Moderate evidence
More
AI scan analysis predicted heart events better than calcium score in a 1,702-patient study
Study title: Fully Automated, Explainable AI Analysis of Coronary CTA Improves Prediction of Major Cardiovascular Events
- Evidence type
- Observational study
- Sample
- n = 1,702
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-3087cite or check it
Study title: Fully Automated, Explainable AI Analysis of Coronary CTA Improves Prediction of Major Cardiovascular Events
Researchers developed and validated a fully automated, explainable AI system for analysing coronary CT angiography scans, called AI-ACA, and tested it on a retrospective cohort of 1,702 patients. It predicted major adverse cardiovascular events with an AUC of 0.81, compared with 0.76 for the Agatston score and 0.67 for demographic markers.
Read the study abstract
Coronary artery disease (CAD) remains a major global cause of morbidity and mortality. Coronary computed tomography angiography (CCTA) is central to noninvasive diagnosis and risk assessment, yet current workflows are labor-intensive and depend on semiquantitative scoring systems that do not fully leverage prognostic information in CCTA. The objectives of the study were to develop and validate a fully automated, explainable artificial intelligence (AI) system (AI-automated coronary analysis [ACA]) for comprehens…
The gain
In a retrospective study of 1,702 patients, a fully automated AI analysis of coronary CT scans predicted major cardiovascular events with an AUC of 0.81, versus 0.76 for the Agatston calcium score and 0.67 for demographic risk markers.
In brief
The system also separated patients' event-free survival more clearly than the Agatston score: its Harrell C-index was 0.791 versus 0.742, and the Kaplan-Meier log-rank statistic was 31.79 versus 13.71. Its performance matched an AI-derived CAD-RADS score at an AUC of 0.81.
Current CCTA workflows are labor-intensive and rely on semiquantitative scoring. The authors say the platform offers transparent, biomarker-based risk assessment and scalable analysis that may help bring quantitative CCTA into wider clinical use.
Main points
- The cohort was retrospective and included 1,702 patients.
- AI-ACA reached an AUC of 0.81 for major adverse cardiovascular events, against 0.76 for the Agatston score and 0.67 for demographic markers.
- Harrell C-index was 0.791 for AI-ACA versus 0.742 for the Agatston score.
- Kaplan-Meier log-rank statistic for event-free survival was 31.79 for AI-ACA versus 13.71 for the Agatston score.
- The system extracts stenosis severity, plaque burden and composition, AI-derived CAD-RADS, coronary tortuosity and subtended myocardial mass.
- Shapley Additive Explanations analysis ranked total plaque volume and burden, maximum stenosis severity and AI-CADRADS as the strongest contributors to risk estimates.
The full story
The study set out to build a fully automated system that analyses coronary CT angiography without manual intervention. The AI-ACA tool pulls out a set of imaging biomarkers, including stenosis severity, plaque burden and composition, an AI-derived CAD-RADS category, coronary tortuosity and the mass of heart muscle supplied by each vessel, and combines them into interpretable risk estimates.
In the 1,702-patient retrospective cohort, AI-ACA was compared with the Agatston score, standard demographic risk markers and AI-CADRADS. It outperformed the Agatston score and demographic markers on AUC, C-index and survival separation, but matched AI-CADRADS at an AUC of 0.81.
Explainability was assessed with Shapley Additive Explanations. Total plaque volume and burden, maximum stenosis severity and AI-CADRADS contributed most to the risk estimates, which the authors say makes the assessment transparent.
Sources
1 on file- JACC: Advancesdoi:10.1016/j.jacadv.2026.103461
Talk it over
The AI-ACA system reached an AUC of 0.81 versus 0.76 for the Agatston score, and matched AI-CADRADS at 0.81. How would you weigh that gap against the effort of adopting a new tool?
AI scan analysis predicted heart events better than calcium score in a 1,702-patient studyhttps://truvace.com/item/fully-automated-explainable-ai-analysis-of-coronary-cta-improves-prediction-of-m
The AI-ACA system reached an AUC of 0.81 versus 0.76 for the Agatston score, and matched AI-CADRADS at 0.81. How would you weigh that gap against the effort of adopting a new tool?
Answer in the threadThe study looked back at 1,702 patients rather than following new ones forward. What can a retrospective cohort tell us about heart events, and where might it mislead?
Answer in the threadTotal plaque volume and burden, maximum stenosis severity and AI-CADRADS drove the risk estimates. Which result from a prospective trial would change your mind about whether AI-ACA should replace the Agatston score?
Answer in the thread
If you have more time
Look up how the Agatston calcium score is used in a clinic you know of. What would a cardiologist need to see before using AI-ACA instead?
Answer in the threadWhich patients might the 1,702-patient cohort not represent, and how could that change the AUC of 0.81?
Answer in the thread
ace
What readers say