In vivo human sock‐mapping validation of a simple model that explains unipolar electrogram morphology in relation to conduction‐repolarization dynamics

Abstract Introduction The unipolar electrogram (UEG) provides local measures of cardiac activation and repolarization and is an important translational link between patient and laboratory. A simple theoretical model of the UEG was previously proposed and tested in silico. Method and results The aim of this study was to use epicardial sock‐mapping data to validate the simple model's predictions of unipolar electrogram morphology in the in vivo human heart. The simple model conceptualizes the UEG as the difference between a local cardiac action potential and a position‐independent component representing remote activity, which is defined as the average of all action potentials. UEGs were recorded in 18 patients using a multielectrode sock containing 240 electrodes and activation (AT) and repolarization time (RT) were measured using standard definitions. For each cardiac site, a simulated local action potential was generated by adjusting a stylized action potential to fit AT and RT measured in vivo. The correlation coefficient (cc) measuring the morphological similarity between 13,637 recorded and simulated UEGs was cc = 0.89 (0.72–0.95), median (Q1–Q3), for the entire UEG, cc = 0.90 (0.76–0.95) for QRS complexes, and cc = 0.83 (0.58–0.92) for T‐waves. QRS and T‐wave areas from recorded and simulated UEGs showed cc> 0.89 and cc> 0.84, respectively, indicating good agreement between voltage isochrones maps. Simulated UEGs accurately reproduced the interaction between AT and QRS morphology and between RT and T‐wave morphology observed in vivo. Conclusions Human in vivo whole heart data support the validity of the simple model, which provides a framework for improving the understanding of the UEG and its clinical utility.


INTRODUCTION
The unipolar electrogram (UEG) measures cardiac extracellular potentials as the difference between the signal recorded by an exploring electrode in contact with the tissue and the potential of a remote reference often taken as the potential of the Wilson central terminal, the inferior vena cava, 1 or other distant nonexcitable sites. 2,3 The UEG is widely used in electrophysiological research and in the catheter lab since it allows simultaneous multisite assessment of fundamental parameters such as local activation (AT) and repolarization times (RT) as well as action potential duration, tissue viability, used theoretical models of cardiac electrophysiology, 8 by performing mathematical simplifications based on the assumption of isotropic and homogeneous conductivity. The validity of the simple model was demonstrated in silico by comparing its output with that of a more complex and detailed multiscale 3D bidomain ionic model. 7,9 In the present study, we aim to provide unique human in vivo data to formally validate the simple model and demonstrate its validity as a useful tool to relate the morphology of the signals recorded in the catheter lab to cardiac activation-repolarization dynamics.
We performed high-density cardiac mapping in patients undergoing cardiac surgery using a multielectrode epicardial sock covering both ventricles from apex to the base, 2 and we compared the morphology of UEGs recorded at each cardiac site with that of simulated UEGs generated by the simple model 7 using as input the local AT and RT measured in vivo but no information regarding the UEG morphology.
Results demonstrate good correlation between the morphology of simulated and recorded UEGs, suggesting that despite its simplicity the proposed model provides a sound description of the morphology of the UEGs in terms of activation-repolarization dynamics.

METHODS
The validation scheme implemented in this study is described in Figure 1. Local UEGs were recorded from up to 240 epicardial sites in intact human hearts during cardiac surgery and local AT and RT were measured ( Figure 1A). These were used to generate stylized APs having AT and RT as measured in vivo ( Figure 1B). A position-independent component representing remote activity is computed as the average of all APs ( Figure 1C), and finally the local UEG is obtained as a rescaled difference between the position-independent component and the local AP ( Figure 1D). Morphological features of recorded and simulated UEGs from the same site were then compared for validation. Importantly, no information related to the morphology of the UEGs recorded in vivo was used as model's input.

