Author + information
- Received May 4, 2018
- Revision received July 19, 2018
- Accepted August 23, 2018
- Published online December 17, 2018.
- Brian J. Hansen, BSca,b,
- Jichao Zhao, PhDc,
- Ning Li, MD, PhDa,b,
- Alexander Zolotarev, BSca,d,
- Stanislav Zakharkin, PhDa,
- Yufeng Wang, BScc,
- Josh Atwalc,
- Anuradha Kalyanasundaram, PhDa,b,
- Suhaib H. Abudulwaheda,
- Katelynn M. Helfrich, BSca,
- Anna Bratasz, PhDb,
- Kimerly A. Powell, PhDb,
- Bryan Whitson, MDb,e,
- Peter J. Mohler, PhDb,f,
- Paul M.L. Janssen, PhDa,b,
- Orlando P. Simonetti, PhDb,g,
- John D. Hummel, MDb,f and
- Vadim V. Fedorov, PhDa,b,∗ ()
- aDepartment of Physiology and Cell Biology, The Ohio State University Wexner Medical Center, Columbus, Ohio
- bDavis Heart and Lung Research Institute, The Ohio State University Wexner Medical Center, Columbus, Ohio
- cAuckland Bioengineering Institute, The University of Auckland, Auckland, New Zealand
- dPhystech School of Biological and Medical Physics, Moscow Institute of Physic and Technology, Dolgoprudny, Russian Federation
- eDepartment of Surgery, The Ohio State University Wexner Medical Center, Columbus, Ohio
- fDepartment of Internal Medicine, The Ohio State University Wexner Medical Center, Columbus, Ohio
- gDepartment of Biomedical Engineering, The Ohio State University Wexner Medical Center, Columbus, Ohio
- ↵∗Address for correspondence:
Dr. Vadim V. Fedorov, Department of Physiology and Cell Biology, The Ohio State University Wexner Medical Center, 5196 Graves Hall, 333 West 10th Avenue, Columbus, Ohio 43210-1218.
Objectives This study sought to improve atrial fibrillation (AF) driver identification by integrating clinical multielectrode mapping with driver fingerprints defined by high-resolution ex vivo 3-dimensional (3D) functional and structural imaging.
Background Clinical multielectrode mapping of AF drivers suffers from variable contact, signal processing, and structural complexity within the 3D human atrial wall, raising questions on the validity of such drivers.
Methods Sustained AF was mapped in coronary-perfused explanted human hearts (n = 11) with transmural near-infrared optical mapping (∼0.3 mm2 resolution). Simultaneously, custom FIRMap catheters (∼9 × 9 mm2 resolution) mapped endocardial and epicardial surfaces, which were analyzed by Focal Impulse and Rotor Mapping activation and Rotational Activity Profile (Abbott Labs, Chicago, Illinois). Functional maps were integrated with contrast-enhanced cardiac magnetic resonance imaging (∼0.1 mm3 resolution) analysis of 3D fibrosis architecture.
Results During sustained AF, near-infrared optical mapping identified 1 to 2 intramural, spatially stable re-entrant AF drivers per heart. Driver targeted ablation affecting 2.2 ± 1.1% of the atrial surface terminated and prevented AF. Driver regions had significantly higher phase singularity density and dominant frequency than neighboring nondriver regions. Focal Impulse and Rotor Mapping had 80% sensitivity to near-infrared optical mapping–defined driver locations (16 of 20), and matched 14 of 20 driver visualizations: 10 of 14 re-entries seen with Rotational Activity Profile; and 4 of 6 breakthrough/focal patterns. Focal Impulse and Rotor Mapping detected 1.1 ± 0.9 false-positive rotational activity profiles per recording, but these regions had lower intramural contrast-enhanced cardiac magnetic resonance imaging fibrosis than did driver regions (14.9 ± 7.9% vs. 23.2 ± 10.5%; p < 0.005).
Conclusions The study revealed that both re-entrant and breakthrough/focal AF driver patterns visualized by surface-only clinical multielectrodes can represent projections of 3D intramural microanatomic re-entries. Integration of multielectrode mapping and 3D fibrosis analysis may enhance AF driver detection, thereby improving the efficacy of driver-targeted ablation.
