Détection d’anomalies à faible contraste par deep learning : Application à l’imagerie mammographique

Abstract

Breast cancer is one of the most common cancers among women worldwide. Early detection plays an essential role in improving treatment outcomes. In this work, we propose a system based on Deep Learning techniques for the automatic classification of breast abnormalities from mammographic images. The study was conducted using the RSNA Screening Mammog- raphy Breast Cancer Detection dataset after a preprocessing phase aimed at improving image quality. Two convolutional neural network architectures, resnet-50 and efficientnet-B4, were trained and evaluated using several performance metrics. The obtained results show that the efficientnet-B4 model achieves better performance than resnet-50. This study highlights the potential