AHA Scientific Sessions 2026 Abstract · Nov 2, 2026
Hidden in Plain Sight: AI-Derived Coronary Calcium from Non-Gated Chest CT Predicts Long-Term Mortality Across Age, Sex, and Race
By Phan J, Reddy N, Nguyen LC et al.
Introduction/Background: Coronary artery calcium (CAC) is a established marker of atherosclerotic cardiovascular risk. Advances in artificial intelligence (AI) has enabled automated CAC quantification from non-gated chest computed tomography (CT) scans across diverse clinical settings. Research Questions/Hypothesis: We hypothesize that AI-derived CAC provides prognostic value for long-term all-cause mortality across heterogenous real-world imaging protocols and diverse patient populations. Methods/Approach: We analyzed 31,262 adults undergoing non-contrast, non-gated chest CT scans from 2010 to 2024 which vastly increased the dataset size of chest CTs compared to the original validation study. A previously validated deep learning was applied to quantify CAC. Patients were categorized as CAC 0, 1-100, 101-400, or > 400. A Cox proportional hazards regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (CI) for all-cause mortality across CAC categories. Kaplan-Meier analysis evaluated survival across CAC stratified across age, sex, and race. Results/Data (descriptive and inferential statistics): Among 31,262 patients, we demonstrated a graded rise in all-cause mortality across AI-CAC categories over the long-term. Compared to CAC of 0, a higher CAC burden was associated with increased mortality risk: CAC 1-100 (HR 1.79 [ 95% CI 1.66-1.92]), CAC 101-400 (HR 2.46 [ 2.3-2.64)], and CAC > 400 (HR 3.73 [3.51-3.96] with all p < 0.005. Kaplan-Meier analysis revealed early divergence with persistent separation of survival curves across our CAC strata. At 10 years, survival decreased stepwise from CAC 0 at 70% to 50.5%, 40.3%, and 25.5% in CAC 1-100, 101-400, and > 400 groups, respectively. CAC prevalence increased markedly with age and consistently higher in men than women across age group. CAC refines risk within age groups as higher CAC categories are associated with progressively increased mortality within each age group. Similarly, age-dependent increases in CAC prevalence were observed across European White and African American, with a higher CAC burden in men compared to women. CAC prevalence was generally higher among White patients compared to African American patients at younger and middle ages, but less so with advancing age. Conclusion(s): In conclusion, AI-derived CAC from routine non-gated CT was associated with increased mortality and improved risk stratification across diverse populations. This supports its use in clinical practice to identify high-risk individuals who may benefit from earlier preventive interventions.