- atrial fibrillation
- contrast-enhanced cardiac magnetic resonance
- multi-electrode mapping
- optical mapping
Atrial fibrillation (AF) affects over 33 million people worldwide and is a leading cause of hospitalization and mortality (1,2). Recent evidence from experimental and clinical studies suggests that AF may be maintained by localized AF drivers, which are finite regions of fast, repetitive conduction in both the left atrium (LA) and right atrium (RA) (3–5). Targeted ablation of AF drivers guided by panoramic multielectrode mapping techniques, such as Focal Impulse and Rotor Mapping (FIRM) (Abbott Labs, Chicago, Illinois) and electrocardiographic imaging (ECGI), is used at health care centers worldwide with considerable success rates in AF treatment (5,6). However, some studies using the same techniques have reported much lower success rates (6,7). Furthermore, recent high-density local multielectrode mapping during open-chest procedures did not visualize any AF drivers (8). Therefore, resolving controversies in identifying as well as clarifying the role of drivers in human AF is of utmost importance to the field of AF ablation (9).
The complex 3-dimensional (3D) structure of human atria (10,11) remains a critical obstacle in identifying AF drivers for all clinical electrode mapping approaches, which are limited to endocardial/epicardial surface-only recordings. Moreover, unvalidated analysis of fractionated clinical electrograms may incompletely represent or miss intramural AF drivers (12). Nonetheless, some of these electrogram limitations could be mitigated by phase mapping, which bypasses the need for annotating local activation times (13) and has been utilized in clinical mapping, such as ECGI (4) and FIRM, to highlight rotation by phase singularities (PS) and Rotational Activity Profiles (RAP) (Abbott Labs, Chicago, Illinois) (13). However, phase processing can create false-positive artificial “drivers,” which obscure AF driver detection in clinical settings (11,14).
Therefore, the accuracy of clinical mapping may be reliably determined from direct validation of transmural conduction within the complex 3D human atria. Our recent studies have revealed that dual-sided high-resolution near-infrared optical mapping (NIOM) of the explanted, coronary-perfused human atria, integrated with high-resolution 3D contrast-enhanced cardiac magnetic resonance (CE-CMR) imaging, can “interpolate” conduction patterns of intramural re-entrant AF drivers that appear as breakthrough or focal patterns projected onto endocardial or epicardial surfaces (15–17).
Hence, the goals of this ex vivo human atrial study are to: 1) simultaneously map ex vivo human atria with the clinically used electrode mapping system, FIRM, and dual-sided transmural NIOM, to assess whether re-entrant or breakthrough/focal activation patterns in human AF, seen clinically, represent AF drivers; 2) identify unique patterns or “fingerprints” of AF driver regions by integrating functional and structural features from NIOM and multielectrode mapping with 3D CE-CMR analyses to distinguish true-positive AF drivers from false-positive “drivers.”
An expanded methods section can be found in the Online Appendix.
Explanted human hearts and inclusion criteria
Deidentified, coded human hearts were obtained from The Ohio State University Cardiac Transplant team and LifeLine of Ohio in accordance with The Ohio State University Institutional Review Board. Patient-specific data can be found in Online Table 1 and the procedures performed on each atrial preparation and the type of preparation are listed in Online Table 2. Only atrial preparations with sustained AF (>1 min) induced by burst pacing and localized drivers confirmed by NIOM (15–17) were included in this study’s driver analysis (n = 11). The current study presents entirely new analyses and techniques, such as integrated phase singularity density (PSD), multielectrode mapping, and 3D CE-CMR fibrosis analysis.
NIOM of coronary-perfused human atria to define AF driver regions
Human intact biatrial (n = 2), RA (n = 8), and LA (n = 1) preparations were isolated, coronary-perfused, immobilized, and stained, as previously described (Online Figure 1) (15,16). Biatrial or dual-sided NIOM was conducted during sustained AF and data were analyzed to identify AF drivers by maximum signal derivative (dV/dtmax) activation mapping and dominant frequency (DF) analysis (15,16). Driver activation patterns were classified as re-entrant or breakthrough, both patterns were shown by dual-sided NIOM to be caused by intramural re-entry (Online Figure 2) (16). Online Table 3 lists the conditions under which sustained AF was induced, number of drivers, and driver visualization pattern for each AF episode. Hilbert transform of optical action potentials (OAPs) was used to calculate the local phase (18), PS, and PSD, which was calculated for each pixel as the percentage of total temporal frames (Figure 1B).
Termination of AF to normal sinus rhythm or atrial tachycardia by targeted radiofrequency ablation applied endocardially to re-entrant tracks (Online Figure 3) was used to confirm NIOM-defined AF drivers as the mechanism of sustained AF (15). The endocardial surface area for targeted ablation (that which terminated AF, converted to atrial tachycardia, or slowed AF by ≥10%), untargeted ablation (a negative control that was outside the driver region and had no effect on driver dynamics or global AF pattern), and ablation of atrial tachycardia was calculated relative to total endocardial surface area.
