Spatial and temporal variability of rotational, focal, and irregular activity: Practical implications for mapping of atrial fibrillation

Abstract Background Charge density mapping of atrial fibrillation (AF) reveals dynamic localized rotational activation (LRA), irregular activation (LIA) and focal firing (FF). Their spatial stability, conduction characteristics and the optimal duration of mapping required to reveal these phenomena and has not been explored. Methods Bi‐atrial mapping of AF propagation was undertaken using AcQMap (Acutus Medical) and variability of activation patterns quantified up to a duration of 30 s. The frequency of each pattern was quantified at each unique point of the chamber over two separate 30‐s recordings before ablation and R 2 calculated to quantify spatial stability. Regions with the highest frequency were identified at increasing time durations and compared to the result over 30 s using Cohen's kappa. Properties of regions with the most stable patterns were assessed during sinus rhythm and extrastimulus pacing. Results In 21 patients, 62 paired LA and RA maps were obtained. LIA was highly spatially stable with R 2 between maps of 0.83 (0.71–0.88) compared to 0.39 (0.24–0.57), and 0.64 (0.54–0.73) for LRA and FF, respectively. LIA was most temporally stable with a kappa of >0.8 reached by 12 s. LRA showed greatest variability with kappa >0.8 only after 22 s. Regions of LIA were of normal voltage amplitude (1.09 mv) but showed increased conduction heterogeneity during extrastimulus pacing (p = .0480). Conclusion Irregular activation patterns characterized by changing wavefront direction are temporally and spatially stable in contrast with LRA that is transient with least spatial stability. Focal activation appears of intermediate stability. Regions of LIA show increased heterogeneity following extrastimulus pacing and may represent fixed anatomical substrate.


| INTRODUCTION
The limited efficacy of pulmonary vein isolation for the ablation of persistent atrial fibrillation (persAF) has resulted in concerted efforts to identify nonpulmonary vein mechanisms responsible for AF maintenance. This has led to the development of techniques to facilitate mapping of the underlying atrial electrophysiology with the aim of revealing fibrillatory mechanisms and guiding targeted ablation. [1][2][3][4][5][6] Noncontact charge-density mapping allows visualization of whole chamber activation. Ultrasound is used to generate a high resolution and focal firing (FF) (see Figure 1). 7 To be classed as LRA, a smooth depolarization wavefront has to rotate 360°around a central point. LIA is characterized by a difference in angle between conduction that enters and leaves a confined region exceeding a threshold of 90°(and not meeting the criteria for LRA above). In contrast, FF is defined as activation of a primary vertex that precedes adjoining neighbors and extends centrifugally from this primary vertex.
A catheter ablation approach aimed at targeting these zones has been evaluated in one prospective observational study. 6 Within this study, mapping durations of approximately 5 s were used to identify ablation targets, but little work has been done exploring the spatial and temporal stability of these electrophysiological phenomena and the properties of these regions in sinus rhythm, knowledge of which is crucial in developing an optimal approach. We have performed the first study employing simultaneous mapping of AF activation in both the left and right atria using the AcQMap system and sought to investigate the spatial stability between two separate 30-s recordings of left and right atrial AF propagation and the effects of increasing duration of AF recording length on the degree of variability in mechanisms observed. Properties of atrial regions with the most stable F I G U R E 1 Dynamic AcQTrack analysis identifies each activation pattern including focal firing (FF) (A) characterized by radial activation from a central earliest point, localized rotational activation (LRA) (B) where smooth rotational activation of >270°is observed; and localized irregular activation (LIA) (C). LIA includes a range of specific patterns of activation, all characterized by changing wavefront direction of >90°(C-F, dynamic AcQTrack detection not shown). Color scale depicts the leading (red) to trailing edge (purple) of a wavefront with the full color spectrum occupying 80 ms. LAA, left atrial appendage; LUPV, left upper pulmonary vein; MV, mitral valve patterns were explored using long and short cycle length pacing and electroanatomic voltage mapping in sinus rhythm.

