Search

Sudden cardiac death from arrhythmia is a major cause of mortality worldwide. In this study, we developed a novel deep learning (DL) approach that blends neural networks and survival analysis to predict patient-specific survival curves from contrast-enhanced cardiac magnetic resonance images and clinical covariates for patients with ischemic heart disease. The DL-predicted survival curves offer accurate predictions at times up to 10 years and allow for estimation of uncertainty in predictions. The performance of this learning architecture was evaluated on multi-center internal validation data and tested on an independent test set, achieving concordance indexes of 0.83 and 0.74 and 10-year integrated Brier scores of 0.12 and 0.14. We demonstrate that our DL approach, with only raw cardiac images as input, outperforms standard survival models constructed using clinical covariates. This technology has the potential to transform clinical decision-making by offering accurate and generalizable predictions of patient-specific survival probabilities of arrhythmic death over time.

Background: The ability to increase heart rate during exercise and other stressors is a key homeostatic feature of the sinoatrial node (SAN). When the physiological heart rate response is blunted, chronotropic incompetence limits exercise capacity, a common problem in patients with heart failure with preserved ejection fraction (HFpEF). Despite its clinical relevance, the mechanisms of chronotropic incompetence remain unknown. Methods: Dahl salt-sensitive rats fed a high-salt diet and C57Bl6 mice fed a high-fat diet and an inhibitor of constitutive nitric oxide synthase (Nω-nitro-L-arginine methyl ester [L-NAME]; 2-hit) were used as models of HFpEF. Myocardial infarction was created to induce HF with reduced ejection fraction. Rats and mice fed with a normal diet or those that had a sham surgery served as respective controls. A comprehensive characterization of SAN function and chronotropic response was conducted by in vivo, ex vivo, and single-cell electrophysiologic studies. RNA sequencing of SAN was performed to identify transcriptomic changes. Computational modeling of biophysically-detailed human HFpEF SAN was created. Results: Rats with phenotypically-verified HFpEF exhibited limited chronotropic response associated with intrinsic SAN dysfunction, including impaired β-adrenergic responsiveness and an alternating leading pacemaker within the SAN. Prolonged SAN recovery time and reduced SAN sensitivity to isoproterenol were confirmed in the 2-hit mouse model. Adenosine challenge unmasked conduction blocks within the SAN, which were associated with structural remodeling. Chronotropic incompetence and SAN dysfunction were also found in rats with HF with reduced ejection fraction. Single-cell studies and transcriptomic profiling revealed HFpEF-related alterations in both the “membrane clock” (ion channels) and the “Ca 2+ clock” (spontaneous Ca 2+ release events). The physiologic impairments were reproduced in silico by empirically-constrained quantitative modeling of human SAN function. Conclusions: Chronotropic incompetence and SAN dysfunction were seen in both models of HF. We identified that intrinsic abnormalities of SAN structure and function underlie the chronotropic response in HFpEF.

Accurate quantification of left atrium (LA) scar in patients with atrial fibrillation is essential to guide successful ablation strategies. Prior to LA scar quantification, a proper LA cavity segmentation is required to ensure exact location of scar. Both tasks can be extremely time-consuming and are subject to inter-observer disagreements when done manually. We developed and validated a deep neural network to automatically segment the LA cavity and the LA scar. The global architecture uses a multi-network sequential approach in two stages which segment the LA cavity and the LA Scar. Each stage has two steps: a region of interest Neural Network and a refined segmentation network. We analysed the performances of our network according to different parameters and applied data triaging. 200+ late gadolinium enhancement magnetic resonance images were provided by the LAScarQS 2022 Challenge. Finally, we compared our performances for scar quantification to the literature and demonstrated improved performances.

Despite the global COVID-19 pandemic, during the past 2 years, there have been numerous advances in our understanding of arrhythmia mechanisms and diagnosis and in new therapies. We increased our understanding of risk factors and mechanisms of atrial arrhythmias, the prediction of atrial arrhythmias, response to treatment, and outcomes using machine learning and artificial intelligence. There have been new technologies and techniques for atrial fibrillation ablation, including pulsed field ablation. There have been new randomized trials in atrial fibrillation ablation, giving insight about rhythm control, and long-term outcomes. There have been advances in our understanding of treatment of inherited disorders such as catecholaminergic polymorphic ventricular tachycardia. We have gained new insights into the recurrence of ventricular arrhythmias in the setting of various conditions such as myocarditis and inherited cardiomyopathic disorders. Novel computational approaches may help predict occurrence of ventricular arrhythmias and localize arrhythmias to guide ablation. There are further advances in our understanding of noninvasive radiotherapy. We have increased our understanding of the role of His bundle pacing and left bundle branch area pacing to maintain synchronous ventricular activation. There have also been significant advances in the defibrillators, cardiac resynchronization therapy, remote monitoring, and infection prevention. There have been advances in our understanding of the pathways and mechanisms involved in atrial and ventricular arrhythmogenesis.

