Assessment of Micro T-Wave Alternans Using T-Wave Morphology-Based Methods

Abstract

Microvolt T-wave alternans (MTWA) is a risk marker for life threatening arrhythmias, defined as beat-by-beat alternation of the T-wave amplitudes. Sensitive to noise, MTWA classical method (CM) requires accurate determination of T-wave peaks. This study assessed alternative methods prescinding this limitation. Fifty ECG recordings from PhysioNet T-Wave Alternans Database were assessed (channels 1 and 2). A sequence of 128 sinus beats was extracted from each signal. The FFT of the 128 T-wave peak value series was employed to perform CM, taken as reference. Further, T-waves were isolated and consecutively concatenated, composing an artificial continuous (AC) signal. Over AC two methods were investigated: (i) Hilbert Transform approach (HT)—assessed by the FFT of the AC envelope, calculated by HT; (ii) Short-time spectral approach (STS)—assessed in the average of 96 FFT obtained from moving windows with 32 consecutive T-waves, displaced one by one. MTWA was calculated as ‘alternans ratio’, (alternans peak divided by the SD of 10 vicinal harmonics). CM versus HT and CM versus STS were compared by correlation coefficient, Bland-Altman charts and ROC. Correlation coefficient were, respectively, channel 1: 0.787 and 0.797 and channel 2: 0.835 and 0.836 (p < 0.001 for all). Bland-Altman plots showed nonsignificant (NS) differences for both channels. The area under ROC curves of CM, HT and STS were, respectively, 0.75, 0.80 and 0.71 (channel 1) and 0.77, 0.80 and 0.70 (channel 2) (p = NS for all). As a conclusion, quantification of MTWA based on T-wave morphology analysis is feasible, accurate and reproducible, and have potential clinical application.