Beatwise ECG Classification for the Detection of Atrial Fibrillation with Deep Learning.

Atrial fibrillation (AF) is the most common, sustained cardiac arrhythmia. Early intervention and treatment could have a much higher chance of reversing AF. An electrocardiogram (ECG) is widely used to check the heart's rhythm and electrical activity in clinics. The current manual processing of ECGs and clinical classification of AF types (paroxysmal, persistent and permanent AF) is ill-founded and does not truly reflect the seriousness of the disease. In this paper, we proposed a new machine learning method for beat-wise classification of ECGs to estimate AF burden, which was defined by the percentage of AF beats found in the total recording time. Both morphological and temporal features for categorizing AF were extracted via two combined classifiers: a 1D U-Net that evaluates fiducial points and segmentation to locate each heartbeat; and the other Recurrent Neural Network (RNN) to enhance the temporal classification of an individual heartbeat. The output of the classifiers had four target classes: Normal Sinus Rhythm (SN), AF, Noises (NO), and Others (OT). The approach was trained and validated on the Icentia11k dataset, with 1001 and 250 patients' ECGs, respectively. The testing accuracy for the four classes was found to be 0.86, 0.81, 0.79, and 0.75, respectively. Our study demonstrated the feasibility and superior performance of combing U-net and RNN to conduct a beat-wise classification of ECGs for AF burden. However, further investigation is warranted to validate this deep learning approach.Clinical relevance- This paper proposes a novel machine learning network for ECG beatwise classification, specifically for aiding AF burden determination.

Chamber-specific wall thickness features in human atrial fibrillation.

Persistent atrial fibrillation (AF) is not effectively treated due to a lack of adequate tools for identifying patient-specific AF substrates. Recent studies revealed that in 30-50% of patients, persistent AF is maintained by localized drivers not only in the left atrium (LA) but also in the right atrium (RA). The chamber-specific atrial wall thickness (AWT) features underlying AF remain elusive, though the important role of AWT in AF is widely acknowledged. We aimed to provide direct evidence of the existence of distinguished RA and LA AWT features underlying AF drivers by analysing functionally and structurally mapped human hearts ex vivo. Coronary-perfused intact human atria (n = 7, 47 ± 14 y.o.; two female) were mapped using panoramic near-infrared optical mapping during pacing-induced AF. Then the hearts were imaged at approximately 170 µm3 resolution by 9.4 T gadolinium-enhanced MRI. The heart was segmented, and 3D AWT throughout atrial chambers was estimated and analysed. Optical mapping identified six localized RA re-entrant drivers in four hearts and four LA drivers in three hearts. All RA AF drivers were anchored to the pectinate muscle junctions with the crista terminalis or atrial walls. The four LA AF drivers were in the posterior LA. RA (n = 4) with AF drivers were thicker with greater AWT variation than RA (n = 3) without drivers (5.4 ± 2.6 mm versus 5.0 ± 2.4 mm, T-test p < 0.05; F-test p < 0.05). Furthermore, AWT in RA driver regions was thicker and varied more than in RA non-driver regions (5.1 ± 2.5 mm versus 4.4 ± 2.2 mm, T-test p < 0.05; F-test p < 0.05). On the other hand, LA (n = 3) with drivers was thinner than the LA (n = 4) without drivers. In particular, LA driver regions were thinner than the rest of LA regions (3.4 ± 1.0 mm versus 4.2 ± 1.0 mm, T-test p < 0.05). This study demonstrates chamber-specific AWT features of AF drivers. In RA, driver regions are thicker and have more variable AWT than non-driver regions. By contrast, LA drivers are thinner than non-drivers. Robust evaluation of patient-specific AWT features should be considered for chamber-specific targeted ablation.

Mechanistic insight into the functional role of human sinoatrial node conduction pathways and pacemaker compartments heterogeneity: A computer model analysis.