The field of cardiac electrophysiology (EP) had adopted simple artificial intelligence (AI) methodologies for decades. Recent renewed interest in deep learning techniques has opened new frontiers in electrocardiography analysis including signature identification of diseased states. Artificial intelligence advances coupled with simultaneous rapid growth in computational power, sensor technology, and availability of web-based platforms have seen the rapid growth of AI-aided applications and big data research. Changing lifestyles with an expansion of the concept of internet of things and advancements in telecommunication technology have opened doors to population-based detection of atrial fibrillation in ways, which were previously unimaginable. Artificial intelligence-aided advances in 3D cardiac imaging heralded the concept of virtual hearts and the simulation of cardiac arrhythmias. Robotics, completely non-invasive ablation therapy, and the concept of extended realities show promise to revolutionize the future of EP. In this review, we discuss the impact of AI and recent technological advances in all aspects of arrhythmia care., Graphical Abstract Artificial intelligence-enhanced arrhythmia care.

Atrial fibrillation (AF) is the most common human arrhythmia, affecting millions of patients worldwide. A combination of risk factors and comorbidities results in complex atrial remodeling, which increases AF vulnerability and persistence. Insights from animal models, clinical studies, and computational modeling have advanced the understanding of the mechanisms and pathophysiology of AF. Areas of heterogeneous pathological remodeling, as well as altered electrophysiological properties, serve as a substrate for AF drivers and spontaneous activations. The complex and individualized presentation of this arrhythmia suggests that mechanisms-based personalized approaches will likely be needed to overcome current challenges in AF management. In this paper, we review the insights on the mechanisms of AF obtained from animal models and clinical studies and how computational models integrate this knowledge to advance AF clinical management. We also assess the challenges that need to be overcome to implement these mechanistic models in clinical practice.

Despite significant advances in its detection, understanding and management, atrial fibrillation (AF) remains a highly prevalent cardiac arrhythmia with a major impact on morbidity and mortality of millions of patients. AF results from complex, dynamic interactions between risk factors and comorbidities that induce diverse atrial remodelling processes. Atrial remodelling increases AF vulnerability and persistence, while promoting disease progression. The variability in presentation and wide range of mechanisms involved in initiation, maintenance and progression of AF, as well as its associated adverse outcomes, make the early identification of causal factors modifiable with therapeutic interventions challenging, likely contributing to suboptimal efficacy of current AF management. Computational modelling facilitates the multilevel integration of multiple datasets and offers new opportunities for mechanistic understanding, risk prediction and personalized therapy. Mathematical simulations of cardiac electrophysiology have been around for 60 years and are being increasingly used to improve our understanding of AF mechanisms and guide AF therapy. This narrative review focuses on the emerging and future applications of computational modelling in AF management. We summarize clinical challenges that may benefit from computational modelling, provide an overview of the different in silico approaches that are available together with their notable achievements, and discuss the major limitations that hinder the routine clinical application of these approaches. Finally, future perspectives are addressed. With the rapid progress in electronic technologies including computing, clinical applications of computational modelling are advancing rapidly. We expect that their application will progressively increase in prominence, especially if their added value can be demonstrated in clinical trials., Graphical Abstract

Aims Direct remuscularization with pluripotent stem cell-derived cardiomyocytes (PSC-CMs) seeks to address the onset of heart failure post-myocardial infarction (MI) by treating the persistent muscle deficiency that underlies it. However, direct remuscularization with PSC-CMs could potentially be arrhythmogenic. We investigated two possible mechanisms of arrhythmogenesis-focal vs reentrant-arising from direct remuscularization with PSC-CM patches in two personalized, human ventricular computer models of post-MI. Moreover, we developed a principled approach for evaluating arrhythmogenicity of direct remuscularization that factors in the VT propensity of the patient-specific post-MI fibrotic substrate and use it to investigate different conditions of patch remuscularization. Methods & results Two personalized, human ventricular models of post-MI (P1 & P2) were constructed from late gadolinium enhanced (LGE)-magnetic resonance images (MRI). In each model, remuscularization with PSC-CM patches were simulated under different treatment conditions that included patch engraftment, patch myofibril orientation, remuscularization site, patch size (thickness and diameter), and patch maturation. To determine arrhythmogenicity of treatment conditions, VT burden of heart models was quantified prior to and after simulated remuscularization and compared. VT burden was quantified based on inducibility (i.e., weighted sum of pacing sites that induced) and severity (i.e., the number of distinct VT morphologies induced). Prior to remuscularization, VT burden was significant in P1 (0.275) and not in P2 (0.0, not VT inducible). We highlight that reentrant VT mechanisms would dominate over focal mechanisms; spontaneous beats emerging from PSC-CM grafts were always a fraction of resting sinus rate. Moreover, incomplete patch engraftment can be particularly arrhythmogenic, giving rise to particularly aberrant electrical activation and conduction slowing across the PSC-CM patches along with elevated VT burden when compared to complete engraftment. Under conditions of complete patch engraftment, remuscularization was almost always arrhythmogenic in P2 but certain treatment conditions could be anti-arrhythmogenic in P1. Moreover, the remuscularization site was the most important factor affecting VT burden in both P1 and P2. Complete maturation of PSC-CM patches, both ionically and electrotonically, at the appropriate site could completely alleviate VT burden. Conclusion We identified that reentrant VT would be the primary VT mechanism in patch remuscularization. To evaluate the arrhythmogenicity of remuscularization, we developed a principled approach that factors in the propensity of the patient-specific fibrotic substrate for VT. We showed that arrhythmogenicity is sensitive to the patient-specific fibrotic substrate and remuscularization site. We demonstrate that targeted remuscularization can be safe in the appropriate individual and holds the potential to nondestructively eliminate VT post-MI in addition to addressing muscle deficiency underlying heart failure progression. Translational perspective If safety from ventricular arrhythmias can be addressed, direct remuscularization with PSC-CMs-achieved either through engineered myocardial patches or intramyocardial injections-holds the potential to halt heart failure progression post-MI. Using personalized 3 D models of the post-MI ventricles derived from LGE-MRI, we provide evidence that arrhythmogenesis following remuscularization with PSC-CM patches is driven by a reentrant as opposed to focal VT mechanism. Moreover, the existing patient-specific fibrotic substrate together with the remuscularization site were primary determinants of arrhythmogenesis. These results suggest that the clinical safety of remuscularization can be achieved through patient-specific optimization guided in-part by computational modeling.