The simple nonionic model of the UEG
Local APs were simulated using established equations. The shape of these simulated APs is dictated by two position-dependent parameters determining AT and RT, respectively, and two position-independent parameters determining the steepness of the upslope and downslope during activation and repolarization, respectively. Mathematically, the local AP is the product of two logistic functions representing electrical excitation and recovery, respectively: In this expression, the subindex i = [1 … M] indicates a given cardiac site, the first factor represents activation and the second repolarization, A is the AP amplitude and V rest the resting potential. Parameters AT and RT determine the steepness of the upslope during activation and the steepness of the downslope during repolarization, respectively. Since within each heartbeat parameters A, V rest , AT , and RT are assumed to be the same for all cardiac sites (parameters are position-independent), all APs exhibit a very similar shape. Parameters AT,i and RT,i represent local AT and RT, respectively, which in the model correspond to the time of the steepest upslope during activation,(dAP i (t)∕dt) max , and the time of the steepest downslope during repolarization, (dAP i (t)∕dt) min , respectively. Following what was F I G U R E 1 Schematic representation of the model. UEG were recorded using a multielectrode sock measuring 240 UEGs. A, Activation (AT, circle) and repolarization times (RT, cross) were measured from the unipolar electrogram (UEG). B, For each cardiac site, a local action potential (red line) was generated by adjusting a stylized action potential to fit AT and RT measured in vivo. C, The position-independent component (blue bold line) representing remote activity was computed as the mean of all APs (only a representative subset of the local APs is shown dashed black lines). D, The simulated UEGs, generated as a rescaled difference between the remote and local components, shows similar morphology as the corresponding recorded one shown in panel A [Color figure can be viewed at wileyonlinelibrary.com] The local UEG measured at a given site i is obtained by inverting and rescaling the difference between the local AP and the positionindependent remote component (see Figure 1): The scaling factor is equal to = g i ∕(g i + g e ) and represents the balance between intracellular (g i ) and extracellular (g e ) conductivities, which are assumed to be constant in space and time. 7 Mathematically, it is possible to demonstrate that within this theoretical framework, standard measures of AT and RT derived from UEGs, usually timed to (dUEG(t)∕dt) min and (dUEG(t)∕dt) max , respectively, are equal to local AT, AT,i , and RT, RT,i , as long as the remote component AP(t) changes much more slowly than the local AP, AP i (t). The amplitude of the QRS complexes and T-waves depends on parameters AT and RT , as well as on spatial dispersion of both depolarization and repolarization.
Importantly, as shown in Figure 2, the model predicts that sites that activate before (or after) the remote component exhibit QS complexes (or prominent R waves), and sites that repolarize before (or after) the remote component exhibit a positive (or negative) T-wave.

Clinical study and data analysis
Whole-heart epicardial contact mapping was performed during openheart cardiac surgery using a multielectrode sock, 2,10 enabling the acquisition of 240 UEGs. The multielectrode sock was made of a flexible material allowing to fit over hearts of different sizes and cover both ventricles from apex to base. Eighteen patients undergoing either coronary artery bypass grafting (n = 14) or aortic valve replacement (n = 1) or both (n = 3) were studied. The study was approved by the local Ethics Committee and all patients gave written informed consent. S1 drive trains of 30-50 beats were delivered from one of the epicardial electrodes at cycle lengths decreasing from 600 to 350 ms.
UEGs were recorded at a sampling rate of 1 KHz within 0.05-500 Hz, and referenced to the rib retractor. Hemodynamic stability during the pacing protocol was closely monitored and pacing discontinued if appropriate. Data were exported and analyzed offline with bespoke algorithms as in previous studies. [10][11][12] Signal averaging of beats with same cycle length was conducted to reduce background noise. UEGs showing beat-to-beat morphological variability, i.e., with mean correlation coefficient between the median UEG and UEGs from any single beat lower than 0.98, were not considered. Signal-to-noise ratio (SNR) was assessed using a spectral-based measure where spectral bands for signal and noise were defined within 1-40 Hz and 40-100 Hz, respectively. Signal-averaged UEGs with SNR < 10 dB were not considered.
AT and RT were measured on the signal-averaged UEGs using standard definitions, i.e., AT and RT was measured as the intervals between the pacing stimulus and the time of the minimum of the first derivative during the QRS complex and the maximum of the first derivative of the T-wave independently of its polarity, 13 respectively. Activation recovery interval (ARI), a standard surrogate for the local AP duration, 3 was measured as ARI = RT -AT.

Statistical analysis
The Pearson's correlation coefficient was used to assess the similarity between recorded and simulated AT and RT sequences, as well as between morphological features of recorded and simulated UEG from the same cardiac site. The morphological similarity was assessed considering the entire UEGs, as well as the QRS complex and T-wave separately. Standard box-plots were used to describe data distribution, where central line is the median, the edges of the box are the first (Q 1 ) and third (Q 3 ) quartiles, and the whiskers extend to the most extreme data points not considered outliers. Values lower than Q 1 -1.5*(Q 3 -Q 1 ) and higher than Q 3 + 1.5*(Q 3 -Q 1 ) are considered outliers. Nonnormally distributed data are described by the median, median absolute deviation, i.e., median(X-median(X)), and interquartile range. Statistical analysis was performed both pooling the data together and on a patient-by-patient basis.

RESULTS
In total, after applying the aforementioned automatic inclusion criteria based on morphological stability and signal quality, 13,637 UEGs from 18 patients were utilized (Table 1). Cardiac intervals  Note: ΔAT, ΔRT and ΔARI: median differences between recorded and simulated cardiac intervals. cc-AT, cc-RT and cc-ARI: correlation coefficients between ATs, RTs and ARI from recorded and simulated data. ccAT-QRSa and ccRT-TWa: correlation coefficients between AT and the QRS area, and between RT and T-wave area, within recorded and model data. cc-QRSa and cc-TWa: correlation coefficients comparing the QRS area and T-wave area in the recorded and simulated data, respectively. All results are given as median ± median absolute deviation.

