Abstract
Breast cancer remains the most frequently diagnosed malignancy and a leading cause of cancer death among women worldwide, with GLOBOCAN 2022 estimating 2.30 million new cases and 666,103 deaths globally (Bray et al., 2024; Gu et al., 2026). Artificial intelligence (AI), and deep learning (DL) in particular, has become central to efforts to improve early detection, diagnostic accuracy, and outcome prediction across the breast-imaging pipeline. This review synthesizes recent, peer-reviewed literature across four interconnected stages of the AI pipeline for breast cancer: (a) image augmentation, principally through generative adversarial networks (GANs) that mitigate small and imbalanced datasets; (b) semantic segmentation of tumor regions in ultrasound, mammographic, and magnetic resonance images using U-Net–based architectures; (c) diagnosis and classification of lesions as benign or malignant using convolutional neural networks (CNNs), transfer learning, and AI-assisted computer-aided diagnosis (CAD) systems applied to mammography, ultrasound, and histopathology; and (d) prognosis and survival prediction using multimodal and multi-omics deep learning. Reported performance metrics are summarized, including Dice similarity coefficients above 0.87 for state-of-the-art segmentation networks, classification accuracies exceeding 99% on benchmark histopathology datasets, and improvements in radiologist area-under-the-curve (AUC) from 0.84 to 0.91 with AI assistance (Abu Abeelh & Abuabeileh, 2025). The review also discusses persistent challenges — dataset scarcity, generalizability, interpretability, and clinical validation — and highlights emerging directions such as multimodal fusion and explainable AI. Findings indicate that while AI systems increasingly match or exceed human-level performance on narrow imaging tasks, translation into routine clinical workflows requires larger prospective trials, standardized reporting, and attention to fairness across breast densities and populations.