Machine Learning to Identify Patients at Risk of Developing New-Onset Atrial Fibrillation after Coronary Artery Bypass

BACKGROUND

This study aims to get an effective machine learning (ML) prediction model of new-onset postoperative atrial fibrillation (POAF) following coronary artery bypass grafting (CABG) and to highlight the most relevant clinical factors.

METHODS

Four ML algorithms were employed to analyze 394 patients undergoing CABG, and their performances were compared: Multivariate Adaptive Regression Spline, Neural Network, Random Forest, and Support Vector Machine. Each algorithm was applied to the training data set to choose the most important features and to build a predictive model. The better performance for each model was obtained by a hyperparameters search, and the Receiver Operating Characteristic Area Under the Curve metric was selected to choose the best model. The best instances of each model were fed with the test data set, and some metrics were generated to assess the performance of the models on the unseen data set. A traditional logistic regression was also performed to be compared with the machine learning models.

RESULTS

Random Forest model showed the best performance, and the top five predictive features included age, preoperative creatinine values, time of aortic cross-clamping, body surface area, and Logistic Euro-Score.

CONCLUSIONS

The use of ML for clinical predictions requires an accurate evaluation of the models and their hyperparameters. Random Forest outperformed all other models in the clinical prediction of POAF following CABG.

Synergistic antiarrhythmic effect of inward rectifier current inhibition and pulmonary vein isolation in a 3D computer model for atrial fibrillation

AIMS

Recent clinical studies showed that antiarrhythmic drug (AAD) treatment and pulmonary vein isolation (PVI) synergistically reduce atrial fibrillation (AF) recurrences after initially successful ablation. Among newly developed atrial-selective AADs, inhibitors of the G-protein-gated acetylcholine-activated inward rectifier current (IKACh) were shown to effectively suppress AF in an experimental model but have not yet been evaluated clinically. We tested in silico whether inhibition of inward rectifier current or its combination with PVI reduces AF inducibility.

METHODS AND RESULTS

We simulated the effect of inward rectifier current blockade (IK blockade), PVI, and their combination on AF inducibility in a detailed three-dimensional model of the human atria with different degrees of fibrosis. IK blockade was simulated with a 30% reduction of its conductivity. Atrial fibrillation was initiated using incremental pacing applied at 20 different locations, in both atria. IK blockade effectively prevented AF induction in simulations without fibrosis as did PVI in simulations without fibrosis and with moderate fibrosis. Both interventions lost their efficacy in severe fibrosis. The combination of IK blockade and PVI prevented AF in simulations without fibrosis, with moderate fibrosis, and even with severe fibrosis. The combined therapy strongly decreased the number of fibrillation waves, due to a synergistic reduction of wavefront generation rate while the wavefront lifespan remained unchanged.

CONCLUSION

Newly developed blockers of atrial-specific inward rectifier currents, such as IKAch, might prevent AF occurrences and when combined with PVI effectively supress AF recurrences in human.

Persistent Atrial Fibrillation: The Role of Left Atrial Posterior Wall Isolation and Ablation Strategies

Atrial fibrillation (AF) is a global disease with rapidly rising incidence and prevalence. It is associated with a higher risk of stroke, dementia, cognitive decline, sudden and cardiovascular death, heart failure and impairment in quality of life. The disease is a major burden on the healthcare system. Paroxysmal AF is typically managed with medications or endocardial catheter ablation to good effect. However, a large proportion of patients with AF have persistent or long-standing persistent AF, which are more complex forms of the condition and thus more difficult to treat. This is in part due to the progressive electro-anatomical changes that occur with AF persistence and the spread of arrhythmogenic triggers and substrates outside of the pulmonary veins. The posterior wall of the left atrium is a common site for these changes and has become a target of ablation strategies to treat these more resistant forms of AF. In this review, we discuss the role of the posterior left atrial wall in persistent and long-standing persistent AF, the limitations of current endocardial-focused treatment strategies, and future perspectives on hybrid epicardial-endocardial approaches to posterior wall isolation or ablation.

