Theoretical and Empirical Estimates of V-index Variability

Abstract

Spatial heterogeneity of the ventricular repolarization (SHVR) might lead to life-threatening events when pathologically increases. The V-index is a metric meant to assess SHVR from the surface ECG and is computed on a set of ECG beats. The number of beats required to ensure low variability estimates is still matter of investigation. In this study, we investigated the role of the variability of the V-index estimator as function of the number of beats, using computerized simulations. We generated 100 sets of 1000 beats with seven different values of SHVR (from 10 to 70 ms) using the ECGSIM model. Using such dataset, we tested two approaches to estimate the variability of the V-index estimator. First, we analytically derived the probability density function of the estimator under independence and normality assumptions. Second, we used the bootstrap technique and derived that 1000 bootstrap iterations ensured stable estimates. We obtained that the standard deviation (STD) of V-index increased with SHVR values and decreased with the number of beats considered, with a rate similar to that of the sample mean. The independence assumption overestimated the STD by a factor of about 2. To conclude, the variability of the V-index estimator is below 5 ms when at least 100 beats are considered. The number of beats is in line with other techniques meant to estimate SHVR.

Spatial Correlation Between Myocyte's Repolarization Times and Their Alternans Drives T-Wave Alternans on the ECG

“…However, the variance of the R index was found relatively high even at high SNRs. This result is in line with what previously observed with the V-index, in which the standard deviation of its estimator (ratio of two sample variances) was in the order of few milliseconds when computed with 128 beats [34]. Perhaps, an alternative selection of beats and a different algorithm for estimating the lead factors might improve the performance of the sampling scheme and thus obtaining a better effective sample size [35].…”