Loss of side-to-side electrical connections between atrial muscle bundles is thought to underlie conduction disturbances predisposing to atrial fibrillation (AF). Putatively, disruption of electrical connections occurs not only within the epicardial layer but also between the epicardial layer and the endocardial bundle network, thus impeding transmural conductions (‘breakthroughs’). However, both clinical and experimental studies have shown an enhancement of breakthroughs during later stages of AF. We tested the hypothesis that endo-epicardial uncoupling enhances endo-epicardial electrical dyssynchrony, breakthrough rate (BTR), and AF stability.
In a novel dual-layer computer model of the human atria, 100% connectivity between the two layers served as healthy control. Atrial structural remodelling was simulated by reducing the number of connections between the layers from 96 to 6 randomly chosen locations. With progressive elimination of connections, AF stability increased. Reduction in the number of connections from 96 to 24 resulted in an increase in endo-epicardial dyssynchrony from 6.6 ± 1.9 to 24.6 ± 1.3%, with a concomitant increase in BTR. A further reduction to 12 and 6 resulted in more pronounced endo-epicardial dyssynchrony of 34.4 ± 1.15 and 40.2 ± 0.52% but with BTR reduction. This biphasic relationship between endo-epicardial coupling and BTR was found independently from whether AF was maintained by re-entry or by ectopic focal discharges.
Loss of endo-epicardial coupling increases AF stability. There is a biphasic relation between endo-epicardial coupling and BTR. While at high degrees of endo-epicardial connectivity, the BTR is limited by the endo-epicardial synchronicity, at low degrees of connectivity, it is limited by the number of endo-epicardial connections.
Using this model, we could simulate transmural conduction between the endocardial and the epicardial layers.
Decrease in the degree of coupling between endocardium and epicardium increased the degree of endo-epicardial electrical dissociation and increased the incidence of transmural conduction until a certain degree of connectivity.
Endo-epicardial electrical dissociation was the strongest determinant of atrial fibrillation persistence in all simulations.
Introduction
Several mechanisms have been suggested to explain the increasing stability of atrial fibrillation (AF) over time. Disruption of electrical coupling between muscle bundles, resulting in narrower and thus more fibrillation waves, is considered as one of the main mechanisms contributing to AF stability in structurally remodelled atria.1–4 The anatomy of the atrial wall has been demonstrated to significantly determine conduction patterns of fibrillation waves during AF.5 For example, Gray et al.6 showed that the direction of atrial muscle bundles correlates well with the location of block lines and breakthrough in right atrium. Similarly, Wu et al.7 demonstrated an effect of orientation of pectinate muscles on fibrillation patterns and observed transmural conduction in the area of pectinate muscles. De Groot et al.8 found increases both in the number of fibrillation waves and in incidence of epicardial breakthroughs in patients with longstanding AF compared with patients with acute AF, suggesting that the development of the substrate for AF goes along with increasing incidence of conduction from the sub-epicardial layer to the endocardial bundle network and vice versa. This concept of three-dimensional (3D) conduction of fibrillation waves during AF was recently confirmed by several animal studies using simultaneous mapping of the epicardial layer and the endocardial bundle network.6,9–11 While these studies conclusively demonstrate transmural conduction in the atrial wall, they leave open several important conceptual questions.
In particular, they could not answer the question what drives 3D conduction during AF. In structurally remodelled atria, disruption of electrical connections not only occurs between the muscle bundles within the epicardial layer but also between the epicardial layer and the endocardial bundle network. This should enhance endo-epicardial conduction block and thus impede endo-epicardial conduction. However, the epicardial breakthrough rate (BTR) is higher in goats and patients with persistent AF compared with acute AF.2,11,12 In a previous simulation study,13 we used a new dual-layer computer model of AF to address the question whether dissociation of endo-epicardial electrical activity results in increased AF stability. In the present study, we test the hypothesis that progressive loss of endo-epicardial electrical connections increases the degree of endo-epicardial dissociation of electrical activity and that enhanced degree of endo-epicardial dissociation results in an enhanced BTR and a more complex fibrillation pattern.
