Session S84.1

T-Wave Alternans: A Comparison of Different Measurement Techniques

D Zheng*, S Stevens, P Langley, K Wang,
AJ Haigh, A Murray

Newcastle University
Newcastle upon Tyne, UK

T-wave alternans (TWA) is characterized by a pattern of beat-to-beat alternations in the amplitude of the T-wave of the electrocardiogram (ECG). TWA has been shown to be a predictor of ventricular tachyarrhythmias and sudden cardiac death. However, TWA is usually very small and needs computational algorithms to detect or measure it. Our aim was to develop and compare three automated TWA measurement algorithms.
100 ECG recordings from the 2008 PhysioNet/Computers in Cardiology Challenge were analysed. ECGs were pre-processed with a bandpass filter (0.5-20 Hz) to remove baseline movement and noise. This was followed by QRS and T-wave detection.
The three algorithms were moving window technique, FFT spectral analysis and principal component analysis (PCA). For the moving window method, paired alternating T-wave amplitude changes in a four-beat window were calculated and averaged as the window moved through the data. For the FFT method, the power spectrum of T-wave amplitude at 0.5 cycle/beat was calculated. For the PCA method, the first principal component of T-waves was extracted and the FFT calculated for beat-to-beat change in coefficients of the first principal component.
The correlation of results between moving window and FFT was 0.67, between FFT and PCA 0.33, and between moving window and PCA 0.24. When a threshold was set to exclude potentially unreliable results, the number of ECGs with TWA was for moving window 29, FFT 29, and PCA 38. In only 7 of the ECG records was TWA detected by all three techniques. We have shown that different analysis techniques produced substantially different results.

(Abstract Control Number: 386)