Incidental breast findings identified on non-breast imaging: A multi-facility real-world analysis and implications for early detection.

e12565 Background: Breast cancer screening relies primarily on age- and risk-based criteria and dedicated breast imaging. However, breast abnormalities are frequently visible on non-breast imaging performed for unrelated clinical indications. The prevalence and clinical implications of these incidental breast findings across large health systems remain incompletely understood. Methods: Radiology reports from 93 hospitals within Lifepoint Health were retrospectively analyzed over an 18-month period. Approximately 4.9 million hospital-based imaging reports across all modalities and anatomic regions were evaluated using a validated computational linguistics (CL) model designed to identify breast and lymph node findings. Model performance was assessed through dual human annotation of over 1,000 reports, demonstrating > 96% accuracy, with a positive predictive value of 95% and negative predictive value of 97%. Identified breast findings were classified as known (dedicated breast imaging or documented follow-up) or unknown (potential incidental findings). Incidental findings were further stratified by patient age, sex, imaging modality, and anatomic region. Results: Among 4.9 million reports, 506,500 (10.3%) referenced the breast. The CL model identified 36,338 examinations (7.2%) with positive breast lesion measurements, the majority arising from mammography, breast or axillary ultrasound, or breast MRI. Incidental breast findings accounted for 4.5% of all positive breast lesions, with 86% identified on CT imaging of the neck, chest, or abdomen. A total of 1,518 patients were identified with incidental breast abnormalities. Of these, 945 (62.3%) fell within the American Cancer Society average-risk screening age range (45–80 years), while 573 (37.7%) did not meet screening criteria, including younger patients, older patients, and men. Conclusions: Incidental breast findings identified on non-breast imaging represent a definitive and under-recognized opportunity for earlier cancer detection, including among populations not captured by current screening guidelines. Following completion of this analysis, a multi-facility pilot was implemented across 59 hospitals using AI-enabled detection and follow-up technology to systematically identify incidental breast findings and re-engage patients in screening and diagnostic workflows. Through this approach, a subset of incidental findings was confirmed to represent breast cancer, including early-stage disease. These findings support the integration of systematic incidental finding detection and management into routine imaging workflows as a complementary pathway to guideline-based breast cancer screening.