Combined mammography AI, polygenic, and clinical risk scores improve 10-year breast cancer risk discrimination

In 82,957 women with negative screening mammograms (Kaiser Permanente cohort), combining Mirai AI, a 313-SNP polygenic score, and BCSCv3 clinical risk reached C-index 0.70 (95% CI 0.69–0.71) vs 0.66, 0.62, and 0.61 for each alone — supporting integrated risk-stratified screening.

• Prospective cohort (2003–2020); n = 82,957 women across four racial/ethnic groups; 2,471 breast cancers over 10 years; Cox models with 5-fold cross-validation for C-index estimation.
• Mirai AI alone (C-index 0.66, 95% CI 0.65–0.67) outperformed both BCSCv3 clinical risk (0.62, 0.61–0.63) and PRS313adj polygenic risk (0.61, 0.60–0.62) individually; gains were additive when all three were combined.
• Key limitation: single integrated health system (Kaiser Permanente), ~75% non-Hispanic White; external and prospective validation in more diverse, community-based settings is needed before broad implementation.

RadPigeon summaries are original and for information only. They are not clinical advice.