Paper Title
Convolutional Neural Network Visualization for Medical Images Classification

Abstract
Convolutional Neural Networks (CNNs) have been largely used in medical image analysis for detection, classification, and diagnostic of diseases. However, interpretability is one of the shortcomings of these methods. Although the CNNs are effective in computer-aided detection and diagnosis systems (CADe/CADx), the physicians need clarification and interpretation about the decisions given by them, especially the image's regions used. In this work, we were interested in visualizing the regions of interest used by CNN to detect malignancy in the breast histological images. In the experiments carried out, we tested the Gradient and Layer-Wise Relevance Propagation (LRP) visualization methods to extracted the regions of interest, then a comparison with the pathologist's annotations is achieved. Keywords - Convolutional Neural Networks, Breast histological image classification, visualization methods, Benign, Malignant.