Automated Aortic Regurgitation Detection and Quantification: A Deep Learning Approach Using Multi-View...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Accurate assessment of aortic regurgitation (AR) severity is crucial for timely intervention and chronic disease management.
• Doppler echocardiography is the standard for AR assessment, but its accuracy can be limited by image quality and the need for multi-view integration.
• Existing methods require expert interpretation and can be subjective, highlighting the need for automated, objective assessment tools.

Purpose of the Study:

• To develop and validate a deep learning model for automated assessment of aortic regurgitation severity.
• To evaluate the model's performance using multi-view color Doppler echocardiography videos.
• To assess the generalizability of the model in an external validation cohort.

Main Methods:

• A convolutional neural network (R2+1D) was developed to classify AR severity from five standard echocardiographic views.
• The model was trained on a large dataset of 47,638 videos from 32,396 studies at Cedars-Sinai Medical Center.
• External validation was performed on 3,369 videos from 1,504 studies at Stanford Healthcare Center.

Main Results:

• The EchoNet-AR model demonstrated high accuracy in identifying at least moderate AR (AUC 0.95) and severe AR (AUC 0.97) in the training cohort.
• Consistent performance was observed in the external validation cohort, with AUCs of 0.92 for at least moderate AR and 0.94 for severe AR.
• The model showed robust performance across varying image quality, valve morphologies, and patient demographics, focusing on hemodynamically significant regions.

Conclusions:

• The EchoNet-AR model accurately classifies AR severity by synthesizing information from multiple echocardiographic views.
• The model exhibits robust generalizability and potential as an automated clinical decision support tool for AR assessment.
• Clinical interpretation remains essential, especially for complex cases involving multiple valve pathologies or altered hemodynamics.