Artificial intelligence clinical applications in breast diagnostic imaging

Breast cancer is the most diagnosed cancer in women worldwide, causing significant morbidity and mortality. Imaging techniques play a pivotal role in the early detection of breast cancer; digital mammography (DM) and digital breast tomosynthesis are commonly used for screening average-risk women, while magnetic resonance imaging is employed for high-risk women. Although several progresses have been made in early diagnosis, the number of breast cancer-related deaths remains high, especially among younger women and those diagnosed at advanced stages. To address this problem, new tools are needed that can enable personalized screening or new early diagnosis strategies. Artificial intelligence (AI)-base techniques can assist radiographers and radiologists in various aspects of breast cancer management, including image quality optimization, breast density evaluation, risk assessment and lesion characterization. The level of maturity of the AI technologies currently available in breast imaging is variable. Computer-aided detection (CADe) and computer-aided diagnosis (CADx) were the first AI models introduced to aid radiologists in interpreting DM; CADe marked suspicious areas, while CADx assisted in characterizing findings. However, large-scale studies revealed limited utility and potential negative impacts on mammography interpretation. Conventional CAD systems suffered from low specificity and frequent false positives, failing to address human image perception limitations. The new generation of AI algorithms aims to overcome these limitations and assist radiologists in identifying hidden lesions. This review provides an overview of the current contributions of AI in breast cancer diagnosis, focusing on achieved results, potential objectives, and limitations in clinical practice application.