The sinoatrial node (SAN), the primary pacemaker of the heart, is responsible for the initiation and robust regulation of sinus rhythm. 3D mapping studies of the ex-vivo human heart suggested that the robust regulation of sinus rhythm relies on specialized fibrotically-insulated pacemaker compartments (head, center and tail) with heterogeneous expressions of key ion channels and receptors. They also revealed up to five sinoatrial conduction pathways (SACPs), which electrically connect the SAN with neighboring right atrium (RA). To elucidate the role of these structural-molecular factors in the functional robustness of human SAN, we developed comprehensive biophysical computer models of the SAN based on 3D structural, functional and molecular mapping of ex-vivo human hearts. Our key finding is that the electrical insulation of the SAN except SACPs, the heterogeneous expression of If, INa currents and adenosine A1 receptors (A1R) across SAN pacemaker-conduction compartments are required to experimentally reproduce observed SAN activation patterns and important phenomena such as shifts of the leading pacemaker and preferential SACP. In particular, we found that the insulating border between the SAN and RA, is required for robust SAN function and protection from SAN arrest during adenosine challenge. The heterogeneity in the expression of A1R within the human SAN compartments underlies the direction of pacemaker shift and preferential SACPs in the presence of adenosine. Alterations of INa current and fibrotic remodelling in SACPs can significantly modulate SAN conduction and shift the preferential SACP/exit from SAN. Finally, we show that disease-induced fibrotic remodeling, INa suppression or increased adenosine make the human SAN vulnerable to pacing-induced exit blocks and reentrant arrhythmia. In summary, our computer model recapitulates the structural and functional features of the human SAN and can be a valuable tool for investigating mechanisms of SAN automaticity and conduction as well as SAN arrhythmia mechanisms under different pathophysiological conditions.

Mechanisms of pulmonary arterial hypertension-induced atrial fibrillation: insights from multi-scale models of the human atria.

This study aimed to use multi-scale atrial models to investigate pulmonary arterial hypertension (PAH)-induced atrial fibrillation mechanisms. The results of our computer simulations revealed that, at the single-cell level, PAH-induced remodelling led to a prolonged action potential (AP) (ΔAPD: 49.6 ms in the right atria (RA) versus 41.6 ms in the left atria (LA)) and an increased calcium transient (CaT) (ΔCaT: 7.5 × 10-2 µM in the RA versus 0.9 × 10-3 µM in the LA). Moreover, heterogeneous remodelling increased susceptibility to afterdepolarizations, particularly in the RA. At the tissue level, we observed a significant reduction in conduction velocity (CV) (ΔCV: -0.5 m s-1 in the RA versus -0.05 m s-1 in the LA), leading to a shortened wavelength in the RA, but not in the LA. Additionally, afterdepolarizations in the RA contributed to enhanced repolarization dispersion and facilitated unidirectional conduction block. Furthermore, the increased fibrosis in the RA amplified the likelihood of excitation wave breakdown and the occurrence of sustained re-entries. Our results indicated that the RA is characterized by increased susceptibility to afterdepolarizations, slow conduction, reduced wavelength and upregulated fibrosis. These findings shed light on the underlying factors that may promote atrial fibrillation in patients with PAH.

Patchy fibrosis promotes trigger-substrate interactions that both generate and maintain atrial fibrillation.

Fibrosis has been mechanistically linked to arrhythmogenesis in multiple cardiovascular conditions, including atrial fibrillation (AF). Previous studies have demonstrated that fibrosis can create functional barriers to conduction which may promote excitation wavebreak and the generation of re-entry, while also acting to pin re-entrant excitation in stable rotors during AF. However, few studies have investigated the role of fibrosis in the generation of AF triggers in detail. We apply our in-house computational framework to study the impact of fibrosis on the generation of AF triggers and trigger-substrate interactions in two- and three-dimensional atrial tissue models. Our models include a reduced and efficient description of stochastic, spontaneous cellular triggers as well as a simple model of heterogeneous inter-cellular coupling. Our results demonstrate that fibrosis promotes the emergence of focal excitations, primarily through reducing the electrotonic load on individual fibre strands. This enables excitation to robustly initiate within these single strands before spreading to neighbouring strands and inducing a full tissue focal excitation. Enhanced conduction block can allow trigger-substrate interactions that result in the emergence of complex, re-entrant excitation patterns. This study provides new insight into the mechanisms by which fibrosis promotes the triggers and substrate necessary to induce and sustain arrhythmia.

