Central Illustration.

Development and Testing of a Novel AI Model to Detect HFpEF From Single Echocardiographic Video Clips

A 3-dimensional convolutional neural network was developed (middle) to detect heart failure with preserved ejection fraction using only apical 4-chamber video clips. Discrimination performance was excellent (area under receiver-operating characteristic curve; bottom left), and age-adjusted risk of mortality was higher when patients received from the model a diagnostic output suggestive of heart failure with preserved ejection fraction compared to a diagnostic output not suggestive of heart failure with preserved ejection fraction (bottom right). AI = artificial intelligence; AUROC = area under receiver-operating characteristic curve; HFpEF = heart failure with preserved ejection fraction.