Echocardiographic video-driven multi-task learning model for coronary artery disease diagnosis and severity grading

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Echocardiography is a primary noninvasive method for diagnosing coronary artery disease (CAD).
• Current visual assessments by experts are time-consuming, highlighting the need for automated solutions.

Purpose of the Study:

• To develop and validate an AI model, Intelligent echo for CAD (IE-CAD), for automated CAD screening and stenosis grading.
• To improve the efficiency and accuracy of CAD diagnosis using echocardiographic video analysis.

Main Methods:

• Constructed an echocardiographic video-driven multi-task learning model (IE-CAD) using a 3DdeeplabV3+ backbone.
• Integrated multi-task learning to capture dynamic myocardial contours and extract multifarious features.
• Trained and tested the model on 870 echocardiographic videos from Beijing Hospital and validated on 450 videos from Fuwai Hospital.

Main Results:

• The IE-CAD model achieved an AUC of 0.78 and sensitivity of 0.85 for detecting significant/severe CAD.
• Demonstrated a Pearson correlation coefficient of 0.545 for predicting the Gensini score.
• Achieved an AUC of 0.77 and sensitivity of 0.78 on an external validation dataset.

Conclusions:

• The IE-CAD model shows significant potential for effective CAD diagnosis and grading in patients with clinical suspicion.
• This AI-driven approach can enhance the diagnostic workflow for coronary artery disease.