Automatic ECG Classification Using Continuous Wavelet Transform and Convolutional Neural Network

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Cardiovascular disease (CVD) mortality can be reduced through early arrhythmia detection and effective treatment.
• Current clinical diagnosis of arrhythmia relies on time-consuming, manual electrocardiogram (ECG) analysis.
• Automated diagnostic tools are needed to streamline ECG interpretation and improve patient outcomes.

Purpose of the Study:

• To develop and evaluate an automatic ECG classification method for improved arrhythmia detection.
• To enhance the accuracy and efficiency of ECG analysis compared to traditional methods.
• To explore the utility of Continuous Wavelet Transform (CWT) and Convolutional Neural Networks (CNN) in ECG signal processing.

Main Methods:

• ECG signals were decomposed using Continuous Wavelet Transform (CWT) to generate time-frequency representations (scalograms).
• Convolutional Neural Networks (CNNs) were employed to extract features from these scalograms.
• Additional RR interval features were extracted and combined with CNN features for classification.

Main Results:

• The proposed method achieved high accuracy (98.74%) and a notable F1-score (68.76%) on the MIT-BIH arrhythmia database.
• Compared to existing methods, the F1-score showed a significant increase of 4.75% to 16.85%.
• Positive predictive value and sensitivity were reported at 70.75% and 67.47%, respectively.

Conclusions:

• The developed CWT-CNN based method offers a simple yet highly accurate approach for automatic ECG classification.
• This automated method demonstrates potential as a valuable auxiliary diagnostic tool in clinical settings for arrhythmia detection.
• The findings suggest a promising advancement in leveraging AI for cardiovascular disease management.