ViViEchoformer: Deep Video Regressor Predicting Ejection Fraction.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Heart disease is a leading global cause of mortality.
• Accurate ejection fraction (EF) measurement is crucial for cardiac function assessment and patient outcomes.
• Human interpretation of echocardiograms for EF measurement has inherent variability.

Purpose of the Study:

• To introduce ViViEchoformer, a deep learning model utilizing a video vision transformer.
• To enable fully automatic and accurate prediction of left ventricular ejection fraction (LVEF) from echocardiogram videos.
• To provide a reliable alternative to manual echocardiogram interpretation.

Main Methods:

• Developed ViViEchoformer, a deep learning approach using a video vision transformer.
• Trained the model on 10,030 apical-4-chamber echocardiography videos from Stanford University Hospital.
• The model extracts spatiotemporal tokens to capture spatial information and inter-frame relationships.

Main Results:

• ViViEchoformer achieved a mean absolute error of 6.14% for EF prediction.
• The model demonstrated strong performance in predicting heart failure with reduced ejection fraction (HFrEF) with an AUC of 0.83.
• Achieved a classification accuracy of 87% for identifying HFrEF.

Conclusions:

• ViViEchoformer provides precise, automated left ventricular function quantification from echocardiogram videos.
• The model offers a reliable and accurate alternative to human assessment, reducing reader variance.
• Establishes a foundational tool for enhancing echocardiogram interpretation in clinical practice.