Feature extraction of ECG signal

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Electrocardiogram (ECG) signal analysis is crucial for diagnosing cardiac conditions.
• Accurate feature extraction from ECG signals is often challenging due to noise and variability.
• Existing methods for ECG analysis may lack precision in identifying critical waveform components.

Purpose of the Study:

• To develop and validate a new approach for analyzing ECG signals.
• To extract comprehensive diagnostic features from ECG data.
• To improve the accuracy and reliability of ECG-based diagnostics.

Main Methods:

• Noise elimination using Maximal Overlap Discrete Wavelet Transform (MODWT) and universal thresholding.
• R-peak detection via discrete wavelet transform and thresholding.
• Identification of P, Q, S, T waves, and QRS complex using rule-based algorithms.
• Computation of advanced features based on extracted fiducial points.

Main Results:

• The developed software was validated using the MIT-BIH ECG database.
• Achieved high accuracy in beat detection: 99.98% sensitivity, 99.97% predictivity, and 0.05% error.
• Demonstrated superior performance compared to existing ECG analysis techniques.

Conclusions:

• The proposed method offers a robust and accurate approach for ECG signal analysis.
• Effective noise reduction and precise feature extraction enhance diagnostic capabilities.
• This technique holds significant potential for improving the diagnosis and monitoring of cardiovascular diseases.