Longitudinal changes in image-based risk scores can identify patients predisposed to the disease years before diagnosis, new research shows.
Researchers using artificial intelligence found that image-based risk scores for breast cancer derived from screening mammograms evolve over time and differ between women who develop cancer and those who do not. The study, published in Radiology, suggests these changes could enable dynamic breast cancer risk assessment.
Deep learning models generate these risk scores using the entire mammogram image rather than specific features like density. According to the study, these models have performed better than traditional risk models in estimating a woman’s five-year breast cancer risk.
“Deep learning models have been primarily used to assess cancer risk scores at a static point in time,” says Constance D Lehman, MD, PhD, professor of radiology at Harvard Medical School and CEO of Clairity Inc, in a release. “In this study, we evaluated longitudinal changes in the image-only deep learning breast cancer risk score using serial mammograms from a large screening cohort.”
Study Methodology and Findings
The research cohort included 54,014 women with a median age of 61. This group included 817 patients diagnosed with cancer and 53,197 cancer-free controls. Researchers applied a validated, open-source image-only deep learning model to 158,807 mammograms to generate five-year risk scores. No demographic, clinical, or historical imaging data were used in the model.
The study found that artificial intelligence risk scores for women who developed cancer increased progressively over the six years preceding their diagnosis. The median score for these patients rose from 2.1 in the first five to six years of the study to 6.6 at the final exam before diagnosis. In contrast, scores for cancer-free women remained stable, with medians ranging from 1.8 to 2.2.
“The increase in scores among cancer patients was detectable as early as six years prior to diagnosis and became more pronounced over time,” says Lehman in a release.
Clinical Implications for Personalized Care
The findings are significant because approximately 85% of women diagnosed with breast cancer have no known genetic mutations or family history of the disease, according to the researchers. Most breast cancer cases are sporadic, meaning they are not driven by familial inheritance.
“AI-derived risk scores can identify patients who are otherwise predisposed to the disease, and our findings demonstrate that image-based AI risk scores evolve over time and that changes in those scores may provide additional information about future breast cancer risk,” says Lehman in a release.
The risk trajectory for cancer patients increased most sharply in the two years immediately preceding a diagnosis. These trends remained consistent across different subgroups defined by age and breast density.
Lehman says the findings support using image-based risk models as dynamic biomarkers to guide personalized strategies. “With the power of AI, computer vision, and the ability to extract predictive data, we are able to apply the power of imaging to risk assessment and preventing disease from developing,” says Lehman in a release.
These image-based risk scores are included in the 2026 National Comprehensive Cancer Network guidelines. The guidelines recommend that women aged 35 and older with an elevated five-year risk score consider breast magnetic resonance imaging in addition to annual mammography. An FDA-approved version of this risk-scoring model is currently in use at select healthcare institutions in the US.
ID 243946737 © Beritk | Dreamstime.com