An Adapted Deep Convolutional Neural Network for Automatic Measurement of Pancreatic Fat and Pancreatic Volume in Clinical Multi-Protocol Magnetic Resonance Images: A Retrospective Study with Multi-Ethnic External Validation.

Pancreatic volume and fat fraction are critical prognoses for metabolic diseases like type 2 diabetes (T2D). Magnetic Resonance Imaging (MRI) is a required non-invasive quantification method for the pancreatic fat fraction. The dramatic development of deep learning has enabled the automatic measurement of MR images. Therefore, based on MRI, we intend to develop a deep convolutional neural network (DCNN) that can accurately segment and measure pancreatic volume and fat fraction. This retrospective study involved abdominal MR images from 148 diabetic patients and 246 healthy normoglycemic participants. We randomly separated them into training and testing sets according to the proportion of 80:20. There were 2364 recognizable pancreas images labeled and pre-treated by an upgraded superpixel algorithm for a discernible pancreatic boundary. We then applied them to the novel DCNN model, mimicking the most accurate and latest manual pancreatic segmentation process. Fat phantom and erosion algorithms were employed to increase the accuracy. The results were evaluated by dice similarity coefficient (DSC). External validation datasets included 240 MR images from 10 additional patients. We assessed the pancreas and pancreatic fat volume using the DCNN and compared them with those of specialists. This DCNN employed the cutting-edge idea of manual pancreas segmentation and achieved the highest DSC (91.2%) compared with any reported models. It is the first framework to measure intra-pancreatic fat volume and fat deposition. Performance validation reflected by regression R2 value between manual operation and trained DCNN segmentation on the pancreas and pancreatic fat volume were 0.9764 and 0.9675, respectively. The performance of the novel DCNN enables accurate pancreas segmentation, pancreatic fat volume, fraction measurement, and calculation. It achieves the same segmentation level of experts. With further training, it may well surpass any expert and provide accurate measurements, which may have significant clinical relevance.

How synergy between mechanistic and statistical models is impacting research in atrial fibrillation.

Atrial fibrillation (AF) with multiple complications, high morbidity and mortality, and low cure rates, has become a global public health problem. Although significant progress has been made in the treatment methods represented by anti-AF drugs and radiofrequency ablation, the therapeutic effect is not as good as expected. The reason is mainly because of our lack of understanding of AF mechanisms. This field has benefited from mechanistic and (or) statistical methodologies. Recent renewed interest in digital twin techniques by synergizing between mechanistic and statistical models has opened new frontiers in AF analysis. In the review, we briefly present findings that gave rise to the AF pathophysiology and current therapeutic modalities. We then summarize the achievements of digital twin technologies in three aspects: understanding AF mechanisms, screening anti-AF drugs and optimizing ablation strategies. Finally, we discuss the challenges that hinder the clinical application of the digital twin heart. With the rapid progress in data reuse and sharing, we expect their application to realize the transition from AF description to response prediction.

Non-Contact Intracardiac Potential Mapping Using Mesh-Based and Meshless Inverse Solvers.

Atrial fibrillation (AF) is the most common cardiac dysrhythmia and percutaneous catheter ablation is widely used to treat it. Panoramic mapping with multi-electrode catheters has been used to identify ablation targets in persistent AF but is limited by poor contact and inadequate coverage of the left atrial cavity. In this paper, we investigate the accuracy with which atrial endocardial surface potentials can be reconstructed from electrograms recorded with non-contact catheters. An in-silico approach was employed in which "ground-truth" surface potentials from experimental contact mapping studies and computer models were compared with inverse potential maps constructed by sampling the corresponding intracardiac field using virtual basket catheters. We demonstrate that it is possible to 1) specify the mixed boundary conditions required for mesh-based formulations of the potential inverse problem fully, and 2) reconstruct accurate inverse potential maps from recordings made with appropriately designed catheters. Accuracy improved when catheter dimensions were increased but was relatively stable when the catheter occupied >30% of atrial cavity volume. Independent of this, the capacity of non-contact catheters to resolve the complex atrial potential fields seen in reentrant atrial arrhythmia depended on the spatial distribution of electrodes on the surface bounding the catheter. Finally, we have shown that reliable inverse potential mapping is possible in near real-time with meshless methods that use the Method of Fundamental Solutions.

