Mechanism and magnitude of bipolar electrogram directional sensitivity: characterizing underlying determinants of bipolar amplitude

Abstract

Background

The amplitude of bipolar electrograms (EGMs) is directionally sensitive, decreasing when measured from electrode pairs oriented oblique to a propagating wavefront.

Objective

Use computational modeling and clinical data to establish the mechanism and magnitude of directional sensitivity.

Methods

Simulated EGMs were created using a computational model with electrode pairs rotated relative to a passing wavefront. A clinical database of 18,740 EGMs with varying electrode separation and orientations was recorded from the left atrium of 10 atrial fibrillation patients during pacing. For each EGM, the angle of incidence between the electrodes and the wavefront was measured using local conduction velocity (CV) mapping.

Results

A theoretical model was derived describing the effect of changing angle of incidence, electrode spacing, and CV on the local activation time (LAT) difference between a pair of electrodes. Model predictions were validated using simulated and clinical EGMs. Bipolar amplitude measured by an electrode pair is decreased (directionally sensitive) at angles of incidence resulting in LAT differences shorter than unipolar downstroke duration. Directional sensitivity increases with closer electrode spacing, faster CV, and longer unipolar EGM duration. For narrowly spaced electrode pairs (<5mm) it is predicted at all orientations.

Conclusions

Directional sensitivity occurs because bipolar amplitude is reduced when component unipolar EGMs overlap, such that neither electrode is “indifferent.” At the electrode spacing of clinical catheters, this is predicted to occur regardless of catheter orientation. This suggests that bipolar directional sensitivity can be lessened but not overcome by recently introduced catheters with additional, rotated electrode pairs.

INTRODUCTION

The features and locations of recorded intracardiac electrograms (EGMs) are used to target therapy in many cardiac ablation procedures. So-called “unipolar” EGMs are calculated as the potential difference between an electrode and an indifferent electrode that is assumed to contribute negligible voltage. Interpretation of these signals is, however, often challenged by unwanted summation with signals due to respiration, cardiac motion, and far-field tissue activation. For this reason, “bipolar” EGMs are often used, and measured (or calculated) as the extracellular potential difference between two adjacent electrodes. Signal to noise ratio is increased in bipolar EGMs by virtue of rejection of noise and far-field potentials. Bipolar EGMs are therefore easier to interpret and often considered more representative of electrophysiologic activity localized to the recording site.

In spite of limited theoretical understanding of the determinants of the bipolar EGM,¹ their amplitude has come to form the backbone of clinical “substrate mapping” approaches. Anatomical areas with low bipolar voltage amplitude are taken to represent unhealthy or “scarred” tissue and used to target ablation.^2,3 The efficacy of such ablation strategies remains disappointing, in part because of the complexity of EGM interpretation and analysis. The amplitude and morphology of a bipolar EGM is complex, with multiple underlying determinants.⁴ Bipolar EGMs are known to be influenced by tissue proximity,⁵ inter-electrode spacing,^6,7, conduction velocity,⁸ and more.¹

Directional sensitivity describes the influence of the angle of incidence between a propagating wavefront and an electrode pair on the resulting bipolar EGM morphology. ^8–12 Although well recognized, it remains incompletely characterized. We hypothesized that an improved understanding of the mechanism and extent of bipolar directional sensitivity would allow more selective and reproducible mapping of underlying physiology and improved treatment outcomes.

METHODS

Computational model

We developed a computational model in which simulated electrograms can be recorded during wavefront propagation. A 20 × 20mm isotropic sheet of tissue was simulated using a human atrial cell model. Simulated electrograms were calculated during a planar wave of excitation from electrode pairs centered in the middle of the domain with electrode separations of 2, 6, 8, and 10mm. Results were recorded with each electrode pair rotated relative to the wavefront at angles from 0° to 180°. Extracellular potentials at these sites were output at a precision of 0.001mV and sampling frequency of 10kHz. These simulated unipolar electrograms were band pass filtered from 2–240Hz prior to analysis and bipolar EGM calculation to recreate clinical filtering. See Supplementary Material for model details.

