Author + information
- Received March 9, 2012
- Revision received July 12, 2012
- Accepted August 1, 2012
- Published online December 1, 2012.
- Flavia Ravelli, PhD⁎,⁎ (, )
- Michela Masè, PhD⁎,
- Alessandro Cristoforetti, PhD⁎,
- Maurizio Del Greco, MD†,
- Maurizio Centonze, MD‡,
- Massimiliano Marini, MD† and
- Marcello Disertori, MD†
- ↵⁎Reprint requests and correspondence:
Dr. Flavia Ravelli, Laboratory of Biophysics and Biosignals, Department of Physics, University of Trento, Via Sommarive 14, Povo, 38123 Trento, Italy
Objectives The aim of this study was to investigate the anatomic distribution of critical sources in patients with atrial fibrillation (AF) by fusion of biatrial computed tomography (CT) images with cycle length (CL) and wave similarity (WS) maps.
Background Experimental and clinical studies show that atrial fibrillation (AF) may originate from rapid and repetitive (RR) sources of activation. Localization of RR sources may be crucial for an effective ablation treatment. Atrial electrograms showing rapid and repetitive activations can be identified by combining WS and CL analysis.
Methods Patients with persistent AF underwent biatrial electroanatomic mapping and pre-procedural CT cardiac imaging. WS and CL maps were constructed in 17 patients by calculating the degree of repetitiveness of activation waveforms (similarity index [S]) and the cycle length at each atrial site. WS/CL maps were then integrated with biatrial 3-dimensional CT reconstructions by a stochastic approach.
Results Repetitive sources of activation (S ≥0.5) were present in most patients with persistent AF (94%) and were mainly located at the pulmonary veins (82% of patients), at the superior caval vein (41%), on the anterior wall of the right atrium (23%), and at the left atrial appendage (23%). Potential driver sources showing both rapid and repetitive activations (CL = 140.7 ± 25.1 ms, S = 0.65 ± 0.15) were present only in a subset of patients (65%) and were confined to the pulmonary vein region (47% of patients) and left atrial appendage (12%). Differently, the repetitive activity of the superior caval vein was characterized by a slow activation rate (CL = 184.7 ± 14.6 ms).
Conclusions The identification and localization of RR sources is feasible by fusion of biatrial anatomic images with WS/CL maps. Potential driver sources are present only in a subset of patients with persistent AF and are mainly located in the pulmonary vein region.
Chronic atrial fibrillation (AF) remains one of the greatest challenges in cardiac arrhythmia therapy. Despite the high success rate of pulmonary vein (PV) ablation in paroxysmal forms, effective ablation therapy in patients with chronic AF remains elusive (1), due to the still incomplete understanding of the mechanisms sustaining the arrhythmia.
Potential driver sources, characterized by rapid and repetitive activations, as well as areas with fractionated electrograms, have been proposed as critical regions for the maintenance of AF (2–4). The identification and localization of electrograms with these features may improve the success rate of targeted ablation. In order to identify potential AF drivers from single atrial electrograms, a new time-domain approach has been proposed based on the combined analysis of atrial cycle length (CL) and wave similarity (WS) (5). WS analysis is a technique to assess directional organization during fibrillation (6,7), whose performances have been thoroughly tested in previous AF studies (6,8,9). Specifically, WS analysis measures the degree of repetitiveness over time of electrogram activation waveforms, thus allowing the identification of repetitive (R) sources of activation. Complementarily, CL analysis provides a local measure of the fibrillatory rate, determining among organized sources those with the shortest CL.
In the present work, we set out to construct anatomically detailed WS and CL maps to investigate the anatomic distribution of potential drivers in patients with persistent AF. This task was performed by a set of innovative signal and imaging techniques directed to the integration of WS/CL maps with highly resolved 3-dimensional (3D) anatomic reconstructions of the atrial chambers (5,10,11). Accurate 3D computed tomography (CT) reconstructions of the atrial anatomy were obtained by a semiautomatic segmentation process based on a marker-controlled watershed segmentation (11). The spatial distribution of WS/CL indexes was assessed by applying WS analysis and CL measure to each atrial electrogram recorded by an electroanatomic mapping system. Sparse-point WS and CL maps were then integrated on CT atrial reconstructions by a fully automated registration and fusion procedure based on a stochastic approach (10).
