Performance Evaluation of Time-Frequency Distributions for ECG Signal Analysis

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Biomedical signals like ECG are non-stationary and multi-frequency, necessitating time-frequency analysis (TFA).
• TFA offers simultaneous time and frequency domain interpretations crucial for comprehensive ECG signal analysis.
• A variety of Time-Frequency Distributions (TFDs) exist, each with unique properties, highlighting the need to identify the optimal TFD for ECG analysis.

Purpose of the Study:

• To evaluate the performance of five different TFDs for detecting abnormalities in ECG signals.
• To determine the most reliable TFD for ECG abnormality detection, particularly for automated systems.

Main Methods:

• Extracted features from the QRS complex of ECG signals to establish detection criteria for normal and abnormal cases.
• Estimated the energy concentration magnitude of ECG signal components using five TFDs: Dual-Tree Wavelet Transform, Spectrogram, Pseudo Wigner-Ville, Choi-Williams, and Born-Jordan.
• Analyzed one-minute ECG signal intervals from the MIT-BIH normal sinus rhythm database (18 long-term records at 128 Hz), randomly selecting ten-minute intervals per record.

Main Results:

• The Choi-Williams distribution demonstrated the highest reliability in detecting ECG abnormalities.
• Achieved a detection accuracy of 99.44% across 15,735 analyzed ECG beats.
• The analysis of one-minute intervals, a novel approach, proved effective for long-term ECG monitoring.

Conclusions:

• The Choi-Williams distribution is the most reliable TFD for detecting heart problems using ECG signals.
• This TFD is particularly suitable for automated systems requiring continuous, long-term cardiac monitoring.
• The proposed one-minute interval analysis method enhances the effectiveness of TFDs for ECG abnormality detection.