BASIC SCIENCE
Europace (2014) 16, 766–773 doi:10.1093/europace/euu003
Influence of atrial substrate on local capture induced by rapid pacing of atrial fibrillation Alexandru Rusu 1, Vincent Jacquemet 2, Jean-Marc Vesin 1, and Nathalie Virag 3* 1 Applied Signal Processing Group, Swiss Federal Institute of Technology, CH-1015 Lausanne, Switzerland; 2Department of Physiology, University of Montreal and Centre de Recherche, Hoˆpital du Sacre´-Coeur, Montre´al, (QC) H4J 1C5, Canada; and 3Medtronic Europe, CH-1131 Tolochenaz, Switzerland
Received 15 July 2013; accepted after revision 3 January 2014
Aims
----------------------------------------------------------------------------------------------------------------------------------------------------------Keywords
Atrial fibrillation † Septal pacing † Computer modelling † Pacemakers
Introduction Atrial fibrillation (AF) is the most common form of sustained cardiac arrhythmia. Its polymorphic dynamical nature prevents the development of a single therapy effective in all individual patients.1 Several guidelines have been proposed for the management of AF, from rate control strategies to sinus rhythm restoration.2 Pharmacological therapy is an important part for any rhythm control strategy and is often the first-line approach. However, due to the side-effects of drugs, there has been a substantial development of nonpharmacological therapies in the last 15 years. Ablation techniques, generally performed percutaneously using a catheter, have proved successful in reducing the burden associated with AF. External electrical cardioversion has a very high initial AF termination rate, but few patients remain in sinus rhythm on the long term. Some of the recent pacemakers and defibrillators have incorporated different features to prevent or terminate atrial tachycardia via pacing.3,4 These pacing
algorithms deliver rapid pacing bursts at a cycle length shorter than that of the detected arrhythmia and then gradually decrease the pacing rate. As compared with electrical cardioversion, pacing has the advantage of being painless, safe, and energy efficient in implantable devices. However, in the case of a more complex arrhythmia such as AF, while the possibility of local atrial capture by rapid pacing has been demonstrated in animal or humans,5,6 no pacing therapy has proved so far to be effective in terminating AF and no clinical study has yet proven the long-term clinical benefit. Computer modelling has gained increased importance in the development of therapeutic strategies for AF, overcoming some of the limitations encountered in experimental and clinical research.7 Integrated functional computer modelling is considered today a potentially effective solution for the development of therapeutic options for AF.8 Following a computer modelling approach, previous simulations permitted a systematic study of pacing algorithms currently used in pacemakers and allowed the search for optimal
* Corresponding author. Tel: +41 21 802 73 35; Fax: +41 21 802 79 29. Email: [email protected] Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2014. For permissions please email: [email protected].
Downloaded from by guest on November 16, 2015
Preliminary studies showed that the septum area was the only location allowing local capture of both the atria during rapid pacing of atrial fibrillation (AF) from a single site. The present model-based study investigated the influence of atrial substrate on the ability to capture AF when pacing the septum. ..................................................................................................................................................................................... Methods Three biophysical models of AF with an identical anatomy from human atria but with different AF substrates were used: (i) AF based on multiple wavelets, (ii) AF based on heterogeneities in vagal activation, (iii) AF based on heterogeneities in and results repolarization. A fourth anatomical model without Bachmann’s bundle (BB) was also implemented. Rapid pacing was applied from the septum at pacing cycle lengths in the range of 50 –100% of AF cycle length. Local capture was automatically assessed with 24 pairs of electrodes evenly distributed on the atrial surface. The results were averaged over 16 AF simulations. In the homogeneous substrate, AF capture could reach 80% of the atrial surface. Heterogeneities degraded the ability to capture during AF. In the vagal substrate, the capture tended to be more regular and the degradation of the capture was not directly related to the spatial extent of the heterogeneities. In the third substrate, heterogeneities induced wave anchorings and wavebreaks even in areas close to the pacing site, with a more dramatic effect on AF capture. Finally, BB did not significantly affect the ability to capture. ..................................................................................................................................................................................... Conclusion Atrial fibrillation substrate had a significant effect on rapid pacing outcomes. The response to therapeutic pacing may therefore be specific to each patient.
767
Influence of atrial substrate on local capture
What’s new? † While the polymorphic nature of atrial fibrillation (AF) is difficult to study in biological experiments, the model-based approach proposed in this paper offers a new method to study in detail the influence of atrial substrate during rapid pacing of AF and to see how much the inter-patient variability affects the efficacy of therapeutic pacing. † Changes in atrial substrate have a significant effect on rapid pacing outcomes. Furthermore, this influence is different in the presence of the heterogeneities in vagal activation compared with the heterogeneities in repolarization. This suggests that AF pacing therapies should be specific to each patient’s atrial substrate. † Even if the Bachmann’s bundle plays a fundamental role on electrical propagation in the atria by providing a fast interatrial conduction pathway, our simulations suggest that it does not significantly influence the ability to capture during rapid pacing of AF.
Modelling different atrial substrates
Modelling atrial geometry
A model of cell electrophysiology based on membrane ion kinetics was assigned to each atrial unit of the mesh. Two different cellular models were implemented: (i) a variant of the Luo– Rudy electrophysiological model adapted to atrial cellular properties (LR model),18 (ii) the Courtemanche – Ramirez – Nattel human atrial cell model (CRN model).20 The CRN model has become popular for the simulation of arrhythmias because quantitative data are available to include the effect of regional heterogeneities, AF-induced electrical remodelling, or vagal stimulation through acetylcholine. The LR model was used as a generic multiplewavelet model of AF substrate to examine the relevance and the limitations of former computational studies9,10 and to enable a systematic comparison with more recent models. While using the same atrial geometry of Figure 1A, different types of sustained AF dynamics were obtained by modifying either the electrophysiology (three models) or the anatomy (one model), as summarized in Figure 2. The first three models correspond to different postulated AF mechanisms (multiple wavelets, mother rotor, heterogeneous remodelling). Electrophysiological changes were simulated by modifying the local membrane kinetics and therefore creating different types of atrial substrates or AF dynamics. These substrates were implemented by following the same method as in the previous studies18,21,22 but using the new geometry of Figure 1A and the zones of the heterogeneities depicted in Figure 1B. In this paper, we compared the following AF substrates (Figure 2):
The geometry of the biophysical model of the human atria was reconstructed from computed tomography (CT) scans of a patient in AF
† AF Substrate #1: Multiple-wavelet AF. This substrate was based on a homogeneous and isotropic tissue in which the LR model was adjusted
Methods
Downloaded from by guest on November 16, 2015
pacing sites and pacing cycle lengths (PCLs) leading to a local capture of AF.9 As a result, a better ability to sustain capture was found in the right atrial free wall, the left atrial appendage, and the pulmonary veins, where the wavefronts induced by pacing encompassed the major part of the paced atrium. This capture was accompanied by residual reentrant waves outside the area of capture such that AF re-initiation was generally observed after stopping the pacing protocol, with no pace termination of AF. This study suggested that rapid pacing of AF could indeed induce a local capture of the atrial tissue and that the septum area was the only location allowing the capture of both the atria from a single site. Based on these observations, a novel pacing scheme based on pacing from the septum area was developed and showed promising results for AF termination using a model-based approach.10 Atrial fibrillation can have different clinical forms corresponding to different patient-specific atrial substrates. While the polymorphic nature of AF is difficult to study in biological experiments, computer modelling allows a systematic study of the influence of atrial substrate. To get some insights into how much inter-patient variability may affect the efficacy of septum pacing, we used a computer model to assess the influence of heterogeneities in atrial substrate on the ability to capture AF with rapid pacing from the septum area. We have chosen to consider in this paper only substrates with electrophysiological remodelling and no structural remodelling such as fibrosis or slow conduction areas. This would correspond to applying the pacing therapy at an early stage in the progression of AF. We hypothesized that the pacing therapy would be more likely to be effective under these conditions.
