Author + information
- Received March 31, 2017
- Revision received June 1, 2017
- Accepted June 9, 2017
- Published online January 15, 2018.
- Bhupesh Pathik, MBBSa,b,
- Jonathan M. Kalman, MBBS, PhDa,b,
- Tomos Walters, MBBS, PhDa,b,
- Pawel Kuklik, PhDc,
- Jichao Zhao, PhDd,
- Andrew Madry, PhDa,
- Sandeep Prabhu, MBBSa,b,e,
- Chrishan Nalliah, MBBSa,b,
- Peter Kistler, MBBS, PhDb,e and
- Geoffrey Lee, MBChB, PhDa,b,∗ ()
- aRoyal Melbourne Hospital, Parkville, Australia
- bUniversity of Melbourne, Parkville, Australia
- cUniversity Medical Center Hamburg-Eppendorf, Hamburg, Germany
- dAuckland Bioengineering Institute, University of Auckland, Auckland, New Zealand
- eAlfred Hospital and Baker IDI, Melbourne, Australia
- ↵∗Address for correspondence:
Dr. Geoffrey Lee, Department of Cardiology, Royal Melbourne Hospital, Grattan Street, Parkville, Victoria 3052, Australia.
Objectives This study sought to validate a 3-dimensional (3D) phase mapping system and determine the distribution of dominant propagation patterns in persistent atrial fibrillation (AF).
Background Currently available systems display phase as simplified 2-dimensional maps. We developed a novel 3D phase mapping system that uses the 3D location of basket catheter electrodes and the patient’s 3D left atrial surface geometry to interpolate phase and create a 3D representation of phase progression.
Methods Six-min AF recordings from the left atrium were obtained in 14 patients using the Constellation basket catheter and analyzed offline. Exported signals underwent both phase and traditional activation analysis and were then visualized using a novel 3D mapping system. Analysis involved: 1) validation of phase analysis by comparing beat-to-beat AF cycle length calculated using phase inversion with that determined from activation timing in the same 20-s segment; 2) validation of 3D phase by comparing propagation patterns observed using 3D phase with 3D activation in the same 1-min segment; and 3) determining the distribution of dominant propagation patterns in 6-min recordings using 3D phase.
Results There was strong agreement of beat-to-beat AF cycle length between activation analysis and phase inversion (R2 = 0.91). There was no significant difference between 3D activation and 3D phase in mean percentage of propagation patterns classified as single wavefronts (p = 0.99), focal activations (p = 0.26), disorganized activity (p = 0.76), or multiple wavefronts (p = 0.70). During prolonged 3D phase, single wavefronts were the most common propagation pattern (50.2%). A total of 34 rotors were seen in 9 of 14 patients. All rotors were transient with mean duration of 1.0 ± 0.6 s. Rotors were only observed in areas of high electrode density where the interelectrode distance was significantly shorter than nonrotor sites (7.4 [interquartile range: 6.3 to 14.6] vs. 15.3 mm [interquartile range: 10.1 to 22.2]; p < 0.001).
Conclusions During prolonged 3D phase mapping, transient rotors were observed in 64% of patients and reformed at the same anatomic location in 44% of patients. The electrode density of the basket catheter may limit the detection of rotors.
The mechanisms that sustain persistent atrial fibrillation (AF) remain uncertain and are the subject of ongoing debate and investigation. Using differing mapping approaches, a variety of potential mechanisms have been reported including multiple wavefronts maintained by longitudinal dissociation (1) and by epicardial-endocardial dissociation (2,3), focal drivers (4), and rotors either transient (5–7) or sustained (8). Although these high-density epicardial mapping techniques provided excellent spatial resolution, a potential limitation has been the narrower field-of-view of the mapping plaques and the possibility that rotors may be present outside this field. Recent approaches taking a more global mapping approach have included the use of body surface phase mapping, which demonstrated transient rotors or driver domains (5); or alternately the use of a 64-electrode basket catheter and phase mapping, which demonstrated the presence of 1 to 2 persistent rotors per atrial chamber (8).
