Aortic and Pulmonary Artery Diameters Measured by AI on Coronary Calcium Scans and Incident Cerebrovascular Events, Atrial Fibrillation and Heart Failure Independent of Ventricular Geometry

Aortic and Pulmonary Artery Diameters Measured by AI on Coronary Calcium Scans and Incident Cerebrovascular Events, Atrial Fibrillation and Heart Failure Independent of Ventricular Geometry

Submitted manuscript | Scientific Publications

Study overview

This submitted MESA analysis examined whether AI-enabled measurements of the ascending aorta, descending aorta, and pulmonary artery on coronary artery calcium (CAC) scans were associated with future cerebrovascular events, atrial fibrillation (AF), and heart failure (HF), after accounting for conventional risk factors and cardiac ventricular geometry.

Design and methods

The study included 5,538 participants. Arterial diameters were measured using AI and indexed to body surface area. Adjusted Cox models estimated hazard ratios per standard-deviation increase, accounting for left ventricular volume index (LVVI) and the left-to-right ventricular volume ratio. The analyses were repeated in 4,153 participants whose LVVI was below the sex-specific 75th percentile.

Key findings

• During median follow-up of 16.9–19.6 years, the cohort experienced 424 cerebrovascular events, 1,111 AF events, and 417 HF events.

• A larger ascending aortic diameter was associated with cerebrovascular events (HR 1.21, 95% CI 1.06–1.39) and AF (HR 1.10, 95% CI 1.01–1.20).

• A larger pulmonary artery diameter was associated with AF (HR 1.07, 95% CI 1.02–1.13) and HF (HR 1.11, 95% CI 1.02–1.21).

• Among participants below the LVVI 75th percentile, associations between ascending aortic diameter and cerebrovascular events (HR 1.19, 95% CI 1.01–1.41) and AF (HR 1.16, 95% CI 1.05–1.29) remained.

Interpretation

The findings suggest that AI-derived aortic and pulmonary artery measurements from routine CAC scans may provide outcome-specific prognostic information beyond standard risk factors and ventricular geometry. As an observational analysis of a submitted manuscript, the results show associations and do not establish causation or clinical utility.

Download Abstract (DOCX)