ECG data enhancement method using generate adversarial networks based on Bi-LSTM and CBAM

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Electrocardiogram (ECG) classification algorithms struggle with imbalanced data, leading to poor performance on minority classes.
• This data imbalance negatively impacts the overall accuracy and reliability of ECG diagnostic models.
• Existing methods often fail to adequately address the challenge of limited data for rare cardiac conditions.

Purpose of the Study:

• To develop and validate a Generative Adversarial Network (GAN)-based data augmentation method for imbalanced ECG datasets.
• To improve the classification performance of ECG analysis models, particularly for underrepresented cardiac arrhythmias.
• To enhance the overall accuracy and robustness of ECG classification systems.

Main Methods:

• Proposed a novel ECG data augmentation approach using a Generative Adversarial Network (GAN).
• Integrated bidirectional long short-term memory (Bi-LSTM) networks and a convolutional block attention mechanism (CBAM) within the GAN framework.
• Validated the method on the MIT-BIH arrhythmia (MIT-BIH-AR) and Chinese cardiovascular disease (CCDD) databases, assessing generated ECG signal quality using metrics like Frechet distance and DTW.

Main Results:

• Achieved 99.46% accuracy for a 15-type heartbeat classification task on the MIT-BIH-AR database.
• Improved ventricular premature contraction (PVC) detection accuracy to 99.15% on the CCDD database.
• Demonstrated significant improvements in ECG classification performance through data enhancement.

Conclusions:

• The proposed GAN-based data augmentation method effectively addresses data imbalance in ECG classification.
• The integration of Bi-LSTM and CBAM enhances the generation of high-quality synthetic ECG data.
• This approach offers a promising solution for improving the diagnostic accuracy of ECG analysis tools.