Integration of FIRM
To compare NIOM and clinical multielectrode mapping, 2 64-electrode (8 × 8) FIRMap catheters (Abbott Labs, Chicago, Illinois) with the standard resolution of ∼9 × 9 mm2 were customized and flattened so they could map both epicardial and endocardial lateral atrial surfaces. The flattened FIRMap catheter has the same electrode size and interelectrode distances as its clinical counterpart (Online Figure 1). RhythmView (version 6.1; Abbott Labs) processed AF activation patterns. Blinded to NIOM and structural data, unipolar activation times for a 4-s FIRM epoch within NIOM recordings were marked manually by Abbott engineers in accordance with the RhythmView criteria. RAP marked by RhythmView version 6.1, which combines unipolar electrogram-derived PSD and activation maps, was used to define re-entrant patterns of AF drivers by FIRM (13). Focal patterns were defined as activation beginning at 1 or more electrodes simultaneously and spreading outward. The location of a driver was congruent between NIOM and FIRM if the location differed by ≤1 interelectrode distance. RAP locations not matched with a NIOM-defined AF driver were considered false-positives. For electrogram DF analysis, each unipolar atrial electrogram from the catheter was filtered using a Butterworth filter, and then DF was estimated using a fast Fourier transform.
Gadolinium-based CE-CMR (9.4 T) was used to define atrial anatomy (Online Figure 1) and 3D fibrosis distribution at a resolution of 0.09 to 0.18 mm3, which is significantly higher than could be obtained by current clinical CE-CMR (∼1.25 mm3 by 3-T), allowed the subdivision of the atrial wall into subepicardium, intramural wall, and subendocardium, as previously described (17).
CE-CMR–detected fibrosis is reported as the percentage of gadolinium-enhanced voxels, based on intensity thresholds defined by comparing CE-CMR with histology, within the subvolume of the atrial wall.
Data are presented as mean ± SD. Pairwise comparisons were done with Tukey adjustment. Analysis was done in R version 3.4.1 using packages lme4 and lsmeans. Pearson correlation was used to evaluate agreement between NIOM and FIRM results. p values ≤0.05 were considered significant.
AF driver identification by NIOM in the human heart ex Vivo
AF driver dynamics were analyzed in 17 episodes of sustained AF in 2 intact atria, 8 isolated lateral RA, and 1 isolated LA. Thirty-four AF driver visualizations were recorded, either simultaneously from the subendocardium and subepicardium in the isolated RA/LA (n = 14 × 2 = 28) or recorded from the subepicardium in the intact atria (n = 6). During sustained AF, NIOM wave front activation maps revealed 1.2 ± 0.4 AF drivers per episode (Figure 1A). All drivers were spatially stable, but in 3 hearts (nos. 947200, 402879, and 728878), when 2 drivers were present in the same episode, the temporal stability of each could be <100%, yet AF drivers repeatedly returned to and remained at the same anatomical location (16,17). Re-entrant and breakthrough patterns were both seen in driver regions (Online Table 3). Similar to earlier findings (15), localized intramural re-entry (n = 14) mapped by dual-sided NIOM was seen more often as re-entry from the endocardium, and the same driver visualized from the epicardium could show a stable breakthrough pattern (n = 7). Radiofrequency ablation targeting the pivot points of NIOM-defined re-entrant drivers confirmed they were the sustaining mechanism in 10 of 10 AF episodes each lasting ≥10 min (Online Figure 3, Online Table 2). Driver-targeted ablation covered 2.2 ± 1.1% of the endocardial surface, and AF terminated or converted to atrial tachycardia 92 ± 36 s after the last targeted AF ablation application. Atrial tachycardia ablation, necessary in 4 atria, covered 1.2 ± 1.0%, and untargeted ablation, used in 4 atria, covered 1.3 ± 0.5%, but had no effect of AF dynamics (Online Table 2). A total of 3.5 ± 1.4% of the endocardial surface was ablated until arrhythmia could not be induced.
Identifying AF driver regions by NIOM phase mapping
Phase processing was applied to NIOM OAP to assess whether AF drivers in the human heart could be visualized by their resulting PS distribution and density (Figure 1). Figure 1A shows an activation map of a temporally and spatially stable re-entrant driver, and Figure 1B shows PS were present primarily near the pivoting points. Whereas PS were intermittent and discontinuous throughout 1 re-entry cycle, PSD mapping showed the highest density along the re-entry track, formed by several partially insulated pectinate bundles seen by 3D CE-CMR. PS could also be present outside the re-entrant track due to wave breaks, primarily at discontinuous or twisted myobundles, but they occurred with a lower PSD. All AF driver regions had significantly higher average and maximum PSD of 0.39 ± 0.26% and 3.14 ± 2.67% versus 0.20 ± 0.18% (p < 0.05) and 0.87 ± 0.74% (p < 0.01) for nondriver regions, respectively. However, in cases where the intramural re-entrant drivers were visualized with a breakthrough pattern from either subendocardial or subepicardial NIOM, phase mapping did not have the distinguishable clusters of high PSD that were present on maps from the other atrial surface where a re-entrant pattern was visualized (Figure 1C, Online Figure 2). Separating AF driver regions into their visualization patterns showed average PSD to be significantly higher only in those drivers visualized as re-entrant and not breakthrough patterns (Figure 1D).