Patient study

10 patients with a history of atrial fibrillation (5 persistent, 5 paroxysmal) undergoing a planned ablation procedure were enrolled. Patients were excluded if they had undergone prior non-pulmonary vein isolation (PVI) left atrial ablation (including ablation of left atrial fractionated electrograms, roof line, floor line or posterior wall isolation). The patients had mean age of 68.4 years (range 57–78) and for all but one patient this was a first-time ablation. The remaining patient had undergone a single PVI procedure at our institution. This study was approved by the Duke University Medical Center Institutional Review Board and all subjects granted informed consent prior to enrollment.

Data collection

High density maps of the left atrial posterior wall (LAPW) were created using the Carto3 electroanatomic mapping system (Biosense Webster, Diamond Bar, CA). Data were collected following PVI, after confirming pulmonary vein entrance and exit block and prior to any extra-pulmonary vein ablation. Patients who presented in atrial fibrillation were first electrically cardioverted to sinus rhythm. A detailed left atrial geometry was collected by fast anatomical mapping (FAM) with a 2–6-2 PentaRay catheter. Pacing was performed from the mid-coronary sinus at 600ms cycle length with output 5mA and pulse duration 2ms. Electrograms were recorded point-by-point using a PentaRay multipolar mapping catheter and points acquired outside of the LAPW or more than 7mm from the collected geometry were discarded. Unipolar electrograms were bandpass filtered at 2–240Hz. There were no complications or arrhythmia induced by this pacing protocol. Following the clinical procedure, data were de-identified and exported using the built-in Carto export functionality. All subsequent analyses were performed offline using a custom software platform written in the Matlab programming environment (Mathworks, Inc).

Electrode definition

Using a PentaRay multipolar mapping catheter with inter-electrode spacing of 2–6-2mm, two electrode pairs with 2mm center-to-center electrode distance are available on each spline. To increase our sampling of activation time differences, we defined extended bipolar pairs consisting of all available electrode pairs on a given PentaRay spline. In this way, each of the five splines yielded the two standard 2mm bipoles in addition to a 6mm, two 8mm, and one 10mm electrode pair. Bipolar pairs spanning splines were not created in order to ensure fixed inter-electrode distance and avoid spatial inaccuracy in inter-electrode distance measurements. Small changes in inter-electrode distance can occur even between electrodes on the same spline due to manufacturing differences or catheter deformation, but the agreement of our results with theoretical predictions suggest this effect is minor. For each electrode pair a bipolar electrogram was calculated by subtracting the more proximal unipolar electrogram from the more distal electrogram of the pair.

Electrogram measurements

Local activation time was annotated for each unipolar electrogram in model and clinical EGMs. LAT was defined as the timing of the most negative unipolar slope (dV/dt). For each unipolar electrogram, the peak to peak unipolar amplitude was measured in a 30ms window surrounding its maximum negative dV/dt. Peak to peak bipolar amplitude was measured for each electrogram and assigned negative amplitude if the absolute bipolar peak occurred with negative voltage.

Local conduction velocity calculation

To determine the wavefront angle relative to bipolar pairs, for each patient we created a local conduction velocity (CV) map. CV maps were created by previously described methods, ^9–11 in which CV vectors are calculated using the position and local activation times of triads of electrodes available during each acquisition of the PentaRay catheter. Detailed methods are included in the Supplement. A mean of 1915.3 CV vectors were calculated per patient.

Electrode-wavefront angle of incidence calculation

The angle of each bipolar electrode pair relative to its local wavefront propagation was calculated using the local CV map. We sought to include only the most local CV vectors for each bipole when calculating the angle of incidence of the propagating wave. To do so, we compared the angle between each bipole pair and the mean angle of all CV vectors that were calculated using either electrode in the bipole pair. Note that this means each angle of incidence was therefore calculated using only data from the single beat in which the data was collected.

RESULTS

Theoretical dependence of local activation time difference on electrode-wavefront angle of incidence

We hypothesized that directional sensitivity of bipolar EGM amplitude is mediated through differences in local activation time difference (Δt) as an electrode pair rotates relative to a propagating wavefront. Δt will be maximal when the electrode pair is oriented parallel and minimal when oriented orthogonal to a passing wavefront. The distance travelled by a planar wavefront with velocity $\overrightarrow{v}$ between the activation time of electrode and the activation time of electrode U₂is the scalar projection of this pair in the direction of wavefront propagation

$$d^{\prime }=d\cos \theta ,$$ (1)

where θ is the electrode-wavefront angle of incidence and d is the inter-electrode distance (Fig. 1A). The local activation time difference is therefore $\text{Δ}t=d^{\prime }/|\overrightarrow{v}|$, or

$\text{Δt}=\frac{d}{|\overrightarrow{v}|}\cos \theta$ (2)

Figure 1.