before an ablation procedure. Since the patient was in chronic AF, the atria were slightly dilated. A three-dimensional (3D) cubic mesh representing the atrial myocardium and the atrial fast conducting system was created (1 040 000 nodes, 0.33 mm resolution) as published previously.7,11 In our anatomically simplified geometry, two interatrial connections were represented among the several connections existing between the right and the left atria:12 the Bachmann’s bundle (BB) and the rim of the septum. Bachmann’s bundle could only be partially segmented, therefore its structure was extrapolated to replicate the other anatomical models.13 – 15 The septum area was represented as a normally conducting tissue connecting both the atria through the septal rim, with a hole in the middle representing the fossa ovalis. The resulting geometry is shown in Figure 1A, with BB and the septum area highlighted in red. Rule-based fibre orientation was included, based on the previous works.7,13 – 17 Since investigating the different pacing protocols presented in this paper involved a total of 1600 simulations of 30 s, computational load was a major issue that was addressed here by further simplifying the atrial geometry: a coarser surface model (triangular mesh with 117 000 nodes, 0.45 mm resolution similar to previous models14,18) was derived from our previously published 3D anatomical model.7 Fibre orientation of the 3D model was projected onto the atrial surface. This simplified anatomical surface model matches our previous 3D model to facilitate the comparison. As a result, the number of nodes (and thus computational time) was reduced by a factor of 9. The full 3D model was used in a subset of the simulations for the validation of the surface model. Electrical propagation in the atrial tissue was simulated using the monodomain equation solved using operator splitting, explicit time integration, and spatial discretization based on a finite volume method19 in the surface model and finite difference method in the cubic 3D model. Time step was adaptive with 20 ms for diffusion, 12.5 – 100 ms for reaction, and 100 ms as operator splitting time step.18
768
A. Rusu et al.
A Left atrium
Right atrium Bachmann’s bundle
Right atrium
Pulmonary veins Superior vena cava
Zones of heterogeneities
B
10%
50%
20%
60%
30%
70%
40%
80%
Mitral valve Coronary artery
Tricuspid valve
Inferior vena cava
Septum area (Left atrium)
Figure 1 (A) Anterior and posterior view of the atrial geometry obtained from CT. The Bachmann’s bundle is highlighted in red. The bottom
Multiple wavelets
Anisotropy
Heterogeneities
Cellular model
1:1
Homogeneous
LR for AF
Action potentials
[ms] 0
Cholinergic AF
3:1
Heterogeneities in acetylcholine
CRN for AF
+ 0
Heterogeneous AF
4:1
Heterogeneities in repolarization
1:1
Homogeneous BB removed
100 200 300 0
[ms] 100 200 300
+
CRN for AF 0
AF without BB
100 200 300
100 200 300 0
[ms] 100 200 300
LR for AF [ms] 0
100 200 300
Figure 2 Summary of the different AF models corresponding to electrophysiological changes (three substrate models in blue) or anatomical changes (one anatomical model in red). For the second and the third models, baseline action potentials are represented on the left and the action potential inside the zones of the heterogeneities are on the right.
Downloaded from by guest on November 16, 2015
images show the left atrium separate from the right atrium with the septum area highlighted in red. (B) Introduction of random patchy heterogeneities in the atrial substrate.
769
Influence of atrial substrate on local capture
to mimic electrical remodelling as observed in patients suffering from permanent AF, such as the shortening of the action potential duration (APD). This was simulated by setting the channel conductance GNa, GK, and Gsi to 16, 0.423, and 0.55 mS/cm2, respectively.18 The resulting AF dynamics revealed multiple wavelets continuously changing in size and duration due to functional or anatomical reentries. † AF Substrate #2: Cholinergic AF. This substrate was based on an atrial tissue with a 3 : 1 anisotropy and the CRN model in which ICaL was reduced by 70% and Ito and IKur by 50%. An acetylcholine-dependent K+ current IKACh was added.23 Spatial variations in acetylcholine concentrations were introduced: IKACh was globally increased by 250% while it was decreased by 30% inside the red areas shown in Figure 1B. The resulting AF dynamics was maintained by the stable rotors. † AF Substrate #3: Heterogeneous AF or AF due to repolarization heterogeneities. This substrate was based on an atrial tissue with a 4 : 1 anisotropy and the CRN model in which the current ICaL was reduced 30%, Ito by 80%, IKur by 90%, and IKr increased by 50%.22 In addition, patchy heterogeneities were introduced as shown in Figure 1B, where ICaL was decreased by 50% and Ito and IKr were increased by 200% relative to the original model. Wavebreaks were created by the repolarization gradients and this contributed to the maintenance of AF.
Modelling rapid pacing of atrial fibrillation Atrial fibrillation was initiated in the biophysical models in a similar way to clinical experiments, using a programmed stimulation protocol or a burst pacing protocol.18,21,23 To avoid any timing adjustment of the stimulation protocol, burst pacing at 10 Hz was applied from one to five selected locations on the atrial surface until AF was observed (from 0.5 to 5 s). To perform computer simulations of AF therapies equivalent to animal or clinical experiments, 16 time instants were selected at 30 s intervals during one simulation of sustained AF and the corresponding tissue states (including membrane potential maps and all other state variables) were stored. This selection was repeated separately for each of the four AF models (three AF substrates and one anatomical model). These tissue states correspond to different, independent conditions of fibrillatory activity in the atrial tissue. They served as initial conditions for the repetitive application of the pacing therapy. Simulations results were averaged over 16 AF simulations for each model, thereby allowing statistical comparison between the different models. Rapid pacing was applied by injecting an intracellular current of 80 mA/ cm2 for the LR model and 100 mA/cm2 for the CRN model with 1 ms pacing pulses inside the cells located in the septum area highlighted in red in Figure 1A. Each pacing sequence was applied during 30 s at PCLs in the range of 50 – 100% of the AF cycle length (AFCL). Pacing cycle length was increased with steps of 2% AFCL. Therefore, in total, 25 simulations of 30 s of pacing have been performed for each of the 16 AF initial conditions. Atrial fibrillation cycle length was computed as a mean of the cycle length values measured on the transmembrane potentials recorded at the 24 electrode pairs shown in Figure 3, and averaged over 5 s. In this
Conditions for capture at each pair 1. Propagation direction from the interatrial junction towards the appendages 2. CL = PCL + 2%
Figure 3 Automatic assessment of atrial capture during pacing. A set of 24 pairs of electrodes was evenly distributed on the atrial surface to assess the direction of wavefront propagation and the CL compared with the PCL.
Downloaded from by guest on November 16, 2015
The patchy patterns of heterogeneities were created by using a region growth algorithm21,22 and were intended to reproduce the type of patterns (repolarization gradients) observed in the experimental models of AF.24 They were generated using a random Markov field approach (described in our previous works21,25,26), which consists in filtering a white-noise field with a Gaussian spatial filter (with a length scale of 7.5 mm) and then thresholding the resulting filtered field to control the area of altered tissue that varied here between 0 (homogeneous) and 80%. The fourth model is an anatomical model that was created based on the atrial geometry of Figure 1A using the AF Substrate #1 (with no heterogeneities) with BB removed (Figure 2). The objective was to study the impact
of BB, which is a major pathway for interatrial conduction, on the ability to capture AF during rapid pacing from the septum. Finally, to get an idea about the influence of atrial thickness on capture, we performed one set of simulations using the 3D version of our model7 based on the AF Substrate #1.
770
A. Rusu et al.
study, we did not take into account the additional activity originating from the sinoatrial node during or after pacing.
Capture with rapid pacing during multiple-wavelet atrial fibrillation
Automatic assessment of atrial fibrillation capture Capture was defined as the ability of the pacing burst to take control over a given area away from the pacing site. It relates to the local capture around the pacing site, not to the global capture of both the atria. During rapid pacing, the transmembrane potentials of 24 pairs of electrodes evenly distributed on the atrial surface (Figure 3) were analysed to determine the percentage of captured atrial tissue. At each pair of electrodes, the cycle length and the direction of propagating waves were assessed: we considered that a pair was captured if the direction of propagation was from the interatrial junction to the appendages and the cycle length was equal to the PCL + 2%. For each pacing sequence, a period of 30 s was analysed and the percentage of capture atrial tissue was computed as the ratio of the pairs where capture was observed divided by the total number of pairs. The capture window was defined as the range of PCL for which capture was .50%.
In the multiple-wavelet AF with no heterogeneities (AF Substrate #1), the maximum percentage of captured tissue was 78.9% for a PCL of 89% of AFCL (Figure 4) and the capture window was 85 – 95% AFCL. Atrial fibrillation cycle length was 74 ms. This means that for optimal parameters, AF capture could be obtained on almost 80% of the atrial surface in both the atria (highlighted in red in Figure 4), with residual reentrant wavelets present in the extremities.