Methods
Model
Model structure
A recent study by Eckstein et al.10 revealed an increase in electrical activity dissociation between epicardial layer and endocardial bundle network due to progressive uncoupling between these two layers. In order to model this phenomenon, we chose a novel dual-layer computer model with varying degrees of electrical coupling between two layers (each layer is a representation of either the endocardial or the epicardial layer). Each layer was modelled with a mono-domain reaction-diffusion model of a size of 4 × 4 cm and composed of 400 × 400 segments with a size of 0.01 × 0.01 cm. Since the specific connectivity between remodelled epicardial and endocardial layers is not known, we chose a randomly distributed set of circular connection points between both layers. However, as connection point distribution may have a large impact on conduction patterns, we performed simulations with more heterogeneous connection distributions (see Supplementary material online).
Dual-layer model of the atrial wall. (A) Model structure with two layers (epi- and endocardium) and transmural connections (grey cylinders). (B) Dissociation of electrical activity between the two layers and transmural conduction resulting in a breakthrough wave (black arrow). (C) Simulation protocol (see the Methods section for details).
Atrial cell model
Simulation protocols
To demonstrate that the quantitative relation between the BTR, the degree of endo-epicardial dyssynchrony, and the number of endo-epicardial connections is independent from the mechanisms driving AF, the arrhythmia was maintained by (A) spiral waves and (B) random focal activation mimicking continuous ectopic focal discharges.
AF driven by spiral waves
In one layer, a spiral wave was initiated using an S1–S2 protocol,15–17 while the other layer was quiescent, as described in our previous study.13 The simulation was continued for 1 s.
One second after the start of the simulation, six connection points were added at randomly chosen sites. These sites were chosen such that two connection points were at least 0.15 cm apart. To exclude possible bias resulting from a particular geometry of connection points, eight different sets of six randomly chosen sites were created. For each of the eight configurations, the simulation of Step 1 was continued for 6 more seconds. This resulted in eight different simulations that were used to initialize the simulations in the next step.
The eight simulations from Step 2 were continued for another 6 s, either without changing the connection points or after adding randomly chosen connection points (at least 0.15 cm apart). Thus, the total number of connections was 6, 12, 24, 48, or 96. In addition, each simulation from Step 2 was continued with 100% connectivity, i.e. all opposing segments in the two layers were connected to each other.
Atrial fibrillation driven by random ectopy
Both layers were paced at randomly chosen pacing cites distributed over both layers. The model was paced at random time steps. The frequency of ectopy was chosen in a way that an AF cycle length (AFCL) of ∼200 ms was achieved.
Step 1 was done for all different number of connections (6, 12, 24, 48, 96, and 100% connectivity) for 8 randomly distributed connections and simulations continued for 6 s.
Analysis
In all simulations, the following parameters were analysed (for algorithms used for these analyses, see the Supplementary material online, Supplemental methods):
Number of phase singularities (PSs)
Phase singularity lifespan
Breakthrough rates
Breakthrough lifespans
Number of fibrillation waves
Wave front lifespans
Atrial fibrillation cycle length
Wave front conduction velocity (WFCV)
Excitable area (EA)
Degree of endo-epicardial dyssynchrony
Degree of endo-epicardial dissociation
Atrial fibrillation duration
Results
Colour-coded membrane potentials of representative spiral wave-driven simulations with different degrees of endo-epicardial electrical coupling. The epicardial layer is shown in the top, the endocardial in the bottom panel. Snapshots are shown for every second of simulation time.
In the examples with 96 and 48 connections (Figure 2B and C), the two layers showed rapid synchronization by a high number of BTs. With increasing synchronization of the activity in the two layers, the number of waves, PSs, and BTR decreased. After 2000 ms the two layers showed synchronized activity and shortly thereafter, the episode terminated.
In the example with 24 connections (Figure 2D), dyssynchrony decreased between 1000 and 2000 ms, which resulted in synchronized activity between 3000 and 4000 ms and termination of the episode at 4100 ms.
In the examples with 12 and 6 connections (Figure 2E and F), the initial rate of BTs is lower, no resynchronization of the two layers occurred, dyssynchrony of electrical activation between the two layers remained high, and the fibrillation patterns in both layers remained complex.
Colour-coded membrane potentials of representative of AF simulations driven by random ectopy with different degrees of endo-epicardial electrical coupling. The epicardial layer is shown in the top and the endocardial in the bottom panel. Snapshots are shown for every second of simulation time. The black arrow in the upper panel at 2000 ms is showing an ectopy. The arrow in the lower panel at 1000 ms is showing a BT.