Sensitivity and false-positive rate of AF driver detection by clinical multielectrode catheters compared with NIOM
In 6 hearts, we estimated the correlation between FIRM and NIOM for activation patterns and AF driver visualization. Figure 2A shows that activation patterns during 500-ms cycle length pacing were similar between the 2 methods, and the total activation time across the optical field of view correlated. During sustained AF, driver DF seen by FIRM and NIOM was also strongly correlated (Figure 2B, Online Table 3). Next, we compared AF driver visualization between NIOM and FIRM, which were analyzed blinded to each other (Online Table 3). Figure 3A shows an activation map of 2 competing re-entrant drivers identified by NIOM that were characterized by clusters of high PSD. Clusters of RAP were present within or near NIOM re-entry tracks (Online Video 1). Figure 3B shows OAP and FIRM electrograms on the re-entrant driver track confirming repetitive re-entrant activity. During sustained AF, FIRM identified 16 of 20 NIOM-defined AF driver locations within 1 interelectrode distance. RhythmView visualized RAP at 10 of 14 re-entrant AF driver regions. Additionally, 4 of 6 breakthrough patterns defined by NIOM were also visualized by FIRM as focal activation (Figure 4A, Online Video 2, right panel). Figure 4B shows 1 of the 2 instances when a re-entrant driver identified by epicardial NIOM was seen as a focal activation pattern by epicardial FIRM. This observation was consistent as shown in Online Figure 4 by 5 consecutive activation maps for this case from both epicardium and endocardium seen by both NIOM and FIRM simultaneously. Online Figure 5 shows FIRM correctly identified the NIOM-defined post-ablation AF driver in the isolated LA preparation as well. Furthermore, 2 re-entrant and 2 breakthrough/focal NIOM AF driver patterns were unidentified by FIRM and seen only as chaotic activation (Online Video 2, left panel). In 2 of these cases, the electrodes overlying these missed AF drivers were >2 mm away from the tissue. In some instances, more organized re-entrant arrhythmias, such as post-ablation atrial tachycardia (n = 4) (Online Figure 3) were also induced, and multielectrode mapping consistently reproduced the macro-re-entry pattern and location identified by NIOM.
Furthermore, FIRM could identify false-positive RAP outside NIOM-defined driver regions along with the RAP at true-positive drivers. Figure 5A shows correctly defined RAP at electrodes CD23 and EF23 overlying the 2 pivot points of the NIOM-defined re-entrant AF driver (Figure 5B). Moreover, targeted ablation at EF23 successfully terminated AF, whereas the remaining 3 RAP locations were not involved in AF maintenance and represent false-positive RAP. In total, 18 false-positive RAP locations were identified (1.1 ± 0.9 per AF visualization), 4 of which had 1 or more electrodes with no underlying atrial tissue, and therefore were excluded from NIOM analysis, just as false-positives identified at electrodes overhanging the atrioventricular annuli would be ignored by clinicians. In Figure 5B, NIOM activation maps show that false-positive RAP resulted from passive rotation, where activation rotates but does not make a self-sustaining circuit. Moreover, true-positive RAP locations had a NIOM-defined average and maximum PSD significantly higher than false-positive RAP locations (Figure 5C). There were no multielectrode maps where false-positive RAP was marked without identifying at least 1 NIOM-defined AF driver in the same map (Online Table 3). False-positive RAP locations did not have a significantly lower FIRM DF or NIOM DF than AF drivers did. Importantly, there was no significant difference in the intensity of RAP between true-positive drivers and false-positives, which prevents FIRM alone from distinguishing them.
Fibrosis architecture distinguishes true-positive from false-positive RAP
High-resolution (∼100 μm3) CE-CMR shows the 3D distribution of intramural fibrosis in the lateral RA (Figure 6A) with histological analysis confirming the layers of epicardial and endocardial fibrosis as well as the specific intramural fibrosis-insulated myobundles (Figure 6B). Whereas the percentage of intramural fibrosis was significantly higher in NIOM-defined AF driver regions than in false-positive RAP and true-negative nondriver regions (Figure 6C), subepicardial, subendocardial, and total fibrosis values were not associated with true-positive AF drivers.
