Introduction
Cardiovascular disease (CVD) is the leading cause of death in women yet is frequently underdiagnosed (1–3). Risk assessment for CVD is typically performed using a combination of laboratory markers, family history, and other risk factors that are combined in various risk prediction models, with the PREVENT calculator currently recommended by guidelines(4–6). However, a key limitation of these models is their reliance on clinical and laboratory data alone as they do not include any anatomic assessment of the vasculature.
Because of its high resolution and routine use in breast cancer screening, mammography provides an opportunity to directly visualize the vasculature in almost all adult women. Arterial calcification is readily apparent on screening mammograms, correlates with calcification in other arterial beds(7), and can predict CVD events in the general population(8,9). Unlike coronary artery calcification (CAC), which primarily reflects intimal-layer atherosclerotic changes leading to vessel narrowing, BAC occurs in the medial layer, resulting in increased arterial stiffness and decreased pulsatility rather than luminal obstruction(7,10), Although BAC shares certain risk factors with atherosclerosis (such as age and diabetes), it has also independently predicts CVD beyond these factors (11,12). Thus, medial arterial calcification represents a distinct contribution to CVD risk, suggesting that incorporating BAC evaluation can enhance overall cardiovascular risk assessment beyond traditional risk factors or measures of atherosclerosis alone.
Previous studies have typically only considered presence or absence of breast arterial calcification (BAC) as a binary variable [refs], with limited demonstration of reliable automated BAC segmentation and quantification at scale. We previously developed patch-based segmentation models capable of quantifying BAC with adequate sensitivity to detect disease progression and identify risk factors (13) (14); however, these models suffered from false positives due to imaging artifacts in breast tissue. To address this, we subsequently developed a more advanced transformer-based segmentation model that maintains high sensitivity while reducing false-positive detections related to artifacts or non-vascular calcifications. In this study, we evaluate whether automated quantification of BAC from routine screening mammograms enhances cardiovascular disease (CVD) prediction in a large, multi-racial cohort across two geographically diverse regions. This automated strategy has the potential to provide opportunistic cardiovascular risk assessment in women already undergoing routine screening mammography.
Methods
Study Population
Our study included 74,124 women from Emory Healthcare (Atlanta, GA, USA) in the Emory Breast Imaging Dataset (EMBED)(15) who served as our internal cohort for development and validation, and 49,638 women from Mayo Clinic Enterprise (Arizona, Florida, Mid-West, Rochester) who constituted our external validation cohort (Supplementary Figure 1). Inclusion criteria included age >18 years, at least one CVD event or any face-to-face clinical follow-up following an initial screening mammogram,. For each patient from Emory Healthcare, the earliest screening mammogram was included for evaluation to maximize the follow-up period. At Mayo Clinic, all mammogram samples were from 2017. Therefore, each patient was represented only as a single sample at both sites.
In order to evaluate BAC as an opportunistic screening risk marker, we excluded women with prior known cardiovascular disease, women who had existing records of major adverse cardiovascular events (MACE), and women who had CAC scoring or coronary angiography before the index mammogram. These strict inclusion criteria maximize the likelihood of excluding any women who had extant CVD that could artificially inflate the demonstrated prognostic value of BAC for asymptomatic screening.
Measurement of Breast Arterial Calcification
Model Development
To increase the BAC segmentation accuracy over the previously published patch-based segmentation model(14) that does not fully leverage the spatial location of the patches, we developed a proprietary transformer-based neural network architecture to directly segment and quantify BAC within the mediolateral oblique (MLO) views of high resolution, two-dimensional, full-field digital screening mammograms (FFDM). MLO images are used since this plane is more perpendicular to breast arteries compared to the craniocaudal (CC) view. To capture hierarchical information, we created a set of Mix Transformer (MiT) encoders that use the same design but come in different sizes. Inspired by the structure and attention method used in Pyramid vision transformer network(16), we added new features like overlapping patch merging and removing positional encoding. We applied a decoder made only of MLP layers, avoiding the complex and hand-crafted components. This simple design is effective for BAC segmentation because our hierarchical transformer encoder has a much larger effective receptive field compared to traditional CNN encoders. The model was trained, validated, and tested using screening mammograms annotated by an expert radiologist (HT) and divided into 90%–5%–5% training, validation, and test exams.
The final model output is a binary mask with detected BAC pixels. Post-processing is applied to differentiate between calcified pixels in a vessel and background non-calcified vessel. BAC area is quantified by computing the number of segmented pixels multiplied by the pixel size defined in the DICOM metadata to yield an area in mm2. Model performance compared to expert-annotation for area quantification resulted in a mean absolute error of 3.94 mm2, R2 of 0.91, and Pearson Correlation of 0.95, which supersedes previously reported inter-observer variability for BAC which ranges from 0.48–0.53(8).
BAC Severity Classification
BAC severity was categorized into four groups based on the quantified BAC area: Zero BAC (0mm2), Mild (>0–10 mm2), Moderate (>10–25 mm2) and Severe (>25 mm2) (Figure 1). These thresholds were established through empirical analysis that maximized between-group hazard ratio differences in the Emory cohort (Supplementary Figure 2). The zero BAC was validated through visual inspection of 500 exams (Supplementary Figure 6).
Figure 1.
Examples of mammograms with mild, moderate and severe BAC quantified by our AI model. Left: original image; Right: AI model heatmap with BAC score shown above.
Identification of Cardiovascular Events, Risk Factors, and PREVENT Score
Using a previous expert-curated list, the primary/principal diagnostic ICD codes were extracted from the electronic health record (EHR) with timestamp data from both clinical sites to identify MACE(17) (Supplementary Table 1) and all-cause mortality. For women without events at Emory, a face-to-face visit following the mammogram date was required, defined by an in-person visit to a healthcare provider at the institution.
To compare BAC to the PREVENT risk score(5), risk factors including total and HDL cholesterol, systolic blood pressure, BMI, eGFR, diabetes (Supplementary Table 2), smoking, use of anti-hypertensive medication and use of statins were extracted from the EHR as ICD codes or lab values. In only 4,876 of the 74,124 Emory patients (6.58%) was all required data available to calculate the PREVENT score. PREVENT scores were calculated by implementing the algorithm from Khan, et al. (5). We validated our implementation by manually comparing computed scores of 40 random patients against the online American Heart Association PREVENT calculator (Supplementary Figure 3) (18). The PREVENT score calculator was then applied to all eligible Emory and Mayo patients with complete risk factor data.
Statistical Analysis
We compared the prevalence of various demographics and risk factors between BAC severity groups and presence or absence of MACE using several statistical tests: chi-square for discrete variables and t-tests or one-way ANOVA for continuous variables. The Kruskal-Wallis test was used to assess the differences in median follow-up duration between groups.
