An Artificial Intelligence Model to Detect Abnormal Ejection Fraction from Non-Contrast Chest Computed Tomography: The CT-LVEF study

Heart failure (HF), a major global health challenge, affects millions worldwide and poses substantial healthcare and economic burdens. The left ventricular ejection fraction (LVEF) is a critical dynamic parameter used to characterize HF and guide treatment. In this study, we developed and validated an artificial intelligence (AI) model capable of predicting abnormal LVEF directly from static, non-gated, non-contrast chest computed tomography (CT) scans, a novel application for an imaging modality typically used for unrelated indications. Using a multi-institutional dataset of 34,058 paired CT and echocardiogram studies from two academic centers, we trained our model on over 25,000 studies and validated it on 8,110 studies from a separate institution. Remarkably, our model demonstrated robust performance, achieving an area under the receiver operating characteristic curve (AUROC) of 0.786 on the hold-out test set and 0.755 on external validation. Our findings are particularly promising given the widespread availability of CT scans—over 80 million performed annually in the U.S.—making this opportunistic screening approach highly practical. Beyond strong predictive performance, the AI model outperformed expert radiologists in both accuracy and efficiency and provided interpretable visualizations highlighting imaging features linked to reduced LVEF. By enabling the identification of HF from routine chest CTs performed for other indications, this technology holds significant promise for early detection, reducing the diagnostic gap, and improving outcomes in asymptomatic HF.