Use of the basket catheter to identify persistent rotors has resulted in some notable ablation success rates in persistent AF (9–11) but good results have not been uniformly reported (12–14). One of the limitations of the current iteration of this technology has been the 2-dimensional (2D) display of the atria with all of the inherent assumptions that this engenders. In the current study, we used a 64-electrode basket catheter to map the left atrium (LA) in persistent AF with activation patterns displayed in 3-dimensional (3D) format onto patient-specific LA anatomy obtained by pre-procedural cardiac computerized tomography (CT) scan.
The aims of the study were to validate the novel 3D phase mapping approach (RMHeartMap3D) by comparing 3D phase maps with 3D activation maps derived from the same 1-min segment of AF data; and to analyze continuous 6-min segments of AF data using 3D phase mapping to determine the relative frequency and location of rotors and other dominant propagation patterns in LA recordings of human persistent AF.
A total of 14 patients with persistent AF undergoing catheter ablation were studied. Persistent AF was defined as continuous AF sustained for >7 days (15). All patients gave written informed consent and the study protocol was approved by the Melbourne Health Research and Human Ethics Committee.
Antiarrhythmic medications were discontinued for 5 half-lives before the procedure. All patients underwent cardiac CT. Transesophageal echocardiography was performed in all patients immediately before the commencement of the procedure to exclude LA thrombus. All cases were performed under general anesthesia. A Decapole catheter was placed in the coronary sinus and a Hexapole catheter at the His location. For patients in sinus rhythm at the time of procedure, burst atrial pacing was used to induce AF.
Double transseptal punctures were performed under transesophageal echocardiography guidance to optimize the septal crossing at the mid portion of the septum. SL1 sheaths of 8F and 8.5F catheter were advanced into the LA for the circular mapping and ablation, catheters respectively. All patients were administered intravenous heparin to maintain activated clotting time (ACT) >300 s before transeptal puncture. 3D electroanatomic mapping was performed with the Ensite Velocity System (NavX, St. Jude Medical, St. Paul, Minnesota).
Creation and registration of the 3D LA surface geometry
The LA CT was segmented using the Ensite segmentation tool Verismo to create a 3D polygonal mesh model of the LA surface anatomy. After transseptal puncture, a circular mapping catheter was used to create a patient-specific LA geometry. A scaling algorithm (Field Scaling) was applied to the completed detailed geometry to adjust for the nonlinearity of the geometry that occurs as a result of local changes in impedance fields. Field scaling was based on the measured interelectrode spacing for all locations within the geometry. The 3D polygonal mesh model was then registered with the NavX-created LA geometry and NavX coordinate system using operator-defined fiducial points as previously described (16). Once the registration process was completed, the Constellation basket catheter (Boston Scientific, Natick, Massachusetts) was introduced into the LA.
Basket catheter selection and deployment
The 64-pole basket catheter was advanced into the LA using either the 8.5-F catheter SL1 sheath or a steerable sheath (Agilis, St. Jude Medical) to guide basket catheter positioning in the LA. The Constellation basket catheter is spherically shaped and consists of 8 splines with a total of 64 unipoles or 56 bipoles. The 60-, 48-, and 38-mm basket catheters were used in the study. The size of the basket catheter selected was based on the LA size in the pre-procedural cardiac CT and the intraoperative transesophageal echocardiography. In all patients, a larger basket size was preferentially selected and deployed in the LA. If the basket catheter was found to be too large to fit into the LA and allow full deployment, the basket catheter was subsequently replaced with a smaller sized catheter. Considerable amount of time was taken to optimize the positioning of the basket catheter to improve electrode contact and coverage of the LA.
Recording of AF data and signal processing
Continuous 6-min recordings of AF were obtained from the basket catheter in each patient. On completion of the clinical case, raw unipolar and bipolar electrogram signals were exported for off-line signal processing (activation and phase analysis). In addition, the Cartesian coordinates of the Constellation catheter unipolar electrodes in 3D space relative to the registered 3D polygonal mesh model during the time of the AF recording were exported. The 3D polygonal mesh model of the LA surface was also exported in a file that contained the 3 critical elements that define the geometry (polygonal vertices, vertice normals, and polygonal surface data).