Our study provides the first validation of clinically visualized re-entrant and focal patterns as AF drivers in human hearts. Our novel approach, integrating NIOM and 3D CE-CMR with clinical multielectrode mapping, demonstrates that stable or repetitive re-entrant and focal surface visualizations of AF drivers may stem from intramural microanatomic re-entry and could represent targets for ablation treatment. False-positive “drivers” could also be visualized by clinical multielectrode mapping because of passively rotating fibrillatory waves or phase processing of electrograms with poor signal quality. Furthermore, by defining structural and electrophysiologic fingerprints of human AF drivers, our results emphasize that further integration of clinical multielectrode mapping with intramural 3D fibrosis distribution may help differentiate true-positive AF drivers from false-positives, which could improve the success of driver-targeted ablation.
Validation of clinically visualized AF drivers
FIRM mapping surface visualization of AF drivers matched pattern and location seen by NIOM in 70% and 80% of cases, respectively. These discrepancies in driver visualization may be due to differences in the nature of signal collection. NIOM collects signals at high resolution (up to 0.3 mm2) and uses near-infrared voltage sensitive dyes to record optical transmembrane action potentials over a depth of 4 mm and does not rely on contact with the tissue (15). On the other hand, conventional unipolar and bipolar multielectrode mapping have relatively low resolution and rely on direct contact with the tissue to detect extracellular potentials, which may be contaminated by far-field signals (19,20). Thus, a re-entry circuit could be misinterpreted as a focal pattern (Figure 4, Online Figure 4) or completely missed due to low resolution of panoramic electrode mapping (∼9 mm2 for FIRM). Even the highest resolution (5.6 to 7 mm) panoramic biatrial epicardial clinical multielectrode mapping study to date by the Waldo group (21) visualized predominantly focal/breakthrough patterns rather than re-entry. However, current clinical high-resolution multielectrode arrays (2 to 2.5 mm2) (8,22) have only small tissue coverage; thus, the lack of a stable driver signal in these studies may result from missing the driver entirely.
Additionally, 2 of the 4 AF drivers that were missed by FIRM were due to poor tissue contact in driver regions. Clinically, the spherical shape of the basket catheter used for mapping may not make good tissue contact with the nonspheroidal atrial chambers. We recently reviewed (23) the challenge of ensuring good contact across the human atria wall, which varies from 0.1- to 12-mm thick (17). Dual-sided multielectrode arrays (8) may have direct contact only with the thickest pectinate bundles; thus, driving re-entry along small, intramural bundles could be visualized at the surface as breakthrough or mistaken entirely as multiwavelets.
A recent clinical study by de Groot et al. (8) used local dual-sided multielectrode arrays, inserted from an incision in the RA appendage to map activation patterns in 14 patients with and without AF. Their analysis of AF patterns demonstrated dissociation of endocardial–epicardial activation and focal waves/breakthroughs in all patients, but they did not report re-entrant activity. Although this contact multielectrode study would suffer from the same obstacles facing FIRM contact electrodes, the discrepancies in driver identification may stem from factors other than those listed herein. First, the high-resolution dual-sided array only covers a 14 × 30 mm2 section of the RA versus the panoramic FIRM, and waves entering from the periphery of the array were not eligible for their driver definition. Second, the de Groot et al. study (8) placed restrictions on what were considered continuous waves by activation time delays between electrodes whereas FIRM has no such restrictions. Third, other factors that varied between the 2 approaches but have unknown effects on driver detection are electrode sizes, interpolation, and filtering in electrograms, as we recently reviewed (10). However, only the direct recording of FIRM-defined AF drivers by the dual-sided multielectrode array used in the de Groot et al. (8) study in the same heart, or validation of the dual-sided array by optical mapping ex vivo, could reconcile the differences in AF patterns revealed between these techniques.
Early ex vivo animal studies found that differences between optical and electrogram activation times stemmed primarily from the increased depth of interrogation by optical mapping (20,24). Our recent dual-sided NIOM study of the human atria showed epicardial–endocardial dissociation up to 105 ms during AF (15). Moreover, atrial wall thickness variation, myofiber orientation misalignment, and intramural fibrosis (8,15) could lead to asynchronous activation of the partially discontinuous neighboring myocardial layers, thereby causing activation discordance and fractionation on a bipolar electrogram (15,23). Nevertheless, clinical studies using both FIRM (5) and CARTOFINDER (Biosense Webster, Diamond Bar, California) (25) found spatially stable AF drivers, which could be visualized with both focal and re-entrant patterns; likewise, the microanatomic re-entrant AF drivers observed in our ex vivo studies were visualized by FIRM with similar patterns, which suggests that the mechanism of AF maintenance could be similar in both in vivo and ex vivo human atria.
