Abstract
Breast cancer is the second-leading cause of mortality among women worldwide. However, early detection and diagnosis significantly improve treatment outcomes. In recent years, Computer-Aided Diagnosis (CAD) systems, which leverage Artificial Intelligence (AI) techniques, have emerged as valuable tools for assisting radiologists in the accurate and efficient analysis of medical images. Following the PRISMA guidelines, this study presents the first meta-review that synthesizes evidence from 48 systematic reviews published between 2015 and January 2025. In contrast to previous reviews, which often focus on a single imaging modality or clinical task, our work provides a comprehensive overview of imaging techniques, publicly available datasets, AI methods, and clinical tasks employed in CAD systems for breast cancer diagnosis and treatment. Our analysis shows that mammography is the most frequently applied imaging modality, while DDSM, MIAS, and INBreast are the most commonly used datasets. Among clinical tasks, the detection and classification of breast lesions are the most extensively studied, with deep learning approaches being increasingly prevalent. However, current CAD systems face notable limitations, including the lack of large and diverse datasets, limited transparency and interpretability of AI-based decisions, and restricted clinical integration. By highlighting both the achievements and the limitations, this systematic review aims to support medical professionals and technical researchers in understanding the current state of CAD systems in breast cancer care and to provide guidance for future research directions.