Parameters determining stability of atrial fibrillation
Atrial fibrillation persistence. The Kaplan–Meier curve showing AF persistence in different sets of simulations with 6, 12, 24, 48, and 96 connections and 100% connectivity (n = 8 for all different sets of simulations).
In order to study how loss of connectivity stabilizes AF, the effect of the number of connections on a variety of parameters describing the substrate for AF was studied.
Electrophysiological parameters as a function of the degree of connectivity. Averages for the whole simulation period (6000 ms) are shown for all parameters. (A) average number of PSs, (B) PS lifespan, (C) average BTRs, (D) breakthrough lifespan, (E) average number of waves, (F) wave front lifespan, (G) average AFCLs, (H) average WFCV, (I) average EA, and (J) average degree of dyssynchrony.
The number of waves was significantly larger at 12, 24, 48, and 96 connections than at 100% connectivity (Figure 5E), while the wave front lifespans were shorter than at 100% connectivity (Figure 5F). There was no significant difference in AFCL, WFCV, and EA between simulations with different numbers of connections (see Figure 5G–I). Dyssynchrony progressively and strongly increased with the elimination of the endo-epicardial connections (Figure 5J).
Differences in electrophysiological parameters between non-sustained (<6 s) and sustained AF episodes. All parameters were averaged for the entire durations of AF episodes. (A) Average number of PSs, (B) PS lifespan, (C) average BTRs, (D) breakthrough lifespan, (E) average number of waves, (F) wave front lifespan, (G) average AFCLs, (H) average WFCV, (I) average EA, and (J) average degree of dyssynchrony.
In Figure 6A and B, the numbers of PSs and PS lifespans are shown. There were no significant differences between the two groups in these parameters. There were neither differences in BTR and BT lifespan (Figure 6C and D) nor in the number of waves or wave front lifespan (Figure 6E and F). There were no differences in AFCL and WFCV, but EA was slightly but significantly lower in sustained vs. non-sustained AF (Figure 6G–I). The most striking difference was that the degree of dyssynchrony was much higher in the group with sustained AF compared with non-sustained AF (Figure 6, group J).
In order to investigate which parameter is the strongest determinant of AF duration, we used multilevel logistic regression analysis among measured parameters. Three parameters were significantly correlated to AF stability (see Table 1). Of these significant predictive parameters, dyssynchrony was by far the strongest determinant of AF persistence in this set of simulations.
Association between electrophysiological parameters and AF stability determined by multilevel logistic regression analysis
| EP parameters | P-values | Coefficients |
|---|---|---|
| PSs | 0.639 | 0.0519 |
| PS lifespan | 0.270 | 0.0053 |
| BTs | 0.369 | −0.1644 |
| BT lifespan | 0.244 | −0.0035 |
| Waves | 0.720 | 0.1157 |
| Wave front lifespan | 0.957 | 0.0015 |
| EA | 0.027* | −0.0013* |
| AFCL | 0.245 | −0.0348 |
| WFCV | 0.2070 | −65.441 |
| Dyssynchrony | 0.004* | 0.1315* |
| Connections | 0.01* | −0.0628* |
| EP parameters | P-values | Coefficients |
|---|---|---|
| PSs | 0.639 | 0.0519 |
| PS lifespan | 0.270 | 0.0053 |
| BTs | 0.369 | −0.1644 |
| BT lifespan | 0.244 | −0.0035 |
| Waves | 0.720 | 0.1157 |
| Wave front lifespan | 0.957 | 0.0015 |
| EA | 0.027* | −0.0013* |
| AFCL | 0.245 | −0.0348 |
| WFCV | 0.2070 | −65.441 |
| Dyssynchrony | 0.004* | 0.1315* |
| Connections | 0.01* | −0.0628* |
*P-value < 0.05.
