Breast Tumor Detection In Mammogram Images Using Convolutional Neural Networks

Abstract

Breast cancer is the second leading cause of death among women. Mammogram images are the widely utilized method to identify breast cancer at an early stage. In this study, we implemented a convolutional neural network that classifies mammogram images into normal and abnormal(tumor) with 100% accuracy. The dataset was collected from the Mammographic Image Analysis Society MiniMammographic Database (MiniMIAS) and due to a shortage of abnormal mammography, 92 images were added from the Chinese Mammography Database (CMMD), which only contains abnormal mammogram images. The dataset was pre-processed using several filters in order to extract the ROI (Region Of Interest) and eliminate any noises, resulting in better images for training, which were shown to be effective based on the results. The dataset was split into 75%, 5%, and 20% as training, validation, and testing sets respectively. The proposed model was trained, then evaluated using a test set with 100% accuracy.