Our study observed AF drivers in both the RA and LA. Ever since the pulmonary veins were shown to be sources of triggers in the initiation of AF, AF treatment has focused mainly on LA ablation. However, recent panoramic atrial mapping studies, which have found AF drivers using a variety of techniques, observed drivers in both the LA and RA (4,5,21). The data suggest that patients with more persistent forms of AF may have more RA drivers than those with paroxysmal AF, which suggests that patient-specific AF drivers could be more efficient targets in the persistent AF population rather than pulmonary triggers (5).
Pros and cons of phase mapping to identify re-entrant AF drivers
To bypass the challenge of marking activation times during AF (13,14), instantaneous phase analysis has been applied to identify PS or the center of rotational activity in clinical AF mapping. This is the first study that identifies intramural re-entrant AF drivers by phase analysis of high-resolution NIOM in the ex vivo human heart. Because the highest PSD clusters were found at re-entrant AF driver sites, our study supports the cautious use of phase processing of multielectrode mapping in AF patients.
Noticeably, PS seen clinically are suggested to represent the functional core of “rotors” capable of meandering across the atrium (3,26), which raises concerns regarding targeted ablation of “rotor”-driven AF (27). However, clinical multielectrode mapping is unable to distinguish between microanatomic re-entries, rotors, or focal AF drivers. Indeed, our study observed that FIRM could visualize NIOM re-entry as focal activity (Figure 4, Online Figure 4). Furthermore, our data also suggest that clinically visualized drifting/meandering “rotors” could be a misinterpretation by phase processing of an elongated microanatomic re-entry track formed by partially discontinuous myobundles. Our findings are supported by clinical ECGI studies that also observed intermittent yet recurring PS in driver regions (4). The recent 64-electrode basket catheter mapping study (25) processed with CARTOFINDER activation mapping of unipolar electrograms, also identified ablatable localized drivers in almost all patients in the form of intermittent but repetitive focal or rotational activation patterns.
Our findings could also explain the lack of temporal stability or intermittency of AF drivers, a critical limitation of clinical recordings. First, as we described, surface activation does not always portray intramural activation. Thus, surface visualization of microanatomic 3D re-entrant drivers may be obscured as beat-to-beat fibrillatory conduction changes around the continuously active driver. Second, PS temporal stability at microanatomic re-entrant drivers may also be intermittent or interrupted due to the elongated structure of the re-entrant track and not the existence of a functional “rotor” (Online Figure 2). Lastly, if more than 1 driver was identified (Figure 3), each may not be continuously present and may instead disappear and reoccur at the same atrial region; however, at least 1 driver was always active during sustained AF (16).
Even though phase-based mapping may provide an opportunity to visualize re-entrant AF drivers in the clinical setting, it is also biased toward rotational visualization, which may lead to false-positive RAP identification (11,14,28) as observed in our study. False-positive targets might explain some of the discrepancies between clinical studies using FIRM (3,5,6). Thus, exclusion criteria identified by ex vivo studies might provide a valuable approach to accurately distinguish between true-positive and false-positive RAP.
Interestingly, there were no apparent differences between true-positive drivers and false-positive RAP by either RAP intensity or NIOM and FIRM DF. However, we found that false-positive RAP locations were defined by lower maximum NIOM PSD than by drivers. Whereas RAP is partly based on PSD analysis (13), further studies are required to understand how differences between these 2 methods, such as resolution, may affect AF driver identification.
One possible explanation for the clinically visualized false-positive “drivers” may be fibrillatory conduction block causing repetitive passive rotation (Figure 5). Vijayakumar et al. (11) recently reported that phase processing can create PS at regions of conduction block, which do not represent self-sustaining re-entrant activity and instead could be mistaken as false-positive “rotors” (14,28). Our results also revealed that false-positive RAP could form at structural/functional conduction blocks (Figure 5) with less intramural fibrosis than NIOM-confirmed driver regions (Figure 6). These important findings suggest that mapping intramural atrial fibrosis may be particularly useful in identifying AF drivers.