| Patient selection
Patients between the ages 18-80 undergoing first time catheter ablation for either paroxysmal (n = 5) or persistent (n = 17) AF were recruited following appropriate ethical approval (REC reference 18/SC/0409, clinicaltrials.gov NCT03812601). Exclusion criteria included prior cardiac surgery, congenital cardiac abnormalities and severe valvular heart disease.

| Electrophysiological mapping and ablation
Procedures were carried out under general anesthetic. With the exception of amiodarone, antiarrhythmic drugs were stopped a minimum of 5 days before the procedure. Venous access was obtained via bilateral femoral vein puncture under direct ultrasound guidance. Heparin boluses were administered before trans-septal puncture followed by continuous heparin infusion to maintain an ACT >350 s. A decapolar catheter (Inquiry; Abbott Medical) was inserted into the coronary sinus through an AcQRef (Acutus Medical) sheath which includes a distal electrode used as a unipolar reference. The first AcQMap catheter was advanced over a 0.032 guide wire into the RA via an AcQGuide (Acutus Medical) sheath and ultrasound used to reconstruct the right atrial chamber anatomy as previously described. 7 The ablation catheter (Tacticath; Abbott Medical) was advanced via an Agilis sheath into the left atrium across the single transseptal puncture site. The second AcQGuide sheath was then exchanged for the transseptal access sheath over the guide wire and a second AcQMap catheter advanced into the LA (see Figure 2). The LA anatomy was then generated with ultrasound. A circular mapping catheter (Inquiry Optima or Advisor Variable Loop; Abbott Medical) was used to undertake electroanatomic voltage mapping and guide pulmonary vein isolation.
In a subset of nine patients, activation maps were obtained using Supermap during pacing from up to three atrial sites (left atrial appendage, high right atrium, and proximal coronary sinus) with direct cardioversion used to restore sinus rhythm where necessary. Pacing consisted of a 4-beat drive train at 800 ms cycle length followed by a single extrastimulus with coupling interval 20 ms above the effective refractory period and was mapped using the AcQMap Supermap algorithm. Additional details are provided in the supplementary methods.
In patients attending the procedure in sinus rhythm AF was induced using burst atrial pacing, otherwise all AF recordings were obtained before DCCV. Once AF was established, recordings of 2 min duration were generated and time alignment between systems facilitated using a below threshold pacing stimulus from the coronary sinus catheter of 4 beats at 1000, 800, and 600 ms intervals and 12 s rest period. Pulmonary vein F I G U R E 2 Fluoroscopic image of catheters positioned for simultaneous bi-atrial noncontact mapping POPE ET AL. | 2395 isolation was performed using contact force guided radiofrequency ablation using 40-50 W (Tacticath ablation catheter; Abbott Medical). Simultaneous biatrial AF mapping was repeated following completion of PVI. Additional ablation and AF mapping was undertaken at the discretion of the operator.

| Propagation map calculation and data export
Raw AcQMap electrode biopotential signals were visually inspected to identify outlying or corrupted signals, which were then manually excluded. A low pass filter at 100 Hz as well as a 50 Hz notch filter and smoothing algorithm were applied followed by selection of a QRS-T wave template for subtraction. The AcQMap system allows operators to define separate mapping segments of any duration up to a maximum of approximately 14-15 s, limited by software processing capability. For this study, three consecutive 10-s mapping segments were created for simultaneous left and right atrial mapping, synchronized using the coronary sinus low amplitude pacing spike. This process was repeated to create two 30-s AF maps following splicing together of each 10 s segment. Propagation maps were calculated using the default timing method (based on -dv/dt of dipole signals), window width (for isochronal color bars) of 80 ms and time threshold of 70 ms (representing presumed minimum refractoriness). AcQTrack propagation patterns were calculated for each segment and data exported for offline analysis.

