AI Imaging Model Predicts Five-Year Breast Cancer Risk from 3D Mammograms

Breast cancer screening programs often struggle to stratify women by near-term risk, which can lead to missed cancers or unnecessary testing. Forecasting who will develop disease within five years is difficult when assessments rely on a single exam or demographic calculators. Earlier identification of higher-risk patients could enable more targeted supplemental imaging while sparing lower-risk women from added procedures. To meet this need, researchers have developed an AI imaging model that analyzes multiple years of 3D mammograms to personalize risk assessment.

NYU-DRP, created at NYU Langone Health (New York, NY, USA) and its Perlmutter Cancer Center, is a deep-learning system that processes longitudinal digital breast tomosynthesis (DBT) studies. The model evaluates annual 3D mammograms across multiple years to capture temporal changes in breast tissue. By modeling these longitudinal patterns, it estimates a woman’s probability of developing breast cancer within five years. The findings were published in the American Journal of Roentgenology on August 12, 2026.

The developers trained and evaluated NYU-DRP using 313,531 yearly 3D mammograms from 161,165 women without breast cancer who were imaged at NYU Langone hospitals between 2016 and 2020. Study follow-up continued through 2025, during which less than 3% of women developed the disease. A separate comparison involved 432 women matched by age and background, including those who did and did not develop cancer within five years.

NYU-DRP correctly ranked women at higher risk 72% of the time. A model limited to a single DBT exam achieved 70%, and an artificial intelligence model analyzing 2D mammography reached 68%. Against the widely used Tyrer-Cuzick risk assessment, NYU-DRP correctly predicted higher-risk women after five years 67% of the time, compared with 56% for Tyrer-Cuzick.

The analysis indicated that breast density alone did not correspond to predicted risk. Among women with extremely dense breasts, the model classified 37.6% as having average risk, while actual five-year cases were 0.7%. Among women with less-dense, fatty breasts, the model classified 15.5% as high risk, while actual five-year cases were 2.5%.

All testing in the current study used breast imaging equipment manufactured by Hologic Inc. The team plans to share and cross-check NYU-DRP with datasets from other academic health centers and with images from different 3D mammography manufacturers. Prospective use of the longitudinal DBT approach is also planned to track breast health over time and evaluate real-world impact on screening decisions.

"Our findings demonstrate that repeated 3D mammograms contain information about a woman's future breast cancer risk that is not fully captured by either breast density or a single mammogram on its own," said study senior investigator Yiqiu "Artie" Shen, Ph.D., an assistant professor in the Department of Radiology at NYU Grossman School of Medicine and Perlmutter Cancer Center.

"If future experiments in other women with breast cancer prove successful, then AI-assisted 3D mammograms like NYU-DRP could help physicians better tailor screening to a woman's actual risk by identifying those women who may benefit from additional screening while avoiding unnecessary supplemental tests for those at lower risk," said study co-investigator Laura Heacock, M.D., an associate professor in the Department of Radiology at NYU Grossman School of Medicine and Perlmutter Cancer Center.

Related Links
NYU Langone Health
NYU Grossman School of Medicine