A: Schematic of bipolar electrode pair, consisting of two electrodes (U₁ and U₂) separated by inter-electrode distance d. The angle between the wavefront vector of propagation ($\overrightarrow{v}$) and the vector connecting the electrode pair describes the electrode-wavefront angle of incidence (θ). B: Schematic of electrode pair rotation, with θ = 90° defined as perpendicular to the wavefront and θ = 0° or 180° as parallel.

This predicts that directional sensitivity of local activation time difference is dependent on the electrode separation and the wavefront speed. This also predicts a sinusoidal relationship between electrode-wavefront angle of incidence and Δt, with a larger effect seen when electrode separation widens. This prediction is shown in Fig. 2 (red), for which $|\overrightarrow{v}|$ was set to the measured computational model wavefront speed of 71.43 cm/sec. Interelectrode time delay when the electrode pair is parallel to the wavefront (θ = 0° and θ = 180°) and minimal when the electrode pair is orthogonal to the wavefront (θ = 90°; as diagrammed in Fig. 1B).

Figure 2.

Effect of electrode-wavefront angle of incidence on activation time difference (Δt) for a 2mm bipole pair. Model results (blue) compared to theoretical prediction by Eq. 2 (red).

Model electrogram characteristics

Simulated electrograms were created during planar wavefront propagation, with modeled electrode pairs separated by 2, 3, 6, and 10mm. For each electrode separation, the pair was rotated relative to the wavefront from 0° to 180°. In this way a total of 364 bipole pairs (728 unipolar electrograms) were modelled. The model CV was 71.43cm/sec. Example model EGMs are seen in Fig. 3A−B for rotating 2mm and 6mm bipoles. See Supplementary Tables S1 and S2 for full model EGM characteristics. For all model unipolar electrograms, mean R and S wave amplitudes were 0.197±0.004mV and −0.238±0.003mV, respectively (n=728). The mean downstroke duration of unipolar EGMs was 4.314±0.471ms. Model EGMs were shorter in duration and smaller in amplitude than clinically recorded electrograms. This is a consequence of the use of a two-dimensional model, which was chosen for computational efficiency. Absolute peak to peak bipolar amplitudes averaged 0.454±0.139mV (n=364). See Supplementary Table S2 for bipolar EGM characteristics by electrode separation.

Figure 3.

Representative unipolar (red and blue) and bipolar (black) EGMs showing the effect of rotation relative to the wavefront. Shown are examples of model data with 2mm (A) and 6mm (B) electrode spacing and clinical data with 2mm (C) and 6mm (D) electrode spacing. The bipolar amplitude is dependent on the degree of unipolar overlap, which varies according to electrode-wavefront angle of incidence.

Underlying determinants of bipolar amplitude in the computational model.

We next sought to explore the influence of wavefront angle of incidence, Δt, and electrode spacing on the peak to peak bipolar amplitude. The relationship between wavefront-electrode angle of incidence and bipolar amplitude is seen in Fig. 4A. For all electrode separations, bipolar amplitude is 0 mV when the electrode pair is oriented perpendicular to the wavefront (θ = 90°). For extended bipole pairs (6,8, and 10mm), as the electrode pair is rotated from perpendicular to parallel (θ = 0° or 180°) bipolar amplitude increases within a small range of rotation angles before reaching a plateau amplitude. In contrast, for 2mm bipoles there is a monotonically increasing bipolar amplitude as the electrode pair is rotated. It is important to note that for each angle of incidence, the bipolar amplitude cannot be predicted from the angle of incidence without knowing the electrode separation.

Figure 4.

Underlying determinants of bipolar EGM directional sensitivity in the computational model (A-C) and clinical EGMs (D-F). See text for full description.

