Abstract:

Breast cancer remains one of the leading causes of mortality among women worldwide, with early detection significantly improving treatment outcomes and survival rates. Mammography is a fundamental screening tool; however, its effectiveness depends on radiologists' expertise and is subject to variability in interpretation. This research presents a comprehensive overview of artificial intelligence (AI)-based systems, particularly deep learning models, for automated breast cancer diagnosis using mammographic images. We conducted a systematic review of recent advancements in convolutional neural networks (CNNs), hybrid architectures, and explainable AI (XAI) techniques applied to mammographic analysis. Our findings demonstrate that modern AI systems achieve diagnostic accuracies exceeding 99% on benchmark datasets, with area under the receiver operating characteristic curve (AUC) values above 0.95. Additionally, we present an analysis of model performance metrics, including growth rates in accuracy improvements and transformation ratios across different architectures. The integration of attention mechanisms, multi-view fusion, and transfer learning has substantially enhanced model robustness and generalizability across diverse populations and imaging protocols. However, significant challenges remain regarding data bias, model interpretability, external validation, and clinical integration. This paper provides evidence that AI-assisted mammographic analysis has transformative potential to augment radiologist decision-making, reduce diagnostic errors, and improve accessibility to breast cancer screening in resource-limited settings.

Keywords:

Breast cancer, artificial intelligence, mammography, deep learning, convolutional neural networks, computer-aided diagnosis, explainable AI, transfer learning, medical image analysis, BI-RADS classification.

Aim/Objectives

Methods