A hybrid approach for machine learning based beat classification of ECG using different digital differentiators and...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Automated electrocardiogram (ECG) analysis is crucial for diagnosing cardiac conditions.
• Accurate ECG beat classification aids in identifying various arrhythmias.
• Existing methods may require improvements in feature extraction and QRS detection.

Purpose of the Study:

• To develop and evaluate a systematic approach for ECG beat classification using machine learning.
• To enhance the discriminative power of ECG signal features through novel filtering techniques.
• To improve the accuracy and robustness of automated cardiac arrhythmia detection.

Main Methods:

• Manual feature extraction using Dual-Tree Complex Wavelet Transform (DTCWT) for ECG signal analysis.
• Application of four novel digital filters for ECG signal differentiation and enhancement of QRS complex detection.
• Integration of DTCWT-derived morphological features with statistical features for a comprehensive feature set.
• Training and evaluation of various machine learning classifiers on the MIT-BIH Arrhythmia Database.

Main Results:

• The proposed methodology demonstrated high accuracy in classifying ECG beats into six distinct classes.
• Enhanced QRS detection was achieved through the integration of novel digital differentiators with the Pan-Tompkins algorithm.
• The comprehensive feature set improved the performance of machine learning classifiers for arrhythmia identification.
• Experimental validation on the complete MIT-BIH Arrhythmia Database confirmed the approach's robustness.

Conclusions:

• The developed machine learning approach offers an effective solution for automated ECG beat classification.
• The combination of DTCWT, novel filters, and statistical features significantly advances ECG signal processing.
• This research contributes to more precise and reliable automated analysis of cardiac signals for clinical applications.