Twenty consecutive patients with persistent AF (duration ≥6 months), who underwent catheter ablation for drug-refractory, symptomatic AF, were studied. Patient characteristics are shown in Table 1. The duration of continuous AF was more than 1 year in 18 patients (long-standing persistent AF), whereas in 2 patients, AF lasted 6 and 10 months, respectively. Fifteen of the 20 patients had underlying heart diseases, including 5 patients with hypertensive heart disease. All patients provided written informed consent. The study was approved by the local institution's research ethics board.
Biatrial CT imaging and segmentation
All patients underwent cardiac multidetector CT imaging within 24 h before the ablation procedure. Scans were performed with a 16-slice helical scanner (Philips 8000 Mx IDT, Philips Medical System, Best, the Netherlands) during an inspiratory breath-hold, after intravenous injection of 125 ml of contrast medium (Iomeron 400, Bracco, Italy). Whenever possible, motion artifacts were minimized by retrospective electrocardiogram gating reconstruction at the end-diastolic phase, whereas in patients with high-frequency AF, cardiac gating was not applied.
The isolation of the 3D anatomy of the left atrium (LA), PV, and right atrium (RA) from CT images was performed by a custom-made semiautomatic algorithm based on marker-controlled watershed segmentation. Technical details on the segmentation procedure have been provided elsewhere (11). The 3D reconstructed biatrial surfaces were used in the subsequent registration procedure.
Electrophysiological study and electroanatomic mapping
A quadripolar catheter was placed at the His position and a 20-pole deflectable catheter was inserted into the coronary sinus (CS) for continuous monitoring. Real-time 3D left and right atrial maps were reconstructed by an electroanatomic mapping system (CARTO XP, Biosense Webster, Inc., Diamond Bar, California), sampling evenly distributed endocardial sites with a fill-threshold of 15 to 20 mm. The catheter (NAVISTAR THERMOCOOL, Biosense Webster, Inc.) was held stationary in each location, allowing the acquisition of a bipolar endocardial electrogram of 10-s length at each site. Atrial bipolar electrograms, surface electrocardiograms, and CS signals were continuously monitored and stored on a computer-based digital amplifier/recorder system (C.R. Bard Inc., Bard Electrophysiology Division, Lowell, Massachusetts). Intracardiac electrograms were band-pass filtered (30 to 500 Hz) and digitized at 1-kHz sampling frequency. All electrograms were exported for off-line measurement and analysis. Electrograms presenting an unsatisfying signal-to-noise ratio were excluded from subsequent analysis.
Measurement of CL and WS
The 10-s bipolar electrograms recorded at each mapped location were analyzed to obtain local measures of CL and WS. For CL measurement (5), bipolar electrograms were first pre-processed to remove ventricular interference by an averaging technique. Local atrial activation waves were then identified by signal filtering and comparison with an adaptive threshold, accounting for variations in waveform amplitude. Following wave recognition, atrial activation times were estimated by measuring the barycenter of local activation waves, defined as the time that divided into 2 equal parts the local area of the modulus of the signal. The method provided a reliable estimation of activation times even in presence of fragmented potentials and complex wave morphologies, since the barycenter was computed from the whole shape of the activation wave and thus was less sensitive to local modifications of the waveform (5).
The degree of repetitiveness of the fibrillatory process at each recording site was quantified by measuring the level of similarity of the activation waves (similarity index [S]) (6). S was determined by comparing the morphologies of all possible pairs of atrial activation waves extracted from a single recording. Details on the calculation of S, as well as on its performance in the determination of the spatiotemporal complexity of AF have been published elsewhere (6,8,9). Figure 1A shows representative examples of the application of WS analysis to bipolar electrograms recorded at different atrial sites and presenting various degree of morphological complexity.
The temporal stability of WS/CL indexes was checked by analyzing the CS electrograms over the entire duration of the mapping study. No significant variations were observed either in S (S = 0.21 ± 0.13 vs. 0.21 ± 0.14 vs. 0.21 ± 0.11; p = NS) nor in CL (CL = 169.7 ± 18.8 ms vs. 165.8 ± 21.8 ms vs. 165.6 ± 20.0 ms; p = NS) between the beginning, middle, and end of the study.