Introduction We recently developed two noninvasive methodologies to help guide VT ablation: population-derived automated VT exit localization (PAVEL) and virtual-heart arrhythmia ablation targeting (VAAT). We hypothesized that while very different in their nature, limitations, and type of ablation targets (substrate-based vs. clinical VT), the image-based VAAT and the ECG-based PAVEL technologies would be spatially concordant in their predictions. Objective The objective is to test this hypothesis in ischemic cardiomyopathy patients in a retrospective feasibility study. Methods Four post-infarct patients who underwent LV VT ablation and had pre-procedural LGE-CMRs were enrolled. Virtual hearts with patient-specific scar and border zone identified potential VTs and ablation targets. Patient-specific PAVEL based on a population-derived statistical method localized VT exit sites onto a patient-specific 238-triangle LV endocardial surface. Results Ten induced VTs were analyzed and 9-exit sites were localized by PAVEL onto the patient-specific LV endocardial surface. All nine predicted VT exit sites were in the scar border zone defined by voltage mapping and spatially correlated with successful clinical lesions. There were 2.3 ± 1.9 VTs per patient in the models. All five VAAT lesions fell within regions ablated clinically. VAAT targets correlated well with 6 PAVEL-predicted VT exit sites. The distance between the center of the predicted VT-exit-site triangle and nearest corresponding VAAT ablation lesion was 10.7 ± 7.3 mm. Conclusions VAAT targets are concordant with the patient-specific PAVEL-predicted VT exit sites. These findings support investigation into combining these two complementary technologies as a noninvasive, clinical tool for targeting clinically induced VTs and regions likely to harbor potential VTs.

Pulmonary hypertension is a complex disorder characterized by pulmonary vascular remodeling and right ventricular hypertrophy, leading to right heart failure. The mechanisms underlying this process are not well understood. We hypothesize that the structural remodeling occurring in the cardiomyocytes of the right ventricle affects the cytosolic Ca 2+ handling leading to arrhythmias. After 12 days of monocrotaline-induced pulmonary hypertension in rats, epicardial mapping showed electrical remodeling in both ventricles. In myocytes isolated from the hypertensive rats, a combination of high-speed camera and confocal line-scan documented a prolongation of Ca 2+ transients along with a higher local Ca 2+ -release activity. These Ca 2+ transients were less synchronous than in controls, likely due to disorganized transverse-axial tubular system. In fact, following pulmonary hypertension, hypertrophied right ventricular myocytes showed significantly reduced number of transverse tubules and increased number of axial tubules; however, Stimulation Emission Depletion microscopy demonstrated that the colocalization of L-type Ca 2+ channels and RyR2 (ryanodine receptor 2) remained unchanged. Finally, Stimulation Emission Depletion microscopy and super-resolution scanning patch-clamp analysis uncovered a decrease in the density of active L-type Ca 2+ channels in right ventricular myocytes with an elevated open probability of the T-tubule anchored channels. This may represent a general mechanism of how nanoscale structural changes at the early stage of pulmonary hypertension impact on the development of the end stage failing phenotype in the right ventricle.