Directed graph mapping exceeds phase mapping in discriminating true and false rotors detected with a basket catheter in a complex in-silico excitation pattern

Atrial fibrillation (AF) is the most frequently encountered arrhythmia in clinical practise. One of the major problems in the management of AF is the difficulty in identifying the arrhythmia sources from clinical recordings. That difficulty occurs because it is currently impossible to verify algorithms which determine these sources in clinical data, as high resolution true excitation patterns cannot be recorded in patients. Therefore, alternative approaches, like computer modelling are of great interest. In a recent published study such an approach was applied for the verification of one of the most commonly used algorithms, phase mapping (PM). A meandering rotor was simulated in the right atrium and a basket catheter was placed at 3 different locations: at the Superior Vena Cava (SVC), the Crista Terminalis (CT) and at the Coronary Sinus (CS). It was shown that although PM can identify the true source, it also finds several false sources due to the far-field effects and interpolation errors in all three positions. In addition, the detection efficiency strongly depended on the basket location. Recently, a novel tool was developed to analyse any arrhythmia called Directed Graph Mapping (DGM). DGM is based on network theory and creates a directed graph of the excitation pattern, from which the location and the source of the arrhythmia can be detected. Therefore, the objective of the current study was to compare the efficiency of DGM with PM on the basket dataset of this meandering rotor. The DGM-tool was applied for a wide variety of conduction velocities (minimal and maximal), which are input parameters of DGM. Overall we found that DGM was able to distinguish between the true rotor and false rotors for both the SVC and CT basket positions. For example, for the SVC position with a CVmin=0.01cmms, DGM detected the true core with a prevalence of 82% versus 94% for PM. Three false rotors where detected for 39.16% (DGM) versus 100% (PM); 22.64% (DGM) versus 100% (PM); and 0% (DGM) versus 57% (PM). Increasing CVmin to 0.02cmms had a stronger effect on the false rotors than on the true rotor. This led to a detection rate of 56.6% for the true rotor, while all the other false rotors disappeared. A similar trend was observed for the CT position. For the CS position, DGM already had a low performance for the true rotor for CVmin=0.01cmms (14.7%). For CVmin=0.02cmms the false and the true rotors could therefore not be distinguished. We can conclude that DGM can overcome some of the limitations of PM by varying one of its input parameters (CVmin). The true rotor is less dependent on this parameter than the false rotors, which disappear at a CVmin=0.02cmms. In order to increase to detection rate of the true rotor, one can decrease CVmin and discard the new rotors which also appear at lower values of CVmin.

Time resolution for wavefront and phase singularity tracking using activation maps in cardiac propagation models

The dynamics of cardiac fibrillation can be described by the number, the trajectory, the stability, and the lifespan of phase singularities (PSs). Accurate PS tracking is straightforward in simple uniform tissues but becomes more challenging as fibrosis, structural heterogeneity, and strong anisotropy are combined. In this paper, we derive a mathematical formulation for PS tracking in two-dimensional reaction-diffusion models. The method simultaneously tracks wavefronts and PS based on activation maps at full spatiotemporal resolution. PS tracking is formulated as a linear assignment problem solved by the Hungarian algorithm. The cost matrix incorporates information about distances between PS, chirality, and wavefronts. A graph of PS trajectories is generated to represent the creations and annihilations of PS pairs. Structure-preserving graph transformations are applied to provide a simplified description at longer observation time scales. The approach is validated in 180 simulations of fibrillation in four different types of substrates featuring, respectively, wavebreaks, ionic heterogeneities, fibrosis, and breakthrough patterns. The time step of PS tracking is studied in the range from 0.1 to 10 ms. The results show the benefits of improving time resolution from 1 to 0.1 ms. The tracking error rate decreases by an order of magnitude because the occurrence of simultaneous events becomes less likely. As observed on PS survival curves, the graph-based analysis facilitates the identification of macroscopically stable rotors despite wavefront fragmentation by fibrosis.

Effect of Na+-channel blockade on the three-dimensional substrate of atrial fibrillation in a model of endo-epicardial dissociation and transmural conduction

Aims

Atrial fibrillation (AF) is a progressive arrhythmia characterized by structural alterations that increase its stability. Both clinical and experimental studies showed a concomitant loss of antiarrhythmic drug efficacy in later stages of AF. The mechanisms underlying this loss of efficacy are not well understood. We hypothesized that structural remodelling may explain this reduced efficacy by making the substrate more three-dimensional. To investigate this, we simulated the effect of sodium (Na+)-channel block on AF in a model of progressive transmural uncoupling.

