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

Summary

In a 2,022-participant cohort with 13.9 years follow-up, AI-quantified left ventricular mass from routine non-contrast CT independently predicted MACE beyond coronary calcium score and improved discrimination and reclassification. LV mass, but not CAC, predicted cardiovascular death.

Key Findings

• AI-quantified LV mass from non-contrast CT independently predicted long-term MACE (HR 2.76, p<0.001) after multivariable adjustment.
• Adding LV mass to CAC improved AUC (0.753→0.767; p=0.031) and achieved a continuous NRI of 18% (p=0.011).
• LV mass predicted cardiovascular death (HR 3.89, p<0.001), whereas CAC did not.

Clinical Implications

Incorporating AI-derived LV mass into reporting for non-contrast CAC scans can refine long-term risk assessment and potentially guide preventive therapy intensity beyond CAC alone.

Limitations

• Single-trial cohort; external validation in diverse populations and scanners was not reported.
• Observational design limits causal inference; clinical utility thresholds for LV mass need prospective testing.

Future Directions

Prospective, multi-center validation and randomized implementation studies testing AI-LV mass–guided prevention strategies, integration with other AI-derived phenotypes, and cost-effectiveness analyses.