Influence of heterogeneities in atrial substrate Figure 4 presents the capture results for the AF substrates including heterogeneities. Atrial fibrillation cycle length in the cholinergic model (AF Substrate #2) ranged from 69 ms (0% heterogeneities)
Cholinergic AF
Multiple wavelets 100 90 80 70 60 50 40 30 20 10 0
Heterogeneous AF
100
100
90
90
80
0% 10% 20% 30% 40% 50% 60% 70% 80%
70 60 50 40 30 20 10 50
60
70 80 90 PCL [%AFCL]
100
0 50
Downloaded from by guest on November 16, 2015
Capture [%]
Results
80 70 60 50 40 30 20 10
60
70 80 90 PCL [%AFCL]
100
0 50
60
70 80 90 PCL [%AFCL]
PCL: 91% AFCL
PCL: 79% AFCL 30% Heterogeneities
PCL: 67% AFCL 30% Heterogeneities
AF
AF
AF –80 mV
Capture of AF
Capture of AF
100
20 mV
Capture of AF
Figure 4 Capture as a function of the PCL for the different AF substrates. For each PCL, the capture percentage was computed as an average on 16 simulations. For the cholinergic and heterogeneous AF, the percentage of patchy heterogeneities corresponds to those presented in Figure 1. The middle graph represent the spatial extend (in red) of the best capture results obtained for each model. The two bottom graphs show example of transmembrane potential maps during AF and during capture with rapid pacing of AF for each AF substrate.
771
Influence of atrial substrate on local capture
A
100
Capture [%]
80 60
Without BB
40
With BB
20 0 60
90 70 80 PCL [%AFCL]
B Activation maps during
100
rapid pacing with BB
Activation maps during rapid pacing without BB
0 ms
0 ms
15 ms
15 ms
30 ms
30 ms
45 ms
45 ms
Influence of Bachmann’s bundle The AFCL in the anatomical model of AF without BB was 72 ms, which is comparable with the value obtained in the model with BB (AFCL ¼ 74 ms). Figure 5 shows the capture window obtained for both the models: 83 –100% AFCL with BB and 81– 100% AFCL without BB. The removal of BB did not significantly improve the capture window. The activation maps of the simulations presented in Figure 5 however showed that the BB induced multiple wavebreaks close to the pacing site. This did not however affect the fibrillatory activity away from the pacing site.
Influence of atrial thickness Figure 6 presents a comparison of AF capture results in the surface model and the corresponding 3D model in the case of multiple-wavelet AF with no heterogeneities. The overall shape of the curve of capture as a function of PCL was similar in both the cases. However, the capture values were slightly decreased in the 3D model (the maximum capture percentage was decreased from 78.9 to 63.9% of the atrial surface). Due to the heavy computational load, simulations in the 3D model could only be performed in one AF substrate.
Discussion The model-based results obtained in this study suggest that the heterogeneities in atrial substrate greatly influence the ability to capture AF with rapid pacing from the septum area. This behaviour was expected since we already observed that the heterogeneities in atrial substrate play a role in the perpetuation of AF by promoting
–80 mV 20 mV PCL = 91% AFCL
Figure 5 (A) Capture as a function of the PCL with and without the BB, using AF Substrate #1. For each PCL, the capture percentage was computed as an average on 16 simulations. (B) Example of transmembrane potential maps during rapid pacing of AF at a PCL of 91% of AFCL with and without BB.
AF reentries.21,22 However, this influence was different in the presence of heterogeneities in vagal activation or heterogeneities in repolarization. In the first case, the capture patterns tended to be more regular and they were fewer anchored wavelets than in other AF models. We could not establish a direct link between the percentage of heterogeneities and their anatomical location and the capture results. In some cases, vagal activation even facilitated AF capture by rapid pacing compared with multiple-wavelet AF. In the second case, the heterogeneities in repolarization had a more dramatic impact on AF capture, with the creation of wave anchoring and wavebreaks. Note that AFCL in the first model (69 –103 ms) was much lower (AF faster) than in the second model (180 –252 ms), as in the previous works.22
Downloaded from by guest on November 16, 2015
to 103 ms (80% heterogeneities). The cholinergic model with no heterogeneities had therefore an AFCL comparable with the multiplewavelet AF. The capture window was also similar (89–100% AFCL), except that for lower values of PCL where capture in about 10% of the atrial tissue could be observed (no capture was observed in the multiple-wavelet model for these values). In the cholinergic model, the capture patterns tended to be more regular than in the multiplewavelet model (Figure 4), BB did not affect the pacing-induced wavefronts and there were fewer anchored wavelets. We could not observe a direct link between the increase of the percentage of heterogeneities in vagal activation and the ability to capture AF. For example, the best capture results were obtained in the cholinergic model with 30% of heterogeneities, with a very broad capture window (70–100% AFCL). The worst capture results were obtained with 50% of heterogeneities. We were not able to link the capture results to the anatomical location of heterogeneities. The AFCL in the AF Substrate #3 ranged from 180 ms (0% heterogeneities) to 252 ms (80% heterogeneities). These values are higher than for the first two substrates. The capture window for the substrate with no heterogeneities was the broadest so far (70–100% AFCL). However, the introduction of patchy heterogeneities in repolarization progressively degraded the capture results. In this AF substrate, there was a clearer link between the percentage of heterogeneities and the degradation of capture. These heterogeneities significantly influenced the capture patterns by creating wave anchoring and wavebreaks, even in areas close to the pacing site (Figure 4). Even if AF was slower, the spatial extend of the capture was greatly reduced compared with the first two substrates (as shown in red in Figure 4).
772
A. Rusu et al.
100 90
Capture [%]
80 70 60
Surface model 3D model
50 40 30 20 10 0 50
60
70 80 90 PCL [% AFCL]
100
Figure 6 Capture as a function of the PCL for the surface and for the 3D model, using AF Substrate #1.
Conflict of interest: N.V. is a full time employee of Medtronic Europe (Tolochenaz, Switzerland).
Downloaded from by guest on November 16, 2015
Our previous studies on rapid pacing during AF were based only the first substrate presented in this paper, namely the multiplewavelet AF model using the LR model in a homogeneous tissue.9,10 This previous work was based on a simplified atrial geometry that did not include BB or any inhomogeneities. We had studied the influence of atrial tissue dynamics on the ability to capture AF with rapid pacing by changing the APD and the conduction velocity (CV) on the entire atrial surface. The capture window obtained for the baseline model (with no changes in APD and CV) was comparable with what we obtained in this paper for the multiple-wavelet model (LR model) and for the cholinergic AF model (CRN model) with 0% heterogeneities (Figure 4). It is also comparable with the value found in humans by Capucci et al.,27 who published a capture window of 85 –105% of AFCL. In the previous modelling study, decreasing CV or APD globally in the atrial tissue led to a significant increase of the capture window and vice versa,9 this effect being more dramatic for the changes in APD. In the present study, we followed a different approach for modifying the atrial substrate since the changes were introduced as a patchy pattern of a heterogeneity of gradually increasing density. As a result, we observed in the model of cholinergic AF that an increase of APD in patchy patterns could either decrease (10% of heterogeneities) or increase (30% of heterogeneities) the capture window (Figure 4). Similar observations could be made for the model of heterogeneous AF. Therefore, when APD is changed in a patchy pattern rather than globally in the whole atrial tissue, the capture outcome is difficult to predict. Our previous studies also showed that AFCL is not predictive of the ability to capture AF by rapid pacing suggesting that a measure proportional to wavelength (estimated by the product of CV and APD) could give an estimation of the capture abilities in a specific patient.9 The present work also confirms the non-predictive value of AFCL. For example, in the model of cholinergic AF, increasing the percentage of the patchy heterogeneities from 0 to 10% did not significantly change AFCL while it significantly affected the capture window (Figure 4). This was also observed in clinical data in which subjects with the same AFCL had different capture results.27 However, these results further suggest that an overall measure of wavelength is probably not sufficient to assess the ability to capture
during AF and that a more detailed assessment of each patient’s atrial substrate would be needed. We investigated the role of BB during pacing since it is considered to play a fundamental role in interatrial conduction. When observing the activation maps during capture in the multiple-wavelet AF, we could observe many wavebreaks initiated due to conduction through BB that significantly affected the capture patterns. However, our study showed that even if BB greatly influenced locally the electrical propagation by providing a fast accessory interatrial conduction pathway, it did not significantly affect the global ability to capture during rapid pacing of AF from the septum area. This may be related to the fact that the septal pacing site was relatively close to both sides of BB. Significant progresses have been made in the last 10 years in the modelling of human atria and today most existing models include a detailed atrial geometry based on human imaging with fibre orientation and anisotropy, detailed cellular models of atrial electrophysiology as well as pathological components such as fibrotic tissue or atrial remodelling.12 Our objective was to focus on how pacing could be developed as a clinical therapy for AF. Compared with the existing computer models of human atria, the model presented here represents a tradeoff between complexity and the ability to simulate long runs of AF (in total more than 13 h of AF have been simulated for the results presented in this paper). Therefore, the work we presented here is based on human data, but to keep the computational load within a reasonable range we had to simplify the AF models into surface models. However, we performed a set of simulations in a 3D model to assess the effect of atrial thickness, which was a reduction of about 15% of the atrial surface that could be captured, while the pacing frequencies leading to capture remained comparable. It should be noted that we studied here different atrial substrates without the inclusion of external triggers such as rapid foci originating from the pulmonary veins. We expect that the conditions we have simulated correspond to a best-case scenario in which the pacing therapy is most likely to be effective. Additional pathways for (micro-) reentry created by fibrosis or fibre dissociation within the atrial wall may hinder the ability of pacing therapy to achieve local capture.28 Multilayer 3D models with a finer discretization could be used to test the hypothesis. In summary, changes in the atrial substrate have a significant effect on the rapid pacing outcomes, implying that patients may respond in very different manners to the rapid pacing of AF. This might be one explanation to the fact that so far no clinical study on pacing for AF termination has been conclusive:3,4 the characteristics of an individual undergoing a specific therapy is in general different from the mean of the population, therefore the therapy may not benefit every patient. This suggests that a successful pacing therapy for AF should be specific to each patient’s atrial substrate. While this polymorphic nature of AF is difficult to handle in clinical experiments, a modelbased approach offers a new method to study in detail the influence of atrial substrate. Personalized computational cardiac models are emerging today as an important tool for studying cardiac arrhythmia mechanisms, and have the potential to become powerful instruments for guiding clinical anti-arrhythmia therapy.29
Influence of atrial substrate on local capture
Funding This work was supported by the Theo-Rossi di Montelera Foundation (Lausanne, Switzerland) and Medtronic Europe (Tolochenaz, Switzerland).
