An arrhythmia classification using a deep learning and optimisation-based methodology.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Electrocardiogram (ECG) signal analysis is crucial for diagnosing cardiac conditions.
• Traditional methods face challenges with noise and accurate classification.
• Deep learning offers potential for improved ECG interpretation.

Purpose of the Study:

• To develop a robust methodology for classifying five types of ECG signals.
• To enhance ECG signal quality by removing noise and baseline wander.
• To achieve high accuracy in automated ECG beat classification.

Main Methods:

• ECG signal preprocessing using moving average filter and discrete wavelet transformation.
• Image formation (grayscale and scalograms) from preprocessed signals.
• Feature extraction via EfficientNet-B0 deep learning model.
• Feature selection using a hybrid approach combining filter methods and Self Adaptive Bald Eagle Search (SABES) optimization.
• Z-score normalization for feature scaling.

Main Results:

• Successful segmentation of ECG signals using R-peak detection.
• Effective noise reduction, including baseline wandering and powerline interference.
• High-accuracy classification of five ECG signal classes, achieving 99.31% accuracy.
• Demonstration of the efficacy of the EfficientNet-B0 model and SABES optimization.

Conclusions:

• The proposed methodology provides an effective and accurate approach for ECG signal classification.
• The integration of deep learning and advanced feature selection significantly improves diagnostic capabilities.
• This work contributes a valuable tool for automated cardiac rhythm analysis.