The association between BAC severity and MACE incidence was examined using Kaplan-Meier curves. For our primary analysis, we used time-to-event as the timescale. A secondary analysis was performed using age as the primary timescale, consistent with methodologies such as PREVENT, to account for age as a strong predictor of CVD and to adjust for left-truncation due to varying patient ages at entry into the study.
To investigate potential differential effects, subgroup analyses were performed. First, given the potential prognostic value in younger populations, a dedicated analysis was conducted in patients under 50 years of age, specifically examining MACE outcomes in those with moderate to severe BAC compared to those without BAC. Second, to align with established risk stratification, additional subgroup analyses were conducted within different CVD risk categories (low, borderline, intermediate, and high) as defined by PREVENT. Within these risk categories, event-free survival was also compared between moderate to severe BAC and zero BAC patients. Differences in event-free survival rates between BAC severity groups were assessed using the log-rank test.
To assess the predictive value of BAC and rigorously compare the performance of various Cox Proportional Hazards (CPH) models, three distinct models were examined and validated:
Age + BAC Model: a CPH model incorporating BAC and age;
PREVENT Model: a CPH model utilizing the established PREVENT risk scores;
PREVENT + BAC Model: a CPH model combining PREVENT with BAC.
For Models 1 and 3, we examined BAC separately as categorical (Zero BAC, Mild, Moderate, and Severe) and continuous variables. For completeness, we explored non-linear associations by transforming BAC using log2(BAC+1) as was done in studies of coronary artery calcium (19,20), allowing us to examine how a doubling of the BAC measurement affects event risk. Adding 1 to BAC before transformation enables us to include zero BAC values in our analysis.
To robustly evaluate and compare the discriminative performance of competing CPH models, a 5×5-fold cross-validation scheme was employed. The C-index was computed for each repetition and fold for each model. To determine if there was a statistically significant difference in performance, the paired Wilcoxon signed-rank test was applied. A two-sided p-value was calculated, and a significance level of α=0.05 was used for hypothesis testing. The proportional hazards assumption for all CPH models was formally tested using Schoenfeld residuals. No significant violations were detected across the models.
Results
Study Characteristics
The internal and external cohorts consisted of women with a mean age of 55.5 years (SD: 10.4) and 59.5 (SD: 10.2) years, respectively (Table 1). Over a median follow-up of 7 years (IQR: 4–10), the incidence of MACE was 4.8% in the internal cohort and 12.1% in the external cohort (median follow-up: 7 years, IQR: 6–7). Across both cohorts, patients who experienced MACE were significantly older than those who did not (mean age difference: ~8 years, p<0.001).
Table 1.
Baseline demographic characteristics and risk factors based on the BAC severity and whether a study participant had any subsequent MACE event
| BAC SEVERITY | MACE | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||
| Overall | Zero | Mild | Moderate | Severe | P-Value | No Event | Event | P-Value | ||
|
| ||||||||||
| INTERNAL COHORT | n | 74124 | 62187 (83.9) | 10369 (14.0) | 854 (1.2) | 714 (1.0) | 70530 (95.2) | 3594 (4.8) | ||
| Age, mean (SD) | 55.5 (10.4) | 54.6 (10.0) | 58.9 (10.9) | 68.4 (8.6) | 67.8 (9.1) | <0.001 | 55.1 (10.3) | 62.8 (9.8) | <0.001 | |
| Race (%) | Asian | 4940 (6.7) | 4172 (6.7) | 688 (6.6) | 52 (6.1) | 28 (3.9) | <0.001 | 4835 (6.9) | 105 (2.9) | <0.001 |
| Black | 31971 (43.1) | 27153 (43.7) | 4068 (39.2) | 338 (39.6) | 412 (57.7) | 29805 (42.3) | 2166 (60.3) | |||
| Other | 1423 (1.9) | 1173 (1.9) | 223 (2.2) | 15 (1.8) | 12 (1.7) | 1401 (2.0) | 22 (0.6) | |||
| Unknown | 5747 (7.8) | 4886 (7.9) | 771 (7.4) | 52 (6.1) | 38 (5.3) | 5688 (8.1) | 59 (1.6) | |||
| White | 30043 (40.5) | 24803 (39.9) | 4619 (44.5) | 397 (46.5) | 224 (31.4) | 28801 (40.8) | 1242 (34.6) | |||
| Hispanic, n (%) | 4010 (5.4) | 3311 (5.3) | 602 (5.8) | 57 (6.7) | 40 (5.6) | 0.079 | 3917 (5.6) | 93 (2.6) | <0.001 | |
| Follow up (y), median [Q1,Q3] | 7.0 [4.0,10.0] | 7.0 [4.0,10.0] | 7.0 [4.0,9.0] | 6.0 [4.0,9.0] | 5.0 [3.0,8.0] | <0.001 | 7.0 [4.0,10.0] | 4.0 [2.0,6.0] | <0.001 | |
| Smoking, n (%) | 3881 (22.4) | 3145 (22.2) | 589 (22.9) | 78 (25.7) | 69 (23.8) | 0.413 | 3318 (21.3) | 563 (31.4) | <0.001 | |
| Diabetes, n (%) | 12663 (17.1) | 9951 (16.0) | 2077 (20.0) | 307 (35.9) | 328 (45.9) | <0.001 | 10859 (15.4) | 1804 (50.2) | <0.001 | |
| Anti-HT medication, n (%) | 4910 (6.6) | 3723 (6.0) | 858 (8.3) | 147 (17.2) | 182 (25.5) | <0.001 | 3955 (5.6) | 955 (26.6) | <0.001 | |
| Statins, n (%) | 3381 (4.6) | 2453 (3.9) | 629 (6.1) | 132 (15.5) | 167 (23.4) | <0.001 | 2453 (3.5) | 928 (25.8) | <0.001 | |
| Total Cholesterol, mean (SD) | 193.3 (34.4) | 193.5 (34.3) | 193.2 (34.8) | 189.6 (34.1) | 187.4 (36.3) | 0.159 | 193.6 (34.1) | 190.2 (36.8) | 0.01 | |