Methods and results

In a computer model consisting of two cross-connected atrial layers, with realistic atrial membrane behaviour, structural remodelling was simulated by reducing the number of connections between the layers. 100% of endo-epicardial connectivity represented a healthy atrium. At various degrees of structural remodelling, we assessed the effect of 60% sodium channel block on AF stability, endo-epicardial electrical activity dissociation (EED), and fibrillatory conduction pattern complexity quantified by number of waves, phase singularities (PSs), and transmural conduction ('breakthrough', BT). Sodium channel block terminated AF in non-remodelled but not in remodelled atria. The temporal excitable gap (EG) and AF cycle length increased at all degrees of remodelling when compared with control. Despite an increase of EED and EG, sodium channel block decreased the incidence of BT because of transmural conduction block. Sodium channel block decreased the number of waves and PSs in normal atrium but not in structurally remodelled atrium.

Conclusion

This simple atrial model explains the loss of efficacy of sodium channel blockers in terminating AF in the presence of severe structural remodelling as has been observed experimentally and clinically. Atrial fibrillation termination in atria with moderate structural remodelling in the presence of sodium channel block is caused by reduction of AF complexity. With more severe structural remodelling, sodium channel block fails to promote synchronization of the two layers of the model.

Understanding the Beat-to-Beat Variations of P-Waves Morphologies in AF Patients During Sinus Rhythm: A Scoping Review of the Atrial Simulation Studies

The remarkable advances in high-performance computing and the resulting increase of the computational power have the potential to leverage computational cardiology toward improving our understanding of the pathophysiological mechanisms of arrhythmias, such as Atrial Fibrillation (AF). In AF, a complex interaction between various triggers and the atrial substrate is considered to be the leading cause of AF initiation and perpetuation. In electrocardiography (ECG), P-wave is supposed to reflect atrial depolarization. It has been found that even during sinus rhythm (SR), multiple P-wave morphologies are present in AF patients with a history of AF, suggesting a higher dispersion of the conduction route in this population. In this scoping review, we focused on the mechanisms which modify the electrical substrate of the atria in AF patients, while investigating the existence of computational models that simulate the propagation of the electrical signal through different routes. The adopted review methodology is based on a structured analytical framework which includes the extraction of the keywords based on an initial limited bibliographic search, the extensive literature search and finally the identification of relevant articles based on the reference list of the studies. The leading mechanisms identified were classified according to their scale, spanning from mechanisms in the cell, tissue or organ level, and the produced outputs. The computational modeling approaches for each of the factors that influence the initiation and the perpetuation of AF are presented here to provide a clear overview of the existing literature. Several levels of categorization were adopted while the studies which aim to translate their findings to ECG phenotyping are highlighted. The results denote the availability of multiple models, which are appropriate under specific conditions. However, the consideration of complex scenarios taking into account multiple spatiotemporal scales, personalization of electrophysiological and anatomical models and the reproducibility in terms of ECG phenotyping has only partially been tackled so far.

The effect of complex intramural microstructure caused by structural remodeling on the stability of atrial fibrillation: Insights from a three-dimensional multi-layer modeling study

BACKGROUND

Recent researches have suggested that the complex three-dimensional structures caused by structural remodeling play a key role in atrial fibrillation (AF) substrates. Here we aimed to investigate this hypothesis using a multi-layer model representing intramural microstructural features.

METHODS

The proposed multi-layer model was composed of the endocardium, connection wall, and epicardium. In the connection wall, intramural fibrosis was simulated using fibrotic patches randomly scattered in the myocardial tissue of fibrotic layers, while endo-epicardial dissociation was simulated using myocardial patches randomly scattered in the fibrotic tissue of isolation layers. Multiple simulation groups were generated to quantitatively analyze the effects of endo-epicardial dissociation and intramural fibrosis on AF stability, including a stochastic group, interrelated groups, fibrosis-degree-controlled groups, and dissociation-degree-controlled groups.

