Abstract

Breast cancer is the most prevalent cancer globally and a leading cause of cancer-related deaths, with over 2.3 million new cases reported annually. It is the leading cancer in women and also significantly affects men. Early detection through routine mammography is critical, as it significantly reduces mortality. The Breast Imaging Reporting and Data System (BI-RADS) is a standard classification system used to assess mammography findings, categorizing lesions based on their likelihood of malignancy. Recent advancements in deep learning and computer-aided detection (CADe) systems have improved BI-RADS classification, aiding radiologists in identifying suspicious findings more effectively. This review explores the application of deep learning, particularly convolutional neural networks (CNNs), for BI-RADS category prediction. It discusses the strengths and limitations of existing models, highlighting the use of public datasets and the integration of mammography images and radiology reports. Additionally, it suggests a novel multi-modal approach for more accurate predictions, offering insights into the future of breast cancer detection and classification.