Intramural fibrosis as a distinguishing “fingerprint” of re-entrant AF drivers
Clinical and experimental studies have shown that increased fibrosis is a significant contributor to the dysfunctional effects of atrial remodeling (9,23,29). Our current study suggests that AF drivers can be distinguished from nondriver areas by the extent of intramural fibrosis. Furthermore, we could eliminate 5 of 14 false-positive RAP if a threshold of >12% intramural fibrosis by ex vivo CE-CMR (Figure 6) is applied for a RAP region to qualify as an AF driver region in our dataset. Our recent experimental (15,17) and computational studies from other groups (30) suggest that fibrotic architecture in the border zones of dense fibrosis may be more arrhythmogenic than the dense patch. Moreover, our human atria studies (17,23) also suggested that only fibrosis affecting the path of conduction could be arrhythmogenic, in contrast to fibrosis that naturally covers the epicardial or endocardial surfaces (Figure 6). Based on our recent (17) and current data (Figure 6), we suggest that arrhythmogenic intramural fibrosis may be considered a unique fingerprint of intramural re-entrant AF drivers.
However, assessing the 3D architecture of intramural fibrosis in human atria may prove difficult for clinical CMR. Although our CE-CMR methodology mimics that used clinically (29) with resolution of 1.25 mm3, the higher magnet strength allowed us to acquire the considerably higher resolution of ∼0.1 mm3 in the ex vivo human atria. As such, the RA wall with an average thickness of ∼4 mm can contain ∼22 voxels even at the lowest resolution done in our study (∼0.180 mm3), which successfully allowed us to differentiate among subendocardial, subepicardial, and intramural fibrosis (Figure 6). Conflicting results have been reported from clinical studies that have looked for a correlation between late gadolinium enhanced (LGE) CMR, a surrogate for fibrosis, and clinical AF driver locations. One study found that areas with high re-entrant activity by ECGI showed higher local LGE-CMR density compared with other areas (31), whereas no correlation between FIRM-defined AF drivers and LGE-CMR was found in another study (32). These discrepancies could be explained by the limitation of clinical multielectrode approaches as well as atrial LGE-CMR, which require further validation by ex vivo CE-CMR and an increase in resolution to resolve 3D architecture of intramural fibrosis.
Recent clinical CMR (31) and postmortem histological (33) studies indicate that fibrosis can be similarly up-regulated in both LA and RA in AF patients. Although several clinical multielectrode mapping studies have shown that patients have not only LA drivers but often RA drivers as well (4,21,34), it is currently unknown whether RA drivers have similar or different structural substrates or fingerprints as LA drivers. This study and our previous high-resolution CE-CMR study of the intact human atria (17) suggest that similar structural fingerprints might underlie drivers in both atria. However, rigorous studies directly comparing structural substrates of LA and RA drivers are required to elucidate differences of clinical relevance.
It may be possible to bridge the gap between high-resolution NIOM and current clinical electrode mapping by complementing low-resolution panoramic catheters with high-resolution catheters. Furthermore, combining multiple analyses of electrode signals (i.e., phase, activation, and RAP) may provide stronger evidence of AF drivers (13). Our data indicate that clinical multielectrode mapping alone may lack the ability to distinguish true-positive drivers from false-positives. Thus, systematic clinical and experimental studies integrating 3D electrical and structural approaches are required to address the patient-specific 3D fibrotic architecture to better identify re-entrant AF drivers for targeted ablation (Figure 7).
Our study includes a small sample of ex vivo human hearts with varied disease history; extrapolation of results to a wider AF population requires further in vivo and ex vivo studies. Due to the limited number of samples, this study cannot definitively state what approach is superior for clinical AF mapping and ablation. We found no difference in fibrotic content between LA and RA drivers; however, further studies are needed to uncover any subtle differences that may affect their identification or ablation. Our denervated and mechanically arrested ex vivo atrial preparations are in the absence of both autonomic and metabolic stresses at baseline. Therefore, pharmacologic autonomic (adenosine or isoproterenol) (16) or metabolic (pinacidil) stimulations (15) were used to include autonomic and metabolic stresses known to play a critical role in AF and recapitulate clinical conduction and repolarization (9). For the direct comparison of NIOM and FIRM, clinical FIRMap basket catheters were flattened; however, electrode size and data analysis remained the same.
This study provides the first validation and mechanistic explanation of localized re-entrant and/or focal human AF drivers as visualized by clinical multielectrode mapping and identifies conditions when drivers may potentially be misinterpreted. Integrated electrophysiological and structural mapping can identify unique structural fingerprints of AF drivers to distinguish true-positive from false-positive AF drivers, but warrants further validation in the clinical setting. If validated in patients, our findings could lead to a more comprehensive and precise understanding of the fundamental mechanisms of AF in patients and improve the success rate of driver-targeted catheter ablation and minimizing the amount of tissue damage.
COMPETENCY IN MEDICAL KNOWLEDGE 1: Whereas clinical multielectrode mapping may successfully detect localized re-entrant AF drivers, it also has the propensity to detect false-positive re-entrant patterns; hence, caution should be taken to prevent unnecessary ablation in AF patients.