References
15. Vigmond EJ, Ruckdeschel R, Trayanova N. Reentry in a morphologically realistic atrial model. J Cardiovasc Electrophysiol 2001;12:1046 –54. 16. Krueger M, Schmidt V, Tobo´n C, Weber FM, Lorenz C, Keller DUJ et al. Modeling atrial fiber orientation in patient-specific geometries: a semi-automatic rule-based approach. In Metaxas DM, Axel L (eds). Functional Imaging and Modeling of the Heart. Springer-Verlag Berlin Heidelberg; 2011. 223 –32. 17. McDowell KS, Vadakkumpadan F, Blake R, Blauer J, Plank G, Macleod RS et al. Mechanistic inquiry into the role of tissue remodeling in fibrotic lesions in human atrial fibrillation. Biophys J 2013;104:2764 – 73. 18. Virag N, Jacquemet V, Henriquez CS, Zozor S, Blanc O, Vesin J-M et al. Study of atrial arrhythmias in a computer model based on magnetic resonance images of human atria. Chaos 2002;12:754–63. 19. Zozor S, Blanc O, Jacquemet V, Virag N, Vesin J-M, Pruvot E et al. A numerical scheme for modeling wavefront propagation on a monolayer of arbitrary geometry. IEEE Trans Biomed Eng 2003;50:412 –20. 20. Courtemanche M, Ramirez RJ, Nattel S. Ionic mechanisms underlying human atrial action potential properties: insights from a mathematical model. Am J Physiol 1998; 275:H301–21. 21. Jacquemet V, Virag N, Ihara Z, Dang L, Blanc O, Zozor S et al. Study of unipolar electrogram morphology in a computer model of atrial fibrillation. J Cardiovasc Electrophysiol 2003;14:172. 22. Jacquemet V, Virag N, Kappenberger L. Wavelength and vulnerability to atrial fibrillation: insights from a computer model of human atria. Europace 2005;7(Suppl 2): 83 – 92. 23. Kneller J, Zou R, Vigmond EJ, Wang Z, Leon LJ, Nattel S. Cholinergic atrial fibrillation in a computer model of a two-dimensional sheet of canine atrial cells with realistic ionic properties. Circ Res 2002;90:73. 24. Fareh S, Villemaire C, Nattel S. Importance of refractoriness heterogeneity in the enhanced vulnerability to atrial fibrillation induction caused by tachycardia-induced atrial electrical remodeling. Circulation 1998;98:2202 –9. 25. Jacquemet V, Henriquez CS. Loading effect of fibroblast-myocyte coupling on resting potential, impulse propagation and repolarization: insights from a microstructure model. Am J Physiol Heart Circ Physiol 2008;294:H2040 –52. 26. Jacquemet V. A biophysical model of atrial fibrillation and electrograms: formulation, validation and applications. Ph.D. Thesis. E´cole Polytechnique Fe´de´rale de Lausanne (EPFL), Suisse, 2004. 27. Capucci A, Ravelli F, Nollo G, Montenero AS, Biffi M, Villani GQ. Capture window in human atrial fibrillation: evidence of excitable gap. J Cardiovasc Electrophysiol 1999;10: 319 –27. 28. de Groot NMS, Houben RPM, Smeets JL, Boersma E, Schotten U, Schalij MJ et al. Electropathological substrate of longstanding persistent atrial fibrillation in patients with structural heart disease: epicardial breakthrough. Circulation 2010;122: 1674 –82. 29. McDowell KS, Vadakkumpadan F, Blake R, Blauer J, Plank G, MacLeod RS et al. Methodology for patient-specific modeling of atrial fibrosis as a substrate for atrial fibrillation. J Electrocardiol 2012;45:640 – 5.
Downloaded from by guest on November 16, 2015
1. Nattel S. New ideas about atrial fibrillation 50 years on. Nature 2002;415:219 – 26. 2. European Heart Rhythm Association, European Association for Cardio-Thoracic Surgery, Camm AJ, Kirchhof P, Lip GY, Schotten U, Savelieva I, Ernst S et al. Guidelines for the management of atrial fibrillation: the task force for the management of atrial fibrillation of the European Society of Cardiology (ESC). Eur Heart J 2010;31: 2369– 429. 3. Redfearn DP, Yee R. Pacing delivered rate and rhythm control for atrial fibrillation. Curr Opin Cardiol 2006;21:83–7. 4. Ellenbogen KA. Pacing therapy for prevention of atrial fibrillation. Heart Rhythm 2007;4:S84 –7. 5. Allessie M, Kirchhof C, Scheffer GJ, Chorro F, Brugada J. Regional control of atrial fibrillation by rapid pacing in conscious dogs. Circulation 1991;84:1689 –97. 6. Pandozi C, Bianconi L, Villani M, Castro A, Altamura G, Toscano S et al. Local capture by atrial pacing in spontaneous chronic atrial fibrillation. Circulation 1997;95: 2416– 22. 7. Virag N, Jacquemet V, Kappenberger L. Modeling of Atrial Fibrillation. In Shenasa M, Hindricks G, Borggrefe M, Breithardt G, Josephson ME (eds). Cardiac Mapping, 4th ed. Blackwell Publishing Ltd.; 2013.131–9. 8. Kirchhof P, Bax J, Blomstrom-Lundquiest C, Calkins H, Camm AJ, Cappato R et al. Early and comprehensive management of atrial fibrillation: proceedings from the 2nd AFNET/EHRA consensus conference on atrial fibrillation entitled ‘research perspective in atrial fibrillation’. Europace 2009;11:860 –85. 9. Uldry L, Virag N, Vesin J-M, Kappenberger L. Optimizing local capture of atrial fibrillation: a model based study of the influence of tissue dynamics on the outcome of rapid pacing. Ann Biomed Eng 2010;38:3664 – 73. 10. Uldry L, Virag N, Lindemans F, Vesin J-M, Kappenberger L. Atrial septal pacing for the termination of atrial fibrillation: study in a biophysical model of human atria. Europace 2012;14:v112 –20. 11. Herlin A, Jacquemet V. Eikonal-based initiation of fibrillatory activity in thin-walled cardiac propagation models. Chaos 2011;21:043136. 12. Do¨ssel O, Krueger MW, Weber FM, Wilhelms M, Seemann G. Computational modeling of the human atrial anatomy and electrophysiology. Med Biol Eng Comput 2012; 50:773 –99. 13. Harrild D, Henriquez C. A computer model of normal conduction in the human atria. Circ Res 2000;87:E25 –36. 14. Tobon C, Ruiz-Villa CA, Heidenreich E, Romero L, Hornero F, Saiz J. A threedimensional human atrial model with fiber orientation. Electrograms and arrhythmic activation patterns relationship. PLoS ONE 2013;8:e50883.
