Artificial Intelligence for Breast Cancer Screening: Trade-offs between Sensitivity and Specificity
Manisha Bahl, Synho Do · Radiology Artificial Intelligence · 2024
M ammography is the only screening modality that has been shown to decrease mortality from breast cancer (1); however, screening mammography does have shortcomings.Its sensitivity is approximately 87% and even lower in women with dense breast tissue (2).Interval cancers, which are diagnosed based on symptoms after a negative screening examination but before the next screening round, represent up to 30% of breast cancers (3).In addition to the inherent limitations of mammography, a current challenge facing screening programs is the lack of specialized breast imaging radiologists and wide variability in recall rates and other performance metrics (2,4).These limitations can potentially be addressed through the use of artificial intelligence (AI)-based computerassisted detection (CAD) and diagnosis algorithms for mammography.Although traditional CAD algorithms for mammography, approved by the U.S. Food and Drug Administration in the 1990s, showed promise in early studies, subsequent postimplementation research concluded that traditional CAD does not lead to improved screening mammography performance, largely due to high rates of false-positive marks and poor specificity (5).In recent years, interest in CAD for mammography has been rejuvenated by the development of new-generation AI