Gray zone, a measure of tissue heterogeneity on late gadolinium enhanced-cardiac magnetic resonance (LGE-CMR) imaging, has been shown to predict ventricular arrhythmias (VAs) in ischemic cardiomyopathy (ICM) patients. However, no studies have described whether left ventricular (LV) tissue heterogeneity and intramyocardial fat mass on contrast-enhanced computed tomography (CE-CT), which provides greater spatial resolution, is useful for assessing the risk of VAs in ICM patients with LV systolic dysfunction and no previous VAs.The purpose of this proof-of-concept study was to determine the feasibility of measuring global LV tissue heterogeneity and intramyocardial fat mass by CE-CT for predicting the risk of VAs in ICM patients with LV systolic dysfunction and no previous history of VAs.Patients with left ventricular ejection fraction ≤35% and no previous VAs were enrolled in a prospective, observational registry and underwent LGE-CMR. From this cohort, patients with ICM who additionally received CE-CT were included in the present analysis. Gray zone on LGE-CMR was defined as myocardium with signal intensity (SI)peak SI of healthy myocardium but50% maximal SI. Tissue heterogeneity on CE-CT was defined as the standard deviation of the Hounsfield unit image gradients (HU/mm) within the myocardium. Intramyocardial fat on CE-CT was identified as regions of image pixels between -180 and -5 HU. The primary outcome was VAs, defined as appropriate implantable cardioverter-defibrillator shock or sudden arrhythmic death.The study consisted of 47 ICM patients, 13 (27.7%) of whom experienced VA events during mean follow-up of 5.6 ± 3.4 years. Increasing tissue heterogeneity (per HU/mm) was significantly associated with VAs after multivariable adjustment, including for gray zone (odds ratio [OR] 1.22;In ICM patients, CE-CT-derived LV tissue heterogeneity was independently associated with VAs and may represent a novel marker useful for risk stratification.

Developing highly efficient non-viral gene delivery reagents is still difficult for many hard-to-transfect cell types and, to date, has mostly been conducted via brute force screening routines. High throughput in silico methods of evaluating biomaterials can enable accelerated optimization and development of devices or therapeutics by exploring large chemical design spaces quickly and at low cost. This work reports application of state-of-the-art machine learning algorithms to a dataset of synthetic biodegradable polymers, poly(beta-amino ester)s (PBAEs), which have shown exciting promise for therapeutic gene delivery in vitro and in vivo. The data set includes polymer properties as inputs as well as polymeric nanoparticle transfection performance and nanoparticle toxicity in a range of cells as outputs. This data was used to train and evaluate several state-of-the-art machine learning algorithms for their ability to predict transfection and understand structure-function relationships. By developing an encoding scheme for vectorizing the structure of a PBAE polymer in a machine-readable format, we demonstrate that a random forest model can satisfactorily predict DNA transfection in vitro based on the chemical structure of the constituent PBAE polymer in a cell line dependent manner. Based on the model, we synthesized PBAE polymers and used them to form polymeric gene delivery nanoparticles that were predicted in silico to be successful. We validated the computational predictions in two cell lines in vitro, RAW 264.7 macrophages and Hep3B liver cancer cells, and found that the Spearman's R correlation between predicted and experimental transfection was 0.57 and 0.66 respectively. Thus, a computational approach that encoded chemical descriptors of polymers was able to demonstrate that in silico computational screening of polymeric nanomedicine compositions had utility in predicting de novo biological experiments. STATEMENT OF SIGNIFICANCE: Developing highly efficient non-viral gene delivery reagents is difficult for many hard-to-transfect cell types and, to date, has mostly been explored via brute force screening routines. High throughput in silico methods of evaluating biomaterials can enable accelerated optimization and development for therapeutic or biomanufacturing purposes by exploring large chemical design spaces quickly and at low cost. This work reports application of state-of-the-art machine learning algorithms to a large compiled PBAE DNA gene delivery nanoparticle dataset across many cell types to develop predictive models for transfection and nanoparticle cytotoxicity. We develop a novel computational pipeline to encode PBAE nanoparticles with chemical descriptors and demonstrate utility in a de novo experimental context.

Aims Multiple wavefront pacing (MWP) and decremental pacing (DP) are two electroanatomic mapping (EAM) strategies that have emerged to better characterize the ventricular tachycardia (VT) substrate. The aim of this study was to assess how well MWP, DP, and their combination improve identification of electrophysiological abnormalities on EAM that reflect infarct remodelling and critical VT sites. Methods and results Forty-eight personalized computational heart models were reconstructed using images from post-infarct patients undergoing VT ablation. Paced rhythms were simulated by delivering an initial (S1) and an extra-stimulus (S2) from one of 100 locations throughout each heart model. For each pacing, unipolar signals were computed along the myocardial surface to simulate substrate EAM. Six EAM features were extracted and compared with the infarct remodelling and critical VT sites. Concordance of S1 EAM features between different maps was lower in hearts with smaller amounts of remodelling. Incorporating S1 EAM features from multiple maps greatly improved the detection of remodelling, especially in hearts with less remodelling. Adding S2 EAM features from multiple maps decreased the number of maps required to achieve the same detection accuracy. S1 EAM features from multiple maps poorly identified critical VT sites. However, combining S1 and S2 EAM features from multiple maps paced near VT circuits greatly improved identification of critical VT sites. Conclusion Electroanatomic mapping with MWP is more advantageous for characterization of substrate in hearts with less remodelling. During substrate EAM, MWP and DP should be combined and delivered from locations proximal to a suspected VT circuit to optimize identification of the critical VT site.

