Delineation of Electrocardiograms Using Multiscale Parameter Estimation.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Unobtrusive electrocardiography (ECG) monitoring demands robust algorithms to handle increased signal artifacts.
• Previous methods for ECG analysis, while accurate for synthetic data, struggled with real-world signal classification.
• Accurate delineation of heartbeat waves and fiducial points is essential for reliable ECG interpretation.

Purpose of the Study:

• To develop a general, mathematically sound delineation method for electrocardiogram (ECG) signals.
• To improve the accuracy of parameter estimation for heartbeat waves and fiducial points in ECG.
• To overcome limitations of previous classifiers in identifying relevant signal features for robust analysis.

Main Methods:

• A novel line classification approach using domain-specific pre-filtering was implemented.
• An exhaustive search identified optimal zero-crossing lines based on a mathematical error measure.
• The parameter estimation framework was adapted to compute all nine fiducial points with high precision.

Main Results:

• The proposed method achieved high sensitivity for P, QRS, and T waves (all >99.89%).
• Mean errors for onset and offset fiducial points were consistently below 1 millisecond.
• The algorithm demonstrated robust performance on the expert-annotated QT database.

Conclusions:

• The developed delineation method provides a robust and accurate approach for analyzing ECG signals.
• This framework effectively compensates for artifacts, enabling reliable parameter estimation in unobtrusive monitoring.
• The combination of advanced line classification and parameter estimation is highly suitable for ECG delineation.