Integration and fusion of WS/CL maps with biatrial CT images
Anatomically detailed WS/CL maps were constructed by determining S and CL indexes at each mapping site, and subsequently, registering mapping points and fusing the corresponding indexes on LA and RA anatomic reconstructions (Fig. 1). The registration step was performed by a custom-made, landmark-free algorithm, based on a stochastic approach, which has been detailed elsewhere (10). Briefly, a parameterized rigid geometric transformation was repeatedly applied to the mapping points, searching for the best parameter set that minimized the misalignment between transformed points and atrial reconstructions. The stochastic procedure was fully automated and proved accurate and reproducible data integration (10). Following registration, WS/CL values associated with the mapping points were fused on the registered 3D CT atrial surfaces by a radial basis function interpolation.
After the construction of WS/CL maps, a detailed spatial analysis was performed to evaluate the presence of regional index differences. The RA was divided into superior caval vein (SCV) and anterior, lateral, posterior, and septal regions, whereas the LA was separated into the 4 pulmonary veins (PVs) and posterior wall, septum, roof, and appendage regions.
Continuous data are presented as mean ± SD, and categorical data as numbers or percentages. One-way analysis of variance for repeated measures was performed to compare S/CL values among atrial regions, followed by post hoc analysis with Bonferroni correction. Variable normality distribution was evaluated by Shapiro-Wilk test. A p value <0.05 was considered statistically significant. Statistical analysis was performed with Origin 8.1 Pro (OriginLab Corporation, Northampton, Massachusetts).
WS/CL mapping and registration
To construct WS/CL biatrial maps 123 ± 14 mapping points per patient were acquired (47 ± 6 in the RA and 76 ± 14 points in the LA, respectively). Map analysis was accomplished in 17 patients. Three patients were excluded due to the high number of electrograms with poor signal-to-noise ratio.
In all patients, RA/LA electroanatomic maps were successfully registered on the corresponding 3D CT surface reconstructions with good accuracy (mean point-to-surface distance of 2.37 ± 0.68 mm and 2.22 ± 0.44 mm for RA and LA, respectively) and precision (point scattering over 20 parallel solutions of 1.57 ± 0.48 mm and 1.43 ± 0.75 mm for RA and LA, respectively).
Anatomic distribution of WS
Regional analysis of fused wave similarity maps showed the existence of clear regional differences in wave similarity in both left and right atria (Figs. 2 and 3).⇓⇓ On average (Fig. 4), regions with highly organized electrical activity (high-similarity indexes) were found in the PVs (S = 0.46 ± 0.13) and LA appendage (S = 0.43 ± 0.19), whereas disorganized areas, characterized by low-similarity values, were located on the posterior wall, roof, and septum (S = 0.17 ± 0.08). Focusing on the PV region, the levels of similarity were slightly higher at the superior than the inferior PVs (S = 0.49 ± 0.15 vs. S = 0.42 ± 0.16, p = NS) and at the distal portion of the veins than around the ostium (S = 0.47 ± 0.14 vs. S = 0.31 ± 0.15, p<0.001). In the RA, high-similarity values were found at the SCV (S = 0.47 ± 0.18) and on the anterior wall (S = 0.42 ± 0.13), whereas posterior and septal regions were the most disorganized sites (S = 0.27 ± 0.10).
Anatomic distribution of R sources
In each patient, regions showing S ≥0.5 were selected and marked as repetitive (R) sources. The regional distribution of R sources for the whole population of patients is shown in Figure 4. In 17 patients, a total of 49 R sources were found, with an average of 2.9 ± 1.6 sources per patient (range 0 to 6). R sources were present in almost all patients (94%) and were mainly located in the LA (35 of 49). As evidenced by the representative examples of Figure 2 and by the cumulative data in Figure 4, R sources in the LA were exclusively present at the PV level (31 sources in 82% of patients) and LA appendage (4 in 23% of patients). Focusing on the 4 PVs, the incidence of repetitive activity was higher at the superior than the inferior PVs (18 vs 13). R sources on the RA (Figs. 3 and 4) were mostly located at the SCV (7 of 14) and on the anterior walls (4 of 14).
