AI-Enabled CT Cardiac Chamber Volumetry Predicts Atrial Fibrillation and Stroke Comparable to MRI

Published research | JACC: Advances, November 2024

This study evaluates whether AI-enabled automated left atrial (LA) volumetry from routine noncontrast coronary artery calcium (CAC) CT scans can predict atrial fibrillation (AF) and stroke as effectively as expert-measured LA volume on cardiac MRI (CMRI). The work also compares these imaging measures with established risk markers.

Study design

The analysis used 15-year outcome data from 3,552 asymptomatic MESA participants who underwent both CAC CT and CMRI at baseline. CMRI LA volume had been measured by human experts. The investigators compared time-dependent discrimination for incident AF and stroke using AI-CAC LA volume, CMRI LA volume, CHARGE-AF risk score, Agatston CAC score, and NT-proBNP.

Key findings

During 15 years of follow-up, 562 AF cases and 140 strokes occurred. Discrimination was similar for AI-CAC and CMRI volumetry: AUC 0.802 versus 0.798 for AF, and 0.762 versus 0.751 for stroke; the differences were not statistically significant. Adding AI-CAC LA volume improved reclassification for 5-year AF prediction when added to CHARGE-AF (continuous NRI 0.23), NT-proBNP (0.37), or Agatston score (0.44; P<0.0001 for all).

Why it matters

The findings suggest that LA volume can be quantified from noncontrast CAC scans already obtained in clinical care, with prognostic performance comparable to expert CMRI measurements in this cohort. The authors note that further studies are needed to establish the clinical utility of AI-CAC for AF and stroke prediction.

Publication

Naghavi M, Reeves AP, Atlas KC, et al. AI-Enabled CT Cardiac Chamber Volumetry Predicts Atrial Fibrillation and Stroke Comparable to MRI. JACC: Advances. 2024;3(11):101300. doi:10.1016/j.jacadv.2024.101300. Open access under CC BY-NC-ND.

Download the full article

Download Full Article (PDF)