T wave alternans evaluation using adaptive time-frequency signal analysis and non-negative matrix factorization

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Sudden cardiac death (SCD) affects 400,000 North Americans annually, with risk identification remaining a challenge.
• T wave alternans (TWA) on electrocardiograms (ECGs) is a promising tool for stratifying cardiac patients at risk.
• Accurate TWA detection is crucial due to its microvolt signal range and susceptibility to noise and non-stationarity.

Purpose of the Study:

• To address limitations of existing TWA estimation methods (Spectral Method and Modified Moving Average).
• To propose novel, robust TWA quantification frameworks for improved sudden cardiac death risk assessment.
• To enhance the accuracy of TWA detection in noisy and non-stationary ambulatory ECG recordings.

Main Methods:

• Development of an Adaptive Spectral Method (SM) using non-linear time-frequency distribution (TFD).
• Introduction of a Non-Negative Matrix Factorization (NMF)-Adaptive SM technique for enhanced robustness.
• Evaluation using synthetic TWA signals in simulated and real-world ambulatory ECG data.

Main Results:

• The proposed Adaptive SM and NMF-Adaptive SM methods demonstrate effectiveness in TWA analysis.
• Numerical simulations confirm the accuracy of the novel approaches under noisy and non-stationary conditions.
• The developed algorithms show potential for improving the reliability of TWA detection.

Conclusions:

• The novel TWA quantification framework offers improved accuracy and robustness.
• These advanced methods hold promise for more effective sudden cardiac death risk stratification.
• Further application of these techniques could lead to better patient management and outcomes.