| HDL, mean (SD) | 59.2 (15.3) | 59.2 (15.2) | 59.6 (15.5) | 60.6 (15.6) | 56.7 (15.9) | 0.16 | 59.6 (15.2) | 55.8 (15.3) | <0.001 | |
| SBP, mean (SD) | 128.8 (18.8) | 128.4 (18.6) | 130.0 (19.3) | 133.7 (19.9) | 137.9 (21.3) | <0.001 | 128.3 (18.5) | 135.4 (21.5) | <0.001 | |
| BMI, mean (SD) | 28.4 (5.4) | 28.5 (5.4) | 27.8 (5.5) | 28.5 (5.3) | 29.6 (5.4) | <0.001 | 28.3 (5.4) | 29.6 (5.6) | <0.001 | |
| eGFR, mean (SD) | 83.5 (20.4) | 84.8 (19.5) | 80.6 (21.6) | 68.6 (23.3) | 60.7 (28.6) | <0.001 | 84.7 (19.5) | 72.6 (25.1) | <0.001 | |
|
| ||||||||||
| EXTERNAL COHORT | n | 49638 | 36737 (74.0) | 10725 (21.6) | 1582 (3.2) | 604 (1.2) | 43622 (87.9) | 6016 (12.1) | ||
| Age, mean (SD) | 59.5 (10.2) | 58.5 (9.9) | 63.3 (10.2) | 70.4 (6.6) | 71.1 (6.9) | <0.001 | 58.5 (10.0) | 66.4 (8.8) | <0.001 | |
| Race, n (%) | Asian | 893 (1.8) | 808 (2.2) | 139 (1.3) | 14 (0.9) | 8 (1.4) | <0.001 | 828 (1.9) | 60 (1.0) | <0.001 |
| Black | 546 (1.1) | 330 (0.9) | 128 (1.2) | 17 (1.1) | 9 (1.5) | 479 (1.1) | 60 (1.0) | |||
| Other | 943 (1.9) | 734 (2.0) | 203 (1.9) | 25 (1.6) | 10 (1.7) | 872 (2.0) | 54 (0.9) | |||
| White | 47255 (95.2) | 34863 (94.9) | 10242 (95.5) | 1525 (96.4) | 576 (95.4) | 41440 (95.0) | 5841 (97.1) | |||
| Hispanic, n (%) | 878 (1.77) | 750 (1.92) | 175 (1.64) | 19 (1.26) | 10 (1.68) | <0.001 | 802 (1.84) | 67 (1.13) | <0.001 | |
| Follow up (y), median [Q1,Q3] | 7.0 [6.0, 7.0] | 7.0 [6.0, 7.0] | 7.0 [6.0, 7.0] | 6.0 [5.0,7.0] | 6.0 [3.0, 7.0] | 7.0 [6.0, 7.0] | 3.0 [2.0, 5.0] | <0.001 | ||
| Smoking, n (%) | 5629 (41.5) | 4545 (40.6) | 790 (41.4) | 204 (42.4) | 112 (45.5) | 0.42 | 3,998 (40.0) | 1,771 (49.6) | <0.001 | |
| Diabetes, (%) | 14.44 | 17.08 | 42.23 | 72.15 | 68.98 | <0.001 | 10.86 | 34.41 | <0.01 | |
| Anti-HT medication, (%) | 0.52 | 0.7 | 1.58 | 2.45 | 2.49 | <0.001 | 0.32 | 2.19 | <0.001 | |
| Statins, (%) | 2.82 | 3.36 | 8.94 | 14.59 | 16.73 | <0.001 | 1.44 | 14.47 | <0.001 | |
| Total Cholesterol, mean (SD) | 195.8 (43.7) | 198.58 (43.11) | 195.57 (44.2) | 192.63 (43.74) | 184.79 (43.63) | <0.001 | 198 (42.99) | 183.17 (44.96) | <0.001 | |
| HDL, mean (SD) | 63.9 (19.77) | 65.96 (20.44) | 62.98 (19.29) | 60.98 (18.22) | 60.49 (18.62) | <0.001 | 64.68 (19.72) | 60.02 (19.56) | ||
| SBP, mean (SD) | 131.3 (19.5) | 130.2 (19.5) | 133.2 (19.1) | 134.5 (19.3) | 136.4 (21.9) | <0.001 | 131.4 (19.2) | 136.4 (21.5) | ||
| BMI, mean (SD) | 24.7 (9.8) | 23.4 (10.4) | 25.5 (9.8) | 28.2 (10.5) | 27.4 (10.8) | <0.001 | 24.2 (10.2) | 27.4 (10.8) | ||
| eGFR, mean (SD) | 93.2 (20.4) | 103.4 (19.3) | 98.7 (19.2) | 93.3 (19.5) | 84.5 (19.1) | <0.001 | 104.5 (19.5) | 86.5 (20.1) | <0.001 | |
Calculated on patients who have the corresponding value
Breast arterial calcification was detected in 16.1% of women in the internal cohort and 26.0% of women in the external cohort. BAC severity was strongly associated with age, with a mean age difference of approximately 13 years between patients with zero BAC and those with severe BAC (p<0.001). Furthermore, several cardiometabolic risk factors, including diabetes, use of antihypertensive medications and statins, higher systolic blood pressure, and higher body mass index, as well as lower eGFR, were positively correlated with BAC severity. In contrast, smoking was not found to be associated with BAC in either cohort.
BAC as an Independent Risk Factor for MACE
Kaplan-Meier analysis of the combined cohorts revealed distinct graded separation between zero, mild, moderate, and severe BAC for all outcomes (acute myocardial infarction, stroke, heart failure, and all-cause mortality) (Figure 2). Corresponding analyses using age as the timescale for each individual cohort are provided in the supplementary (Supplementary Figure 7).
Figure 2.
Kaplan-Meier estimates of event-free survival stratified by breast arterial calcification (BAC) severity (Zero, Mild, Moderate, Severe). The tables below the plots indicate the number of patients at risk for each BAC severity group at 2-year intervals. Differences in event-free survival rates among the BAC severity groups were assessed using the multivariate log-rank test.
Notably, in patients younger than 50 years (Figure 3 and Supplementary Figure 8), where BAC is less frequently observed, moderate to severe BAC was associated with significantly lower event-free survival for MACE when compared to zero BAC,. This highlights that BAC in a younger population is a potent indicator of elevated cardiovascular risk.
Figure 3.
Kaplan-Meier estimates of event-free survival in patients under 50 years of age, stratified by Zero BAC and Moderate to Severe BAC. Years to event represents the time since the initial mammogram to the occurrence of a MACE event or last follow-up.
In multivariable analysis adjusting for age, both moderate and severe BAC were still significant predictors of all adverse outcomes (Table 2). A clear dose-response relationship was observed, where the hazard ratio (HR) for all clinical outcomes increased with BAC severity. The most pronounced effects were seen in heart failure for the internal cohort (severe BAC HR: 2.66, 95%CI: 1.92–3.67, p<0.005) and in all-cause mortality for the external cohort (severe BAC HR: 2.25, 95%CI: 1.62–3.13, p<0.005).
Table 2.
