DWT-CNNTRN: a Convolutional Transformer for ECG Classification with Discrete Wavelet Transform

Area of Science:

• Cardiology and Artificial Intelligence
• Biomedical Signal Processing

Background:

• Cardiovascular diseases are a leading global cause of mortality.
• Electrocardiograms (ECGs) are crucial for diagnosing cardiac conditions.
• Accurate and automated ECG classification is essential for clinical practice.

Purpose of the Study:

• To develop a novel, high-performance automated classifier for 12-lead ECG recordings.
• To improve diagnostic accuracy and efficiency in cardiovascular disease detection.
• To create a computationally efficient ECG classification model.

Main Methods:

• A hybrid deep learning model combining Convolutional Neural Networks (CNN) and transformer modules.
• Utilizing discrete wavelet transform of ECG signals as input features.
• Incorporating a global hybrid pooling layer for feature condensation.

Main Results:

• Achieved an average accuracy of 0.86 and an average F1-score of 0.83 on the China Physiological Signal Challenge 2018 dataset.
• Demonstrated strong performance in classifying specific conditions like Left Bundle Branch Block (LBBB), Right Bundle Branch Block (RBBB), and Incomplete Atrioventricular Block (I-AVB).
• The model exhibits a significantly smaller parameter count compared to existing individual and ensemble models.

Conclusions:

• The proposed model offers a highly accurate and efficient ECG classification solution.
• Its reduced model size enhances accessibility for automatic ECG analysis, particularly in resource-limited settings.
• This work contributes to advancing automated diagnostic tools for cardiovascular diseases.