Scott, C.

Abstract Background Artificial intelligence (AI)–based, mammography breast cancer risk prediction models show improved discriminatory accuracy relative to clinical risk models. However, data on their calibration are limited. This study compared model performance of 3 clinical breast cancer risk models: Gail, Tyrer–Cuzick v8, and Breast Cancer Surveillance Consortium v3 with the MIRAI AI-risk model. Methods Digital mammograms were ascertained from a screening mammography cohort of 12 308 women within the Mayo Clinic Biobank with 250 incident breast cancers (176 invasive) within 5 years. We predicted 5-year breast cancer risk, estimated discriminatory accuracy (C-index), and calibration (observed-to-expected ratio) of both overall and invasive breast cancer and compared estimates using bootstrapping approaches. Results MIRAI demonstrated similar or improved discriminatory accuracy of overall breast cancer (C-index = 0.71, 95% confidence interval [CI] = 0.68 to 0.74) and invasive breast cancer (C-index = 0.71, 95% CI = 0.67 to 0.75) compared with clinical models (overall breast cancer: C-index = 0.59-0.68; invasive breast cancer: C-index = 0.60-0.68). MIRAI’s calibration for risk of overall breast cancer (observed-to-expected ratio = 0.96, 95% CI = 0.85 to 1.08) was improved compared with Gail (observed-to-expected ratio = 1.22, 95% CI = 1.07 to 1.38) and Breast Cancer Surveillance Consortium (observed-to-expected ratio = 1.38, 95% CI = 1.22 to 1.56) but similar to TC with volumetric percent density and polygenic risk score (observed-to-expected ratio = 0.99, 95% CI = 0.87 to 1.13). However, for low-risk women (approximately 50%), MIRAI overestimated risk of overall breast cancer. MIRAI also overestimated risk of invasive breast cancer across the risk spectrum (observed-to-expected ratio = 0.68, 95% CI = 0.58 to 0.78), whereas clinical models had good calibration (observed-to-expected ratio = 0.86–0.99). Conclusion MIRAI demonstrated stronger discriminatory accuracy than clinical models for 5-year overall and invasive breast cancer risk prediction but overestimated risk for both breast cancer endpoints. AI-based risk models should consider discriminatory accuracy and calibration for invasive cancer before implementation.

Abstract Background: Artificial intelligence (AI) algorithms based on deep learning approaches show promise in improving breast cancer (BC) detection on mammography and may also improve prediction of future BC risk compared with clinical risk prediction models. Historical clinical risk prediction models underperform among women with BBD; however, AI models for BC detection on mammography have not been tested among women at elevated BC risk due to benign breast disease (BBD). We evaluated if an AI cancer detection algorithm can aid BC risk prediction among women with BBD. Methods: We examined women without a prior BC who had an initial BBD biopsy at Mayo Clinic, Rochester, between 2002 and 2013. Incident BC occurring after BBD was identified using the Mayo Tumor Registry and supplemented by follow-up questionnaires. Only women with a full field digital mammogram at least 2 years prior to BC diagnosis were eligible. A breast pathologist evaluated BBD according to increasing severity, as non-proliferative (NP), proliferative disease without atypia (PDWA) or atypical hyperplasia (AH). AI malignancy scores derived from the Transpara detection algorithm (1-10) and volumetric percent density (VPD) from Volpara (per one standard deviation, SD) were assessed from mammograms close to BBD diagnosis. Cox proportional hazards regression models were applied to estimate hazard ratios (HRs) with 95% Confidence Intervals (CIs), adjusted for age and BMI. C-statistics were estimated to assess the contribution of each factor to BC risk prediction. Likelihood ratio tests (LRTs) and bootstrapping methods were used to compare model performance with the addition of AI malignancy score. Results: The BBD cohort with mammography consisted of 3,125 women followed for a median 12.9 years, with 250 incident BC. Of these, 221 BC occurred at least two years after the mammogram and were used in analyses. As expected, increased BC risk was associated with BBD severity [for AH HR=3.18 (95%CI: 2.04, 4.95) and for PDWA HR=1.59 (95%CI: 1.17, 2.14) compared to NP] and with higher VPD [HR=1.24 per SD (95% CI: 1.05, 1.45)] (Table ). The AI-malignancy score alone was also associated with BC risk [HR=1.16 per 1 unit score (95%CI: 1.09,1.23)] and showed similar discriminatory accuracy [C-statistic=0.626 (95%CI: 0.586, 0.665)] as the model with BBD severity and VPD combined [C-statistic=0.627, (95%CI: 0.585, 0.669)] (Table ). In models with BBD severity and VPD, the AI-malignancy score was an independent risk factor for BC [HR=1.14 (95%CI: 1.07, 1.21); PLRT<0.001] but only achieved a marginally significant improvement in discriminatory accuracy [C-statistic=0.651 (0.609, 0.693), ΔC-statistic=0.024 (95% CI: 0.000, 0.047)]. Conclusion: In this preliminary study, an AI BC detection algorithm achieved similar performance in predicting future BC risk compared to established risk factors. Citation Format: C. M. Vachon, S. Winham, M. Jensen, D. Hursh, L. Pacheco-Spann, A. Norman, J. Fischer, S. Schrup, L. Seymour, D. Gehling, S. Nyante, M. Troester, N. Karssemeijer, R. Vierkant, D. Radisky, C. Scott, S. I. Maimone, A. Degnim, M. Sherman. An AI algorithm for breast cancer detection improves future BC risk prediction among women with benign breast disease [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PD4-02.