is critical for treatment and patient care. In this paper, we propose a fast and interpretable deep learning-based pipeline for automatic detection of the metastatic tissues in breast histopathological images. Firstly, the proposed pipeline uses multiple preprocessing and data augmentation techniques to reduce over-fitting. Then, the proposed pipeline employs one - cycle policy technique in the pre-trained convolutional neural networks model in shallow and deep fine-tuning phases to find the optimal values. Finally, gradient-weighted class activation mapping (Grad-CAM) technique is utilized to produce a coarse localization map of the important regions in the image. Experiments on the PatchCamelyon dataset demonstrate the superior classification performance of the proposed method over the state-of-the-art.