A comparison of three ECG-derived respiration techniques during cardiac magnetic resonance image acquisition

Abstract

myocytes level, involving both action potential duration and morphology (mainly in phases 2 and 3). While this might happen in a specific region (as after myocardial ischemia), it can also involve the entire myocardium. It is still unclear how alternations at the myocytes level are reflected on surface ECGmodification of T waves, especially when in vivo human hearts are considered. A few studies employed numerical methods to describe the extent of TWA on ECG as a function of action potential duration (APD) alternans. While providing a nice sensitivity study, the proposed approaches did not reveal which factors are responsible for the numerical findings. We have recently proposed a simple stochastic model of ventricular repolarization (IEEE Trans. Biomed. Eng., 2011), which takes into account both repolarization heterogeneity across the myocardium as well as random beat-to-beat variations in cells' activity. In this work, we generalized that model incorporating a term that describes myocytes alternans related to even and odd beats T-wave variability. Starting from the model and using the electrophysiological formulation developed by van Oosterom (J. Electrocardiol., 2001), we derived an analytical formula relating surface ECG to variations at the myocytes' level. Three factors play a role: heterogeneity of repolarization, the volume conductor, and beat-to-beat variability. The latter term is responsible both for TWA and small random changes in the shape of the T waves (a confounding factor for the TWA to be measured). This study provides some theoretical results that must be taken into account when quantifying TWA. They are as follows: (i) temporal averaging is necessary not just to remove additive broadband or muscular noise, but also to compensate for the inherent temporal small random variations in repolarization heterogeneity; (ii) a multileads configuration offers a theoretical advantage for which a mathematical estimate can be established; and (iii) the usage of a physiological model, even if simple, enables confidence intervals' computation for any TWA estimator. Such intervals, while computed on synthetic ECGs, surely offer insight on what happens with real ambulatory ECGs (i.e., setting theoretical limits to the minimum number of beats averaged).