Summary of analysis
A summary of the analysis performed is shown in Figure 1. Detailed off-line analysis of the data involved the following 3 components:
1. Validation of sinusoidal recomposition and phase analysis by comparing beat-to-beat AF cycle length (AFCL) calculated from activation timing with that determined using phase inversion in the same 20-s segment of AF for each patient.
2. Validation of the novel 3D phase mapping approach by comparing propagation patterns and wavefront directionality observed using 3D phase maps with 3D activation maps in the same 1-min segment of AF for each patient. The observer was blinded to the 3D activation map when analyzing the 3D phase map.
3. Determination of the relative frequency of rotors and other dominant propagation patterns using 3D phase mapping in continuous 6-min recordings of AF for each patient. In terms of rotors, their location and the spatial characteristics of the basket catheter at rotor sites was also determined.
Phase analysis was performed using Matlab version R2015 (Mathworks Inc., Natick, Massachusetts). Sinusoidal recomposition and instantaneous phase analysis have previously been described in detail (17) and can be summarized in the following steps: 1) electrogram is transformed using sinusoidal recomposition; 2) Hilbert transformation of the recomposed sinusoids; 3) followed by calculation of instantaneous phase. Far-field ventricular signals were digitally subtracted from the raw electrograms using a template-matching algorithm before sinusoidal recomposition.
Activation analysis was performed using customized software (Cardiac Electrophysiology Analysis System [CEPAS], Cuoretech, Sydney, Australia). CEPAS has specific user-defined characteristics to identify electrogram activations. These include: 1) a baseline noise threshold; 2) electrogram width criterion to avoid detection of broad far-field activations; 3) electrogram slope; and 4) electrogram-refractory periods to avoid multiple detections within the same activation. Based on previous work (6,18,19), a noise threshold of 0.1 mV, width criterion of 10 ms, and refractory period of 50 ms were used for the analysis. All automated electrogram analyses were visually verified to ensure accurate annotation of activation times. Activations were manually corrected if automated annotation was incorrect. Complex fractionated atrial electrograms were defined as electrograms displaying continuous electrical activity over the entire recording period. Given the difficulty in assigning activation times at sites of continuous electrical activity, these sites were excluded from the animation. Multicomponent electrograms were defined as ≥3 deflections of >50-ms duration separated by a discrete isoelectric baseline between successive electrograms. These were manually annotated at the onset of electrogram activation.
Validation of sinusoidal recomposition and phase
The accuracy of sinusoidal recomposition and instantaneous phase analysis has been validated with contact unipolar electrogram recordings (17). Comparison of AFCL based on intrinsic deflection and based on phase inversion shows high correlation (R2 = 0.99 for paroxymal atrial fibrillation (PAF) and R2 = 0.90 for persistent AF electrograms) (17). Using similar methodology we compared beat-to-beat AFCL calculated from activation timing with AFCL determined by phase inversion. Because the timing of a phase inversion (transition from −π to π) corresponds to the intrinsic deflection in the raw electrogram signal, we used the timings of the phase inversions to reconstruct a beat-to-beat AFCL compared with the previously validated cycle length assessment based on activation timing using the CEPAS software (6). In each patient, a random 20-s segment of AF was chosen for analysis. The same 20-s electrogram recording was subjected to both activation analysis using CEPAS and sinusoidal recomposition and phase analysis using Matlab. Beat-to-beat AFCL during this 20-s segment calculated by traditional activation was then plotted against beat-to-beat AFCL calculated by phase inversion to determine an R2 value.
Novel 3D animation software (RMHeartMap3D)
Phase data, activation data, and catheter location data and the 3D polygonal mesh model were imported into customized 3D animation software (RMHeartMap3D) written using the open source library The Visualization Toolkit (20,21). The 3D polygonal mesh provides the spatial construct on which activation and phase was mapped and animated. Because the 3D polygonal mesh model was registered relative to the location of the Constellation catheter deployed within the LA at the time of geometry registration, the distance of each electrode to the registered surface geometry was known. Two different animation methods were developed to display activation and phase data because activation data is treated as a binary variable (1, 0) whereas phase is treated as a continuous variable (−π, π).