Background Adults with repaired tetralogy of Fallot (rTOF) are at increased risk for ventricular tachycardia (VT) due to fibrotic remodeling of the myocardium. However, the current clinical guidelines for VT risk stratification and subsequent implantable cardioverter-defibrillator deployment for primary prevention of sudden cardiac death in rTOF remain inadequate. Objective The purpose of this study was to determine the feasibility of using an rTOF-specific virtual-heart approach to identify patients stratified incorrectly as being at low VT risk by current clinical criteria. Methods This multicenter retrospective pilot study included 7 adult rTOF patients who were considered low risk for VT based on clinical criteria. Patient-specific computational heart models were generated from late gadolinium enhanced magnetic resonance imaging (LGE-MRI), incorporating the individual distribution of rTOF fibrotic remodeling in both ventricles. Simulations of rapid pacing determined VT inducibility. Model creation and simulations were performed by operators blinded to clinical outcome. Results Two patients in the study experienced clinical VT. The virtual hearts constructed from LGE-MRI scans of 7 rTOF patients correctly predicted reentrant VT in the models from VT-positive patients and no arrhythmia in those from VT-negative patients. There were no statistically significant differences in clinical criteria commonly used to assess VT risk, including QRS duration and age, between patients who did and those who did not experience clinical VT. Conclusion This study demonstrates the feasibility of image-based virtual-heart modeling in patients with congenital heart disease and structurally abnormal hearts. It highlights the potential of the methodology to improve VT risk stratification in patients with rTOF.

Background Visualizing fibrosis on cardiac magnetic resonance (CMR) imaging with contrast enhancement (LGE) is paramount in characterizing disease progression and identifying arrhythmia substrates. Segmentation and fibrosis quantification from LGE-CMR is intensive, manual, and prone to inter-observer variability. There is an unmet need for automated LGE-CMR image segmentation that ensures anatomical accuracy and seamless extraction of clinical features. Objective This study aimed to develop a novel deep learning solution for analysis of contrast-enhanced CMR images which produces anatomically accurate myocardium and scar/fibrosis segmentations and uses these to calculate features of clinical interest. Methods Data sources were 155 2-D LGE-CMR patient scans (1,124 slices) and 246 synthetic “LGE-like” scans (1, 360 slices) obtained from cine CMR using a novel style-transfer algorithm. We trained and tested a 3-stage neural network which identified the left ventricle (LV) region of interest (ROI), segmented ROI into viable myocardium and regions of enhancement, and post-processed the segmentations results to enforce conforming to anatomical constraints. The segmentations were used to directly compute clinical features, such as LV volume and scar burden. Results Predicted LV and scar segmentations achieved 96% and 75% balanced accuracy, respectively, and 0.93 and 0.57 Dice coefficient when compared to trained expert segmentations. The mean scar burden difference between manual and predicted segmentations was 2%. Conclusions We developed and validated a deep neural network for automatic, anatomically accurate expert-level LGE- CMR myocardium and scar/fibrosis segmentation, allowing direct calculation of clinical measures. Given the training set heterogeneity, our approach could be extended to multiple imaging modalities and patient pathologies.

Hypertrophic cardiomyopathy (HCM) is associated with risk of sudden cardiac death (SCD) due to ventricular arrhythmias (VAs) arising from the proliferation of fibrosis in the heart. Current clinical risk stratification criteria inadequately identify at-risk patients in need of primary prevention of VA. Here, we use mechanistic computational modeling of the heart to analyze how HCM-specific remodeling promotes arrhythmogenesis and to develop a personalized strategy to forecast risk of VAs in these patients. We combine contrast-enhanced cardiac magnetic resonance imaging and T1 mapping data to construct digital replicas of HCM patient hearts that represent the patient-specific distribution of focal and diffuse fibrosis and evaluate the substrate propensity to VA. Our analysis indicates that the presence of diffuse fibrosis, which is rarely assessed in these patients, increases arrhythmogenic propensity. In forecasting future VA events in HCM patients, the imaging-based computational heart approach achieved 84.6%, 76.9%, and 80.1% sensitivity, specificity, and accuracy, respectively, and significantly outperformed current clinical risk predictors. This novel VA risk assessment may have the potential to prevent SCD and help deploy primary prevention appropriately in HCM patients.