Anatomic distribution of CL
Map fusion and regional analysis were performed also for atrial CLs. The mean atrial CLs calculated in each atrial region are shown in Table 2. On average, the LA showed a faster rate than the RA (CL = 164.2 ± 20.0 ms vs. 169.9 ± 20.9 ms, p < 0.05). No significant difference in CL was observed among different LA regions (p = NS). The average PV CL was 166.2 ± 24.1 ms, with shorter cycles around the ostium than in the distal area (163.2 ± 23.0 ms vs. 168.7 ± 25.8 ms, p < 0.01). Regional CL differences were observed in the RA (p < 0.01), where the SCV showed the slowest atrial rate (CL = 184.7 ± 14.6 ms, p < 0.01).
Anatomic distribution of RR sources
To identify among R sources those showing a rapid rate, WS maps were systematically compared with CL maps in each patient (Fig. 5). Regions displaying electrical activity with high similarity (S ≥0.5) and high rate of activation (CL < CL of the surrounding tissue) were selected and marked as rapid and repetitive (RR) sources. Only the 33% of the R sources (16 of 49) showed a high activation rate and were thus classified as potential drivers (S = 0.65 ± 0.15; CL = 140.7 ± 25.1 ms, range 104 to 188 ms). As detailed in Table 3, RR sources were present in 11 of 17 patients (65%) and were mainly confined to the LA (12 of 16 sources) at the level of PVs (10 sources) and LA appendage (2 sources) (see example in Fig. 5), whereas they were rarely found in the RA. Focusing on the 4 PVs, the highest incidence of drivers was found at the left superior PV (4 of 10), whereas the right inferior showed the lowest incidence (1 of 10). PV drivers were mostly located at the proximal portion of the veins rather than at the distal area (6 vs. 4). Interestingly (Fig. 6), although the majority of patients (82%) showed R sources in the PVs, R sources displayed the fastest rate in just half of the patients (47%). More importantly, the SCV, which presented highly repetitive activity in 7 of 17 patients (41%), displayed a rapid rate just in 1 patient.
Localization of AF critical sources by anatomically detailed WS/CL maps
This paper introduces a novel strategy for the identification and localization of potential AF drivers, based on the construction of anatomically detailed WS/CL maps. Advanced signal analysis and image processing techniques were combined to extract and quantify the principal electrophysiological features of mapping electrograms, and integrate the electrophysiological information with the highly detailed atrial anatomy provided by CT images.
Previous experimental evidence has pointed out AF drivers to be characterized by 2 principal features, which are a fast activation rate and a high level of electrical organization (12,13). A novel time-domain approach was used to translate this electrophysiological definition into an operative definition of driver, based on quantitative signal properties. Specifically, critical sites were identified by a double-criteria evaluation, which combined a measure of WS with a measure of CL. As shown in a preliminary methodological study (5), the combined evaluation of the 2 indexes conferred to the method a high sensitivity to distinguish potential driver sites from substrate or passive activation regions.
Following identification, AF drivers were localized in a highly detailed anatomic context by integrating the sparse WS/CL maps with CT-derived anatomies via custom-made image processing algorithms (10,11). Integration of multimodal data was aimed to overcome the coarse anatomic detail of the reconstructed map provided by electroanatomic mapping systems. The LA anatomy is indeed complex, and high inter- and intrapatient variability is observed in the number, size, and bifurcation of the PVs (14). Complex anatomic structures cannot be reconstructed from raw mapping data, whereas an accurate description of these aspects can be derived from CT images (15). In the context of AF ablation, integration of electroanatomic data with tomographic images is performed to facilitate the procedure and has been shown to improve ablation outcome (15). In the present paper, the construction of CT-integrated WS/CL maps allowed us to precisely locate the position of AF critical sites, providing a detailed picture of their anatomic distribution in the LA and inside the complex branching structure of the PVs. The accuracy in driver localization was obtained by innovative custom-made algorithms of proved efficiency (10,11), which allowed both a highly resolved reconstruction of the atrial chambers from CT data (11) and an accurate registration and fusion of multimodal data (10). In particular, our landmark-free, fully automated stochastic approach enabled the registration of mapping points on CT reconstructions with accuracy suitable for the detection of regional differences in the WS/CL indexes.