Association between electrophysiological parameters and AF stability determined by multilevel logistic regression analysis
| EP parameters | P-values | Coefficients |
|---|---|---|
| PSs | 0.639 | 0.0519 |
| PS lifespan | 0.270 | 0.0053 |
| BTs | 0.369 | −0.1644 |
| BT lifespan | 0.244 | −0.0035 |
| Waves | 0.720 | 0.1157 |
| Wave front lifespan | 0.957 | 0.0015 |
| EA | 0.027* | −0.0013* |
| AFCL | 0.245 | −0.0348 |
| WFCV | 0.2070 | −65.441 |
| Dyssynchrony | 0.004* | 0.1315* |
| Connections | 0.01* | −0.0628* |
| EP parameters | P-values | Coefficients |
|---|---|---|
| PSs | 0.639 | 0.0519 |
| PS lifespan | 0.270 | 0.0053 |
| BTs | 0.369 | −0.1644 |
| BT lifespan | 0.244 | −0.0035 |
| Waves | 0.720 | 0.1157 |
| Wave front lifespan | 0.957 | 0.0015 |
| EA | 0.027* | −0.0013* |
| AFCL | 0.245 | −0.0348 |
| WFCV | 0.2070 | −65.441 |
| Dyssynchrony | 0.004* | 0.1315* |
| Connections | 0.01* | −0.0628* |
*P-value < 0.05.
In random ectopy-driven simulations, AF was sustained by pacing so that a study on AF stability was not applicable.
Relation between dyssynchrony and breakthrough rate
(A) Relation of BTR and dyssynchrony to the degree of epi-endo connectivity in spiral wave-driven simulations. Average percentage of dyssynchrony (blue circles) over the entire period of each simulation. Average BTR over the entire duration of the episodes (red squares) and within the first 50 ms (green squares, highest degree of dyssynchrony). (B) Relation of BTR and dyssynchrony to the degree of epi-endo connectivity in simulations of AF driven by random ectopy. Average percentage of dyssynchrony (blue circles) over the entire period of each simulation. Average BTR over the entire duration of the episodes (red squares).
In Figure 7B, dyssynchrony (blue line) and BT rate (red line) extracted from random pacing simulations are illustrated. The overall BT rate of the whole AF episode showed a biphasic relation to the number of connections, increasing from 100% connectivity down to 24 connections, but decreasing for 12 and 6 connections. This pattern was similar to that observed in our spiral wave-driven simulations.
Comparison of experimental data and simulation data
Comparison of modelling data with experimental results. Breakthrough rates (as percentage of all waves) vs. the degree of dissociation observed in experimental (red squares) and simulation studies (blue squares). con, number of connections in simulation studies; aAF, goats in acute AF; 3mAF, goats after 3 weeks of AF; 6mAF, goats after 6 months of AF.
Experimental data (open squares) include data derived from goats in acute AF (aAF, n = 7), 3 weeks of AF (3wAF, n = 7), and 6 months of AF (6mAF, n = 7) and were taken from a previous study.18 As illustrated in this figure, the degree of endo-epicardial dissociation increased with AF duration. To some extent of dissociation, increase in degree of dissociation leads to increase in BT both in experimental (3wAF) and modelling (24 connections). Beyond that even with huge increase in dissociation only slight increase in BT rate in experimental (6mAF) and decrease in modelling (12 connections) result was observed. Overall, although there was no point-to-point match between experimental and modelling results, patterns were similar.
Discussion
The mechanisms contributing to an increase in stability of AF are not fully understood. Electrical remodelling, electrical uncoupling between muscle bundles, and heterogeneity of action potential duration all together have been hypothesized to increase stability of AF.2,19 More specifically, disruption of electrical side-to-side connections between muscle bundles occurs as a consequence of structural alterations in the atrial wall.1 This loss of connectivity between muscle bundles leads to an increase in likelihood of conduction block, the number of fibrillation waves, and AF stability.2,12 Disruption of electrical side-to-side connections does not only occur within the epicardial layer but most likely also between the epicardial layer and the endocardial bundle network.10,11 This uncoupling between the epi- and endocardial layers of the atrial wall—at a first glance—should lead to fewer transmural conduction events resulting in a lower incidence of breakthrough during AF. However, in goats as well as in patients, the incidence of breakthroughs has been shown to be much higher in persistent than in acute forms of the arrhythmia.2,8,10,18 The present study was undertaken to clarify how loss of endo-epicardial connectivity might result in more transmural conduction and higher BTRs. We hypothesized that disruption of endo-epicardial connections increases the dissociation of electrical activity between the epicardial layer and the endocardial muscle bundles. As a prerequisite for transmural conduction, increased endo-epicardial dissociation of electrical activity will enhance transmural conduction and result in a higher BTR despite lower availability of endo-epicardial connections and, as a consequence, an increase in AF duration.