| Propagation pattern quantification
Complex propagation patterns described above (LIA, LRA, and FF) were identified using the AcQTrack (Acutus Medical) system during AF to ensure objective classification of activation patterns, as outlined in the supplementary methods. Representative examples of these patterns are seen in Figure 1 and Videos S1-S3. Every wavefront is scrutinized at each vertex of the anatomy. Planar wavefronts are discarded, whilst if the parameters for LIA, LRA, and FF are met within a discreet zone (300 mm 2 for LRA and 200 mm 2 for LIA) the vertices within this zone are highlighted and recorded by the system. These data were exported and analysed using a custom designed program to allow quantification. The process is outlined in Figure 3 and Figure S3. Initially, all AcQTrack data is exported to create a static map quantifying every pattern occurrence at each vertex of the chamber anatomy (approximately 3500 per chamber) for the entire recording duration ( Figure 3A). Each single occurrence is signified by a disc that occurs in a specific region for the duration that pattern is present ( Figure 3B). The number of these unique "discs" equates to the number of occurrences of the specified propagation pattern.
When taken over the duration of the recording, the proportion of time in which these discs are detected on the chamber surface re-  75% of the recording duration are counted, which corresponds to more than six occurrences over the duration of the recording. Similarly, a threshold that corresponds to a 30% relative reduction from the initial maximum duration identifies the region with only the highest number of occurrences. Where discs representing a detection are overlapping at any time point (potentially representing a F I G U R E 3 Method for AcQTrack pattern quantification. A static map is generated (A) quantifying all pattern occurrences at every vertex of the chamber anatomy. Each occurrence is identified in space and time as a single disc allowing calculation of the total number of occurrences, the percentage time they are present, and the proportion of the chamber surface area affected. See also Figure S3 meandering central pivot or rotation point) these are counted as a single pattern detection. Once a cut off is applied, only occurrences with the geometric center of the "disc" within the specified zone are included for quantification and any occurrence detected within 5 ms of a preceding occurrence in the same location is excluded to avoid double counting.

| Temporal stability/optimal mapping duration
We sought to identify the minimum mapping duration required to identify the sites that are shown over the full 30 s analysis to represent regions with the highest frequency and therefore most repetitive activation for each of the specific patterns described.
The region with the highest frequency of each propagation pattern may be considered as reflecting an optimum target for ablation. This was identified using the 30% cut off to identify the relevant zone on maps of increasing duration up to 30 s. Each vertex of the anatomy contained within and outside this region (at 30% cut off) was used to calculate the kappa statistic to quantify the consistency between these zones against the zone identified during a full 30 s segment.

| Statistics
Continuous variables are expressed as mean ± SD or median and interquartile range depending on distribution. Between group comparisons were performed using independent samples t test or Wilcoxon rank sum test depending on distribution as assessed using the Kolmogorov-Smirnov test. A two-sided p value of <.05 was considered significant. Statistical analysis was performed using SPSS (IBM v25) or Matlab (R2019a, MathWorks) and figures created using Matlab.

| Patient characteristics and map segments obtained
The characteristics of all patients recruited to the study are outlined in Table 1 Table 2.
Stability of regions across both maps with the highest frequency patterns identified at 30% cut off was also greatest for LIA in both the LA and RA. Cohen kappa statistic for LIA in the LA and RA, respectively, was 0.75 (0.64-0.78) and 0.75 (0.58-0.79).
Full kappa statistic results for all patients are outlined in Table 3.
The anatomical regions with maximal LIA were the anterior and posterior LA (in 46% and 27% of maps, respectively) and the lateral and septal RA (in 46% and 32%, respectively). LRA showed similar distribution with the zone of highest LRA frequency in the posterior LA in 41% and the anterior LA in 34%. In the RA, the lateral wall was the most common site (in 37%) followed by the septum and posterior walls (each in 24%). A similar distribution was observed for FF, most commonly involving the anterior and posterior LA (in 34% and 29%, respectively), followed by the LA septum (20%). In the RA, the highest frequency of FF was seen in the septum in 59% and the lateral wall in 22%. Note: Data are expressed as n (%), medians, 1st and 3rd quartiles or mean ± SD. Patient 9 was excluded from analysis as only atrial tachycardia could be induced.

