On AI and mammography, several recent studies show that an algorithm’s influence extends beyond whether it detects cancer correctly.
A U.K. reader study found that false-negative AI suggestions caused the sharpest deterioration in mammography performance. Median sensitivity fell from 71% during unassisted reading to 39% when AI incorrectly suggested that cancer cases were negative. Readers also looked less closely at cancers the software failed to identify.
A separate study from Milan examined how often radiologists changed their decisions after receiving deliberately incorrect AI recommendations. Readers followed the manipulated advice in around one-third of cases.
Moving from detection to prediction, a Swedish study found that commercial AI scores began increasing years before breast cancer diagnosis, potentially providing early alerts for women who may benefit from additional assessment.
Another mammography-based model predicted 10-year breast cancer risk more effectively than established clinical and AI models, suggesting that routine images may contain long-term biological risk information that is not yet being fully used.
On the modality side, a prospective Scottish study, contrast-enhanced mammography achieved 100% sensitivity and negative predictive value for breast cancer diagnosis and local staging.
For B3 breast lesions, a 10-year U.K. analysis found that atypical intraductal epithelial proliferation and in situ lobular neoplasia carried the highest risk of later malignancy, suggesting that follow-up intensity may need to reflect the individual lesion type.
On another note, experts have questioned whether imaging is appropriate when breast pain is the only symptom. The discussion highlighted conflicting guidance, the risk of unnecessary biopsies and anxiety, and the importance of distinguishing evidence-based diagnostic imaging from opportunistic screening.
Beyond breast imaging, structured analysis is reshaping the evaluation of endometriosis. Experts advocated compartment-based review of the entire pelvis, standardized terminology, and surgery-oriented reporting to prevent subtle or extrapelvic disease from being overlooked.
This letter features a selection of recent reports from the Women's Imaging Community. Please scroll through the full list of our coverage for more.
Claudia Tschabuschnig
Associate Editor
AuntMinnieEurope.com