Nonlinear adaptive wavelet analysis of electrocardiogram signals

Area of Science:

• Biomedical Engineering
• Signal Processing
• Nonlinear Dynamics

Background:

• Electrocardiogram (EKG) signals are nonstationary, containing complex waveforms vital for diagnosing cardiac conditions.
• Traditional time-frequency analysis faces challenges with the inherent variability and noise in EKG signals.
• Extracting compact, robust features from EKG patterns is crucial for accurate medical diagnosis.

Purpose of the Study:

• To develop an adaptive wavelet-based method for extracting a compact fiducial pattern from EKG signals.
• To leverage nonlinear dynamics for improved EKG signal representation and feature extraction.
• To enhance the efficiency and accuracy of EKG analysis algorithms.

Main Methods:

• Utilized wavelet representation for time-frequency analysis of nonstationary EKG signals.
• Extracted a fiducial EKG pattern based on nonlinear dynamics, specifically eigenfunctions from Poincare sections.
• Adapted wavelet functions to the extracted fiducial pattern.

Main Results:

• Achieved a two-orders-of-magnitude (95%) more compact EKG representation in terms of Shannon signal entropy.
• Demonstrated that the compact representation enhances feature extraction, reducing sensitivity to noise and variations.
• Showcased potential for more efficient algorithms in EKG beat detection, QRS cancellation, and event measurement.

Conclusions:

• The proposed adaptive wavelet approach offers a significantly more compact representation of EKG signals.
• This method facilitates robust feature extraction, crucial for reliable diagnosis of cardiac disorders.
• The technique holds promise for advancing automated EKG analysis and diagnostic tools.