Association of BAC and clinical outcomes in Age+BAC Model
| INTERNAL COHORT | ANY | AMI | STROKE | HF | DEATH | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR | p | HR | p | HR | p | HR | p | HR | p | ||||
|
| |||||||||||||
| BAC CATEGORICAL MODEL | BAC | Zero | Reference | Reference | Reference | Reference | Reference | ||||||
| Mild | 1.09 (1.01–1.18) | 0.02 | 1.05 (0.96–1.14) | 0.28 | 1.06 (0.98–1.15) | 0.17 | 1.09 (1.00–1.18) | 0.05 | 1.11 (1.02–1.20) | 0.02 | |||
| Moderate | 1.60 (1.31–1.97) | <0.005 | 1.29 (1.02–1.65) | 0.04 | 1.49 (1.20–1.84) | <0.005 | 1.44 (1.15–1.82) | <0.005 | 1.70 (1.37–2.12) | <0.005 | |||
| Severe | 2.38 (1.76–3.22) | <0.005 | 1.60 (1.12–2.30) | 0.01 | 1.98 (1.44–2.72) | <0.005 | 2.66 (1.92–3.67) | <0.005 | 2.33 (1.69–3.22) | <0.005 | |||
|
| |||||||||||||
| Age | per 1 year increase | 1.03 (1.03–1.03) | <0.005 | 1.01 (1.01–1.01) | <0.005 | 1.02 (1.02–1.02) | <0.005 | 1.01 (1.01–1.02) | <0.005 | 1.01 (1.01–1.02) | <0.005 | ||
|
| |||||||||||||
| BAC LINEAR MODEL | |||||||||||||
| BAC | per 1mm2 increase | 1.02 (1.01–103) | <0.005 | 1.01 (1.01–1.02) | <0.005 | 1.02 (1.01–1.02) | <0.005 | 1.02 (1.01–1.03) | <0.005 | 1.02 (1.01–1.03) | <0.005 | ||
| Age | per 1 year increase | 1.03 (1.02–1.03 | <0.005 | 1.01 (1.01–1.01) | <0.005 | 1.02 (1.02–1.02) | <0.005 | 1.01 (1.01–1.02) | <0.005 | 1.01 (1.01–1.02) | <0.005 | ||
|
| |||||||||||||
| EXTERNAL COHORT | |||||||||||||
| BAC CATEGORICAL MODEL | BAC | Zero | Reference | Reference | Reference | Reference | Reference | ||||||
| Mild | 1.02 (0.95 – 1.10) | 0.16 | 1.06 (0.97 – 1.15) | 0.18 | 1.06 (0.98 – 1.19) | 0.12 | 1.05 (0.97 – 1.13) | 0.22 | 1.07 (0.99 – 1.16) | 0.10 | |||
| Moderate | 1.38 (1.17 – 1.63) | <0.005 | 1.31 (1.08 – 1.58) | 0.01 | 1.20 (1.00 – 1.45) | 0.05 | 1.47 (1.24 – 1.24) | <0.005 | 1.48 (1.24 – 1.76) | <0.005 | |||
| Severe | 1.81 (1.39 – 2.35) | <0.005 | 1.73 (1.30 – 2.30) | <0.005 | 1.68 (1.27 – 2.23) | <0.005 | 1.81 (1.39 – 2.36) | <0.005 | 2.18 (1.66 – 2.85) | <0.005 | |||
|
| |||||||||||||
| Age | per 1 year increase | 1.05 (1.05 – 1.05) | <0.005 | 1.02 (1.01 – 1.02) | <0.005 | 1.02 (1.02 – 1.02) | <0.005 | 1.04 (1.04 – 1.05) | <0.005 | 1.03 (1.02 – 1.03) | <0.005 | ||
|
| |||||||||||||
| BAC LINEAR MODEL | |||||||||||||
| BAC | per 1mm2 increase | 1.02 (1.01 – 1.02) | <0.005 | 1.01 (1.01 – 1.02) | <0.005 | 1.01 (1.01 – 1.02) | <0.005 | 1.01 (1.01 – 1.02) | <0.005 | 1.02 (1.01 – 1.02) | <0.005 | ||
| Age | per 1 year increase | 1.05 (1.05 – 1.05) | <0.005 | 1.02 (1.01 – 1.02) | <0.005 | 1.02 (1.02 – 1.02) | <0.005 | 1.04 (1.04 – 1.05) | <0.005 | 1.03 (1.02 – 1.03) | <0.005 | ||
Analyses of BAC as a continuous variable further corroborated these findings. Each1 mm2 increase in BAC area was associated with a 1–2% increased risk across all outcomes (p<0.005) (Table 2). Similarly, a non-linear analysis showed that each doubling of the BAC area (a 1-unit increase in log2[BAC+1]) increased the risk of any major cardiovascular event by 7–19% (Supplementary Table 4). Age remained a significant independent predictor across all models.
Incremental Prognostic Value of BAC over PREVENT Scores
The PREVENT risk score was available for 4,876 patients (6.58%) in the internal cohort and 13,564 (27.9%) in the external cohort (Supplementary Table 3). The mean age was 59.1 years (SD: 10.0) for women in the internal cohort and 64.1 (SD: 10.0) in the external cohort. The incidence of MACE in this cohort was higher than the general cohort for both the internal cohort (9.9%) and external cohort (26.3%). Kaplan-Meier survival analysis stratified by PREVENT risk categories, demonstrated that the presence of moderate to severe BAC was associated with significantly lower MACE-free survival compared to those with zero BAC (Figure 4). Even after adjusting for PREVENT risk scores in multivariable Cox models, BAC remained a significant and independent predictor of MACE (Table 3). A dose-dependent relationship was again observed in both cohorts. In the internal cohort, mild, moderate, and severe BAC had hazard ratios of 1.32 (95%CI: 1.10–1.59), 1.75 (95%CI: 1.23–2.50), and 3.29 (95%CI: 2.15–5.05), respectively for any MACE,. And in the external cohort, the hazard ratios were 1.28 (95%CI: 1.17–1.39), 1.79 (95%CI: 1.55–2.06), and 2.80 (95%CI: 2.36–3.32) for mild, moderate, and severe BAC, respectively.
Figure 4.
Kaplan-Meier estimates of event-free survival in Low, Borderline, Intermediate and High Risk patients identified by PREVENT, stratified by Zero BAC and Moderate to Severe BAC
Table 3.
