Analysis of Cardiac Arrhythmias Based on ResNet-ICBAM-2DCNN Dual-Channel Feature Fusion

Area of Science:

• Cardiology and Biomedical Engineering
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Cardiovascular disease (CVD) is a major global health concern, with cardiac arrhythmia being a common complication.
• Accurate arrhythmia classification is crucial for effective CVD management and patient outcomes.
• Current ECG analysis methods face challenges in precision and objectivity.

Purpose of the Study:

• To introduce an innovative dual-channel feature fusion strategy for enhanced Electrocardiogram (ECG) signal processing.
• To improve the accuracy and objectivity of cardiac arrhythmia classification.
• To advance the clinical application of ECG analysis for better patient care.

Main Methods:

• Utilized Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) and wavelet thresholding for noise reduction.
• Employed a ResNet-ICBAM model for feature extraction in the primary channel.
• Transformed 1D ECG signals into Gram angular difference field (GADF), Markov transition field (MTF), and recurrence plot (RP) for 2D-CNN feature extraction in the secondary channel.
• Fused features from both channels for final classification.

Main Results:

• Achieved a classification accuracy of 97.80% on the MIT-BIH arrhythmia database.
• Demonstrated significant improvement in overall performance compared to existing methods.
• The dual-channel feature fusion strategy, incorporating a 2D convolutional network, proved highly effective.

Conclusions:

• The proposed dual-channel feature fusion method offers a substantial advancement in ECG signal processing for arrhythmia classification.
• This approach enhances accuracy and objectivity, holding significant potential for clinical applications.
• The methodology promises to improve the efficiency and precision of patient care in managing cardiovascular diseases.