773
Europace (2014) 16, 766–773 doi:10.1093/europace/euu003
Influence of atrial substrate on local capture induced by rapid pacing of atrial fibrillation Alexandru Rusu 1, Vincent Jacquemet 2, Jean-Marc Vesin 1, and Nathalie Virag 3* 1 Applied Signal Processing Group, Swiss Federal Institute of Technology, CH-1015 Lausanne, Switzerland; 2Department of Physiology, University of Montreal and Centre de Recherche, Hoˆpital du Sacre´-Coeur, Montre´al, (QC) H4J 1C5, Canada; and 3Medtronic Europe, CH-1131 Tolochenaz, Switzerland
Received 15 July 2013; accepted after revision 3 January 2014
Aims
----------------------------------------------------------------------------------------------------------------------------------------------------------Keywords
Atrial fibrillation † Septal pacing † Computer modelling † Pacemakers
Introduction Atrial fibrillation (AF) is the most common form of sustained cardiac arrhythmia. Its polymorphic dynamical nature prevents the development of a single therapy effective in all individual patients.1 Several guidelines have been proposed for the management of AF, from rate control strategies to sinus rhythm restoration.2 Pharmacological therapy is an important part for any rhythm control strategy and is often the first-line approach. However, due to the side-effects of drugs, there has been a substantial development of nonpharmacological therapies in the last 15 years. Ablation techniques, generally performed percutaneously using a catheter, have proved successful in reducing the burden associated with AF. External electrical cardioversion has a very high initial AF termination rate, but few patients remain in sinus rhythm on the long term. Some of the recent pacemakers and defibrillators have incorporated different features to prevent or terminate atrial tachycardia via pacing.3,4 These pacing
algorithms deliver rapid pacing bursts at a cycle length shorter than that of the detected arrhythmia and then gradually decrease the pacing rate. As compared with electrical cardioversion, pacing has the advantage of being painless, safe, and energy efficient in implantable devices. However, in the case of a more complex arrhythmia such as AF, while the possibility of local atrial capture by rapid pacing has been demonstrated in animal or humans,5,6 no pacing therapy has proved so far to be effective in terminating AF and no clinical study has yet proven the long-term clinical benefit. Computer modelling has gained increased importance in the development of therapeutic strategies for AF, overcoming some of the limitations encountered in experimental and clinical research.7 Integrated functional computer modelling is considered today a potentially effective solution for the development of therapeutic options for AF.8 Following a computer modelling approach, previous simulations permitted a systematic study of pacing algorithms currently used in pacemakers and allowed the search for optimal
* Corresponding author. Tel: +41 21 802 73 35; Fax: +41 21 802 79 29. Email: [email protected] Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2014. For permissions please email: [email protected].
Downloaded from by guest on November 16, 2015
Preliminary studies showed that the septum area was the only location allowing local capture of both the atria during rapid pacing of atrial fibrillation (AF) from a single site. The present model-based study investigated the influence of atrial substrate on the ability to capture AF when pacing the septum. ..................................................................................................................................................................................... Methods Three biophysical models of AF with an identical anatomy from human atria but with different AF substrates were used: (i) AF based on multiple wavelets, (ii) AF based on heterogeneities in vagal activation, (iii) AF based on heterogeneities in and results repolarization. A fourth anatomical model without Bachmann’s bundle (BB) was also implemented. Rapid pacing was applied from the septum at pacing cycle lengths in the range of 50 –100% of AF cycle length. Local capture was automatically assessed with 24 pairs of electrodes evenly distributed on the atrial surface. The results were averaged over 16 AF simulations. In the homogeneous substrate, AF capture could reach 80% of the atrial surface. Heterogeneities degraded the ability to capture during AF. In the vagal substrate, the capture tended to be more regular and the degradation of the capture was not directly related to the spatial extent of the heterogeneities. In the third substrate, heterogeneities induced wave anchorings and wavebreaks even in areas close to the pacing site, with a more dramatic effect on AF capture. Finally, BB did not significantly affect the ability to capture. ..................................................................................................................................................................................... Conclusion Atrial fibrillation substrate had a significant effect on rapid pacing outcomes. The response to therapeutic pacing may therefore be specific to each patient.
767
Influence of atrial substrate on local capture
What’s new? † While the polymorphic nature of atrial fibrillation (AF) is difficult to study in biological experiments, the model-based approach proposed in this paper offers a new method to study in detail the influence of atrial substrate during rapid pacing of AF and to see how much the inter-patient variability affects the efficacy of therapeutic pacing. † Changes in atrial substrate have a significant effect on rapid pacing outcomes. Furthermore, this influence is different in the presence of the heterogeneities in vagal activation compared with the heterogeneities in repolarization. This suggests that AF pacing therapies should be specific to each patient’s atrial substrate. † Even if the Bachmann’s bundle plays a fundamental role on electrical propagation in the atria by providing a fast interatrial conduction pathway, our simulations suggest that it does not significantly influence the ability to capture during rapid pacing of AF.
Modelling different atrial substrates
Modelling atrial geometry
A model of cell electrophysiology based on membrane ion kinetics was assigned to each atrial unit of the mesh. Two different cellular models were implemented: (i) a variant of the Luo– Rudy electrophysiological model adapted to atrial cellular properties (LR model),18 (ii) the Courtemanche – Ramirez – Nattel human atrial cell model (CRN model).20 The CRN model has become popular for the simulation of arrhythmias because quantitative data are available to include the effect of regional heterogeneities, AF-induced electrical remodelling, or vagal stimulation through acetylcholine. The LR model was used as a generic multiplewavelet model of AF substrate to examine the relevance and the limitations of former computational studies9,10 and to enable a systematic comparison with more recent models. While using the same atrial geometry of Figure 1A, different types of sustained AF dynamics were obtained by modifying either the electrophysiology (three models) or the anatomy (one model), as summarized in Figure 2. The first three models correspond to different postulated AF mechanisms (multiple wavelets, mother rotor, heterogeneous remodelling). Electrophysiological changes were simulated by modifying the local membrane kinetics and therefore creating different types of atrial substrates or AF dynamics. These substrates were implemented by following the same method as in the previous studies18,21,22 but using the new geometry of Figure 1A and the zones of the heterogeneities depicted in Figure 1B. In this paper, we compared the following AF substrates (Figure 2):
The geometry of the biophysical model of the human atria was reconstructed from computed tomography (CT) scans of a patient in AF
† AF Substrate #1: Multiple-wavelet AF. This substrate was based on a homogeneous and isotropic tissue in which the LR model was adjusted
Methods
Downloaded from by guest on November 16, 2015
pacing sites and pacing cycle lengths (PCLs) leading to a local capture of AF.9 As a result, a better ability to sustain capture was found in the right atrial free wall, the left atrial appendage, and the pulmonary veins, where the wavefronts induced by pacing encompassed the major part of the paced atrium. This capture was accompanied by residual reentrant waves outside the area of capture such that AF re-initiation was generally observed after stopping the pacing protocol, with no pace termination of AF. This study suggested that rapid pacing of AF could indeed induce a local capture of the atrial tissue and that the septum area was the only location allowing the capture of both the atria from a single site. Based on these observations, a novel pacing scheme based on pacing from the septum area was developed and showed promising results for AF termination using a model-based approach.10 Atrial fibrillation can have different clinical forms corresponding to different patient-specific atrial substrates. While the polymorphic nature of AF is difficult to study in biological experiments, computer modelling allows a systematic study of the influence of atrial substrate. To get some insights into how much inter-patient variability may affect the efficacy of septum pacing, we used a computer model to assess the influence of heterogeneities in atrial substrate on the ability to capture AF with rapid pacing from the septum area. We have chosen to consider in this paper only substrates with electrophysiological remodelling and no structural remodelling such as fibrosis or slow conduction areas. This would correspond to applying the pacing therapy at an early stage in the progression of AF. We hypothesized that the pacing therapy would be more likely to be effective under these conditions.
