Abstract P2-06-20: Use of an AI Algorithm to Determine the Prevalence of Breast Arterial Calcifications in Women Undergoing Screening Mammograms Based on Race, Age and Cancer Status
Chirag R. Parghi, Jennifer Pantleo, Jeff Hoffmeister, Julie L. Shisler, Wei Dong Zhang, Avi Sharma, Zi Meng Zhang · Clinical Cancer Research · 2025
Abstract Background: Breast arterial calcifications (BAC) on mammography has been historically overlooked and underreported as an “incidental finding”. Due to the success of mammography as a screening platform and known gaps in cardiac screening for women, BAC presence and extent can potentially identify women that may benefit from enhanced cardiac screening and medical optimization.Methods: A set of 3558 Hologic Digital Breast Tomosynthesis (DBT) screening mammograms, including 394 cancer cases and 3164 noncancer cases from October 7, 2014 to April 16, 2021 across 3 healthcare systems were analyzed using a deep learning AI algorithm trained to detect BAC on 2D images from combo DBT or 2D synthetic images from DBT. Patients ranged from 35 to 94 years of age. The dataset was weighted relative to a screening population based on Breast Cancer Surveillance Consortium based on specific clinical characteristics, namely age, race, and mammographic density with a cancer incidence of 6/1000. The study assessed overall prevalence of BAC as well as distribution among women with cancer and without cancer and by race and age.The AI model was trained using an internal dataset of 2D/synthetic mammograms to detect BAC based on expert annotation and provides a BAC of present or absent. The accuracy of the AI model was validated on a data set of 2D mammograms from 8,881 women. Ther was no overlap in the training, validation and the 3558 women prospective study data sets.Results: The unweighted overall prevalence of BAC in this cancer enriched dataset of screening exams is 17.7%. When normalized by standard age, mammographic density, and racial demographic data with a cancer incidence of 6/1000, the (weighted) prevalence of BAC changed to 15.0%, which was used for subsequent analyses. BAC is present in mammogram exams in 33.9% of women with cancer and 14.8% of women without cancer. BAC prevalence per race in the screening adjusted dataset is 14.6% White, 17.7% Black, 13.9% Asian and 17.4% in other races. BAC prevalence per age group in the screening adjusted dataset is 3.7% in women <50 years old, 8.6% in 50-59, 17.3% in 60-69 and 37.4% in women 70 and older. When the age deciles were consolidated into two groups above and below age 60, the weighted BAC prevalence was 25% in patients age 60 or above and 6.5% below the age of 60.Conclusion: The weighted prevalence and distribution of BAC increases with age as expected in a screening population. Interestingly, BAC prevalence did not vary by race suggesting it could serve as an effective cardiovascular biomarker across racial groups. AI based BAC detection on mammography demonstrates high prevalence of BAC in women with mammographically detected breast cancer. Women with increased BAC and breast cancer may benefit from cardiovascular assessment in addition to undergoing oncological treatment. In that sense, a conventional mammogram can identify cardiac needs of patients prior to or at the time of breast cancer diagnosis. Citation Format: Chirag Parghi, Jennifer Pantleo, Jeff Hoffmeister, Julie Shisler, Wei Zhang, Avi Sharma, Zi Zhang. Use of an AI Algorithm to Determine the Prevalence of Breast Arterial Calcifications in Women Undergoing Screening Mammograms Based on Race, Age and Cancer Status [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P2-06-20.