Automatic cardiac arrhythmias classification using CNN and attention-based RNN network.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Cardiac disease poses a significant public health threat, with 0.29 billion patients in China alone.
• Early diagnosis of cardiac conditions is crucial for reducing mortality and enhancing life quality.
• Electrocardiogram (ECG) signals are a vital, non-invasive, and cost-effective tool for heart disease diagnosis.

Purpose of the Study:

• To develop an automated classification model for distinguishing various cardiac arrhythmias.
• To leverage deep learning techniques, specifically Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN), for ECG analysis.
• To improve the accuracy of arrhythmia classification by incorporating an attention mechanism.

Main Methods:

• Utilized a hybrid model combining CNN for feature extraction and a bidirectional Gated Recurrent Unit (GRU) network for sequence analysis.
• Employed an attention mechanism to emphasize critical features within the ECG signal sequences.
• Evaluated the model on two datasets, including the MIT-BIH arrhythmia database and the China Physiological Signal Challenge 2018 database, addressing class imbalance.

Main Results:

• The proposed model achieved an average F1 score of 0.9110 on a public dataset.
• The model demonstrated strong performance on a subject-specific dataset with an average F1 score of 0.9082.
• The integration of CNN, bidirectional GRU, and attention mechanism proved effective in classifying cardiac arrhythmias.

Conclusions:

• The developed automatic classification model shows significant potential for practical application in diagnosing cardiac arrhythmias.
• The hybrid deep learning approach offers a robust method for analyzing ECG signals and improving diagnostic accuracy.
• Addressing class imbalance in datasets is critical for reliable performance in real-world cardiac monitoring.