T A B L E 1 Patient characteristics
Abbreviations: AF, atrial fibrillation; BMI, body mass index; LA, left atrium; LV, left ventricle; SR, sinus rhythm.

| Voltage and conduction properties
The median distance between merged chamber surfaces was 6.0 mm The aim of technologies designed to facilitate electrophysiological mapping and ablation of AF mechanisms is to identify repetitive patterns within a characteristically disorganized rhythm.
The total duration analysed has a significant impact on how a repetitive pattern is defined and there have been limited efforts previously to determine the optimum duration required. Shi et al. 9 have previously described the distribution and frequency of charge density activation patterns within the LA but used short R-R interval segments within a 30-s recording precluding a robust assessment of either spatial or temporal stability. Studies often do not report the duration of AF mapped but may report that patterns identified are stable over several minutes and separate recordings. 10,11 Other studies have used recording durations of between 10 s and 5 min, 3,5,12 whilst an analysis of the mapping duration required to identify sites at which ablation terminated AF suggested a duration of between 4 and 30 s was required. 13  Traditional electrophysiological assessment has involved mapping of either the endocardial or epicardial surfaces. There is increasing recognition that the remodeling involved in the development and progression of AF is a three-dimensional process resulting in activation time differences between atrial surfaces. 15,16 In this context, epicardial propagation that results in local breakthrough conduction will manifest as a focal activation pattern on the The spatial consistency of LIA detection between separate recordings is illustrated in Figure 5. Bipolar voltage amplitude in these regions is normal, which suggests that the activation properties observed are not the result of dense fibrosis. However, bipolar voltage amplitude is a relatively crude tool and is highly dependent on both rate and vector of activation. 17 Studies using late gadolinium enhanced magnetic resonance imaging reveal patchy areas of fibrosis out of keeping with the burden seen on voltage mapping studies 18,19 suggesting the existence of interstitial fibrosis that is not revealed by measuring bipolar voltage amplitude. The MAT during pacing within LIA zones was longer, suggestive of slower conduction velocity, and short coupled extrastimulus pacing resulted in an increase in CHI in these regions that was not observed in the remainder of the chamber.
Although these sites may represent anatomically normal regions of changing fiber orientation resulting in anisotropic conduction, they may represent disrupted conduction caused by underlying atrial interstitial fibrosis resulting in fiber disarray and rate dependent conduction abnormalities that manifest as local irregular activation patterns during AF. In a study by Walters et al. 20 using surgically placed epicardial plaques in patients with longstanding persistent AF, disorganized activation was frequently observed, which did not satisfy criteria for either rotors or focal activations but was stable over multiple recordings of 10 s duration taken over a period of 10 min.
This disorganized activation may represent similar propagation patterns to the irregular activation observed using charge density mapping, which was similarly stable even at short mapping durations.
Walters also reported that rotors were frequently transient, in keeping with the results outlined here. POPE ET AL.

| 2401
Both the nonhierarchical multiple-wavelet hypothesis and the competing "mother-rotor," or focal driver, hypothesis describe a process of wave-break in the formation of fibrillatory wavefronts involved in maintenance of cardiac fibrillation. 21,22 Tissue homogeneity is thought to play a significant role in the susceptibility to fibrillation 23 with regions of structural inhomogeneity likely responsible for the wave-break that results in fibrillatory conduction. 24 The anatomical regions demonstrating stable LIA patterns identified in this study may therefore reflect sites of structural heterogeneity responsible for wave-break, and therefore play an important role in AF maintenance.
Importantly, this study was not designed to assess ablation strategy or effectiveness and is not able to determine the impact of the phenomena identified on AF maintenance. This requires further

CONFLICT OF INTERESTS
Michael TB Pope has received honoraria and support for con-