Proposing feature engineering method based on deep learning and K-NNs for ECG beat classification and arrhythmia...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Arrhythmia, characterized by abnormal heart rhythms, poses diagnostic challenges due to ECG variability.
• Manual electrocardiogram (ECG) interpretation is prone to errors and requires specialized expertise.
• Computer-aided diagnosis systems are crucial for accessible arrhythmia detection, especially in remote areas.

Purpose of the Study:

• To develop and evaluate a novel feature engineering method for classifying ECG beats for arrhythmia detection.
• To classify four distinct heartbeat classes using the proposed methodology.
• To compare the performance of the proposed method against traditional machine learning and deep learning approaches.

Main Methods:

• A novel feature engineering technique combining deep learning and K-Nearest Neighbors (K-NNs) was developed.
• Extracted features were classified using various algorithms including decision trees, Support Vector Machines (SVMs), and random forests.
• A fivefold cross-validation strategy was employed to assess performance.

Main Results:

• The proposed method achieved high classification performance with an average Accuracy of 99.77%, AUC of 99.99%, Precision of 99.75%, and Recall of 99.30%.
• Demonstrated a favorable balance between sensitivity and specificity, indicating robust feature extraction capabilities.
• Exhibited significantly lower computational time compared to training deep learning models from scratch.

Conclusions:

• The developed feature engineering method provides a highly accurate and computationally efficient solution for ECG beat classification and arrhythmia detection.
• This approach offers a promising alternative to traditional machine learning and full deep learning models for automated cardiac rhythm analysis.
• The method's effectiveness supports its suitability for developing accessible computer-aided diagnosis tools for heart rhythm disorders.