SARS-CoV-2 infection is associated with increased risk for pulmonary embolism (PE), a fatal complication that can cause right ventricular (RV) dysfunction. Serum D-dimer levels are a sensitive test to suggest PE, however lacks specificity in COVID-19 patients. The goal of this study was to identify a model that better predicts PE diagnosis in hospitalized COVID-19 patients using clinical, laboratory, and echocardiographic imaging predictors. We performed a cross-sectional study of 302 adult patients admitted to the Johns Hopkins Hospital (March 2020-February 2021) for COVID-19 infection who underwent transthoracic echocardiography and D-dimer testing; 204 patients had CT angiography. Clinical, laboratory and imaging predictors including, but not limited to, D-dimer and RV dysfunction were used to build prediction models for PE using logistic regression. Model discrimination was assessed using area under the receiver operator curve (AUC) and calibration using Hosmer-Lemeshow

Deep learning (DL) has achieved promising performance in detecting common abnormalities from the 12-lead electrocardiogram (ECG). However, diagnostic redundancy exists in the 12-lead ECG, which could impose a systematic overfitting on DL, causing poor generalization. We, therefore, hypothesized that finding an optimal lead subset of the 12-lead ECG to eliminate the redundancy would help improve the generalizability of DL-based models. In this study, we developed and evaluated a DL-based model that has a feature extraction stage, an ECG-lead subset selection stage and a decision-making stage to automatically interpret multiple common ECG abnormality types. The data analysed in this study consisted of 6877 12-lead ECG recordings from CPSC 2018 (labelled as normal rhythm or eight types of ECG abnormalities, split into training (approx. 80%), validation (approx. 10%) and test (approx. 10%) sets) and 3998 12-lead ECG recordings from PhysioNet/CinC 2020 (labelled as normal rhythm or four types of ECG abnormalities, used as external text set). The ECG-lead subset selection module was introduced within the proposed model to efficiently constrain model complexity. It detected an optimal 4-lead ECG subset consisting of leads II, aVR, V1 and V4. The proposed model using the optimal 4-lead subset significantly outperformed the model using the complete 12-lead ECG on the validation set and on the external test dataset. The results demonstrated that our proposed model successfully identified an optimal subset of 12-lead ECG; the resulting 4-lead ECG subset improves the generalizability of the DL model in ECG abnormality interpretation. This study provides an outlook on what channels are necessary to keep and which ones may be ignored when considering an automated detection system for cardiac ECG abnormalities. This article is part of the theme issue ‘Advanced computation in cardiovascular physiology: new challenges and opportunities’.

Probabilistic estimation of cardiac electrophysiological model parameters serves an important step toward model personalization and uncertain quantification. The expensive computation associated with these model simulations, however, makes direct Markov Chain Monte Carlo (MCMC) sampling of the posterior probability density function (pdf) of model parameters computationally intensive. Approximated posterior pdfs resulting from replacing the simulation model with a computationally efficient surrogate, on the other hand, have seen limited accuracy. In this study, we present a Bayesian active learning method to directly approximate the posterior pdf function of cardiac model parameters, in which we intelligently select training points to query the simulation model in order to learn the posterior pdf using a small number of samples. We integrate a generative model into Bayesian active learning to allow approximating posterior pdf of high-dimensional model parameters at the resolution of the cardiac mesh. We further introduce new acquisition functions to focus the selection of training points on better approximating the shape rather than the modes of the posterior pdf of interest. We evaluated the presented method in estimating tissue excitability in a 3D cardiac electrophysiological model in a range of synthetic and real-data experiments. We demonstrated its improved accuracy in approximating the posterior pdf compared to Bayesian active learning using regular acquisition functions, and substantially reduced computational cost in comparison to existing standard or accelerated MCMC sampling.

Background: Patients with cardiac sarcoidosis (CS) are at increased risk of life-threatening ventricular arrhythmias (VA). Current approaches to risk stratification have limited predictive value. Objectives: To assess the utility of spatial dispersion analysis of LGE-CMR, as a quantitative measure of myocardial tissue heterogeneity, in risk stratifying patients with CS for ventricular VA and death. Methods: 62 patients with CS underwent LGE-CMR. LGE images were segmented and dispersion maps of the left and right ventricles were generated as follows. Based on signal intensity (SI), each pixel was categorized as abnormal (SI {greater than or equal to}3SD above the mean), intermediate (SI 1-3 SD above the mean) or normal (SI

Hypertrophic cardiomyopathy (HCM) causes sudden cardiac death (SCD) due to ventricular arrhythmias (VA) manifesting from myocardial fibrosis proliferation. Current clinical risk stratification criteria inadequately identify at-risk patients in need of primary prevention of VA. Here, we use mechanistic computational modeling of the heart to analyze how HCM-specific remodeling of the heart promotes arrhythmogenesis and to develop a personalized strategy to forecast risk of VAs in these patients. We combine contrast-enhanced cardiac magnetic-resonance (CMR) imaging and T1 mapping data to construct digital replicas of HCM patient hearts that represent the patient-specific distribution of focal and diffuse fibrosis and evaluate the substrate propensity to VA. Our analysis indicates that the presence of diffuse fibrosis, which is rarely assessed in these patients, increases arrhythmogenic propensity. In forecasting future VA events in HCM patients, the imaging-based computational heart approach achieved 84.6%, 76.9%, and 80.1% sensitivity, specificity, and accuracy, respectively, and significantly outperformed current clinical risk predictors. This novel VA risk assessment may have the potential to prevent SCD and help deploy primary prevention appropriately in HCM patients.