Animation of 3D activation
Similar to the previously described 2D wave mapping technique (8), each electrode site was animated independently of each other. Each electrode had 2 states “on” and “off” mapped to the color scale between 1 “on state” and 0 “off state.” The activation information from each Constellation electrode site was projected to the nearest 3D surface point as a sphere of color. When a site is activated (based on the moment of activation annotated using CEPAS), the site projected a sphere of white color onto the closest 3D surface (“on”) and transitions to red (“off”) over user-defined time interval (T1/2). To account for the differences in distance of each electrode site to the surface geometry, the intensity of the sphere at each point was calculated according to a standard formula that projects the intensity of activation as inversely proportional to the electrode distance from the activation position (22). We did not use a minimum threshold value for distance when determining the intensity. The larger the distance factor, the smaller the spread of color. Thus, electrode sites far away from the surface do not add to the overall activation pattern.
Animation of 3D phase
Phase data from each electrode site were projected to the nearest point on the 3D geometry as previously mentioned. Given the scattered nature of the Constellation electrode points in 3D space, phase at each point was interpolated within a radial region around each electrode using Shepard's Inverse Distance Weight algorithm (23). Using this technique, phase values of unknown points on the 3D mesh vertices were assigned values with a weighted average of known points (23). The interpolation has a power parameter, which defines the number of known points used to calculate phase at unknown points. For example, a power parameter of 4 uses the 4 closest known phase points to calculate the interpolated phase value. For the purposes of this study, the power parameter was set to 4. This resulted in known phase values from the Constellation electrode recording sites being interpolated over the 3D mesh vertices and resulted in a 3D phase animation where phase was projected onto a 3D surface. This technique allowed phase data to be visualized over a curved surface.
Propagation pattern analysis
Based on prior studies, propagation patterns were classified into 4 morphologies (6,18): 1) wavefronts; 2) focal activity; 3) rotors; or 4) disorganized activity. The number and direction of wavefront propagation was recorded for each activation. Wavefronts were classified as single or multiple. Single wavefronts were defined as earliest activity at a discrete electrogram site with linear activation of >3 adjacent electrograms ≥2 adjacent electrograms wide. Multiple wavefronts were defined as the presence of ≥2 simultaneous wavefronts activating in a linear fashion fulfilling the same previously mentioned criteria. Focal activations were defined as earliest activation at a discrete electrogram site with centrifugal wavefront activation that spreads radially a distance of >3 electrograms from this site. A rotor was defined as the presence of a rotating wavefront with at least 2 complete revolutions of 360 degrees. Rotors were considered sustained if present for >50 cycles and transient if ≤50 continuous rotations (8). Disorganized activity was defined as a propagation pattern that did not fulfill the criteria for a wavefront and with earliest activation at >2 adjacent electrograms that propagated <3 electrograms or activations that occurred as isolated beats dissociated from activation of adjacent electrograms.
Percentage agreement in wavefront directionality of 3D phase maps with 3D activation maps was determined by simultaneously displaying 3D activation and 3D phase maps from the same time segment of the same patient side by side. The number of activations the wavefront was traveling in the same direction in both the 3D phase and 3D activation maps was divided by the total number of activations for a given time segment.
All statistical analysis was performed using SPSS software version 23.0 (SPSS, IBM, Armonk, New York). Normality of all quantitative data variables was checked using the Shapiro-Wilk test. Continuous variables are reported as mean ± SD and median and interquartile range (IQR), as appropriate. Categorical variables are reported as numbers and percentages. Comparisons of propagation patterns between 3D activation and 3D phase were performed using analysis of variance with patient as a random effect. A p value <0.05 was considered statistically significant.
