An artificial intelligence model to detect abnormal ejection fraction from non-contrast chest computed tomography:...

Area of Science:

• Artificial Intelligence in Medical Imaging
• Cardiology and Cardiovascular Diseases
• Radiology and Diagnostic Imaging

Background:

• Heart failure (HF) is a significant global health burden, with many patients having early systolic dysfunction remaining asymptomatic.
• Current medical therapies can prevent HF progression if initiated early, but diagnosis is often delayed in asymptomatic individuals.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) model for predicting abnormal left ventricular ejection fraction (LVEF) using non-contrast chest CT scans.
• To explore the potential of AI for opportunistic screening of systolic heart failure from routine imaging.

Main Methods:

• A multi-institutional dataset of 34,058 non-contrast chest CT scans and echocardiogram reports was used for training.
• The AI classification model predicted abnormal LVEF (EF < 50%) versus normal.
• External validation was performed on 8,110 additional paired CT and echocardiogram results.

Main Results:

• The AI model achieved an area under the receiver operating characteristic curve (AUROC) of 0.786 on the test set and 0.762 on external validation for detecting abnormal LVEF.
• The AI model demonstrated superior accuracy and efficiency compared to expert radiologists.
• Interpretable visualizations highlighted imaging features associated with reduced LVEF.

Conclusions:

• An AI model can accurately predict abnormal LVEF from static, non-gated, non-contrast chest CT scans.
• This novel application offers a promising method for opportunistic screening and early detection of systolic heart failure.
• The AI technology may help reduce the diagnostic gap and improve outcomes for asymptomatic HF patients.