CADNet: A Lightweight Neural Network for Coronary Artery Disease Classification Using Electrocardiogram Signals

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Coronary Artery Disease (CAD) significantly contributes to global mortality.
• Current diagnostic methods for CAD are often invasive, costly, and inconvenient.
• There is a need for accessible and accurate non-invasive diagnostic tools for CAD.

Purpose of the Study:

• To develop a lightweight deep learning model for classifying Coronary Artery Disease (CAD).
• To utilize non-invasive electrocardiography (ECG) signals for CAD diagnosis.
• To improve the efficiency and accuracy of CAD detection.

Main Methods:

• A novel one-dimensional convolutional neural network, named CADNet, was designed.
• CADNet incorporates Feature Encoding and Compact Pooling for ECG analysis.
• A unique data purification process was implemented to optimize the model.

Main Results:

• CADNet achieved an average accuracy of 99.3% across four diverse datasets.
• The model demonstrated high diagnostic performance with only 2,586 trainable parameters.
• CADNet surpassed the performance of existing state-of-the-art models.

Conclusions:

• The proposed CADNet model offers a highly accurate and efficient method for non-invasive CAD classification using ECG.
• This lightweight deep learning approach has the potential to enhance early detection and management of heart disease.
• CADNet represents a significant advancement in applying AI for cardiovascular diagnostics.