The relationship between angle of incidence and Δt for 2mm electrode pairs is seen in Fig. 2, demonstrating excellent agreement between model results (blue) and theoretical predictions (red). Model results for all electrode spacings are seen in Fig. 4B. For all electrode spacings there is a monotonically decreasing, sinusoidal relationship. Consistent with the theoretical predictions of Eq. 2, the magnitude of Δt is proportional to electrode separation distance. Once again, for each angle of incidence the resulting Δt cannot be predicted without knowing the electrode separation.

As seen in Fig. 4C, there is a 1:1 relationship between Δt and peak to peak bipolar amplitude, at all electrode separations and angles of incidence. The maximal bipolar amplitudes occur at ±4.3ms (gray lines), the downstroke duration of the unipolar EGMs.

The results of Fig. 4A−C together suggest that the effect of wavefront angle of incidence on model bipolar amplitude is mediated through changes in Δt, which is itself dependent on electrode separation and angle of incidence.

Effect of electrode spacing and CV on directional sensitivity

We next sought to generalize the magnitude of directional sensitivity for electrode pairs. As seen in Fig. 4C, bipolar amplitude is most sensitive to changes at small Δt, which is seen more often with closer electrode spacing. Specifically, the steepest dependence of bipolar amplitude on Δt occurs when the component unipolar electrograms overlap, i.e. when Δt is less than the duration of the unipolar downstroke (gray line). The predicted range of Δt with maximal directional sensitivity is therefore

$-\gamma <\text{Δt}<\gamma ,$ (3)

for unipolar downstroke duration, γ.The critical wavefront-electrode angle of incidence that bounds this range of activation time differences, θ_c, can be solved using Eq. 2 as

$${\theta }_C={\cos }^{-1}\left(\frac{v\gamma }{d}\right).$$ (4)

Maximal directional sensitivity (changes in bipolar amplitude due to changes in angle of incidence) occurs between -θ_c < θ < θ_c. Noting that the total range of maximal directional sensitivity has a span of 2θ_c, we quantified the magnitude of directional sensitivity as

$$\epsilon =\left|\left({180}^{\circ }-2{\theta }_C\right)/{180}^{\circ }\right|,$$ (5)

which describes the proportion of angles of incidence within the range of maximal directional sensitivity. Directional sensitivity (changes in bipolar amplitude due to changes in the angle of an electrode pair) will be greater as ε increases. Eq. 4 suggests that directional sensitivity will be more prominent (occur over a wider range of electrode orientations) given faster CV, longer unipolar EGM downstroke duration, or more closely spaced electrodes. Intuitively, each of these increase the degree of overlap of unipolar EGMs.

As seen in Fig. 5, the magnitude of directional sensitivity increases as electrode spacing decreases. Using the model parameters of Fig. 4 (v = 71.4cm/sec), when electrode separations is less than 4mm, bipolar EGM amplitude will be directionally sensitive at all angles of incidence (indicated by ε=1). Atrial CV can range from approximately 30 to 100 cm/sec in patients with atrial fibrillation.¹² When CV is decreased to 30cm/sec or increased to 100cm/sec, this occurs at electrode separations less than 2mm and 5mm, respectively.

Figure 5.

Predicted magnitude of bipolar amplitude directional sensitivity (ε), indicating the proportion of catheter orientations resulting in artifactually decreased (directionally sensitive) bipolar amplitude. The right axis indicates the minimum angle of separation between a pair of bipoles needed to overcome directional sensitivity (θ_c). Directional sensitivity increases as electrode separation distance (d) decreases or conduction velocity (v) increases. Note that ε = 1 (θ_c = 90°) indicates conditions for which bipolar amplitude is predicted to be decreased (directionally sensitive) at all catheter orientations.