Intracardiac Inverse Potential Mapping Using the Method of Fundamental Solutions.

Introduction: Atrial fibrillation (AF) is the most prevalent cardiac dysrhythmia and percutaneous catheter ablation is widely used to treat it. Panoramic mapping with multi-electrode catheters can identify ablation targets in persistent AF, but is limited by poor contact and inadequate coverage. Objective: To investigate the accuracy of inverse mapping of endocardial surface potentials from electrograms sampled with noncontact basket catheters. Methods: Our group has developed a computationally efficient inverse 3D mapping technique using a meshless method that employs the Method of Fundamental Solutions (MFS). An in-silico test bed was used to compare ground-truth surface potentials with corresponding inverse maps reconstructed from noncontact potentials sampled with virtual catheters. Ground-truth surface potentials were derived from high-density clinical contact mapping data and computer models. Results: Solutions of the intracardiac potential inverse problem with the MFS are robust, fast and accurate. Endocardial surface potentials can be faithfully reconstructed from noncontact recordings in real-time if the geometry of cardiac surface and the location of electrodes relative to it are known. Larger catheters with appropriate electrode density are needed to resolve complex reentrant atrial rhythms. Conclusion: Real-time panoramic potential mapping is feasible with noncontact intracardiac catheters using the MFS. Significance: Accurate endocardial potential maps can be reconstructed in AF with appropriately designed noncontact multi-electrode catheters.

Identifying locations susceptible to micro-anatomical reentry using a spatial network representation of atrial fibre maps.

Micro-anatomical reentry has been identified as a potential driver of atrial fibrillation (AF). In this paper, we introduce a novel computational method which aims to identify which atrial regions are most susceptible to micro-reentry. The approach, which considers the structural basis for micro-reentry only, is based on the premise that the accumulation of electrically insulating interstitial fibrosis can be modelled by simulating percolation-like phenomena on spatial networks. Our results suggest that at high coupling, where micro-reentry is rare, the micro-reentrant substrate is highly clustered in areas where the atrial walls are thin and have convex wall morphology, likely facilitating localised treatment via ablation. However, as transverse connections between fibres are removed, mimicking the accumulation of interstitial fibrosis, the substrate becomes less spatially clustered, and the bias to forming in thin, convex regions of the atria is reduced, possibly restricting the efficacy of localised ablation. Comparing our algorithm on image-based models with and without atrial fibre structure, we find that strong longitudinal fibre coupling can suppress the micro-reentrant substrate, whereas regions with disordered fibre orientations have an enhanced risk of micro-reentry. With further development, these methods may be useful for modelling the temporal development of the fibrotic substrate on an individualised basis.

Enhancing the detection of atrial fibrillation from wearable sensors with neural style transfer and convolutional recurrent networks.