RESULTS

1. Stable intramural re-entries were observed to move along complete re-entrant circuits inside the transmural wall in four of 65 simulations in the stochastic group. 2. About 21 of 23 stable simulations in the stochastic group were distributed in the areas with high endo-epicardial dissociation and intramural fibrosis. 3. The difference between fibrosis-degree-controlled groups and dissociation-degree-controlled groups suggested that some distributions of connection areas may affect AF episodes despite low intramural fibrosis and endo-epicardial dissociation. 4. The overview of tracking phase singularities revealed that endo-epicardial dissociation played a visible role in AF substrates.

CONCLUSION

The complex intramural microstructure is positively correlated with critical components of AF maintenance mechanisms. The occurrence of intramural re-entry further indicates the complexity of AF wave-dynamics.

Simulated P wave morphology in the presence of endo-epicardial activation delay

Aims

Evidences of asynchrony between epicardial and endocardial activation in the atrial wall have been reported. We used a computer model of the atria and torso to investigate the consequences of such activation delay on P wave morphology, while controlling for P wave duration.

Methods and results

We created 390 models of the atria based on the same geometry. These models differed by atrial wall thickness (from 2 to 3 mm), transmural coupling, and tissue conductivity in the endocardial and epicardial layers. Among them, 18 were in baseline, 186 had slower conduction in the epicardium layer and 186 in the endocardial layer. Conduction properties were adjusted in such a way that total activation time was the same in all models. P waves on a 16-lead system were simulated during sinus rhythm. Activation maps were similar in all cases. Endo-epicardial delay varied between -5.5 and 5.5 ms vs. 0 ± 0.5 ms in baseline. All P waves had the same duration but variability in their morphology was observed. With slower epicardial conduction, P wave amplitude was reduced by an average of 20% on leads V3-V5 and P wave area decreased by 50% on leads V1-V2 and by 40% on lead V3. Reversed, lower magnitude effects were observed with slower endocardial conduction.

Conclusion

An endo-epicardial delay of a few milliseconds is sufficient to significantly alter P wave morphology, even if the activation map remains the same.

Muscle Thickness and Curvature Influence Atrial Conduction Velocities

Electroanatomical mapping is currently used to provide clinicians with information about the electrophysiological state of the heart and to guide interventions like ablation. These maps can be used to identify ectopic triggers of an arrhythmia such as atrial fibrillation (AF) or changes in the conduction velocity (CV) that have been associated with poor cell to cell coupling or fibrosis. Unfortunately, many factors are known to affect CV, including membrane excitability, pacing rate, wavefront curvature, and bath loading, making interpretation challenging. In this work, we show how endocardial conduction velocities are also affected by the geometrical factors of muscle thickness and wall curvature. Using an idealized three-dimensional strand, we show that transverse conductivities and boundary conditions can slow down or speed up signal propagation, depending on the curvature of the muscle tissue. In fact, a planar wavefront that is parallel to a straight line normal to the mid-surface does not remain normal to the mid-surface in a curved domain. We further demonstrate that the conclusions drawn from the idealized test case can be used to explain spatial changes in conduction velocities in a patient-specific reconstruction of the left atrial posterior wall. The simulations suggest that the widespread assumption of treating atrial muscle as a two-dimensional manifold for electrophysiological simulations will not accurately represent the endocardial conduction velocities in regions of the heart thicker than 0.5 mm with significant wall curvature.

How disruption of endo-epicardial electrical connections enhances endo-epicardial conduction during atrial fibrillation

Aims

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.

Methods and Results

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.

Conclusion

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.

Towards personalized computational modelling of the fibrotic substrate for atrial arrhythmia

: Atrial arrhythmias involving a fibrotic substrate are an important cause of morbidity and mortality. In many cases, effective treatment of such rhythm disorders is severely hindered by a lack of mechanistic understanding relating features of fibrotic remodelling to dynamics of re-entrant arrhythmia. With the advent of clinical imaging modalities capable of resolving the unique fibrosis spatial pattern present in the atria of each individual patient, a promising new research trajectory has emerged in which personalized computational models are used to analyse mechanistic underpinnings of arrhythmia dynamics based on the distribution of fibrotic tissue. In this review, we first present findings that have yielded a robust and detailed biophysical representation of fibrotic substrate electrophysiological properties. Then, we summarize the results of several recent investigations seeking to use organ-scale models of the fibrotic human atria to derive new insights on mechanisms of arrhythmia perpetuation and to develop novel strategies for model-assisted individualized planning of catheter ablation procedures for atrial arrhythmias.