RationalePatients with ischemic cardiomyopathy (ICMP) are at high risk for malignant arrhythmias, largely due to electrophysiological remodeling of the non-infarcted myocardium. The electrophysiological properties of the non-infarcted myocardium of patients with ICMP remain largely unknown.ObjectivesTo assess the pro-arrhythmic behavior of non-infarcted myocardium in ICMP patients and couple computational simulations with machine learning to establish a methodology for the development of disease-specific action potential models based on clinically measured action potential duration restitution (APDR) data.Methods and ResultsWe enrolled 22 patients undergoing left-sided ablation (10 ICMP) and compared APDRs between ICMP and structurally normal left ventricles (SNLVs). APDRs were clinically assessed with a decremental pacing protocol. Using genetic algorithms (GAs), we constructed populations of action potential models that incorporate the cohort-specific APDRs. The variability in the populations of ICMP and SNLV models was captured by clustering models based on their similarity using unsupervised machine learning. The pro-arrhythmic potential of ICMP and SNLV models was assessed in cell- and tissue-level simulations. Clinical measurements established that ICMP patients have a steeper APDR slope compared to SNLV (by 38%, p < 0.01). In cell-level simulations, APD alternans were induced in ICMP models at a longer cycle length compared to SNLV models (385–400 vs 355 ms). In tissue-level simulations, ICMP models were more susceptible for sustained functional re-entry compared to SNLV models.ConclusionMyocardial remodeling in ICMP patients is manifested as a steeper APDR compared to SNLV, which underlies the greater arrhythmogenic propensity in these patients, as demonstrated by cell- and tissue-level simulations using action potential models developed by GAs from clinical measurements. The methodology presented here captures the uncertainty inherent to GAs model development and provides a blueprint for use in future studies aimed at evaluating electrophysiological remodeling resulting from other cardiac diseases.

Probabilistic estimation of cardiac electrophysiological model parameters serves an important step toward model personalization and uncertain quantification. The expensive computation associated with these model simulations, however, makes direct Markov Chain Monte Carlo (MCMC) sampling of the posterior probability density function (pdf) of model parameters computationally intensive. Approximated posterior pdfs resulting from replacing the simulation model with a computationally efficient surrogate, on the other hand, have seen limited accuracy. In this study, we present a Bayesian active learning method to directly approximate the posterior pdf function of cardiac model parameters, in which we intelligently select training points to query the simulation model in order to learn the posterior pdf using a small number of samples. We integrate a generative model into Bayesian active learning to allow approximating posterior pdf of high-dimensional model parameters at the resolution of the cardiac mesh. We further introduce new acquisition functions to focus the selection of training points on better approximating the shape rather than the modes of the posterior pdf of interest. We evaluated the presented method in estimating tissue excitability in a 3D cardiac electrophysiological model in a range of synthetic and real-data experiments. We demonstrated its improved accuracy in approximating the posterior pdf compared to Bayesian active learning using regular acquisition functions, and substantially reduced computational cost in comparison to existing standard or accelerated MCMC sampling.

Patients with myocardial infarction have an abundance of conduction channels (CC); however, only a small subset of these CCs sustain ventricular tachycardia (VT). Identifying these critical CCs (CCCs) in the clinic so that they can be targeted by ablation remains a significant challenge. The objective of this study is to use a personalized virtual-heart approach to conduct a three-dimensional (3D) assessment of CCCs sustaining VTs of different morphologies in these patients, to investigate their 3D structural features, and to determine the optimal ablation strategy for each VT. To achieve these goals, ventricular models were constructed from contrast enhanced magnetic resonance imagings of six postinfarction patients. Rapid pacing induced VTs in each model. CCCs that sustained different VT morphologies were identified. CCCs’ 3D structure and type and the resulting rotational electrical activity were examined. Ablation was performed at the optimal part of each CCC, aiming to terminate each VT with a minimal lesion size. Predicted ablation locations were compared to clinical. Analyzing the simulation results, we found that the observed VTs in each patient model were sustained by a limited number (2.7 ± 1.2) of CCCs. Further, we identified three types of CCCs sustaining VTs: I-type and T-type channels, with all channel branches bounded by scar, and functional reentry channels, which were fully or partially bounded by conduction block surfaces. The different types of CCCs accounted for 43.8, 18.8, and 37.4% of all CCCs, respectively. The mean narrowest width of CCCs or a branch of CCC was 9.7 ± 3.6 mm. Ablation of the narrowest part of each CCC was sufficient to terminate VT. Our results demonstrate that a personalized virtual-heart approach can determine the possible VT morphologies in each patient and identify the CCCs that sustain reentry. The approach can aid clinicians in identifying accurately the optimal VT ablation targets in postinfarction patients.