Electrocardiograms (ECG) provide an effective, non-invasive approach for clinical diagnosis and monitoring treatment in patients with cardiac diseases including the most common cardiac arrhythmia, atrial fibrillation (AF). Portable ECG recording devices including Apple Watch and Kardia devices have been developed for AF detection. However, the efficacy of these smart devices has not been fully validated. We aimed to develop an open-source deep learning framework for automatic AF detection using the largest publicly available single-lead ECG dataset through a mobile Kardia device enhanced with style transfer-driven data augmentation. We developed and validated a 37-layer convolutional recurrent network (CRN) using 5,834 single-lead ECGs with a mean length of 30 seconds from the 2017 PhysioNet Challenge to automatically detect sinus rhythm and AF. To address the challenge of a lack of a large number of AF samples, we proposed a novel style transfer generator that fuses patient-specific clinical ECGs and mathematically modelled ECG features to synthesize realistic ECGs by five-fold. The differences between synthesized and clinical ECGs were analyzed by studying their average ECG morphologies and frequency distributions. Our results indicated the style transfer-driven data augmentation was not classifier-dependent. Validation on 2,917 clinical ECGs showed an F1 score of 96.4%, with the generated ECGs contributing to a 3% improvement in AF detection for the Kardia event recorder. By developing and evaluating our approach on an open-source ECG dataset, we have demonstrated that our framework is both robust and verifiable, and potentially can be used in portable devices for effective AF classification.

Automatic 3D Surface Reconstruction of the Left Atrium From Clinically Mapped Point Clouds Using Convolutional Neural Networks.

Point clouds are a widely used format for storing information in a memory-efficient and easily manipulatable representation. However, research in the application of point cloud mapping and subsequent organ reconstruction with deep learning, is limited. In particular, current methods for left atrium (LA) visualization using point clouds recorded from clinical mapping during cardiac ablation are proprietary and remain difficult to validate. Many clinics rely on additional imaging such as MRIs/CTs to improve the accuracy of LA mapping. In this study, for the first time, we proposed a novel deep learning framework for the automatic 3D surface reconstruction of the LA directly from point clouds acquired via widely used clinical mapping systems. The backbone of our framework consists of a 30-layer 3D fully convolutional neural network (CNN). The architecture contains skip connections that perform multi-resolution processing to maximize information extraction from the point clouds and ensure a high-resolution prediction by combining features at different receptive levels. We used large kernels with increased receptive fields to address the sparsity of the point clouds. Residual blocks and activation normalization were further implemented to improve the feature learning on sparse inputs. By utilizing a light-weight design with low-depth layers, our CNN took approximately 10 s per patient. Independent testing on two cross-modality clinical datasets showed excellent dice scores of 93% and surface-to-surface distances below 1 pixel. Overall, our study may provide a more efficient, cost-effective 3D LA reconstruction approach during ablation procedures, and potentially lead to improved treatment of cardiac diseases.

Modulated Calcium Homeostasis and Release Events Under Atrial Fibrillation and Its Risk Factors: A Meta-Analysis.

Background: Atrial fibrillation (AF) is associated with calcium (Ca2+) handling remodeling and increased spontaneous calcium release events (SCaEs). Nevertheless, its exact mechanism remains unclear, resulting in suboptimal primary and secondary preventative strategies. Methods: We searched the PubMed database for studies that investigated the relationship between SCaEs and AF and/or its risk factors. Meta-analysis was used to examine the Ca2+ mechanisms involved in the primary and secondary AF preventative groups. Results: We included a total of 74 studies, out of the identified 446 publications from inception (1982) until March 31, 2020. Forty-five were primary and 29 were secondary prevention studies for AF. The main Ca2+ release events, calcium transient (standardized mean difference (SMD) = 0.49; I 2 = 35%; confidence interval (CI) = 0.33-0.66; p < 0.0001), and spark amplitude (SMD = 0.48; I 2 = 0%; CI = -0.98-1.93; p = 0.054) were enhanced in the primary diseased group, while calcium transient frequency was increased in the secondary group. Calcium spark frequency was elevated in both the primary diseased and secondary AF groups. One of the key cardiac currents, the L-type calcium current (ICaL) was significantly downregulated in primary diseased (SMD = -1.07; I 2 = 88%; CI = -1.94 to -0.20; p < 0.0001) and secondary AF groups (SMD = -1.28; I 2 = 91%; CI = -2.04 to -0.52; p < 0.0001). Furthermore, the sodium-calcium exchanger (INCX) and NCX1 protein expression were significantly enhanced in the primary diseased group, while only NCX1 protein expression was shown to increase in the secondary AF studies. The phosphorylation of the ryanodine receptor at S2808 (pRyR-S2808) was significantly elevated in both the primary and secondary groups. It was increased in the primary diseased and proarrhythmic subgroups (SMD = 0.95; I 2 = 64%; CI = 0.12-1.79; p = 0.074) and secondary AF group (SMD = 0.66; I 2 = 63%; CI = 0.01-1.31; p < 0.0001). Sarco/endoplasmic reticulum Ca2+-ATPase (SERCA) expression was elevated in the primary diseased and proarrhythmic drug subgroups but substantially reduced in the secondary paroxysmal AF subgroup. Conclusions: Our study identified that ICaL is reduced in both the primary and secondary diseased groups. Furthermore, pRyR-S2808 and NCX1 protein expression are enhanced. The remodeling leads to elevated Ca2+ functional activities, such as increased frequencies or amplitude of Ca2+ spark and Ca2+ transient. The main difference identified between the primary and secondary diseased groups is SERCA expression, which is elevated in the primary diseased group and substantially reduced in the secondary paroxysmal AF subgroup. We believe our study will add new evidence to AF mechanisms and treatment targets.