Model structure
There are two groups of atrial models with respect to geometrical properties: surface models that treat the atria as two-dimensional sheet folded into the shape of the atria (with two main cavities, atrial appendages, and connections between the two atria at the septal rings)20–22 and the true 3D volumetric models. Volumetric models go into greater detail in modelling the atrial wall including such properties as complex anatomy of endocardial bundle networks23–26 and fibre orientations in atrium.27–29 Several mechanisms were investigated using volumetric models such as the effect of ectopies, AP heterogeneities, fibrosis, and meandering of PSs on fibrillatory propagation patterns and AF stability. A few recent studies also investigated the effect of atrial wall thickness on fibrillatory patterns and on AF maintenance.9,30
While all these models were useful to answer many important questions about mechanisms contributing to AF stability, none of them considered endo-epicardial dyssynchrony, transmural conduction, and their effect on AF stability.
The present study was designed to investigate the relationship between breakthrough incidences, endo-epicardial dyssynchrony, and AF stability. We chose a dual-layer model instead of a volumetric 3D model because: (i) The exact location and distribution of muscle bundles providing endo-epicardial electrical connections are very difficult to determine experimentally. Any simulation implementing a detailed anatomy of such endo-epicardial connection points would suggest a level of accuracy that can currently not be achieved even with advanced imaging techniques. (ii) We aimed to exclusively focus on the quantitative determinants of endo-epicardial dissociation and conduction without any confounding effect of complex 3D geometry, which due to scope and complexity could not be addressed in this study. Due to all above-mentioned reasons and to avoid complexities that make us unable to investigate the dynamics of transmural conduction properties, we chose a simplified geometry proof-of-principle model.
Effect of transmural uncoupling on dyssynchrony and breakthroughs
The novelty of our model is that it describes truly 3D (transmural) conduction between the endocardial and the epicardial layers13 and by doing so introduces a new type of simulated fibrillation waves, i.e. breakthroughs. In this study, we demonstrated that a decrease in the degree of coupling between the endocardium and the epicardium increased the degree of dyssynchrony and in contrast increased the incidence of transmural conduction until a certain degree of connectivity. Our results demonstrate that this increase can be explained by the enhanced dyssynchrony between the two layers. If both opposing layers at a connection point are either excited or non-excited, no transmural conduction occurs. Only in the presence of dyssynchronous electrical activity, fibrillation waves can propagate along endo-epicardial connections. As a result, in the presence of a high number of connections, the BTR is low because the degree of endo-epicardial dissociation is low, despite the fact that the number of available endo-epicardial connections is high.
Interestingly, increasing the extent of uncoupling to a larger degree (fewer than 24 connections) showed an opposite effect. The BTR decreased with further decrease in the number of connections, while the degree of dyssynchrony increased. To explain this phenomenon, we measured the BTR during a short period just after introducing more connections (i.e. the start of Step 3 in Figure 2), when the layers have not yet synchronized. During this period, we calculated the number of BTs during the highest achieved degree of dyssynchrony. Under these circumstances, we observed that the BTR strongly and monotonously declined with a reduction in the number of endo-epicardial connections, just as expected based on the simple reasoning that the reduced number of available electrical pathways should reduce the occurrence of endo-epicardial conduction. The results indicate that under these modelling conditions the BT rate not only depends on the degree of dyssynchrony, but also is limited by the actual number of endo-epicardial connection points. In case only a few connections exist between the two layers, breakthroughs will occur less frequently, because of the low number of available endo-epicardial connections, despite a high degree of endo-epicardial dissociation.
We also showed that our conclusions are transferable to circumstances under which AF is maintained by other mechanisms. We tested the proposed relation between endo-epicardial connections, endo-epicardial dissociation, and BTR in a modification of the model in which the arrhythmia is driven by multiple ectopic foci. This mechanism has been shown to perpetuate AF under specific circumstances, such as vagal nerve stimulation.31 Under these conditions, we identified exactly the same pattern in the quantitative relation between the number of endo-epicardial connection points, endo-epicardial dissociation, and BTR.
Effect of transmural uncoupling on atrial fibrillation stability
In this study, we demonstrate that a decrease in transmural coupling increases AF episode stability in spiral waves-driven simulations. Not all of the important determinants of AF stability such as AFCL and conduction velocity were altered by the different degrees of endo-epicardial coupling. Only the EA was slightly but significantly affected. Importantly, all differences observed between small and large degrees of uncoupling were related to 3D conduction. Atrial fibrillation duration increased with decreasing number of connections, and we identified dyssynchrony as the strongest predictor for AF stability. Taken together, these findings clearly suggest that the 3D character of the conduction pattern in itself is an important determinant of the stability of the arrhythmia.
