Weighted Time Warping T-Wave Analysis Robust to Delineation Errors: Clinical Implications

T-wave (TW) morphology has been extensively investigated to develop specific risk prediction markers like $d_{w}$ quantifying the level of warping needed to temporally align a TW respect to a reference one. However, TW boundaries delineation errors may jeopardise $d_{w}$ diagnostic power. Hence, we proposed two TW-based weighting function (WFs) obtained from (i) the reference TW ($\mathcal{T}$), and (ii) the absolute value of its first-derivative ($\mathcal{D}$). We, first, simulated TW boundaries delineation errors by shortening and widening two TWs with morphological variability. We, then, used the variation ratio ($\mathcal{R}$) to compare $d_{w}$ derived applying the two WFs with that obtained in the control case (no weighting, $\mathcal{C}$). Next, we compared the ability of $d_{w}$, with and without WFs, in monitoring blood potassium concentration changes $(\Delta[K^{+}])$ by means of the Pear-son's correlation ($r$) in 29 48-hour Holter recordings from hemodyalisis (HD) patients. The simulations showed that $\mathcal{R}$ values for the two studied TWs, respectively, were 0.17 and 0.19 for $\mathcal{C}$, 0.05 and 0.08 for $\mathcal{T}$ and 0.07 and 0.10 for $\mathcal{D}$. However, similar $r$ median [interquartile range] values were found for $\mathcal{C}$ (0.90 [0.27]), $\mathcal{T}$ (0.90 [0.25]) and $\mathcal{D}$ (0.90 [0.29]) in HD scenario. Thus, using WFs to compute $d_{w}$ reduces the effects of TW boundaries delineation errors, but no improvements in $\Delta[K^{+}]$ monitoring in HD patients were observed. WFs impact on $d_{w}$ risk prediction power remains to be studied