ECG heartbeat classification using progressive moving average transform.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Accurate electrocardiogram (ECG) heartbeat classification is crucial for diagnosing cardiac conditions.
• Traditional signal processing methods may struggle with the complexity and variability of ECG signals.
• The need for robust and efficient automated ECG analysis techniques is growing.

Purpose of the Study:

• To introduce the Progressive Moving Average Transform (PMAT) as a novel method for transforming time-domain signals into 2D representations.
• To integrate PMAT with a 2D-Convolutional Neural Network (CNN) for enhanced ECG heartbeat classification.
• To evaluate the performance and robustness of the PMAT-CNN approach across diverse ECG databases.

Main Methods:

• Developed the Progressive Moving Average Transform (PMAT) to create 2D signal representations using varying window size moving averages.
• Employed a 2D-Convolutional Neural Network (CNN) model to extract features and classify ECG heartbeats from PMAT-generated 2D data.
• Validated the approach using the MIT-BIH and INCART ECG databases, classifying over 6 heartbeat types into 3 main classes.

Main Results:

• Achieved high classification accuracy and F1-scores: 99.09% accuracy and 92.13% F1-score on the MIT-BIH database.
• Obtained 98.37% accuracy and 79.37% F1-score on the INCART database.
• Demonstrated robustness with >95% accuracy when models trained on one database were tested on another, including the ST-T European database.

Conclusions:

• The Progressive Moving Average Transform (PMAT) combined with 2D-CNN is a highly effective method for ECG heartbeat classification.
• The proposed approach exhibits excellent accuracy and stability across different datasets, indicating its reliability.
• PMAT shows significant potential for practical applications in medical diagnostics and healthcare systems for automated cardiac monitoring.