COMPETENCY IN MEDICAL KNOWLEDGE 2: Clinical multielectrode mapping at the atrial surfaces can visualize intramural microanatomic re-entrant AF drivers in the human heart as both re-entrant and focal activation patterns, which represent viable targets for efficient ablation treatment.
TRANSLATIONAL OUTLOOK: AF drivers could be predominantly located in regions of the atria with more intramural fibrosis than in regions with false-positive re-entrant patterns. Further clinical studies are required to assess the benefits of integrating 3D structural imaging with multielectrode mapping for designing successful AF driver-targeted ablation treatments.
The authors thank the Lifeline of Ohio Organ Procurement Organization and the Division of Cardiac Surgery at The Ohio State University Wexner Medical Center for providing the explanted hearts. The human heart repository program is supported by the Davis Heart and Lung Research Institute. The authors thank Ms. Esthela Artiga for critical reading of the manuscript and Abbott engineers Mr. Brian Pederson and Mr. Carey Briggs for their help in integrating FIRM mapping and data analysis of FIRM recordings.
This work was supported by National Institutes of Health grants HL115580 and HL135109 (to Dr. Fedorov) and T32HL134616 (to Dr. Hansen), American Heart Association Grant in Aid 16GRNT31010036 (to Dr. Fedorov), and Health Research Council of New Zealand (to Dr. Zhao). Dr. Simonetti receives research support from Siemens. Dr. Hummel is a consultant to Abbott Laboratories. Dr. Fedorov has received research support from Abbott Laboratories. 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
- contrast-enhanced cardiac magnetic resonance
- dominant frequency
- electrocardiographic imaging
- focal impulse and rotor mapping
- left atrium
- late gadolinium-enhanced cardiac magnetic resonance
- near-infrared optical mapping
- optical action potentials
- phase singularity
- phase singularity density
- right atrium
- rotational activity profile
- Received May 4, 2018.
- Revision received July 19, 2018.
- Accepted August 23, 2018.
- 2018 American College of Cardiology Foundation
- Chugh S.S.,
- Havmoeller R.,
- Narayanan K.,
- et al.
- Calkins H.,
- Hindricks G.,
- Cappato R.,
- et al.
- Narayan S.M.,
- Krummen D.E.,
- Shivkumar K.,
- Clopton P.,
- Rappel W.J.,
- Miller J.M.
- Haissaguerre M.,
- Hocini M.,
- Denis A.,
- et al.
- Miller J.M.,
- Kalra V.,
- Das M.K.,
- et al.
- Benharash P.,
- Buch E.,
- Frank P.,
- et al.
- Jalife J.,
- Filgueiras R.D.,
- Berenfeld O.
- de Groot N.,
- van der Does L.,
- Yaksh A.,
- et al.
- Heijman J.,
- Guichard J.B.,
- Dobrev D.,
- Nattel S.
- Hansen B.J.,
- Csepe T.A.,
- Zhao J.,
- Ignozzi A.J.,
- Hummel J.D.,
- Fedorov V.V.
- Vijayakumar R.,
- Vasireddi S.K.,
- Cuculich P.S.,
- Faddis M.N.,
- Rudy Y.
- Sahli Costabal F.,
- Zaman J.A.B.,
- Kuhl E.,
- Narayan S.M.
- Alhusseini M.,
- Vidmar D.,
- Meckler G.L.,
- et al.
- Rodrigo M.,
- Climent A.M.,
- Liberos A.,
- et al.
- Li N.,
- Csepe T.A.,
- Hansen B.J.,
- et al.
- Zhao J.,
- Hansen B.J.,
- Wang Y.,
- et al.
- van der Does L.J.,
- de Groot N.M.
- Lee S.,
- Sahadevan J.,
- Khrestian C.M.,
- Cakulev I.,
- Markowitz A.,
- Waldo A.L.
- Hansen B.J.,
- Zhao J.,
- Fedorov V.V.
- Honarbakhsh S.,
- Schilling R.J.,
- Dhillon G.,
- et al.
- Weiss J.N.,
- Qu Z.,
- Shivkumar K.
- Kuklik P.,
- Zeemering S.,
- van Hunnik A.,
- et al.
- Cochet H.,
- Dubois R.,
- Yamashita S.,
- et al.
- Chrispin J.,
- Gucuk I.E.,
- Zahid S.,
- et al.
- Platonov P.G.,
- Mitrofanova L.B.,
- Orshanskaya V.,
- Ho S.Y.
- Friedman D.J.,
- Liu P.,
- Barnett A.S.,
- et al.