New Mammography Study Suggests AI May Predict Breast Cancer Detection in Subsequent Screening

Could adjunctive AI help predict breast cancer in subsequent rounds of mammography screening?

In a study involving over 135,372 screening mammograms for a cohort of 67,000 women, researchers assessed the prognostic impact of exam risk scores (ExRS) from a scale of 0-100 with adjunctive AI software (Lunit Insight MMG, Lunit). The study authors noted a 777-day average follow-up between mammography screenings.¹

The study findings, recently presented at the European Congress of Radiology (ECR), revealed a baseline mean exRS of 15.4 and a subsequent screening exRS of 73.9 in 451 women who developed breast cancer. In comparison, women who didn’t develop breast cancer had a mean ExRS of 6.7 for baseline mammography screening and 6.4 for the subsequent screening round.¹

“The AI-derived ExRS is able to differentiate between women with varying levels of risk for developing (breast cancer) at baseline, and this ability was consistent across all (breast density) categories,” pointed out Clauda Maria Weiss, M.D., the director of the UOSD Radiology Department at Treviso Health Authority in Veneto, Italy, and colleagues.

Among women with breast cancer, the study authors found that those with non-dense breast (BI-RADS categories A and B) had a mean exRS of 13.8 in baseline screening and 73.7 in the subsequent mammography screening. Women with dense breasts (BI-RADS categories C and D) had a 17.6 baseline mean ExRS and a 74.1 mean ExRS in the subsequent screening, according to the researchers.¹

“The results show that ExRS can be also used for risk-based stratification in screening,” added Weiss and colleagues.

Reference

1. Weiss CM, Di Gaetano E, Cattarin E., et al. AI-assisted double reading in mammography screening: exam risk score patterns and early cancer risk prediction. European Congress of Radiology. March 4-8, 2026, Vienna, Austria. Available at: https://connect.myesr.org/course/hot-topic-personalised-imaging/ .