Association of BAC and clinical outcomes in PREVENT+BAC Model
| INTERNAL COHORT | ANY | AMI | STROKE | HF | DEATH | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR | p | HR | p | HR | p | HR | p | HR | p | |||
|
|
||||||||||||
| BAC CATEGORICAL MODEL | BAC | Zero | Reference | Reference | Reference | Reference | Reference | |||||
| Mild | 1.32 (1.10–1.59) | <0.005 | 1.31 (0.98–1.75) | 0.06 | 1.26 (1.00–1.58) | 0.05 | 1.24 (0.95–1.61) | 0.12 | 1.21 (0.92–1.57) | 0.17 | ||
| Moderate | 1.75 (1.23–2.50) | <0.005 | 1.31 (0.68–2.52) | 0.43 | 1.74 (1.09–2.77) | 0.02 | 2.18 (1.30–3.67) | <0.005 | 1.79 (1.03–3.10) | 0.04 | ||
| Severe | 3.29 (2.15–5.05) | <0.005 | 1.65 (0.62–4.37) | 0.31 | 2.06 (1.04–4.05) | 0.04 | 3.88 (2.00–7.54) | <0.005 | 5.24 (2.88–9.54) | <0.005 | ||
|
| ||||||||||||
| PREVENT risk | Low | Reference | Reference | Reference | Reference | Reference | ||||||
| Borderline | 0.89 (0.68–1.15) | 0.36 | 0.89 (0.64–1.25) | 0.51 | 0.87 (1.00–1.58) | 0.05 | 0.86 (0.62–1.18) | 0.34 | 0.82 (0.59–1.12) | 0.21 | ||
| Intermediate | 1.79 (1.53–2.09) | <0.005 | 1.07 (0.86–1.34) | 0.53 | 1.74 (1.09–2.77) | 0.02 | 1.15 (0.93–1.41) | 0.19 | 1.32 (1.08–1.62) | 0.01 | ||
| High | 3.69 (2.05–4.45) | <0.005 | 1.95 (1.45–2.61) | <0.005 | 2.43 (1.92–3.08) | <0.005 | 2.82 (2.18–3.66) | <0.005 | 2.27 (1.74–2.96) | <0.005 | ||
|
| ||||||||||||
| BAC LINEAR MODEL | BAC | per 1mm2 increase | 1.02 (1.02–1.03) | <0.005 | 1.02 (1.00–1.04) | 0.09 | 1.02 (1.01–1.03) | 0.01 | 1.03 (1.02–1.05) | <0.005 | 1.03 (1.02–1.05) | <0.005 |
| PREVENT | per 1 point increase | 1.05 (1.04–1.06) | <0.005 | 1.03 (1.02–1.04) | <0.005 | 1.04 (1.03–1.05) | <0.005 | 1.04 (1.03–1.05) | <0.005 | 1.04 (1.03–1.05) | <0.005 | |
|
| ||||||||||||
| EXTERNAL COHORT | ||||||||||||
| BAC CATEGORICAL MODEL | BAC | Zero | Reference | Reference | Reference | Reference | Reference | |||||
| Mild | 1.28 (1.17 – 1.39) | <0.005 | 1.13 (0.97 – 1.32) | 0.12 | 1.16 (1.00 – 1.35) | 0.04 | 1.23 (1.11 – 1.37) | <0.005 | 1.25 (1.09 – 1.43) | <0.005 | ||
| Moderate | 1.79 (1.55 – 2.06) | <0.005 | 1.42 (1.08 – 1.86) | 0.01 | 1.42 (1.09 – 1.83) | 0.01 | 1.68 (1.42 – 1.99) | <0.005 | 1.98 (1.59 – 2.45) | <0.005 | ||
| Severe | 2.80 (2.36 – 3.32) | <0.005 | 2.47 (1.79 – 3.40) | <0.005 | 2.11 (1.54 – 2.90) | <0.005 | 2.55 (2.08 – 3.14) | <0.005 | 3.80 (2.98 – 4.84) | <0.005 | ||
|
| ||||||||||||
| PREVENT risk | Low | Reference | Reference | Reference | Reference | Reference | ||||||
| Borderline | 0.90 (0.78 – 1.05) | 0.18 | 0.91 (0.72 – 1.15) | 0.44 | 0.77 (0.61 – 0.97) | 0.02 | 0.91 (0.76 – 1.09) | 0.29 | 0.83 (0.66 – 1.04) | 0.1 | ||
| Intermediate | 1.31 (1.21 – 1.41) | <0.005 | 1.11 (0.98 – 1.25) | 0.1 | 1.02 (0.90 – 1.14) | 0.79 | 1.32 (1.21 – 1.45) | <0.005 | 1.10 (0.98 – 1.23) | 0.11 | ||
| High | 2.40 (2.22 – 2.60) | <0.005 | 1.48 (1.29 – 1.70) | <0.005 | 1.54 (1.35 – 1.75) | <0.005 | 2.43 (2.22 – 2.67) | <0.005 | 2.08 (1.84 – 2.34) | <0.005 | ||
|
| ||||||||||||
| BAC LINEAR MODEL | BAC | per 1mm2 increase | 1.02 (1.02 – 1.03) | <0.005 | 1.02 (1.01 – 1.03) | <0.005 | 1.02 (1.01 – 1.02) | <0.005 | 1.02 (1.02 – 1.02) | <0.005 | 1.03 (1.03 – 1.03) | <0.005 |
| PREVENT | per 1 point increase | 1.03 (1.03 – 1.04) | <0.005 | 1.02 (1.01 – 1.02) | <0.005 | 1.01 (1.01 – 1.02) | <0.005 | 1.04 (1.03 – 1.04) | <0.005 | 1.03 (1.02 – 1.03) | <0.005 | |
When adjusting for PREVENT, BAC was not a significant predictor for acute myocardial infarction (AMI) in the internal cohort. However, in the external cohort, both moderate (HR: 1.42, 95%CI: 1.08–1.86, p<0.01) and severe BAC (HR: 2.47, 95%CI: 1.79–3.40, p<0.005) remained significantly associated with AMI.
Analysis of BAC as a continuous variable adjusted for PREVENT scores showed that each 1mm2 increase of BAC was associated with 2–3% increased risk across all outcomes (p<0.005), and each doubling of BAC was associated with an 11–27% increased risk across all outcomes (Table 3; Supplementary Table 4).
Model Performance
The discriminative performance of three competing Cox Proportional Hazards (CPH) models was evaluated using a 5×5-fold cross-validation protocol, revealing different patterns of model performance between the two cohorts (Supplementary Figure 11).
In the internal cohort, the PREVENT-based models outperformed the Age+BAC model. The addition of BAC to the PREVENT score resulted in the highest performing model. This was true for both categorical analysis, where the PREVENT+BAC model (median C-index: 0.721, IQR: 0.0030) was superior to the PREVENT-only model (median C-index: 0.737, IQR: 0.028) (p<0.0001), and for continuous analysis, the PREVENT+BAC model (median C-index: 0.741, IQR: 0.029) also significantly improved upon the PREVENT-Only model (median C-index: 0.737, IQR: 0.034) (p<0.01).
