Abstract
Breast cancer is a leading global health concern and the primary cause of female mortality worldwide. With incidence projected to exceed 3 million by 2040, timely and accurate diagnosis is critical to reduce mortality and improve clinical outcomes. There is a growing need to automate breast cancer diagnosis and classification. Artificial intelligence (AI) can enhance the efficiency, reliability, and precision of breast cancer detection, leading to better patient prognoses. This study provides a comprehensive review of recent advancements in automated breast cancer detection using mammograms. This survey examines AI applications in breast cancer diagnosis and classification via mammography over a 15-year period, from 2008 to 2023. A systematic review of 50 journal articles from PubMed, Google Scholar, Scopus, and Institute of Electrical and Electronics Engineers (IEEE) was conducted using keyword combinations including: artificial intelligence, mammograms, and breast cancer diagnosis. Findings revealed that most studies used a single dataset for validation, with limited use of multiple datasets. Thresholding and region-based segmentation were commonly employed. Classification approaches varied widely, including transfer learning, deep learning models, ensemble methods, and hybrid techniques, which often combined architectures to improve accuracy. This review highlights how dataset selection and segmentation techniques influence model performance. AI-based systems show significant potential in supporting radiologists by improving diagnostic accuracy, reducing interpretation time, and ultimately enhancing patient outcomes.
Key words: Artificial Intelligence (AI), mammogram, breast cancer diagnosis.