A total of 14 patients underwent LA mapping using the Constellation basket catheter (Table 1). The mean age was 62 ± 7 years and 50% were male. Median AF duration was 4.7 years (IQR: 2 to 6 years). Mean CHA2DS2VASc score was 1.4 ± 1. All patients either had normal or mild left ventricular systolic dysfunction. The 60-mm basket catheters were used in 3 (21%) patients. The 48-mm catheters were used in 10 (71%) patients. The 38-mm catheter was used in 1 patient. In 7 (50%) patients, a steerable sheath was used to aid optimal basket catheter positioning. The overall mean interelectrode distance of the basket catheter electrodes based on their location in 3D space was 17.0 ± 10.0 mm. A mean of unipolar electrodes within 2 mm of the endocardial surface was 33 ± 8 (52%).
Validation of sinusoidal recomposition and phase analysis
Across all 14 patients, a total of 1,595 activations during persistent AF were used in this analysis. There was strong agreement between activation times determined using traditional activation analysis and those detected by phase inversion (R2 = 0.97). There was also strong agreement between beat-to-beat AFCL using these 2 techniques (R2 = 0.91) (Figure 2).
Validation of 3D phase compared with 3D activation
Our initial analysis focused on validation of the 3D phase mapping technique by comparing propagation patterns and wavefront directionality observed using 3D phase maps with 3D activation maps from the same 1-min segment of AF. Across all 14 patients, a total of 5,038 activation patterns were analyzed. The mean AFCL was 185 ± 107 ms. In each of the 14 patients, both 3D phase and 3D activation mapping showed that AF was characterized by highly dynamic and heterogeneous patterns of activation with transitions between wavefronts, focal activations, and disorganized activity. An example of these highly dynamic activation patterns is shown in Online Video 1. In this patient, single wavefronts were observed arising from different locations in the LA, such as the left atrial appendage (LAA), posterior wall near each of the pulmonary veins, and roof for variable number of activations interspersed with a short period of disorganized activity. Although wavefronts did not have clearly demarcated boundaries, broad wavefronts generally occupied an entire atrial aspect (e.g., posterior wall, Online Video 1).
Using 3D activation mapping, the overall dominant activation pattern was single broad wavefronts (51.1%) followed by apparent focal activations (35.3%), disorganized activity (11.9%), and multiple wavefronts (1.7%). A similar distribution was observed using 3D phase mapping with the dominant propagation pattern also single broad wavefronts (50.2%) followed by apparent focal activations (35.5%), disorganized activity (12.4%), and multiple wavefronts (1.8%). Focal activations were not repetitive.
The distribution of propagation patterns observed in each of the 14 patients is shown in Figure 3 and Table 2. Single wavefronts were the dominant pattern in each patient followed by isolated focal activations. Within each patient, there was no significant difference between 3D activation and 3D phase maps in the mean percentage of propagation patterns classified as single wavefronts (53.2 ± 12.0% vs. 52.6 ± 11.2%; p = 0.99), isolated focal activations (32.7 ± 11.9% vs. 32.7 ± 11.1%; p = 0.26), disorganized activity (11.6 ± 10.7% vs. 11.8 ± 10.0%; p = 0.76), or multiple wavefronts (2.6 ± 6.4% vs. 2.9 ± 7.1%; p = 0.70). Overall, there was a mean 92 ± 3% agreement in wavefront directionality of 3D phase maps with 3D activation maps. An example of this close agreement in wavefront directionality of 3D phase maps with 3D activation maps is shown in Online Video 2.
Propagation patterns during prolonged 3D phase mapping
Propagation patterns during prolonged 6-min recordings of AF were analyzed using 3D phase mapping. Across 14 patients, a total of 30,947 activations were evaluated. Overall, the dominant propagation pattern observed was single wavefronts (50.2%), followed by isolated focal activations (33.4%), disorganized activity (12.4%), transient rotors (2.3%), and multiple wavefronts (1.7%). The distribution of propagation patterns observed in each of the 14 patients is shown in Figure 4 and Table 3. Within each patient, single wavefronts were the dominant pattern followed by focal activations. On a per-patient basis, the mean percentage of propagation patterns classified as single wavefronts was 52.3 ± 10.6%, focal activations 30.5 ± 10.8%, disorganized activity 12.3 ± 10.3%, multiple wavefronts 2.7 ± 6.6%, and rotors 2.3 ± 3.6%.