before an ablation procedure. Since the patient was in chronic AF, the atria were slightly dilated. A three-dimensional (3D) cubic mesh representing the atrial myocardium and the atrial fast conducting system was created (1 040 000 nodes, 0.33 mm resolution) as published previously.7,11 In our anatomically simplified geometry, two interatrial connections were represented among the several connections existing between the right and the left atria:12 the Bachmann’s bundle (BB) and the rim of the septum. Bachmann’s bundle could only be partially segmented, therefore its structure was extrapolated to replicate the other anatomical models.13 – 15 The septum area was represented as a normally conducting tissue connecting both the atria through the septal rim, with a hole in the middle representing the fossa ovalis. The resulting geometry is shown in Figure 1A, with BB and the septum area highlighted in red. Rule-based fibre orientation was included, based on the previous works.7,13 – 17 Since investigating the different pacing protocols presented in this paper involved a total of 1600 simulations of 30 s, computational load was a major issue that was addressed here by further simplifying the atrial geometry: a coarser surface model (triangular mesh with 117 000 nodes, 0.45 mm resolution similar to previous models14,18) was derived from our previously published 3D anatomical model.7 Fibre orientation of the 3D model was projected onto the atrial surface. This simplified anatomical surface model matches our previous 3D model to facilitate the comparison. As a result, the number of nodes (and thus computational time) was reduced by a factor of 9. The full 3D model was used in a subset of the simulations for the validation of the surface model. Electrical propagation in the atrial tissue was simulated using the monodomain equation solved using operator splitting, explicit time integration, and spatial discretization based on a finite volume method19 in the surface model and finite difference method in the cubic 3D model. Time step was adaptive with 20 ms for diffusion, 12.5 – 100 ms for reaction, and 100 ms as operator splitting time step.18
768
A. Rusu et al.
A Left atrium
Right atrium Bachmann’s bundle
Right atrium
Pulmonary veins Superior vena cava
Zones of heterogeneities
B
10%
50%
20%
60%
30%
70%
40%
80%
Mitral valve Coronary artery
Tricuspid valve
Inferior vena cava
Septum area (Left atrium)
Figure 1 (A) Anterior and posterior view of the atrial geometry obtained from CT. The Bachmann’s bundle is highlighted in red. The bottom
Multiple wavelets
Anisotropy
Heterogeneities
Cellular model
1:1
Homogeneous
LR for AF
Action potentials
[ms] 0
Cholinergic AF
3:1
Heterogeneities in acetylcholine
CRN for AF
+ 0
Heterogeneous AF
4:1
Heterogeneities in repolarization
1:1
Homogeneous BB removed
100 200 300 0
[ms] 100 200 300
+
CRN for AF 0
AF without BB
100 200 300
100 200 300 0
[ms] 100 200 300
LR for AF [ms] 0
100 200 300
Figure 2 Summary of the different AF models corresponding to electrophysiological changes (three substrate models in blue) or anatomical changes (one anatomical model in red). For the second and the third models, baseline action potentials are represented on the left and the action potential inside the zones of the heterogeneities are on the right.
Downloaded from by guest on November 16, 2015
images show the left atrium separate from the right atrium with the septum area highlighted in red. (B) Introduction of random patchy heterogeneities in the atrial substrate.
769
Influence of atrial substrate on local capture
to mimic electrical remodelling as observed in patients suffering from permanent AF, such as the shortening of the action potential duration (APD). This was simulated by setting the channel conductance GNa, GK, and Gsi to 16, 0.423, and 0.55 mS/cm2, respectively.18 The resulting AF dynamics revealed multiple wavelets continuously changing in size and duration due to functional or anatomical reentries. † AF Substrate #2: Cholinergic AF. This substrate was based on an atrial tissue with a 3 : 1 anisotropy and the CRN model in which ICaL was reduced by 70% and Ito and IKur by 50%. An acetylcholine-dependent K+ current IKACh was added.23 Spatial variations in acetylcholine concentrations were introduced: IKACh was globally increased by 250% while it was decreased by 30% inside the red areas shown in Figure 1B. The resulting AF dynamics was maintained by the stable rotors. † AF Substrate #3: Heterogeneous AF or AF due to repolarization heterogeneities. This substrate was based on an atrial tissue with a 4 : 1 anisotropy and the CRN model in which the current ICaL was reduced 30%, Ito by 80%, IKur by 90%, and IKr increased by 50%.22 In addition, patchy heterogeneities were introduced as shown in Figure 1B, where ICaL was decreased by 50% and Ito and IKr were increased by 200% relative to the original model. Wavebreaks were created by the repolarization gradients and this contributed to the maintenance of AF.
Modelling rapid pacing of atrial fibrillation Atrial fibrillation was initiated in the biophysical models in a similar way to clinical experiments, using a programmed stimulation protocol or a burst pacing protocol.18,21,23 To avoid any timing adjustment of the stimulation protocol, burst pacing at 10 Hz was applied from one to five selected locations on the atrial surface until AF was observed (from 0.5 to 5 s). To perform computer simulations of AF therapies equivalent to animal or clinical experiments, 16 time instants were selected at 30 s intervals during one simulation of sustained AF and the corresponding tissue states (including membrane potential maps and all other state variables) were stored. This selection was repeated separately for each of the four AF models (three AF substrates and one anatomical model). These tissue states correspond to different, independent conditions of fibrillatory activity in the atrial tissue. They served as initial conditions for the repetitive application of the pacing therapy. Simulations results were averaged over 16 AF simulations for each model, thereby allowing statistical comparison between the different models. Rapid pacing was applied by injecting an intracellular current of 80 mA/ cm2 for the LR model and 100 mA/cm2 for the CRN model with 1 ms pacing pulses inside the cells located in the septum area highlighted in red in Figure 1A. Each pacing sequence was applied during 30 s at PCLs in the range of 50 – 100% of the AF cycle length (AFCL). Pacing cycle length was increased with steps of 2% AFCL. Therefore, in total, 25 simulations of 30 s of pacing have been performed for each of the 16 AF initial conditions. Atrial fibrillation cycle length was computed as a mean of the cycle length values measured on the transmembrane potentials recorded at the 24 electrode pairs shown in Figure 3, and averaged over 5 s. In this
Conditions for capture at each pair 1. Propagation direction from the interatrial junction towards the appendages 2. CL = PCL + 2%
Figure 3 Automatic assessment of atrial capture during pacing. A set of 24 pairs of electrodes was evenly distributed on the atrial surface to assess the direction of wavefront propagation and the CL compared with the PCL.
Downloaded from by guest on November 16, 2015
The patchy patterns of heterogeneities were created by using a region growth algorithm21,22 and were intended to reproduce the type of patterns (repolarization gradients) observed in the experimental models of AF.24 They were generated using a random Markov field approach (described in our previous works21,25,26), which consists in filtering a white-noise field with a Gaussian spatial filter (with a length scale of 7.5 mm) and then thresholding the resulting filtered field to control the area of altered tissue that varied here between 0 (homogeneous) and 80%. The fourth model is an anatomical model that was created based on the atrial geometry of Figure 1A using the AF Substrate #1 (with no heterogeneities) with BB removed (Figure 2). The objective was to study the impact
of BB, which is a major pathway for interatrial conduction, on the ability to capture AF during rapid pacing from the septum. Finally, to get an idea about the influence of atrial thickness on capture, we performed one set of simulations using the 3D version of our model7 based on the AF Substrate #1.
770
A. Rusu et al.
study, we did not take into account the additional activity originating from the sinoatrial node during or after pacing.
Capture with rapid pacing during multiple-wavelet atrial fibrillation
Automatic assessment of atrial fibrillation capture Capture was defined as the ability of the pacing burst to take control over a given area away from the pacing site. It relates to the local capture around the pacing site, not to the global capture of both the atria. During rapid pacing, the transmembrane potentials of 24 pairs of electrodes evenly distributed on the atrial surface (Figure 3) were analysed to determine the percentage of captured atrial tissue. At each pair of electrodes, the cycle length and the direction of propagating waves were assessed: we considered that a pair was captured if the direction of propagation was from the interatrial junction to the appendages and the cycle length was equal to the PCL + 2%. For each pacing sequence, a period of 30 s was analysed and the percentage of capture atrial tissue was computed as the ratio of the pairs where capture was observed divided by the total number of pairs. The capture window was defined as the range of PCL for which capture was .50%.
In the multiple-wavelet AF with no heterogeneities (AF Substrate #1), the maximum percentage of captured tissue was 78.9% for a PCL of 89% of AFCL (Figure 4) and the capture window was 85 – 95% AFCL. Atrial fibrillation cycle length was 74 ms. This means that for optimal parameters, AF capture could be obtained on almost 80% of the atrial surface in both the atria (highlighted in red in Figure 4), with residual reentrant wavelets present in the extremities.
