Benchmarking of a T-wave alternans detection method based on empirical mode decomposition

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• T-wave alternans (TWA) is a key indicator for sudden cardiac death risk.
• Lack of a gold standard for TWA detection necessitates robust benchmarking methods.
• Synthetic signals are often used, limiting real-world applicability.

Purpose of the Study:

• Propose a novel signal model for TWA detection performance analysis.
• Methodologically validate an empirical mode decomposition (EMD) denoising technique.
• Evaluate the combined EMD and spectral method for TWA detection.

Main Methods:

• Developed a test bed using open-source databases and real ECG signals with physiological noise.
• Incorporated randomized TWA episodes for comprehensive testing.
• Analyzed sensitivity and specificity, employing Bootstrap resampling for performance validation.

Main Results:

• The EMD-enhanced method achieved outstanding specificity (0.96 vs. 0.72) in noisy conditions (SNR=8 dB).
• EMD also improved sensitivity in noisy environments (0.57 vs. 0.46), though it decreased in noiseless conditions.
• Specificity consistently surpassed the conventional spectral method alone.

Conclusions:

• The proposed test setting accurately reflects physiological variability for performance analysis.
• The EMD-based denoising significantly enhances TWA detection in noisy ECG signals.
• This approach aids in identifying patients at risk of fatal arrhythmias.