Anatomic distribution of potential drivers in persistent AF
The combined use of CT-integrated WS/CL maps allowed us to demonstrate the presence of clear regional differences in the distribution of potential drivers in patients with persistent AF. Consistent with experimental studies (12,13), high-frequency R sources were not uniformly distributed, but were prevalently confined to the LA and, specifically, to the PV and LA appendage regions. The spatial location of these potential drivers is also in agreement with high-density epicardial mapping studies performed in patients with chronic AF during open heart surgery (16–21). High-frequency repetitive patterns emerging from the posterior LA at or near the PVs, associated with foci or re-entrant circuits, were reported by several authors (16,18,19,21). In a few of these studies, targeted LA ablation at these sites was successfully applied, suggesting the important role of repetitive activation in the perpetuation of chronic AF (16–18).
Our results showed the confinement of potential driver sources in the LA, whereas organized sources in the RA were rarely associated with high-frequency activation. The incidence of AF drivers observed in the PV region (47% of patients) is consistent with the modest outcome of PV ablation in persistent/long-standing persistent AF with respect to paroxysmal AF (1,22). Similarly, the paucity of drivers observed in the RA is in agreement with the secondary role of RA ablation and/or SCV isolation procedures, which were shown to be adjuvant only in a limited subgroup of patients (23).
In this study, we identified other atrial regions with highly repetitive electrical activity but with slow activation rates. A significant example is represented by the SCV. The combination of a high level of organization and a slow activation may indicate the presence of a passively activated region or a secondary source of activation.
Finally, the construction of WS maps provided indications on the spatial distribution of complex fractionated electrograms, identified by low-similarity values. Complex electrogram areas were localized on the posterior wall, roof, and septum of the LA, and on the posterior wall and septum of the RA, which is in agreement with the distribution of fragmented electrograms firstly described by Nademanee et al. (4) and confirmed by several subsequent studies.
The optimal gold standard for the identification of AF critical sources would be targeted ablation. However, due to the retrospective nature of our study, the effects of targeted ablation on the identified driver sites could not be determined. Nonetheless, our work showed the possibility of objectively identifying and precisely localizing regions of rapid and repetitive activity, which have been previously demonstrated to maintain AF. Also, regions with RR sources identified by our approach coincided with potential driver regions identified by other ablation studies (16–18). Our study may thus represent the base for planning future targeted ablations proving the actual role of these areas in AF maintenance.
Other limitations consisted of the evaluation of WS/CL temporal stability on the sole continuous CS recordings and in the off-line analysis of electrograms and images. However, it is worth noticing that the spatiotemporal stability of the 2 indexes during AF has been demonstrated in previous studies (8,9), and the low computational cost of the algorithms is consistent with future online implementation.
The identification and localization of RR sources in AF patients is feasible by fusion of biatrial anatomic images with WS/CL maps. R sources of rapid impulse are present in a subset of patients with persistent AF and are mainly localized in the region of the PVs and LA appendage. The combined use of anatomically integrated WS/CL maps, which allows the detection of 2 critical features of AF sources, may constitute a novel mapping modality for the electrogram-guided ablation of AF.
The authors would like to thank Lucia Magagna for invaluable technical assistance during AF recordings.
This work was supported by a grant from the Fondazione Cassa di Risparmio di Trento e Rovereto. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- cycle length
- coronary sinus
- computed tomography
- left atrium
- pulmonary vein
- right atrium
- R sources
- repetitive sources
- RR sources
- rapid and repetitive sources
- similarity index
- superior caval vein
- wave similarity
- Received March 9, 2012.
- Revision received July 12, 2012.
- Accepted August 1, 2012.
- American College of Cardiology Foundation
- Sanders P.,
- Berenfeld O.,
- Hocini M.,
- et al.
- Masse S.,
- Downar E.,
- Chauhan V.,
- Sevaptsidis E.,
- Nanthakumar K.
- Mandapati R.,
- Skanes A.,
- Chen J.,
- Berenfeld O.,
- Jalife J.
- Kalifa J.,
- Tanaka K.,
- Zaitsev A.V.,
- et al.
- Ho S.Y.,
- Cabrera J.A.,
- Tran V.H.,
- Farre J.,
- Anderson R.H.,
- Sanchez-Quintana D.
- Tops L.F.,
- Schalij M.J.,
- den Uijl D.W.,
- Abraham T.P.,
- Calkins H.,
- Bax J.J.
- Sahadevan J.,
- Ryu K.,
- Peltz L.,
- et al.
- Oral H.,
- Chugh A.,
- Good E.,
- et al.