Comparative analysis of methods for automatic detection and quantification of microvolt T-wave alternans

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Microvolt T-wave alternans (TWA) signifies susceptibility to malignant ventricular arrhythmias.
• Automatic detection techniques are necessary for identifying TWA.
• Several methods exist, but their performance in detecting stationary and time-varying TWA varies.

Purpose of the Study:

• To compare the efficacy of five automatic TWA detection methods: FFTSM, CDM, MMAM, LLRM, and AMFM.
• To evaluate their ability to accurately identify stationary and time-varying TWA.
• To assess their performance in avoiding false-positive TWA detections.

Main Methods:

• Application of five distinct TWA detection algorithms (FFTSM, CDM, MMAM, LLRM, AMFM) to simulated and clinical ECG data.
• Comparative analysis of method performance based on accuracy, sensitivity to TWA variations, and false-positive rates.
• Utilizing simulated ECGs with controlled variability and real Holter recordings from healthy and patient cohorts.

Main Results:

• The modified-moving-average method (MMAM) produced false positives with amplitude variability but accurately quantified stationary TWA.
• The adaptive-match-filter method (AMFM) demonstrated superior performance in identifying time-varying TWA.
• Fast-Fourier-transform spectral method (FFTSM) misclassified non-stationary TWA as stationary; MMAM introduced signal delay; CDM and LLRM were limited to slow-varying TWA.

Conclusions:

• The adaptive-match-filter method (AMFM) offers the best compromise for detecting true TWA while minimizing false positives.
• Simulation results aided in explaining discrepancies in TWA measurements from clinical data across different methods.
• Accurate TWA detection is vital for risk stratification in patients prone to arrhythmias.