Activation and repolarization sequence
As expected owing to the modeling design, AT and RT measured from the simulated UEGs were very similar to those measured from UEGs recorded in vivo, with correlation coefficients higher than 0.97 and minimal absolute differences (Table 1).

Morphological features
The morphology of the simulated UEGs was very similar to the morphology of the corresponding UEGs recorded in vivo. As shown in  (Table 1). In this case, the median correlation coefficients between simulated and recorded UEGs was slightly higher for the QRS complex than for the T-wave for all cycle lengths except 400 ms and 600 ms for which only a trend was observed (0.10 < P < 0.05, Table 1). The first column of Figure 3B shows  In vivo data analysis demonstrates that the interactions between AT distribution and QRS morphology as well as between RT distribution and T-wave morphology described in Figure 2 and predicted by the model exist in the intact human heart. Figure 5A shows that in both recorded and simulated UEGs, cardiac sites that activate early exhibit a QS complex, those that activate around the mean ATs exhibit an RS complex with positive and negative deflections of similar amplitude while those that activate late exhibit a prominent R wave. Similarly, the T-wave is positive for sites that repolarize early, bipolar for sites whose RT is approximately equal to the mean RT and negative for sites that repolarize late (see Figure 5B). Therefore, a positive correlation exists between AT and the area under the QRS complex, while a negative correlation exists between RT and the area under the T-wave (see Figure 5A and B). These correlations were observed in all patients for all cycle lengths (Table 1), with a median correlation coefficients between AT and QRS area and between RT and T-wave area approximately equal to 1 and -1, respectively, in simulated UEGs, and approximately equal to 0.9 and -0.9, respectively, in human UEGs.

DISCUSSION
The  14 based on the assumption that the conductivity tensor fields of both intracellular and extracellular domains were isotropic and homogeneous. In that study, 7 the local component of the UEG corresponded to the local AP generated using a state of the art computational model including ion-currents dynamics. 9,15 The T-wave of simulated UEGs closely correlated to those generated implementing the more complex and realistic model, and the results were used to support the Wyatt method for RT assessment. In this work, the implementation of the simple model by Potse et al. 7 was further simplified by using as local components mathematical functions providing a stylized AP whose upstroke and down-slope are adjusted to match local AT and RT. A similar approach has been adopted in previous theoretical studies. 14,16,17 Of note, in this study we found a high morphological correlation between recorded and simulated UEG even if the model was informed with only epicardial data.
This does not imply that endocardial and septal activity do not contribute to the morphology of epicardial UEG, but it is most likely due to the fact that sock-mapping of both the left and right ventricles provided an accurate measure of the average AT and RT and therefore a reliable evaluation of the position-independent component representing remote activity, which is fundamental to determine the shape of the local UEG. Furthermore, since the simple model assumes homogenous and isotropic conductance, it is best suited to represent UEGs in normal cardiac tissue and its validity in presence of scar should be investigated in future studies.
The possibility of performing multisite RT measurements is a unique feature of the UEG that offers the opportunity of investigating a number of physiological mechanisms linked to ventricular arrhythmia such as spatial inhomogeneity of repolarization, 18 repolarization alternans and variability, 19,20 cardiac restitution, 10,21 and activationrepolarization coupling. 22,23 The investigation of these mechanisms, along with the understanding of the UEG morphology, is critical for prediction as well as ablation of ventricular tachycardias. Indeed, the increased resolution of high-density mapping technologies and new mapping electrode configurations now enable finer mapping of the substrate utilizing unipolar data, so correct interpretation of this information is even more pertinent in the modern electrophysiology arena. 24 Extensive work 3,7,[25][26][27][28] has consistently demonstrated that the time of the steepest upslope during the T-wave is a reliable marker of local RT, as first suggested by Wyatt et al. 13 The simple model predicts that upright and inverted T-waves are associated with early and late repolarization, respectively, and its validation in the in vivo human heart provides further support to the Wyatt method.
The power of this model lies in its simplicity, as it explains the UEG morphology as the difference between only two components, one position-dependent and similar to the local AP and another positionindependent and similar to the average of all APs, representing remote activity. This model could be used to study mechanisms and predict outcome as it provides a simple conceptual framework for linking properties of the UEG to the underlying activation and repolarization spatiotemporal dynamics, generate and test hypotheses, assess new methodologies, and ultimately improve the interpretation and clinical utility of the UEG. Similar analytical approaches have been previously used for similar purposes. 7,29-33

CONCLUSIONS
The similarity between more than 13,000 UEGs recorded with a multielectrode epicardial sock in 18 patients and UEGs generated by the simple theoretical model implemented in this study and first proposed by Potse et al. 7 demonstrates that (1) the UEG can be conceptualized as the difference between a local cardiac AP and a position-independent component representing remote activity and that (2) its morphology is mainly determined by activationrepolarization dynamics. This has important implications in the interpretation of high-density mapping data to advance understanding of arrhythmia mechanisms and enable optimal ablation strategies.