Heartbeat classification based on single lead-II ECG using deep learning.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Electrocardiogram (ECG) signal analysis is crucial for diagnosing cardiovascular diseases.
• Manual ECG interpretation is complex and time-consuming.
• Machine learning offers potential for automated ECG classification.

Purpose of the Study:

• To develop and validate a deep neural network model with residual blocks (DNN-RB) for classifying cardiac cycles into six ECG beat classes.
• To assess the performance of the DNN-RB model against state-of-the-art algorithms.

Main Methods:

• A deep neural network model incorporating residual blocks (DNN-RB) was designed.
• The model was trained and validated using the MIT-BIH dataset.
• Performance metrics included test accuracy, average sensitivity, and average specificity.

Main Results:

• The DNN-RB model achieved a test accuracy of 99.51%.
• Average sensitivity was 99.7%, and average specificity was 98.2%.
• The proposed method demonstrated superior performance compared to other state-of-the-art algorithms on the same dataset.

Conclusions:

• The DNN-RB model is effective for automatic ECG signal classification.
• The method shows promise for clinical and out-of-hospital monitoring using mobile ECG devices.
• A web application integrating the DNN-RB model facilitates ECG analysis and diagnosis.