RSNA: AI Model Tops Breast Density for Predicting Risk for Breast Cancer

MONDAY, Dec. 1, 2025 (HealthDay News) — An image-only artificial intelligence (AI) model is more precise than breast density for predicting the five-year risk for breast cancer, according to a study presented at the annual meeting of the Radiological Society of North America, held from Nov. 30 to Dec. 4 in Chicago.

Constance D. Lehman, M.D., Ph.D., from Harvard Medical School in Boston, and colleagues compared image-only AI model performance with that of breast density in stratifying the five-year breast cancer risk. The AI model was trained on 421,499 mammograms from 27 facilities using a deep convolutional neural network; five-year risk probabilities were generated after calibration on an independent test set. The model was applied to a cohort of 245,344 bilateral two-dimensional screening mammograms acquired from 2011 to 2017. AI-predicted risks were categorized using National Comprehensive Cancer Network thresholds: average, intermediate, and high (<1.7, 1.7 to 3.0, and >3.0 percent, respectively). Radiologist-reported breast density was extracted from medical records.

The researchers observed an association for breast density alone with modestly increased cancer risk (hazard ratio, 1.16). Hazard ratios were 2.06 and 4.49 for intermediate and high risk relative to average risk, irrespective of density. The increased risk for breast density adjusted for AI risk was 1.12 when both factors were considered together; after adjustment for density, the increased risks for AI intermediate- and high-risk categories were 2.04 and 4.47, respectively.

“The model is able to detect changes in the breast tissue that the human eye can”t see,” Lehman said in a statement.

Press Release

More Information