Transient rotors were visualized in 9 of 14 patients (64%). A total of 34 rotors were seen in the 14 patients. The median number of rotors per patient was 1 (IQR: 1 to 6). The median number of revolutions per rotor was 4 (IQR: 3 to 6) with <6 revolutions per rotor in 74% of cases. The mean AFCL during the rotor formation was 178 ± 16 ms. Mean rotor duration was 1.0 ± 0.6 s. Overall, transient rotational activity was only observed for 0.7% of the total recording time. In 4 of 9 (44%) patients with rotational activity, the rotors seemed to reform in the same anatomic location (Online Video 3). The mean time interval from cessation and formation of the rotor at the same anatomic location was 18.2 ± 16.8 s. A total of 3 of 9 patients with rotational activity had geographically separated rotors. The mean time interval from cessation of the rotor at 1 anatomic location to formation at another site was 41.2 ± 36.5 s.
An example of a rotor can be seen in Figure 5 and Online Video 4. Here a rotor lasting 15 clockwise rotations can be seen propagating around the base of the LAA. Immediately before the formation of the rotor, a single wavefront arising from the base of the LAA was observed. Termination of the rotor was followed by disorganized activity. In this particular case, the electrogram at the center of the rotor showed fractionated electrograms; however, this was not always the case. Of the observed rotors, 22 of 34 (65%) had complex fractionated atrial electrograms at the center of the rotor. The remaining 12 of 34 (35%) did not have complex fractionated atrial electrograms at the rotor center.
Spatial distribution of rotors
Rotors were most commonly observed in the regions adjacent to the base of the LAA (14 of 34) and adjacent to the left superior pulmonary vein (12 of 34). Four rotors were seen near the left inferior pulmonary vein, and 4 were observed near the right-sided pulmonary veins (2 rotors each near the right superior pulmonary vein and right inferior pulmonary vein). Rotors were not observed in the septal region possibly because of relatively poor basket coverage.
Spatial analysis of the basket electrodes showed that rotors were usually observed in anatomic areas where there was a higher density of the basket electrode splines. In these areas the median interelectrode distance was significantly less than sites where no rotors were observed (7.4 mm [IQR: 6.3 to 14.6 mm] compared with 15.3 mm [IQR: 10.1 to 22.2 mm]; p < 0.001).
Currently available systems use a 2D technique to project phase. This study is the first to perform 3D phase mapping using a technique that takes into account the actual 3D location of the basket catheter electrodes and the patient’s 3D LA surface geometry. The main findings of this study include:
1. During prolonged 3D phase mapping, rotational activity was observed in 64% of patients. Rotors were transient lasting a median of 4 rotations (IQR: 3 to 6). The median number of rotors per patient was 1 (IQR: 1 to 6).
2. Common sites of rotor visualization were at the base of the LAA (41%), near the ostium of the left superior pulmonary vein (35%), left inferior pulmonary vein (12%), right superior pulmonary vein (6%), and right inferior pulmonary vein (6%).
3. In 4 of 9 (44%) patients with rotors, the transient rotors reformed at the same anatomic location. The mean time interval from the cessation and formation of the rotor at the same anatomic location was 18.2 ± 16.8 s.
4. Rotors were only visualized in areas of relatively higher basket electrode density where the interelectrode distance was shorter compared with nonrotor sites.
Mechanisms of human persistent AF
The mechanisms that sustain persistent AF continue to be debated. Recent studies using different techniques have observed different mechanisms operative in human persistent AF (1–8). High-density epicardial mapping studies have suggested that the mechanism may be caused by multiple wavefronts with both longitudinal dissociation of muscle bundles (1) and epicardial-endocardial dissociation (2,3), whereas another study suggested the presence of multiple focal drivers (4). Neither of these studies detected rotors. One high-density epicardial mapping study demonstrated the presence of transient rotors that tended to reform at the same anatomic location (18).
