Action potential duration restitution (APDR) curves present spatial variations due to the electrophysiological heterogeneities present in the heart. Enhanced spatial APDR dispersion in ventricle has been suggested as an arrhythmic risk marker. In this study, we propose a method to noninvasively quantify dispersion of APDR slopes at tissue level by making only use of the surface electrocardiogram (ECG). The proposed estimate accounts for rate normalized differences in the steady-state <i>T</i>-wave peak to <i>T</i>-wave end interval (<i>Tpe</i>). A methodology is developed for its computation, which includes compensation for the <i>Tpe</i> memory lag after heart-rate (HR) changes. The capability of the proposed estimate to reflect APDR dispersion is assessed using a combination of ECG signal processing, and computational modeling and simulation. Specifically, ECG recordings of control subjects undergoing a tilt test trial are used to measure that estimate, while its capability to provide a quantification of APDR dispersion at tissue level is assessed by using a 2-D ventricular tissue simulation. From this simulation, APDR dispersion, denoted as Δα<sup>SIM</sup>, is calculated, and pseudo-ECGs are derived. Estimates of APDR dispersion measured from the pseudo-ECGs show to correlate with Δα<sup>SIM</sup>, being the mean relative error below 5%. A comparison of the ECG estimates obtained from tilt test recordings and the Δα<sup>SIM</sup> values measured <i>in silico</i> simulations at tissue level show that differences between them are below 20%, which is within physiological variability limits. Our results provide evidence that the proposed estimate is a noninvasive measurement of APDR dispersion in ventricle. Additional results from this study confirm that <i>Tpe</i> adapts to HR changes much faster than the <i>QT</i> interval.
13 Figures and Tables
Fig. 1. Outline of the methods used in this study. Crossed arrow shows a desirable but unaccessible connection. Tasks 1, 2, and 3 represent the different comparison tasks to be done in Section III (see Section III-A– D for details).
Fig. 10. APDR slope dispersion estimates from the tilt test recordings
Fig. 11. APDR slope dispersion, ΔαSIM , for the cell type distribution 80%/20%, and the proposed estimate measured from the pseudo-ECG in pecg3, pecg4, and pecg5.
Fig. 12. APDR slope dispersion, ΔαSIM , computed as a function of RR for
Fig. 13. APDR dispersion, ΔαSIM , for different cell type distributions as
Fig. 14. Rate adaptation of the Tpe and QT interval in a tilt test recording showing two abrupt RR changes.
Fig. 2. Representation of the Tpe interval in terms of APDs and delay of activation times (ΔAT).
Fig. 3. Dynamic restitution curves (APDR) for two regions corresponding to APDmin (dashed line) and APDlast (solid line). Slopes αmin and αlast are estimated for a change in the RR interval.
Fig. 4. Block diagram describing the [RR, Tpe ] relationship consisting of a time invariant FIR filter (impulse response h) and a nonlinear function gk (., a) described by the parameter vector a. v(n) accounts for the output error.
Fig. 5. Two-dimensional tissue slice used in the simulation, with indication of the default cell type distribution across the ventricular wall, and sensor positions used for pseudo-ECG computation.
Fig. 6. Isochronic representation (in milliseconds) of ventricular activation: (a) experiment results reproduced from ; (b) 2-D tissue simulations when pacing at RR intervals of 450, 1000, and 1450 ms.
Fig. 8. Isochronic voltage representation at T -wave peak time instant using three different cell type distributions (mid/epi) and pacing RR intervals of 450, 1000, and 1450 ms.
TABLE III MEAN ± STD ACROSS SUBJECTS OF THE TIME FOR 90% (t90 ), 70% (t70 ), 50% (t50 ) AND 25% (t25 ) OF THE COMPLETE RATE ADAPTATION
Download Full PDF Version (Non-Commercial Use)