Influence of heterogeneities in atrial substrate Figure 4 presents the capture results for the AF substrates including heterogeneities. Atrial fibrillation cycle length in the cholinergic model (AF Substrate #2) ranged from 69 ms (0% heterogeneities)
Cholinergic AF
Multiple wavelets 100 90 80 70 60 50 40 30 20 10 0
Heterogeneous AF
100
100
90
90
80
0% 10% 20% 30% 40% 50% 60% 70% 80%
70 60 50 40 30 20 10 50
60
70 80 90 PCL [%AFCL]
100
0 50
Downloaded from by guest on November 16, 2015
Capture [%]
Results
80 70 60 50 40 30 20 10
60
70 80 90 PCL [%AFCL]
100
0 50
60
70 80 90 PCL [%AFCL]
PCL: 91% AFCL
PCL: 79% AFCL 30% Heterogeneities
PCL: 67% AFCL 30% Heterogeneities
AF
AF
AF –80 mV
Capture of AF
Capture of AF
100
20 mV
Capture of AF
Figure 4 Capture as a function of the PCL for the different AF substrates. For each PCL, the capture percentage was computed as an average on 16 simulations. For the cholinergic and heterogeneous AF, the percentage of patchy heterogeneities corresponds to those presented in Figure 1. The middle graph represent the spatial extend (in red) of the best capture results obtained for each model. The two bottom graphs show example of transmembrane potential maps during AF and during capture with rapid pacing of AF for each AF substrate.
771
Influence of atrial substrate on local capture
A
100
Capture [%]
80 60
Without BB
40
With BB
20 0 60
90 70 80 PCL [%AFCL]
B Activation maps during
100
rapid pacing with BB
Activation maps during rapid pacing without BB
0 ms
0 ms
15 ms
15 ms
30 ms
30 ms
45 ms
45 ms
Influence of Bachmann’s bundle The AFCL in the anatomical model of AF without BB was 72 ms, which is comparable with the value obtained in the model with BB (AFCL ¼ 74 ms). Figure 5 shows the capture window obtained for both the models: 83 –100% AFCL with BB and 81– 100% AFCL without BB. The removal of BB did not significantly improve the capture window. The activation maps of the simulations presented in Figure 5 however showed that the BB induced multiple wavebreaks close to the pacing site. This did not however affect the fibrillatory activity away from the pacing site.
Influence of atrial thickness Figure 6 presents a comparison of AF capture results in the surface model and the corresponding 3D model in the case of multiple-wavelet AF with no heterogeneities. The overall shape of the curve of capture as a function of PCL was similar in both the cases. However, the capture values were slightly decreased in the 3D model (the maximum capture percentage was decreased from 78.9 to 63.9% of the atrial surface). Due to the heavy computational load, simulations in the 3D model could only be performed in one AF substrate.
Discussion The model-based results obtained in this study suggest that the heterogeneities in atrial substrate greatly influence the ability to capture AF with rapid pacing from the septum area. This behaviour was expected since we already observed that the heterogeneities in atrial substrate play a role in the perpetuation of AF by promoting
–80 mV 20 mV PCL = 91% AFCL
Figure 5 (A) Capture as a function of the PCL with and without the BB, using AF Substrate #1. For each PCL, the capture percentage was computed as an average on 16 simulations. (B) Example of transmembrane potential maps during rapid pacing of AF at a PCL of 91% of AFCL with and without BB.
AF reentries.21,22 However, this influence was different in the presence of heterogeneities in vagal activation or heterogeneities in repolarization. In the first case, the capture patterns tended to be more regular and they were fewer anchored wavelets than in other AF models. We could not establish a direct link between the percentage of heterogeneities and their anatomical location and the capture results. In some cases, vagal activation even facilitated AF capture by rapid pacing compared with multiple-wavelet AF. In the second case, the heterogeneities in repolarization had a more dramatic impact on AF capture, with the creation of wave anchoring and wavebreaks. Note that AFCL in the first model (69 –103 ms) was much lower (AF faster) than in the second model (180 –252 ms), as in the previous works.22
Downloaded from by guest on November 16, 2015
to 103 ms (80% heterogeneities). The cholinergic model with no heterogeneities had therefore an AFCL comparable with the multiplewavelet AF. The capture window was also similar (89–100% AFCL), except that for lower values of PCL where capture in about 10% of the atrial tissue could be observed (no capture was observed in the multiple-wavelet model for these values). In the cholinergic model, the capture patterns tended to be more regular than in the multiplewavelet model (Figure 4), BB did not affect the pacing-induced wavefronts and there were fewer anchored wavelets. We could not observe a direct link between the increase of the percentage of heterogeneities in vagal activation and the ability to capture AF. For example, the best capture results were obtained in the cholinergic model with 30% of heterogeneities, with a very broad capture window (70–100% AFCL). The worst capture results were obtained with 50% of heterogeneities. We were not able to link the capture results to the anatomical location of heterogeneities. The AFCL in the AF Substrate #3 ranged from 180 ms (0% heterogeneities) to 252 ms (80% heterogeneities). These values are higher than for the first two substrates. The capture window for the substrate with no heterogeneities was the broadest so far (70–100% AFCL). However, the introduction of patchy heterogeneities in repolarization progressively degraded the capture results. In this AF substrate, there was a clearer link between the percentage of heterogeneities and the degradation of capture. These heterogeneities significantly influenced the capture patterns by creating wave anchoring and wavebreaks, even in areas close to the pacing site (Figure 4). Even if AF was slower, the spatial extend of the capture was greatly reduced compared with the first two substrates (as shown in red in Figure 4).
772
A. Rusu et al.
100 90
Capture [%]
80 70 60
Surface model 3D model
50 40 30 20 10 0 50
60
70 80 90 PCL [% AFCL]
100
Figure 6 Capture as a function of the PCL for the surface and for the 3D model, using AF Substrate #1.
Conflict of interest: N.V. is a full time employee of Medtronic Europe (Tolochenaz, Switzerland).
Downloaded from by guest on November 16, 2015
Our previous studies on rapid pacing during AF were based only the first substrate presented in this paper, namely the multiplewavelet AF model using the LR model in a homogeneous tissue.9,10 This previous work was based on a simplified atrial geometry that did not include BB or any inhomogeneities. We had studied the influence of atrial tissue dynamics on the ability to capture AF with rapid pacing by changing the APD and the conduction velocity (CV) on the entire atrial surface. The capture window obtained for the baseline model (with no changes in APD and CV) was comparable with what we obtained in this paper for the multiple-wavelet model (LR model) and for the cholinergic AF model (CRN model) with 0% heterogeneities (Figure 4). It is also comparable with the value found in humans by Capucci et al.,27 who published a capture window of 85 –105% of AFCL. In the previous modelling study, decreasing CV or APD globally in the atrial tissue led to a significant increase of the capture window and vice versa,9 this effect being more dramatic for the changes in APD. In the present study, we followed a different approach for modifying the atrial substrate since the changes were introduced as a patchy pattern of a heterogeneity of gradually increasing density. As a result, we observed in the model of cholinergic AF that an increase of APD in patchy patterns could either decrease (10% of heterogeneities) or increase (30% of heterogeneities) the capture window (Figure 4). Similar observations could be made for the model of heterogeneous AF. Therefore, when APD is changed in a patchy pattern rather than globally in the whole atrial tissue, the capture outcome is difficult to predict. Our previous studies also showed that AFCL is not predictive of the ability to capture AF by rapid pacing suggesting that a measure proportional to wavelength (estimated by the product of CV and APD) could give an estimation of the capture abilities in a specific patient.9 The present work also confirms the non-predictive value of AFCL. For example, in the model of cholinergic AF, increasing the percentage of the patchy heterogeneities from 0 to 10% did not significantly change AFCL while it significantly affected the capture window (Figure 4). This was also observed in clinical data in which subjects with the same AFCL had different capture results.27 However, these results further suggest that an overall measure of wavelength is probably not sufficient to assess the ability to capture
during AF and that a more detailed assessment of each patient’s atrial substrate would be needed. We investigated the role of BB during pacing since it is considered to play a fundamental role in interatrial conduction. When observing the activation maps during capture in the multiple-wavelet AF, we could observe many wavebreaks initiated due to conduction through BB that significantly affected the capture patterns. However, our study showed that even if BB greatly influenced locally the electrical propagation by providing a fast accessory interatrial conduction pathway, it did not significantly affect the global ability to capture during rapid pacing of AF from the septum area. This may be related to the fact that the septal pacing site was relatively close to both sides of BB. Significant progresses have been made in the last 10 years in the modelling of human atria and today most existing models include a detailed atrial geometry based on human imaging with fibre orientation and anisotropy, detailed cellular models of atrial electrophysiology as well as pathological components such as fibrotic tissue or atrial remodelling.12 Our objective was to focus on how pacing could be developed as a clinical therapy for AF. Compared with the existing computer models of human atria, the model presented here represents a tradeoff between complexity and the ability to simulate long runs of AF (in total more than 13 h of AF have been simulated for the results presented in this paper). Therefore, the work we presented here is based on human data, but to keep the computational load within a reasonable range we had to simplify the AF models into surface models. However, we performed a set of simulations in a 3D model to assess the effect of atrial thickness, which was a reduction of about 15% of the atrial surface that could be captured, while the pacing frequencies leading to capture remained comparable. It should be noted that we studied here different atrial substrates without the inclusion of external triggers such as rapid foci originating from the pulmonary veins. We expect that the conditions we have simulated correspond to a best-case scenario in which the pacing therapy is most likely to be effective. Additional pathways for (micro-) reentry created by fibrosis or fibre dissociation within the atrial wall may hinder the ability of pacing therapy to achieve local capture.28 Multilayer 3D models with a finer discretization could be used to test the hypothesis. In summary, changes in the atrial substrate have a significant effect on the rapid pacing outcomes, implying that patients may respond in very different manners to the rapid pacing of AF. This might be one explanation to the fact that so far no clinical study on pacing for AF termination has been conclusive:3,4 the characteristics of an individual undergoing a specific therapy is in general different from the mean of the population, therefore the therapy may not benefit every patient. This suggests that a successful pacing therapy for AF should be specific to each patient’s atrial substrate. While this polymorphic nature of AF is difficult to handle in clinical experiments, a modelbased approach offers a new method to study in detail the influence of atrial substrate. Personalized computational cardiac models are emerging today as an important tool for studying cardiac arrhythmia mechanisms, and have the potential to become powerful instruments for guiding clinical anti-arrhythmia therapy.29
Influence of atrial substrate on local capture
Funding This work was supported by the Theo-Rossi di Montelera Foundation (Lausanne, Switzerland) and Medtronic Europe (Tolochenaz, Switzerland).
