Coronary Artery Calcium Scans Powered by Artificial Intelligence Predicts Atrial Fibrillation Comparably to Cardiac Magnetic Resonance Imaging: The Multi-Ethnic Study of Atherosclerosis (MESA)

BackgroundApplying artificial intelligence to coronary artery calcium computed tomography scan (AI-CAC) provides more actionable information beyond the Agatston coronary artery calcium (CAC) score. We have recently shown that AI-CAC automated left atrial (LA) volumetry enabled prediction of atrial fibrillation (AF) in as early as one year. In this study we evaluated the performance of AI-CAC automated LA volumetry versus LA volume measured by human experts using cardiac magnetic resonance imaging (CMRI) for predicting AF, and compared them with CHARGE-AF risk score, Agatston score, and NT-proBNP (BNP). MethodsWe used 15-year outcome data from 3552 asymptomatic individuals (52.2% women, ages 45-84 years) who underwent both CAC scans and CMRI in the baseline examination (2000-2002) of the Multi-Ethnic Study of Atherosclerosis (MESA). AI-CAC took on average 21 seconds per scan. CMRI LA volume was previously measured by human experts. Data on BNP, CHARGE-AF risk score and the Agatston score were obtained from MESA. ResultsOver 15 years follow-up, 562 cases of AF accrued. The ROC AUC for AI-CAC versus CMRI and CHARGE-AF were not significantly different (AUC 0.807, 0.808, 0.800 respectively, p=0.60). The AUC for BNP (0.707) and Agatston score (0.694) were significantly lower than the rest (p<.0001). AI-CAC and CMRI significantly improved the continuous Net Reclassification Index (NRI) for prediction of AF when added to CHARGE-AF risk score (0.28, 0.31), BNP (0.43, 0.32), and Agatston score (0.69, 0.41) respectively (p for all<0.0001). ConclusionAI-CAC automated LA volumetry and CMRI LA volume measured by human experts similarly predicted incident AF over 15 years.