In contrast, the findings were different in the external cohort. Here, the Age+BAC model demonstrated higher discriminative ability than both the PREVENT-Only and the PREVENT+BAC models. However, the addition of BAC still provided incremental value, as the PREVENT+BAC model consistently outperformed the PREVENT-Only model. This pattern held for both categorical and continuous analyses.
An analysis of prediction errors of the BAC segmentation model was performed on an independent set of 500 exams. In this set, the model failed to detect BAC in only four cases (false negatives), with a median ground truth BAC of 1.77 mm2 (IQR: 1.60 mm2 – 2.43 mm2). The model produced 24 false positives, which were typically from skin folds (Supplementary Figure 6), with a median predicted area of 0.88 mm2 (IQR: 0.15 – 2.44 mm2). These findings indicate that the model’s prediction errors are small and confined to the lowest end of the mild BAC spectrum.
Discussion
Our study demonstrates that an automated, AI-driven quantification of breast arterial calcification on routine screening mammograms is a strong, independent predictor of adverse cardiovascular events in a large, multi-racial population. We show that this measure not only stratifies risk across a spectrum or severity but also offers significant prognostic value beyond the recently recommended PREVENT score(5,6), highlighting its potential as a widespread, opportunistic screening tool for women.
The prevalence of BAC identified by our model was 16.1% in the internal cohort and 26.0% in the external cohort. The higher prevalence in the external cohort is likely attributable to the older average age of its patients, but both values are consistent with the 12.3% to 29.4% prevalence range reported in the existing literature(21–24). Our findings are also consistent with a growing body of literature establishing the prognostic value of BAC for cardiovascular disease (8,21–23,25), including a recent meta-analysis by Koh et al.(9) which further substantiates the prognostic relevance of BAC by demonstrating significant associations between BAC and several cardiovascular outcomes including ischemic/hemorrhagic stroke, peripheral vascular disease, heart failure, and cardiac death. Interestingly however, that analysis did not find a statistically significant association between BAC and myocardial infarction (MI). Our results reflect this complexity where no significant association was found between BAC and AMI in our internal cohort after adjusting for PREVENT, but a significant association was present in our external cohort. This discrepancy may be due to differences in population characteristics. Nonetheless, our work strengthens the overall evidence linking BAC to a full spectrum of adverse cardiovascular events. This is largely because our quantitative deep learning approach directly answers a recent call to action in the field: towards an automated dedicated measurement of BAC and developing an “Agatston score for mammography”. (26) Our study represents a significant step towards the realization of this concept, demonstrating that an automated, quantitative measurement of BAC can indeed turn “promising but weak signals into more powerful absolute risk predictors.” This quantitative approach moves beyond the moderate inter-observer agreement (κ = 0.48–0.53)(8) of traditional radiologist scoring. While a recent study by Allen et al.(21) also utilized an AI tool, it relied on a relative scoring system in a less diverse, single-center cohort. Our use of absolute BAC area (mm2) offers a more standardized and clinically translatable metric.
This quantitative approach enables analysis of relationships between clinical risk factors and BAC severity. In our internal dataset, we observed that as BAC severity increased, there were significantly higher proportions of patients with diabetes mellitus, smoking history, and use of antihypertensive medications and statins. Systolic blood pressure also showed a positive correlation with BAC severity. These findings align with previous studies that reported higher prevalence of these risk factors in BAC-positive patients(22,27). While Hendriks et al. reported that smoking was associated with lower BAC prevalence(8), we found no association between BAC and smoking prevalence. Their study also found no associations between BAC and hypertension, obesity, or dyslipidaemia. This difference may be attributed to population heterogeneity or explained by treatment effect.
A key finding of our study is the incremental prognostic value of BAC when added to the established PREVENT risk prediction model. In CPH models adjusted for PREVENT, BAC remained a powerful, independent predictor of MACE with a clear dose-response relationship. Notably, our analyses suggest a potential threshold effect, with a disproportionate increase in MACE risk observed in patients transitioning from mild to moderate BAC (Figure 2). This indicates that while any BAC confers risk, a BAC area exceeding 10 mm2 may represent a critical tipping point into a significantly higher-risk state, warranting more aggressive clinical consideration. For instance, in the internal cohort, severe BAC was associated with at least threefold increased risk for any MACE (HR: 3.29, 95%CI: 2.15–5.05, p<0.005) and five-fold increased risk for all-cause death (HR: 5.24; 95%CI: 2.88–9.54, p<0.005) compared to zero BAC. This finding is remarkably consistent with prior work using qualitative radiologist assessment, which also reported an approximately 3-fold increased risk associated with severe BAC independent of traditional risk factors (28). This demonstrates that BAC captures a component of cardiovascular risk not fully accounted for by traditional risk factors. The clinical utility of this added value was confirmed by the significant improvement in model discrimination. Adding categorical BAC to the PREVENT models increased the C-index from 0.701 to 0.721 (p<0.0001) while the continuous PREVENT+BAC model achieved a median C-index of 0.741 (IQR: 0.029), indicating an improvement in the ability to predict which patients will experience an adverse event.
Beyond improving the overall model, BAC also provides crucial risk stratification within the established PREVENT categories. Kaplan-Meier analyses revealed that BAC’s utility extends across the entire risk spectrum (Figure 4 and Supplementary Figure 9), though its impact varied between our cohorts. In the internal cohort, moderate to severe BAC was prognostic for patients already classified as intermediate or high-risk, identifying individuals who may warrant more intensive management. In the external cohort, BAC was prognostic across all PREVENT risk categories (low, borderline, intermediate, and high). This finding is particularly significant, demonstrating that BAC can unmask clinically meaningful cardiovascular risk in women considered low-priority for preventive therapy while also refining risk estimates for those already identified as high-risk. Collectively, these results indicate that BAC is a versatile marker, acting both as a critical alert for hidden risk at the lower end of the spectrum and as a tool for further stratifying at the higher end. This is especially relevant given the practical limitations of current risk calculators. In our internal cohort of over 74,000 women, the data required to calculate a PREVENT score was available for only 6.58% of patients. This challenge in acquiring complete data for risk prediction tools is well-documented.(29) In contrast, screening mammography is performed on approximately 40 million women annually in the United States,(30) offering a readily available opportunity to identify at-risk women without requiring additional tests or data collection.
The prognostic value of BAC is also especially noteworthy in women under 50. While BAC is more prevalent with age, its presence in younger women, where it is far less common, is a particularly potent indicator of elevated cardiovascular risk. Our analysis confirmed that moderate to severe BAC in this younger cohort was associated with significantly lower MACE-free survival, identifying a high-risk group that might otherwise be overlooked by traditional risk models that are heavily weighted by age. This suggests BAC could help in guiding earlier and more aggressive preventive strategies for these younger women.