References
15. Vigmond EJ, Ruckdeschel R, Trayanova N. Reentry in a morphologically realistic atrial model. J Cardiovasc Electrophysiol 2001;12:1046 –54. 16. Krueger M, Schmidt V, Tobo´n C, Weber FM, Lorenz C, Keller DUJ et al. Modeling atrial fiber orientation in patient-specific geometries: a semi-automatic rule-based approach. In Metaxas DM, Axel L (eds). Functional Imaging and Modeling of the Heart. Springer-Verlag Berlin Heidelberg; 2011. 223 –32. 17. McDowell KS, Vadakkumpadan F, Blake R, Blauer J, Plank G, Macleod RS et al. Mechanistic inquiry into the role of tissue remodeling in fibrotic lesions in human atrial fibrillation. Biophys J 2013;104:2764 – 73. 18. Virag N, Jacquemet V, Henriquez CS, Zozor S, Blanc O, Vesin J-M et al. Study of atrial arrhythmias in a computer model based on magnetic resonance images of human atria. Chaos 2002;12:754–63. 19. Zozor S, Blanc O, Jacquemet V, Virag N, Vesin J-M, Pruvot E et al. A numerical scheme for modeling wavefront propagation on a monolayer of arbitrary geometry. IEEE Trans Biomed Eng 2003;50:412 –20. 20. Courtemanche M, Ramirez RJ, Nattel S. Ionic mechanisms underlying human atrial action potential properties: insights from a mathematical model. Am J Physiol 1998; 275:H301–21. 21. Jacquemet V, Virag N, Ihara Z, Dang L, Blanc O, Zozor S et al. Study of unipolar electrogram morphology in a computer model of atrial fibrillation. J Cardiovasc Electrophysiol 2003;14:172. 22. Jacquemet V, Virag N, Kappenberger L. Wavelength and vulnerability to atrial fibrillation: insights from a computer model of human atria. Europace 2005;7(Suppl 2): 83 – 92. 23. Kneller J, Zou R, Vigmond EJ, Wang Z, Leon LJ, Nattel S. Cholinergic atrial fibrillation in a computer model of a two-dimensional sheet of canine atrial cells with realistic ionic properties. Circ Res 2002;90:73. 24. Fareh S, Villemaire C, Nattel S. Importance of refractoriness heterogeneity in the enhanced vulnerability to atrial fibrillation induction caused by tachycardia-induced atrial electrical remodeling. Circulation 1998;98:2202 –9. 25. Jacquemet V, Henriquez CS. Loading effect of fibroblast-myocyte coupling on resting potential, impulse propagation and repolarization: insights from a microstructure model. Am J Physiol Heart Circ Physiol 2008;294:H2040 –52. 26. Jacquemet V. A biophysical model of atrial fibrillation and electrograms: formulation, validation and applications. Ph.D. Thesis. E´cole Polytechnique Fe´de´rale de Lausanne (EPFL), Suisse, 2004. 27. Capucci A, Ravelli F, Nollo G, Montenero AS, Biffi M, Villani GQ. Capture window in human atrial fibrillation: evidence of excitable gap. J Cardiovasc Electrophysiol 1999;10: 319 –27. 28. de Groot NMS, Houben RPM, Smeets JL, Boersma E, Schotten U, Schalij MJ et al. Electropathological substrate of longstanding persistent atrial fibrillation in patients with structural heart disease: epicardial breakthrough. Circulation 2010;122: 1674 –82. 29. McDowell KS, Vadakkumpadan F, Blake R, Blauer J, Plank G, MacLeod RS et al. Methodology for patient-specific modeling of atrial fibrosis as a substrate for atrial fibrillation. J Electrocardiol 2012;45:640 – 5.
Downloaded from by guest on November 16, 2015
1. Nattel S. New ideas about atrial fibrillation 50 years on. Nature 2002;415:219 – 26. 2. European Heart Rhythm Association, European Association for Cardio-Thoracic Surgery, Camm AJ, Kirchhof P, Lip GY, Schotten U, Savelieva I, Ernst S et al. Guidelines for the management of atrial fibrillation: the task force for the management of atrial fibrillation of the European Society of Cardiology (ESC). Eur Heart J 2010;31: 2369– 429. 3. Redfearn DP, Yee R. Pacing delivered rate and rhythm control for atrial fibrillation. Curr Opin Cardiol 2006;21:83–7. 4. Ellenbogen KA. Pacing therapy for prevention of atrial fibrillation. Heart Rhythm 2007;4:S84 –7. 5. Allessie M, Kirchhof C, Scheffer GJ, Chorro F, Brugada J. Regional control of atrial fibrillation by rapid pacing in conscious dogs. Circulation 1991;84:1689 –97. 6. Pandozi C, Bianconi L, Villani M, Castro A, Altamura G, Toscano S et al. Local capture by atrial pacing in spontaneous chronic atrial fibrillation. Circulation 1997;95: 2416– 22. 7. Virag N, Jacquemet V, Kappenberger L. Modeling of Atrial Fibrillation. In Shenasa M, Hindricks G, Borggrefe M, Breithardt G, Josephson ME (eds). Cardiac Mapping, 4th ed. Blackwell Publishing Ltd.; 2013.131–9. 8. Kirchhof P, Bax J, Blomstrom-Lundquiest C, Calkins H, Camm AJ, Cappato R et al. Early and comprehensive management of atrial fibrillation: proceedings from the 2nd AFNET/EHRA consensus conference on atrial fibrillation entitled ‘research perspective in atrial fibrillation’. Europace 2009;11:860 –85. 9. Uldry L, Virag N, Vesin J-M, Kappenberger L. Optimizing local capture of atrial fibrillation: a model based study of the influence of tissue dynamics on the outcome of rapid pacing. Ann Biomed Eng 2010;38:3664 – 73. 10. Uldry L, Virag N, Lindemans F, Vesin J-M, Kappenberger L. Atrial septal pacing for the termination of atrial fibrillation: study in a biophysical model of human atria. Europace 2012;14:v112 –20. 11. Herlin A, Jacquemet V. Eikonal-based initiation of fibrillatory activity in thin-walled cardiac propagation models. Chaos 2011;21:043136. 12. Do¨ssel O, Krueger MW, Weber FM, Wilhelms M, Seemann G. Computational modeling of the human atrial anatomy and electrophysiology. Med Biol Eng Comput 2012; 50:773 –99. 13. Harrild D, Henriquez C. A computer model of normal conduction in the human atria. Circ Res 2000;87:E25 –36. 14. Tobon C, Ruiz-Villa CA, Heidenreich E, Romero L, Hornero F, Saiz J. A threedimensional human atrial model with fiber orientation. Electrograms and arrhythmic activation patterns relationship. PLoS ONE 2013;8:e50883.
773