A novel system for mapping regional electrical properties and characterizing arrhythmia in isolated intact rat atria.

Detailed global maps of atrial electrical activity are needed to understand mechanisms of atrial rhythm disturbance in small animal models of heart disease. To date, optical mapping systems have not provided enough spatial resolution across sufficiently extensive regions of intact atrial preparations to achieve this goal. The aim of this study was to develop an integrated platform for quantifying regional electrical properties and analyzing reentrant arrhythmia in a biatrial preparation. Intact atria from 6/7-mo-old female spontaneously hypertensive rats (SHRs; n = 6) were isolated and secured in a constant flow superfusion chamber at 37°C. Optical mapping was performed with the membrane-voltage dye di-4-ANEPPS using LED excitation and a scientific complementary metal-oxide semiconductor (sCMOS) camera. Programmed stimulus trains were applied from right atrial (RA) and left atrial (LA) sites to assess rate-dependent electrical behavior and to induce atrial arrhythmia. Signal-to-noise ratio was improved by sequential processing steps that included spatial smoothing, temporal filtering, and, in stable rhythms, ensemble-averaging. Activation time, repolarization time, and action potential duration (APD) maps were constructed at high spatial resolution for a wide range of coupling intervals. These data were highly consistent within and between experiments. They confirmed preferential atrial conduction pathways and demonstrated distinct medial-to-lateral APD gradients. We also showed that reentrant arrhythmias induced in this preparation were explained by the spatial variation of these electrical properties. Our new methodology provides a robust means of 1) quantifying regional electrical properties in the intact rat atria at higher spatiotemporal resolution than previously reported, and 2) characterizing reentrant arrhythmia and analyzing mechanisms that give rise to it.NEW & NOTEWORTHY Despite wide-ranging optical mapping studies, detailed information on regional atrial electrical properties in small animal models of heart disease and how these contribute to reentrant arrhythmia remains limited. We have developed a novel experimental platform that enables both to be achieved in a geometrically intact isolated rat bi-atrial preparation.

Understanding PITX2-Dependent Atrial Fibrillation Mechanisms through Computational Models.

Atrial fibrillation (AF) is a common arrhythmia. Better prevention and treatment of AF are needed to reduce AF-associated morbidity and mortality. Several major mechanisms cause AF in patients, including genetic predispositions to AF development. Genome-wide association studies have identified a number of genetic variants in association with AF populations, with the strongest hits clustering on chromosome 4q25, close to the gene for the homeobox transcription PITX2. Because of the inherent complexity of the human heart, experimental and basic research is insufficient for understanding the functional impacts of PITX2 variants on AF. Linking PITX2 properties to ion channels, cells, tissues, atriums and the whole heart, computational models provide a supplementary tool for achieving a quantitative understanding of the functional role of PITX2 in remodelling atrial structure and function to predispose to AF. It is hoped that computational approaches incorporating all we know about PITX2-related structural and electrical remodelling would provide better understanding into its proarrhythmic effects leading to development of improved anti-AF therapies. In the present review, we discuss advances in atrial modelling and focus on the mechanistic links between PITX2 and AF. Challenges in applying models for improving patient health are described, as well as a summary of future perspectives.