Equivalent dipole sources to estimate the influence of extracellular myocardial anisotropy in thin-walled cardiac forward models

The extracellular domain of the heart is anisotropic, which affects volume conduction and therefore body surface potentials. This paper tests the hypothesis that when wall thickness is sufficiently small (such as in the atria), the effect of extracellular anisotropy can be estimated by modifying local dipole current sources. A formula based on the Gabor-Nelson equivalent dipole and on the reciprocity theorem is derived to compute a linear transformation of the dipole sources that approximates in an isotropic volume conductor the far-field of the actual sources in an anisotropic volume conductor. It involves solving three Poisson equation (once for all). The results obtained in an atrial model embedded in a boundary-element torso model suggest that when wall thickness is < 3  mm, simulated P waves are weakly altered by extracellular anisotropy during sinus rhythm: an anisotropy ratio of 4:1 typically reduced the longitudinal component of the dipole sources by < 3%, increased the transverse component by < 5%, and increased the transmural component by ≈ 25% (which may be relevant in case of epicardial-endocardial dissociation). Due to uncertainty on experimental conductivity values, it is proposed that atrial extracellular anisotropy may be neglected when computing P waves.

Anti-arrhythmic strategies for atrial fibrillation: The role of computational modeling in discovery, development, and optimization

Atrial fibrillation (AF), the most common cardiac arrhythmia, is associated with increased risk of cerebrovascular stroke, and with several other pathologies, including heart failure. Current therapies for AF are targeted at reducing risk of stroke (anticoagulation) and tachycardia-induced cardiomyopathy (rate or rhythm control). Rate control, typically achieved by atrioventricular nodal blocking drugs, is often insufficient to alleviate symptoms. Rhythm control approaches include antiarrhythmic drugs, electrical cardioversion, and ablation strategies. Here, we offer several examples of how computational modeling can provide a quantitative framework for integrating multiscale data to: (a) gain insight into multiscale mechanisms of AF; (b) identify and test pharmacological and electrical therapy and interventions; and (c) support clinical decisions. We review how modeling approaches have evolved and contributed to the research pipeline and preclinical development and discuss future directions and challenges in the field.

Cardiac adipose tissue and atrial fibrillation: the perils of adiposity

The amount of adipose tissue that accumulates around the atria is associated with the risk, persistence, and severity of atrial fibrillation (AF). A strong body of clinical and experimental evidence indicates that this relationship is not an epiphenomenon but is the result of complex crosstalk between the adipose tissue and the neighbouring atrial myocardium. For instance, epicardial adipose tissue is a major source of adipokines, inflammatory cytokines, or reactive oxidative species, which can contribute to the fibrotic remodelling of the atrial myocardium. Fibro-fatty infiltrations of the subepicardium could also contribute to the functional disorganization of the atrial myocardium. The observation that obesity is associated with distinct structural and functional remodelling of the atria has opened new perspectives of treating AF substrate with aggressive risk factor management. Advances in cardiac imaging should lead to an improved ability to visualize myocardial fat depositions and to localize AF substrates.

Modeling left and right atrial contributions to the ECG: A dipole-current source approach

This paper presents the mathematical formulation, the numerical validation and several illustrations of a forward-modeling approach based on dipole-current sources to compute the contribution of a part of the heart to the electrocardiogram (ECG). Clinically relevant applications include identifying in the ECG the contributions from the right and the left atrium. In a Courtemanche-based monodomain computer model of the atria and torso, 1000 dipoles distributed throughout the atrial mid-myocardium are found to be sufficient to reproduce body surface potential maps with a relative error <1% during both sinus rhythm and atrial fibrillation. When the boundary element method is applied to solve the forward problem, this approach enables fast offline computation of the ECG contribution of any anatomical part of the atria by applying the principle of superposition to the dipole sources. In the presence of a right-left activation delay (sinus rhythm), pulmonary vein isolation (sinus rhythm) or left-right differences in refractory period (atrial fibrillation), the decomposition of the ECG is shown to help interpret ECG morphology in relation to the atrial substrate. These tools provide a theoretical basis for a deeper understanding of the genesis of the P wave or fibrillatory waves in normal and pathological cases.

