ECG-based multi-class arrhythmia detection using spatio-temporal attention-based convolutional recurrent neural...

Area of Science:

• Cardiology
• Artificial Intelligence
• Signal Processing

Background:

• Automatic arrhythmia detection using electrocardiogram (ECG) is crucial for early cardiac disease diagnosis.
• Deep learning, particularly Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), shows promise but often overlooks feature importance.
• Existing CNN-RNN models for arrhythmia detection can be improved by considering differential contributions of spatial and temporal features.

Purpose of the Study:

• To develop a Spatio-Temporal Attention-based Convolutional Recurrent Neural Network (STA-CRNN) for enhanced multi-class arrhythmia detection.
• To improve classification performance by focusing on representative features across spatial and temporal dimensions of ECG signals.
• To provide a tool that assists cardiologists in diagnosing arrhythmias.

Main Methods:

• Proposed a STA-CRNN architecture integrating CNN and RNN subnetworks with spatio-temporal attention modules.
• The model was trained and evaluated on a public dataset for classifying 8 types of arrhythmias and normal rhythms.
• Feature importance was analyzed using visualization techniques to align with clinical judgment.

Main Results:

• STA-CRNN achieved an average F1 score of 0.835 in classifying 9 cardiac rhythm types.
• Demonstrated significant improvement over state-of-the-art methods on the same dataset for most arrhythmia classifications.
• Visualizations confirmed that learned features correspond to clinically relevant patterns.

Conclusions:

• STA-CRNN offers a promising advancement in automatic arrhythmia detection through effective feature learning.
• The model's ability to focus on salient spatio-temporal features enhances diagnostic accuracy.
• STA-CRNN has the potential to serve as a valuable assistive tool for cardiologists in arrhythmia diagnosis.