Toward Patient-Specific Prediction of Ablation Strategies for Atrial Fibrillation Using Deep Learning.

Atrial fibrillation (AF) is a common cardiac arrhythmia that affects 1% of the population worldwide and is associated with high levels of morbidity and mortality. Catheter ablation (CA) has become one of the first line treatments for AF, but its success rates are suboptimal, especially in the case of persistent AF. Computational approaches have shown promise in predicting the CA strategy using simulations of atrial models, as well as applying deep learning to atrial images. We propose a novel approach that combines image-based computational modelling of the atria with deep learning classifiers trained on patient-specific atrial models, which can be used to assist in CA therapy selection. Therefore, we trained a deep convolutional neural network (CNN) using a combination of (i) 122 atrial tissue images obtained by unfolding patient LGE-MRI datasets, (ii) 157 additional synthetic images derived from the patient data to enhance the training dataset, and (iii) the outcomes of 558 CA simulations to terminate several AF scenarios in the corresponding image-based atrial models. Four CNN classifiers were trained on this patient-specific dataset balanced using several techniques to predict three common CA strategies from the patient atrial images: pulmonary vein isolation (PVI), rotor-based ablation (Rotor) and fibrosis-based ablation (Fibro). The training accuracy for these classifiers ranged from 96.22 to 97.69%, while the validation accuracy was from 78.68 to 86.50%. After training, the classifiers were applied to predict CA strategies for an unseen holdout test set of atrial images, and the results were compared to outcomes of the respective image-based simulations. The highest success rate was observed in the correct prediction of the Rotor and Fibro strategies (100%), whereas the PVI class was predicted in 33.33% of the cases. In conclusion, this study provides a proof-of-concept that deep neural networks can learn from patient-specific MRI datasets and image-derived models of AF, providing a novel technology to assist in tailoring CA therapy to a patient.

Mechanisms underlying pro-arrhythmic abnormalities arising from Pitx2-induced electrical remodelling: an in silico intersubject variability study.

BACKGROUND: Electrical remodelling as a result of the homeodomain transcription factor 2 (Pitx2)-dependent gene regulation induces atrial fibrillation (AF) with different mechanisms. The purpose of this study was to identify Pitx2-induced changes in ionic currents that cause action potential (AP) shortening and lead to triggered activity. METHODS: Populations of computational atrial AP models were developed based on AP recordings from sinus rhythm (SR) and AF patients. Models in the AF population were divided into triggered and untriggered AP groups to evaluate the relationship between each ion current regulated by Pitx2 and triggered APs. Untriggered AP models were then divided into shortened and unshortened AP groups to determine which Pitx2-dependent ion currents contribute to AP shortening. RESULTS: According to the physiological range of AP biomarkers measured experimentally, populations of 2,885 SR and 4,781 AF models out of the initial pool of 30,000 models were selected. Models in the AF population predicted AP shortening and triggered activity observed in experiments in Pitx2-induced remodelling conditions. The AF models included 925 triggered AP models, 1,412 shortened AP models and 2,444 unshortened AP models. Intersubject variability in IKs and ICaL primarily modulated variability in AP duration (APD) in all shortened and unshortened AP models, whereas intersubject variability in IK1 and SERCA mainly contributed to the variability in AP morphology in all triggered and untriggered AP models. The incidence of shortened AP was positively correlated with IKs and IK1 and was negatively correlated with INa , ICaL and SERCA, whereas the incidence of triggered AP was negatively correlated with IKs and IK1 and was positively correlated with INa , ICaL and SERCA. CONCLUSIONS: Electrical remodelling due to Pitx2 upregulation may increase the incidence of shortened AP, whereas electrical remodelling arising from Pitx2 downregulation may favor to the genesis of triggered AP.