Finally, we demonstrate that our deep learning-based tool for automated BAC quantification is robust across a multi-site screening mammography population. The model performed well across Hologic, GE, and LORAD scanners (Supplementary Table 5), and its prognostic value remained significant even after stratifying by breast density, with Kaplan-Meier curves showing significant separations across the BAC severity levels (Supplementary Figure 10). Our model is also robust despite many mammographic artifacts such as mole markers, breast implants, biopsy clips, non-vascular calcifications, and scar markers (Supplementary Figure 4). Occasionally, we observed false positives for very dense ductal calcifications (Supplementary Figure 5), however this was infrequent, with very low predicted scores and can easily be distinguished from BAC (Supplementary Figure 6).
Limitations
This study is conducted using retrospective data from two US institutions and three scanner manufacturers, and further work is needed to ensure performance of the model across all scanners. Ongoing studies are underway to formally assess segmentation performance across all major manufacturers. Similarly, the population was approximately equally divided between Black and White patients for the internal dataset, however there are limited Asian, Hispanic, and Native American populations included in the internal or external datasets; therefore, generalizability to these populations requires further exploration. MACE and risk data was extracted from the EHR using previously validated diagnostic and procedural codes(17), but manual chart review or contact of patients for verification of outcomes was not practical given the large cohort and relatively low prevalence of disease. This could introduce bias from misclassification of events. Cardiovascular mortality is unreliable from EHR data, so all-cause mortality was used instead. Similarly, reproductive history such as menopause status or hormone replacement is associated with BAC(22), however these data were not reliably available in the EHR and therefore not included in the analyses. Because PREVENT score could only be calculated in 6.58% of women, conclusions about the additive value of BAC to traditional risk scores is limited by sample size. It is possible that women with complete data to calculate a PREVENT risk score may differ in underlying risk compared to women who do not have these data available.
While our findings demonstrate that BAC is significantly associated with adverse cardiovascular outcomes, we acknowledge that this retrospective observational study cannot establish causality. Rather, BAC should be considered a risk marker—not a causal risk factor—whose presence identifies women at elevated risk for myocardial infarction, stroke, heart failure, and mortality. Whether BAC is a surrogate for unmeasured cardiovascular burden or reflects distinct vascular pathophysiology, such as medial arterial calcification seen in diabetes or CKD, remains an important area for further study.
Because all-cause mortality was used as a component of our primary endpoint as described above, we were unable to perform a formal competing risk analysis, which would account for non-cardiovascular deaths as competing events for cardiovascular outcomes like myocardial infarction and stroke. Standard survival models, such as the Kaplan-Meier and Cox models used here, may overestimate the cumulative incidence of specific cardiovascular events in the presence of significant competing risks. However, it is important to note that our primary composite endpoint of any MACE included all-cause mortality, which is a robust, unbiased outcome not subject to this specific limitation. Nonetheless, the hazard ratios for individual non-fatal endpoints should be interpreted with this consideration in mind.
Finally, the model functions on full-field digital mammography (FFDM) which is a traditional 2D method of breast imaging. Digital breast tomosynthesis (DBT) is being increasingly used in the United States instead of FFDM(31,32), and therefore a BAC quantification model for DBT is needed to maximize capture of high-risk patients at mammography.
Conclusion
Automated BAC quantification from routine mammography may provide an opportunistic and effective cardiovascular risk assessment method in women, without additional radiation exposure. Its predictive value is independent of traditional risk factors and PREVENT scores suggesting that BAC screening could enhance early cardiovascular risk detection in women undergoing routine mammography.
Supplementary Material
Acknowledgements
We acknowledge the support from the AI Image Extraction Core (AI2EC) (RRID:SCR_026693), an Emory Integrated Core Facility.
REFERENCES
- 1.Mikhail GW. Coronary heart disease in women. BMJ. 2005. Sept 3;331(7515):467–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Wenger NK. Coronary heart disease: the female heart is vulnerable. Prog Cardiovasc Dis. 2003. Nov 1;46(3):199–229. [DOI] [PubMed] [Google Scholar]
- 3.Kelsey SF, James M, Holubkov AL, Holubkov R, Cowley MJ, Detre KM. Results of percutaneous transluminal coronary angioplasty in women. 1985–1986 National Heart, Lung, and Blood Institute’s Coronary Angioplasty Registry. Circulation. 1993. Mar;87(3):720–7. [DOI] [PubMed] [Google Scholar]
- 4.Khan SS, Coresh J, Pencina MJ, Ndumele CE, Rangaswami J, Chow SL, et al. Novel Prediction Equations for Absolute Risk Assessment of Total Cardiovascular Disease Incorporating Cardiovascular-Kidney-Metabolic Health: A Scientific Statement From the American Heart Association. Circulation. 2023. Dec 12;148(24):1982–2004. [DOI] [PubMed] [Google Scholar]
- 5.Khan SS, Matsushita K, Sang Y, Ballew SH, Grams ME, Surapaneni A, et al. Development and Validation of the American Heart Association’s PREVENT Equations. Circulation. 2024. Feb 6;149(6):430–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Zhou H, Zhang Y, Zhou MM, Choi SK, Reynolds K, Harrison TN, et al. Evaluation and Comparison of the PREVENT and Pooled Cohort Equations for 10‐Year Atherosclerotic Cardiovascular Risk Prediction. J Am Heart Assoc. 2025. Feb 18;14(4):e039454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Duhn V, D’Orsi ET, Johnson S, D’Orsi CJ, Adams AL, O’Neill WC. Breast Arterial Calcification: A Marker of Medial Vascular Calcification in Chronic Kidney Disease. Clin J Am Soc Nephrol. 2011. Feb;6(2):377–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Hendriks EJE, de Jong PA, van der Graaf Y, Mali WPTM, van der Schouw YT, Beulens JWJ. Breast arterial calcifications: a systematic review and meta-analysis of their determinants and their association with cardiovascular events. Atherosclerosis. 2015. Mar;239(1):11–20. [DOI] [PubMed] [Google Scholar]