Minimum bipole pair separation angles to overcome directional sensitivity

To address directional sensitivity, newer catheters include overlapping electrode pairs with a variety of rotations to sample the wavefront from multiple angles of incidence.¹³ Directional sensitivity will be overcome in this way if at least one electrode pair has angle of incidence such that its unipolar EGMs do not overlap. To ensure that this is the case, electrode pairs must be included that sample a range of angles of incidence (θ) greater than the range of angles over which directional sensitivity occurs (θ_c). Consider, for example, a catheter with two electrode pairs oriented with a 60° angle relative to one another. If directional sensitivity only occurs at electrode-wavefront angles of incidence less than 30° (θ_c = 30° ), even if one electrode pair is oriented such that directional sensitivity occurs (e.g. θ_c = 10°), the other electrode pair will always be rotated sufficiently to have electrode-wavefront angle of incidence outside of the range (θ = 10° + 60°). Conversely, if directional sensitivity occurs at all electrode-wavefront angles of incidence (θ_c = 90°), both electrode pairs will be affected by directional sensitivity no matter how the electrodes are oriented relative to one another or the wavefront. Fig. 5 shows the influence of electrode separation and CV on this predicted minimum bipole separation angle to avoid directional sensitivity (right axis).

Clinical EGM characteristics

EGMs were recorded from the LAPW of 10 AF patients following PVI. Coronary sinus pacing resulted in an approximately planar wave of excitation ascending the LAPW (Supplementary Fig. S2A). A total of 14168 unipolar electrograms (7084 2mm bipoles) were recorded using the 20-pole catheter, with a mean 708.4 per patient (range 456–1180). By defining extended bipoles with spacing 2, 6, 8, and 10mm on each catheter spline, this initial data set yielded a database of 18740 bipolar electrograms. The mean recorded unipolar peak to peak amplitude was 1.76±0.34mV. See Supplementary Tables S3 and S4 for full clinical EGM characteristics (shown in Supplementary Fig. S1).

Clinical electrode-wavefront angles of incidence measurement by LAPW local CV mapping

We sought to determine the electrode-wavefront angle of incidence of each clinical bipole pair. To do so, a local conduction velocity (CV) map was created for the LAPW of each patient (see example in Supplemental Figure S2A−C). CV was calculated by a triangulation method, resulting in a mean of 2001.8 separate CV vectors for the LAPW of each patient (range 769–3410). Activation and CV maps for all patients depicted the expected, nearly planar wavefront during coronary sinus pacing. The mean calculated speed was 66.4 cm/sec. The mean calculated speed from only the smallest triangles (with length scale less than 5mm) was 80.0cm/sec. The angle of incidence between each bipolar electrode pair and the mean CV vector of all triangles connected to it was calculated. As seen in Supplemental Fig. 2D, angles ranging from 0° to 180° were well sampled by this approach.

Underlying determinants of clinical bipolar electrogram amplitude

Representative example clinical EGMs showing the effect of wavefront angle of incidence are seen in Fig. 3C−D. The pooled clinical results are shown in Fig. 4D−F, demonstrate the directional sensitivity of clinical EGMs. Regardless of electrode spacing, clinically measured peak to peak bipolar amplitude is minimal when bipole pairs are orthogonal to the wavefront (θ_c = 90°) and maximal when parallel (θ_c = 0° or 180°; Fig. 4D). 2mm electrode pairs result in smaller amplitudes than the more widely spaced bipoles for any given bipole electrode orientation and orientation-dependent variation in amplitude over a wider range of angles of incidence. Changes in electrode-wavefront angle of incidence result in changes in Δt consistent with theoretical predictions and computational modeling data (Fig. 4E). Likewise, across all angles of incidence, there is a 1:1 relationship between measured Δt and the resulting peak to peak bipolar amplitude (Fig. 4F). As in the computational model results, the peak bipolar amplitudes occur at approximately Δt > ±5ms (the shortest clinical unipolar mean downstroke durations). These findings demonstrate the in vivo degree of directional sensitivity and suggest that this effect is mediated by changes in Δt. Fig. 6 shows in detail the clinical results seen in Fig. 4D−F. Note that for each electrode separation in Fig. 6C, data points are only included for Δt’s measured in a minimum of 100 EGMs.

Figure 6.

Underlying determinants of clinical bipolar EGM amplitude. For each electrode spacing, the effect of angle of incidence (θ) on peak to peak bipolar amplitude (A) and inter-electrode activation time difference (B) are shown. Shown are means and standard deviation for each 2° angle of incidence bin. C: For each electrode spacing, the relationship between inter-electrode activation time difference and peak to peak bipolar amplitude. D: Pooled results for all electrode spacing.