Other approaches have attempted more global mapping of the atrium to guide an ablation strategy. Haissaguerre et al. (5) using body surface mapping described the presence of driver domains of which 80.5% were re-entry and 19.5% were focal. Re-entry lasted a mean of 2.6 rotations only but tended to recur at the same anatomic site. In a nonrandomized study, ablation of driver domains alone terminated 75% of persistent AF but only 15% of long-lasting persistent (5). Narayan et al. (8) were the first to describe the use of the Constellation basket catheter for mapping persistent AF. Using phase mapping, they made the seminal observation that persistent AF is maintained by 1 to 2 sustained rotors in both the left and right atria (8). Targeting the center of these rotors resulted in successful acute and long-term outcomes in up to 82% of patients with paroxysmal and persistent AF (9,24). Although these data have been reproduced in other laboratories (10,11) this has not been uniform (12–14). One of the limitations of the initial iteration of this technology has been the 2D display of the atria with all the inherent assumptions that this engenders. Assuming that electrodes are evenly distributed around the atrium, that they encompass most of the chamber surface area and can be represented accurately by a uniform 2D grid may not always be the case.
In the current study we performed phase mapping using the actual anatomic location of electrograms in 3D space. In 6-min recordings, the dominant activation patterns were heterogeneous consisting of rotors, single and multiple wavefronts, and isolated focal activations. Although we demonstrated rotors in most patients, these were transient with a median of 4 rotations.
The reasons for the differences in findings between the various studies remains unclear and is likely multifactorial. A recent analysis by Alhusseini et al. (25) observed the presence of sustained rotational activity at sites of AF termination with ablation using a similar phase analysis technique as in our study but with the projection of phase using 2D maps. The difference in findings with our study may be caused by the anatomic and spatial assumptions required of 2D phase mapping. These include the use of an 8 × 8 grid with equidistant electrodes and complete LA coverage. In our technique, we have created maps using the actual electrode locations in 3D space and in relation to patient-specific LA geometry (25). In addition, the way phase data are interpolated and visualized may have a significant impact on the generation and interpretation of the phase maps (26). Berenfeld and Oral (27) observed that because the interpolation algorithm is designed to show predominantly rotational activity, incorrect interpolation of phase may result in overdetection of rotors (27). Cardiac motion may also exacerbate problems associated with interpolation.
Our results showing transient rotors that tend to reform at the same anatomic location are similar to those reported by Haissaguerre et al. (5) and to those we have previously reported using epicardial mapping (18).
Based on work by Kuklik et al. (28) that found that the number of detected phase singularities increased with decreasing electrode density resulting in false-positive detections, we did not perform phase singularity mapping given the possibility of false-positive detection using the low-density Constellation basket catheter.
Spatial distribution of rotors
In the current study, rotors were only visualized in areas where there was a relatively high density of basket electrodes and conversely not seen in regions where there was a paucity of electrodes. Thus, rotors were commonly observed at the base of the LAA and adjacent to the left superior pulmonary vein. They were infrequently seen near the right-sided pulmonary veins and not seen at the septum. The anatomic distribution of rotors may be explained by the better coverage of the basket catheter of the lateral wall and an absence of contact with the septum and the impact of electrode density on rotor detection. The importance of electrode density was reported in a study by Walters et al. (29) who observed that low electrode density overstates the prevalence of simple broad linear wavefront activation while understating the prevalence of complex activation patterns and rotors. The development of new basket electrode catheters with high electrode density and relatively uniform coverage of the atrium seems of critical importance in better understanding the mechanisms of persistent AF and may potentially demonstrate rotors in regions not observed in the current study.
This is a small study designed to validate a novel 3D phase mapping approach to understand the mechanisms underlying human persistent AF. As such, the sample size precludes definitive conclusions regarding AF mechanism. Similarly, because we did not perform radiofrequency ablation at rotor sites, it remains unclear whether the transient rotors observed were critical to the arrhythmia maintenance (25). We defined a rotor as the presence of a rotating wavefront with at least 2 complete revolutions of 360 degrees consistent with prior studies. Distinguishing a rotor compared with leading circle or anatomically determined re-entry was beyond the resolution of our maps. The limitations of the multielectrode basket catheter in terms of electrode density and providing uniform atrial coverage have been discussed previously.
