AI-derived automated quantification of cardiac chambers and myocardium from non-contrast CT: Prediction of major adverse cardiovascular events in asymptomatic subjects

Background and aims: The significance of left ventricular mass and chamber volumes from non-contrast computed tomography (CT) for predicting major adverse cardiovascular events (MACE) has not been studied. Our objective was to evaluate the role of artificial intelligence-enabled multi-chamber cardiac volumetry from non-contrast CT for long-term risk stratification in asymptomatic subjects without known coronary artery disease.

Methods: Our study included 2022 asymptomatic individuals (55.6 ± 9.0 years; 59.2 % male) from the EISNER (Early Identification of Subclinical Atherosclerosis by Noninvasive Imaging Research) trial. Multi-chamber cardiac volumetry was performed using deep-learning algorithms from routine non-contrast CT scans for coronary artery calcium scoring. MACE was defined as cardiac death, acute coronary syndrome, and late (>180 days) revascularization.

Results: A total of 215 individuals (11 %) suffered MACE at a mean follow-up of 13.9 ± 3 years. Individuals with MACE had higher left ventricular mass (115.1g vs. 105.2g, p < 0.001). In a multivariable analysis adjusted for cardiovascular risk factors and medications, left ventricular mass (HR 2.76, p<0.001) and coronary artery calcium score (HR 1.34, p<0.001) were independent predictors of long-term MACE. Adding left ventricular mass to the coronary calcium score improved the Receiver Operating Characteristic Area Under the Curve (AUC 0.753 vs 0.767, p=0.031) with continuous net reclassification index of 18 % (p=0.011). Left ventricular mass (HR 3.89, p<0.001), but not the coronary artery calcium score predicted cardiovascular death.

Conclusions: Left ventricular mass quantified automatically by AI from routine non-contrast CT independently predicted long-term MACE over and above the coronary calcium score in asymptomatic participants without known coronary artery disease.