A comparison of three methodologies for the computation of V-index

Spatial heterogeneity of ventricular repolarization (SHVR) can be assessed from surface ECGs using the V-index. The purpose of this study was to compare three algorithms for its estimate (which, for conciseness, will be termed M0, M1 and M2 in here). In deriving the V-index, the T-wave is modelled as a linear combination of a waveform (dominant T wave, DTW) and its derivatives, through scalar lead factors. M0, M1 and M2 differ in: i) number of DTW derivatives (2 for M0 and M1, 5 for M2); ii) numerical (M0 and M1) or analytical approximation (M2) of the DTW; and iii) a common DTW shared (M1) across beats. Tests were performed on both synthetic and real data. 64-beats synthetic 12-lead ECGs for 40 pseudo-subjects were generated with a direct electrophysiological model for SHVR in the range 10-70 ms. Holter recordings collected on 68 healthy subjects before and after moxifloxacin administration were also considered. M2 obtained the lowest bias on synthetic data for SHVR from 20 to 70 ms (p < 0.05). For SHVR = 10 ms, the three methods provided comparable results (p > 0.05). On Holter data, they were all able to detect the effects of moxifloxacin (p < 0.05) and the drug's peak times were comparable. While the three methodologies were all able to compute the V-index on both synthetic and real data, M2 appeared to obtain more robust estimates.