AI Model Using Multi-Year 3D Mammograms Predicts Breast Cancer Risk

AI model trained on longitudinal mammogram data shows superior predictive power

Researchers at NYU Langone Health have developed an artificial intelligence tool that analyzes multiple years of 3D mammogram data to predict a woman's five-year breast cancer risk more accurately than existing methods. The deep-learning model, called NYU-DRP, correctly identified higher-risk patients 72 percent of the time—outperforming both single 3D mammogram analysis (70 percent) and AI-assisted 2D mammogram evaluation (68 percent).

The findings, published online in the American Journal of Roentgenology on August 12, demonstrate how longitudinal digital breast tomosynthesis (DBT)—3D mammograms taken across multiple annual screenings—captures information about changing breast tissue that single snapshots miss, according to NYU Langone Health.

The research team trained NYU-DRP on 313,531 yearly 3D mammograms from 161,165 women without breast cancer who underwent screening at NYU Langone hospitals between 2016 and 2020. Less than 3 percent of women in the study, which concluded in 2025, developed breast cancer during the observation period.

Outperforming traditional risk assessment

When researchers compared NYU-DRP against the Tyrer-Cuzick risk assessment—a widely used tool that relies on personal and family medical history, genetic mutations, and breast density rather than imaging—the AI model again proved more accurate. NYU-DRP correctly predicted five-year higher-risk cases 67 percent of the time, compared to 56 percent for Tyrer-Cuzick. The comparison involved 432 women, half carefully matched to women of similar age and background who either did or did not develop breast cancer within five years.

The study also revealed that breast density alone does not reliably indicate cancer risk. Among women with extremely dense breasts, NYU-DRP classified 37.6 percent as average risk, with actual cancer incidence at just 0.7 percent after five years. Conversely, the model identified 15.5 percent of women with less dense, fatty breasts as high risk, where actual incidence reached 2.5 percent.

"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 Yiqiu "Artie" Shen, PhD, assistant professor in the Department of Radiology at NYU Grossman School of Medicine.

Why it matters

Personalized screening protocols based on accurate risk prediction could reduce both underdiagnosis in high-risk women and unnecessary testing in lower-risk populations. With more than 43 million mammograms performed in 2025 and an estimated 382,640 new breast cancer diagnoses in 2026, tools that stratify patients by actual risk could make screening programs more efficient and effective. The five-year survival rate exceeds 99 percent when breast cancer is detected early.

Next steps and limitations

The research team plans to validate NYU-DRP prospectively by tracking women's breast health over time and observing who develops cancer. They also intend to test the model with data from other academic medical centers and different 3D mammogram manufacturers—the current study used only equipment from Hologic Inc.

"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," said Laura Heacock, MD, associate professor in the Department of Radiology at NYU Grossman School of Medicine.

The study was funded by the National Science Foundation, National Institutes of Health, and several breast cancer research foundations. Details were first reported by NYU Langone Health.