- 9.Koh TJW, Tan HJH, Ravi PRJ, Sng JWZ, Yeo TC, Tan BYQ, et al. Association Between Breast Arterial Calcifications and Cardiovascular Disease: A Systematic Review and Meta-analysis. Can J Cardiol. 2023. Dec;39(12):1941–50. [DOI] [PubMed] [Google Scholar]
- 10.Hendriks EJE, de Jong PA, Beulens JWJ, Mali WPThM, van der Schouw YT, Beijerinck D. Medial Arterial Calcification: Active Reversible Disease in Human Breast Arteries. JACC Cardiovasc Imaging. 2015. Aug;8(8):984–5. [DOI] [PubMed] [Google Scholar]
- 11.Kemmeren JM, Van Noord PAH, Beijerinck D, Fracheboud J, Banga JD, Van Der Graaf Y. Arterial Calcification Found on Breast Cancer Screening Mammograms and Cardiovascular Mortality in Women: The DOM Project. Am J Epidemiol. 1998. Feb 15;147(4):333–41. [DOI] [PubMed] [Google Scholar]
- 12.Iribarren C, Go AS, Tolstykh I, Sidney S, Johnston SC, Spring DB. Breast Vascular Calcification and Risk of Coronary Heart Disease, Stroke, and Heart Failure. J Womens Health. 2004. May;13(4):381–9. [DOI] [PubMed] [Google Scholar]
- 13.Urooj A, O’Neill WC, Trivedi H, Banerjee I. Multi-Task Learning for Segmentation of Breast Arterial Calcifications in Mammograms. In: Medical Imaging meets NeurIPS. New Orleans; 2023. [Google Scholar]
- 14.Guo X, O’Neill WC, Vey B, Yang TC, Kim TJ, Ghassemi M, et al. SCU-Net: A deep learning method for segmentation and quantification of breast arterial calcifications on mammograms. Med Phys. 2021. Oct;48(10):5851–61. [DOI] [PubMed] [Google Scholar]
- 15.Jeong JJ, Vey BL, Bhimireddy A, Kim T, Santos T, Correa R, et al. The EMory BrEast imaging Dataset (EMBED): A Racially Diverse, Granular Dataset of 3.4 Million Screening and Diagnostic Mammographic Images. Radiol Artif Intell. 2023. Jan 1;5(1):e220047. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wang W, Xie E, Li X, Fan DP, Song K, Liang D, et al. Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV) [Internet]. 2021. [cited 2025 July 15]. p. 548–58. Available from: https://ieeexplore.ieee.org/document/9711179 [Google Scholar]
- 17.Bosco E, Hsueh L, McConeghy KW, Gravenstein S, Saade E. Major adverse cardiovascular event definitions used in observational analysis of administrative databases: a systematic review. BMC Med Res Methodol. 2021. Dec;21(1):241. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.AHA. professional.heart.org. 2023. [cited 2025 May 30]. The American Heart Association PREVENT Online Calculator. Available from: https://professional.heart.org/en/guidelines-and-statements/prevent-calculator [Google Scholar]
- 19.Detrano R, Guerci AD, Carr JJ, Bild DE, Burke G, Folsom AR, et al. Coronary Calcium as a Predictor of Coronary Events in Four Racial or Ethnic Groups. N Engl J Med. 2008. Mar 27;358(13):1336–45. [DOI] [PubMed] [Google Scholar]
- 20.McClelland RL, Jorgensen NW, Budoff M, Blaha MJ, Post WS, Kronmal RA, et al. 10-Year Coronary Heart Disease Risk Prediction Using Coronary Artery Calcium and Traditional Risk Factors: Derivation in the MESA (Multi-Ethnic Study of Atherosclerosis) With Validation in the HNR (Heinz Nixdorf Recall) Study and the DHS (Dallas Heart Study). J Am Coll Cardiol. 2015. Oct 13;66(15):1643–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Allen TS, Bui QM, Petersen GM, Mantey R, Wang J, Nerlekar N, et al. Automated Breast Arterial Calcification Score Is Associated With Cardiovascular Outcomes and Mortality. JACC Adv. 2024. Nov 1;3(11):101283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rotter MA, Schnatz PF, Currier AA, O’Sullivan DM. Breast arterial calcifications (BACs) found on screening mammography and their association with cardiovascular disease. Menopause N Y N. 2008;15(2):276–81. [DOI] [PubMed] [Google Scholar]
- 23.Schnatz PF, Marakovits KA, O’Sullivan DM. The association of breast arterial calcification and coronary heart disease. Obstet Gynecol. 2011. Feb;117(2 Pt 1):233–41. [DOI] [PubMed] [Google Scholar]
- 24.Margolies L, Salvatore M, Hecht HS, Kotkin S, Yip R, Baber U, et al. Digital Mammography and Screening for Coronary Artery Disease. JACC Cardiovasc Imaging. 2016. Apr;9(4):350–60. [DOI] [PubMed] [Google Scholar]
- 25.Iribarren C, Chandra M, Lee C, Sanchez G, Sam DL, Azamian FF, et al. Breast Arterial Calcification: a Novel Cardiovascular Risk Enhancer among Postmenopausal Women. Circ Cardiovasc Imaging. 2022. Mar;15(3):e013526. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Maas AHEM, de Jong PA. Mammograms to catch many birds with one stone. Eur Heart J. 2021. Sept 7;42(34):3371–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Bae MJ, Lee SY, Kim YJ, Lee JG, Jeong DW, Yi YH, et al. Association of breast arterial calcifications, metabolic syndrome, and the 10-year coronary heart disease risk: a cross-sectional case-control study. J Womens Health 2002. 2013. July;22(7):625–30. [DOI] [PubMed] [Google Scholar]
- 28.Hendriks EJE, Beulens JWJ, Mali WPThM, Beijerinck D, van der Graaf Y, de Jong PA, et al. Breast Arterial Calcifications and Their Association With Incident Cardiovascular Disease and Diabetes. JACC. 2015. Mar 3;65(8):859–60. [DOI] [PubMed] [Google Scholar]
- 29.Borden WB, Wang J, Jones P, Tang Y, Contreras J, Daugherty SL, et al. Reducing Cardiovascular Risk in the Medicare Million Hearts Risk Reduction Model: Insights From the National Cardiovascular Data Registry PINNACLE Registry. Circ Cardiovasc Qual Outcomes. 2022. Apr;15(4):e007908. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.American Cancer Society. Cancer Prevention & Early Detection [Internet]. 2024. [cited 2025 Feb 19]. Available from: https://www.cancer.org/research/cancer-facts-statistics/cancer-prevention-early-detection.html
- 31.Richman IB, Hoag JR, Xu X, Forman HP, Hooley R, Busch SH, et al. Adoption of Digital Breast Tomosynthesis in Clinical Practice. JAMA Intern Med. 2019. Sept 1;179(9):1292–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Lee CI, Zhu W, Onega TL, Germino J, O’Meara ES, Lehman CD, et al. The Effect of Digital Breast Tomosynthesis Adoption on Facility-Level Breast Cancer Screening Volume. AJR Am J Roentgenol. 2018. Nov;211(5):957–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.