A new particle swarm optimization-enhanced deep neural network for automatic ECG arrhythmias classification

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Electrocardiogram (ECG) signal analysis is crucial for diagnosing cardiac conditions.
• Automated classification systems require efficient hyperparameter optimization for deep neural networks (DNNs).
• Manual tuning of DNN hyperparameters is time-consuming and suboptimal.

Purpose of the Study:

• To develop an automated ECG classification system using particle swarm optimization (PSO) for DNN hyperparameter tuning.
• To enhance the accuracy and reliability of ECG heartbeat classification.
• To reduce the manual effort involved in optimizing deep learning models for ECG analysis.

Main Methods:

• ECG signals were processed using wavelet decomposition for feature extraction.
• Particle swarm optimization (PSO) was employed to tune five key DNN hyperparameters: neuron counts, dropout rate, learning rate, and optimizer.
• The classification was performed on the MIT-BIH Arrhythmia Database, distinguishing five heartbeat classes.

Main Results:

• The PSO-optimized DNN model achieved a classification accuracy of 99.76%.
• The system demonstrated high precision (99.34%) and F1 score (99.21%).
• The results indicate superior performance compared to traditional methods.

Conclusions:

• Particle swarm optimization is effective for automated hyperparameter tuning in ECG classification.
• The proposed system significantly improves the accuracy and reliability of automated heartbeat analysis.
• This approach offers a robust and efficient solution for clinical ECG interpretation.