A two layers monodomain model of cardiac electrophysiology of the atria

Numerical simulations of the cardiac electrophysiology in the atria are often based on the standard bidomain or monodomain equations stated on a two-dimensional manifold. These simulations take advantage of the thinness of the atrial tissue, and their computational cost is reduced, as compared to three-dimensional simulations. However, these models do not take into account the heterogeneities located in the thickness of the tissue, like discontinuities of the fiber direction, although they can be a substrate for atrial arrhythmia (Hocini et al., Circulation 105(20):2442-2448, 2002; Ho et al., Cardiovasc Res 54(2):325-336, 2002; Nattel, Nature 415(6868):219-226, 2002). We investigate a two-dimensional model with two coupled, superimposed layers that allows to introduce three-dimensional heterogeneities, but retains a reasonable computational cost. We introduce the mathematical derivation of this model and error estimates with respect to the three-dimensional model. We give some numerical illustrations of its interest: we numerically show its convergence for vanishing thickness, introduce an optimization process of the coupling coefficient and assess its validity on physiologically relevant geometries. Our model would be an efficient tool to test the influence of three-dimensional fiber direction heterogeneities in reentries or atrial arrhythmia without using three-dimensional models.

Quantification of the transmural dynamics of atrial fibrillation by simultaneous endocardial and epicardial optical mapping in an acute sheep model

BACKGROUND

Therapy strategies for atrial fibrillation based on electric characterization are becoming viable personalized medicine approaches to treat a notoriously difficult disease. In light of these approaches that rely on high-density surface mapping, this study aims to evaluate the presence of 3-dimensional electric substrate variations within the transmural wall during acute episodes of atrial fibrillation.

METHODS AND RESULTS

Optical signals were simultaneously acquired from the epicardial and endocardial tissue during acute fibrillation in ovine isolated left atria. Dominant frequency, regularity index, propagation angles, and phase dynamics were assessed and correlated across imaging planes to gauge the synchrony of the activation patterns compared with paced rhythms. Static frequency parameters were well correlated spatially between the endocardium and the epicardium (dominant frequency, 0.79 ± 0.06 and regularity index, 0.93 ± 0.009). However, dynamic tracking of propagation vectors and phase singularity trajectories revealed discordant activity across the transmural wall. The absolute value of the difference in the number, spatial stability, and temporal stability of phase singularities between the epicardial and the endocardial planes was significantly >0 with a median difference of 1.0, 9.27%, and 19.75%, respectively. The number of wavefronts with respect to time was significantly less correlated and the difference in propagation angle was significantly larger in fibrillation compared with paced rhythms.

CONCLUSIONS

Atrial fibrillation substrates are dynamic 3-dimensional structures with a range of discordance between the epicardial and the endocardial tissue. The results of this study suggest that transmural propagation may play a role in atrial fibrillation maintenance mechanisms.

A bilayer representation of the human atria

Atrial fibrillation is the most commonly encountered clinical arrhythmia. Despite recent advances in treatment by catheter ablation, its origin is still incompletely understood and it may be difficult to treat. Computer modelling offers an attractive complement to experiment. Simulations of fibrillation, however, are computationally demanding since the phenomenon requires long periods of observation. Because the atria are thin walled structures, they are often modelled as surfaces. However, this may not always be appropriate as the crista terminalis and pectinate muscles are discrete fibrous structures lying on the endocardium and cannot be incorporated into the surface. In the left atrium, there are essentially two layers with an abrupt change in fibre orientation between them. We propose a double layer method, using shell elements to incorporate wall thickness, where fibre direction is independent in each layer and layers are electrically linked. Starting from human multi-detector CT (MDCT) images, we extracted surfaces for the atria and manually added a coronary sinus. Propagation of electrical activity was modelled with the monodomain equation. Results indicate that major features are retained while reducing computation cost considerably. Meshes based on the two layer approach will facilitate studies of AF.