3D phase mapping of persistent AF is a novel technique that allows phase data to be visualized on patient-specific 3D LA surface geometry. During prolonged 3D phase mapping, transient rotors were observed in 64% of patients and reformed at the same anatomic location in 44% of patients. In addition, rotors were only visualized in areas of relatively higher basket electrode density where the interelectrode distance was shorter compared with nonrotor sites.
COMPETENCY IN MEDICAL KNOWLEDGE: Using 3D phase mapping, persistent AF was characterized by heterogeneous patterns of activation consisting predominantly of single wavefronts followed by isolated focal activations, disorganized activity and transient rotational activity. Although rotational activity was seen in most patients, all rotors were transient with a mean duration of 1.0 ± 0.6 s. In 44% of patients with rotors, the rotors reformed in the same anatomic location.
TRANSLATIONAL OUTLOOK: In the current study, rotors were only visualized in areas where there was a relatively high density of basket electrodes and conversely not seen in regions where there was a paucity of electrodes. The detection of rotors may be limited by the electrode density of the Constellation basket catheter.
Dr. Pathik is a recipient of the Postgraduate Research Scholarship from the National Health and Medical Research Council of Australia (NHMRC) and the National Heart Foundation of Australia. Dr. Kalman is supported by a Practitioner Fellowship from the NHMRC; and fellowship support from Medtronic, St. Jude Medical, Biosense Webster, and Boston Scientific. Dr. Kuklik has received lecture fees from Abbott. Dr. Zhao is supported by the Health Research Council of New Zealand. Dr. Kistler is supported by a Practitioner Fellowship from the NHMRC. Dr. Lee is supported by an Early Career Fellowship from the NHMRC. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- atrial fibrillation
- atrial fibrillation cycle length
- Cardiac Electrophysiology Analysis System
- computerized tomography
- interquartile range
- left atrium
- left atrial appendage
- Received March 31, 2017.
- Revision received June 1, 2017.
- Accepted June 9, 2017.
- 2018 American College of Cardiology Foundation
- Allessie M.A.,
- de Groot N.M.,
- Houben R.P.,
- et al.
- de Groot N.M.,
- Houben R.P.,
- Smeets J.L.,
- et al.
- de Groot N.,
- van der Does L.,
- Yaksh A.,
- et al.
- Lee S.,
- Sahadevan J.,
- Khrestian C.M.,
- et al.
- Haïssaguerre M.,
- Hocini M.,
- Denis A.,
- et al.
- Daoud E.,
- Zeidan Z.,
- Hummel J.,
- et al.
- Narayan S.M.,
- Krummen D.E.,
- Shivkumar K.,
- Clopton P.,
- Rappel W.J.,
- Miller J.M.
- Miller J.M.,
- Kalra V.,
- Das M.K.,
- et al.
- Steinberg J.S.,
- Shah Y.,
- Bhatt A.,
- et al.
- January C.T.,
- Wann L.S.,
- Alpert J.S.,
- et al.
- Kuklik P.,
- Zeemering S.,
- Maesen B.,
- et al.
- Walters T.,
- Lee G.,
- Morris G.,
- et al.
- Schroeder N.,
- Martin K.,
- Lorensen W.
- Hanwell M.,
- Martin K.,
- Chaudhary A.,
- Avila L.
- Krane K.S.
- Shepard D.
- Narayan S.M.,
- Baykaner T.,
- Clopton P.,
- et al.
- Alhusseini M.,
- Vidmar D.,
- Meckler G.L.,
- et al.
- Umapathy K.,
- Nair K.,
- Masse S.,
- et al.
- Kuklik P.,
- Zeemering S.,
- van Hunnik A.,
- et al.
- Walters T.E.,
- Lee G.,
- Spence S.,
- Kalman J.M.