Relation between experimental results and the model data
The phenomenon of endo-epicardial dissociation of electrical activity has been experimentally studied by several research groups. Schuessler et al.32 were the first to reveal endo-epicardial dyssynchrony of activation during acetylcholine-induced AF in canine right atria. Gray et al.6 and Wu et al.7 also investigated the 3D structure of the atria and its effect on activation sequences during AF and also observed transmural conduction incidences. Everett et al.33 compared epicardial direct contact electrograms with non-contact mapping endocardial electrograms and found larger endo-epicardial activation time differences in a rapid atrial pacing model than in control animals. Also, the elegant study of Yamazaki et al.9 demonstrated the occurrence of endo-epicardial dissociation during AF in sheep with atrial dilatation. Recently, Eckstein et al.10,11 demonstrated that the degree of endo-epicardial dissociation increases during the course of AF. The mechanism of this increasing dissociation of electrical activity was not related to endo-epicardial dispersion of refractoriness but rather to loss of endo-epicardial coupling between muscle bundles. Very recently, we demonstrated that loss of continuity in the epicardial layer and anatomical rearrangement of atrial muscle bundles also contribute to the increase in endo-epicardial dissociation during the development of persistent AF.5,34
Our modelling data to some extent showed similar pattern with several experimental and clinical studies published during the past years. Both in patients and in goats, enhanced rates of breakthrough have been demonstrated in persistent AF compared with acute AF.8 The relationship between endo-epicardial dissociation and breakthroughs rate found in the simulation study is comparable but not identical with the one reported by Eckstein et al.10,18 The difference in BT rate between experimental and simulation finding could be because of the much greater complexity of the true atrial anatomy compared with our simplified model. Nevertheless, despite differences in absolute values, the nature of the relation is qualitatively similar between the model and experimental results.
Both studies show that at low degrees of endo-epicardial dissociation the BTR is low. At moderate levels of endo-epicardial dissociations, the BTR increases. Interestingly, a further increase in endo-epicardial dissociation (above 50%) only leads to small increase in BTRs in experimental study but decrease in modelling study. Thus, in the range of physiological degrees of endo-epicardial uncoupling, the degree of endo-epicardial dissociation rather than the number of available endo-epicardial electrical connections is the limiting factor for the BTR.
Clinical relevance
Our study strongly supports the importance of a loss of connectivity between epicardial and endocardium that is an integral part of the process of structural remodelling. The resulting increase in endo-epicardial dissociation of electrical activity increases AF complexity by increasing the tortuosity of fibrillatory conduction, allowing more fibrillation waves to coexist in the atrium, thereby increasing AF stability.35 For ablation strategies, this supports the notion that the lesions, whether created by an endocardial or epicardial approach, should be truly transmural. Upstream therapy would be beneficial if it could prevent the loss of connectivity between the epicardial layer and the endocardial bundle network, e.g. by inhibiting fibrosis. This would in effect reduce the area available to fibrillation waves and should therefore strongly reduce AF stability.
Study limitations
Our simplified dual-layer model should be considered as a proof-of-principle model. This model does not reflect complex structure of the atrium with pulmonary veins, valves, heterogeneity in ionic membrane currents, and variability in atrial wall thickness. Despite these limitations, we have shown that our model is well suited to investigate the relation between endo-epicardial dissociation and breakthroughs on AF complexity and stability. How additional complexities in the anatomy of the atrial wall affect endo-epicardial dissociation and transmural conduction should be addressed in future studies.
Supplementary material
Supplementary material is available at Europace online.
Conflict of interest: none declared.
Funding
This work was supported by the Leducq Foundation (07 CVD 03), the Dutch Research Organization (NWO, VIDI-grant 016.086.379), and the European Network for Translational Research in AF (FP7 collaborative project EUTRAF, No. 261057), the Center for Translational Molecular Medicine (CTMM, COHFAR), the Netherlands Heart Foundation (CVON2014-09, RACE V), the European Union (CATCH ME, No. 633196; AFib TrainNet, No. 675351; and the ERACoSysMED H2020 ERA-NET Cofund project Systems medicine for diagnosis and stratification of atrial fibrillation).