DISCUSSION

Our results are consistent with those of prior empiric studies examining differences in bipolar EGM amplitudes as a function of bipole orientation relative to activation wavefronts^13–17. Whereas it is generally understood that EGMs are impacted by bipole orientation, to the best of our knowledge this is the first study to quantitatively describe the determinants of directional sensitivity and characterize its impact over a continuous range of electrode-wavefront angles of incidence.

Directional sensitivity is mediated through changes in the inter-electrode activation time difference due to variable bipole orientation relative to an activation wavefront. This change in activation time difference leads to changes in the resulting bipolar EGM morphology and amplitude whenever unipolar signals overlap (especially their rapid downstrokes). Conceptually, this suggests that directional sensitivity occurs when neither electrode is truly “indifferent.” This effect is increased by closer electrode spacing, faster CV, and longer unipolar EGM duration (which itself is influenced by CV).

These findings highlight limitations in current approaches to overcoming directional sensitivity. Our results suggest that when using catheters with closely spaced electrodes (less than approximately 5mm), directional sensitivity will lead to underestimation of measured bipolar EGM amplitude at all catheter orientations unless conduction velocities are particularly slow. Notably, this describes the electrode spacing of most clinical multi-electrode mapping catheters, such as the PentaRay (2–4mm), HD Grid (2–4mm; Abbott), and Orion catheters (2.5mm; Boston Scientific). To address directional sensitivity, newer multielectrode catheters (such as the HD Grid) include additional, rotated electrode pairs that record bipolar EGMs at multiple orientations.¹³ At these electrode spacings, however, directional sensitivity is predicted to impact measured bipolar amplitude regardless of orientation. This predicts that this approach will not fully overcome directional sensitivity. Alternatively, so-called “omnipolar mapping” overcomes directional sensitivity by mathematically finding the largest amplitude bipolar EGM (at any angle of incidence) from a collection of adjacent electrodes.^18–20 Although this approach measures bipolar EGMs that are independent of catheter orientation, it would not be expected to avoid bipolar EGM morphology (and amplitude) changes due to overlapping unipolar EGMs in the setting of closely spaced electrodes.

This work is motivated by the growing use of bipolar EGM amplitude as a measure of pro-arrhythmic myocardial “substrate.” Multiple clinical approaches to ablation of atrial and ventricular arrhythmias target “unhealthy” myocardium that is identified by bipolar voltage amplitude (for example, Refs. 21−24). Directional sensitivity limits the accuracy of these approaches by causing underestimation of bipolar amplitude, especially when using narrowly spaced bipoles (e.g. 2mm spacing) and in regions with normal conduction velocity. While this does not negate the utility of narrowly spaced bipolar EGMs in annotating local activation time, it complicates their interpretation as a measure of pro-arrhythmic myocardial substrate based on bipolar amplitude. Future work will study whether measure bipolar amplitude can be corrected if the angle of incidence is known. Although unipolar EGMs are not affected by directional sensitivity, substrate assessment by unipolar EGM amplitudes is itself limited by unique considerations.²⁵ 1 We hope that an improved understanding of the multiple underlying determinants of bipolar EGM morphology can allow more informed and specific use of EGMs to characterize myocardial substrate properties.

CONCLUSIONS

Bipolar EGMs demonstrate directional sensitivity due to changes in inter-electrode activation time difference as an electrode pair is rotated relative to a propagating wavefront. The degree of directional sensitivity of a given electrode pair is related to electrode separation, wavefront conduction velocity, and unipolar EGM signal duration. Over a physiologic range of parameters, electrode pairs separated by less than approximately 5mm will result in bipolar EGM amplitudes that are scaled (directionally sensitive) at all electrode-wavefront angles of incidence. Improved understanding of the underlying determinants of bipolar EGM amplitude will inform future clinical substrate mapping approaches.

LIMITATIONS

The complex, multifactorial nature of bipolar EGM morphology necessitates simplified model systems for its study. As such, we employed a computational model with a simplified wavefront, tissue geometry, and resulting signal characteristics. To improve our ability to translate these ideas to the real heart, our clinical study was itself performed in a similarly simplified model system—the LAPW after PVI during approximately planar wavefront propagation. The results of this study therefore need to be confirmed and extended for more complex activation wavefronts, (three dimensional) geometries, and signal features.