The dynamic oscillations of tropomyosin molecules in the azimuthal direction over the surface of the actin filament during thin filament activation are studied here from an energy landscape perspective. A mathematical model based on principles from nonlinear dynamics and chaos theory is derived to describe these dynamical motions. In particular, an energy potential with three wells is proposed to govern the tropomyosin oscillations between the observed regulatory positions observed during muscle contraction, namely the blocked “B”, closed “C” and open “M” states. Based on the variations in both the frequency and amplitude of the environmental (surrounding the thin filament system) driving tractions, such as the electrostatic, hydrophobic, and Ca2+-dependent forces, the tropomyosin movements are shown to be complex; they can change from being simple harmonic oscillations to being fully chaotic. Three cases (periodic, period-2, and chaotic patterns) are presented to showcase the different possible dynamic responses of tropomyosin sliding over the actin filament. A probability density function is used as a statistical measure to calculate the average residence time spanned out by the tropomyosin molecule when visiting each (B, C, M) equilibrium state. The results were found to depend strongly on the energy landscape profile and its featured barriers, which normally govern the transitions between the B-C-M states during striated muscle activation.

Aims Inadequate modification of the atrial fibrotic substrate necessary to sustain re-entrant drivers (RDs) may explain atrial fibrillation (AF) recurrence following failed pulmonary vein isolation (PVI). Personalized computational models of the fibrotic atrial substrate derived from late gadolinium enhanced (LGE)-magnetic resonance imaging (MRI) can be used to non-invasively determine the presence of RDs. The objective of this study is to assess the changes of the arrhythmogenic propensity of the fibrotic substrate after PVI. Methods and results Pre- and post-ablation individualized left atrial models were constructed from 12 AF patients who underwent pre- and post-PVI LGE-MRI, in six of whom PVI failed. Pre-ablation AF sustained by RDs was induced in 10 models. RDs in the post-ablation models were classified as either preserved or emergent. Pre-ablation models derived from patients for whom the procedure failed exhibited a higher number of RDs and larger areas defined as promoting RD formation when compared with atrial models from patients who had successful ablation, 2.6 ± 0.9 vs. 1.8 ± 0.2 and 18.9 ± 1.6% vs. 13.8 ± 1.5%, respectively. In cases of successful ablation, PVI eliminated completely the RDs sustaining AF. Preserved RDs unaffected by ablation were documented only in post-ablation models of patients who experienced recurrent AF (2/5 models); all of these models had also one or more emergent RDs at locations distinct from those of pre-ablation RDs. Emergent RDs occurred in regions that had the same characteristics of the fibrosis spatial distribution (entropy and density) as regions that harboured RDs in pre-ablation models. Conclusion Recurrent AF after PVI in the fibrotic atria may be attributable to both preserved RDs that sustain AF pre- and post-ablation, and the emergence of new RDs following ablation. The same levels of fibrosis entropy and density underlie the pro-RD propensity in both pre- and post-ablation substrates.

Optogenetic defibrillation of hearts expressing light-sensitive cation channels (e.g., ChR2) has been proposed as an alternative to conventional electrotherapy. Past modeling work has shown that ChR2 stimulation can depolarize enough myocardium to interrupt arrhythmia, but its efficacy is limited by light attenuation and high energy needs. These shortcomings may be mitigated by using new optogenetic proteins like Guillardia theta Anion Channelrhodopsin (GtACR1), which produces a repolarizing outward current upon illumination. Accordingly, we designed a study to assess the feasibility of GtACR1-based optogenetic arrhythmia termination in human hearts. We conducted electrophysiological simulations in MRI-based atrial or ventricular models (n = 3 each), with pathological remodeling from atrial fibrillation or ischemic cardiomyopathy, respectively. We simulated light sensitization via viral gene delivery of three different opsins (ChR2, red-shifted ChR2, GtACR1) and uniform endocardial illumination at the appropriate wavelengths (blue, red, or green light, respectively). To analyze consistency of arrhythmia termination, we varied pulse timing (three evenly spaced intervals spanning the reentrant cycle) and intensity (atrial: 0.001–1 mW/mm2; ventricular: 0.001–10 mW/mm2). In atrial models, GtACR1 stimulation with 0.005 mW/mm2 green light consistently terminated reentry; this was 10–100x weaker than the threshold levels for ChR2-mediated defibrillation. In ventricular models, defibrillation was observed in 2/3 models for GtACR1 stimulation at 0.005 mW/mm2 (100–200x weaker than ChR2 cases). In the third ventricular model, defibrillation failed in nearly all cases, suggesting that attenuation issues and patient-specific organ/scar geometry may thwart termination in some cases. Across all models, the mechanism of GtACR1-mediated defibrillation was voltage forcing of illuminated tissue toward the modeled channel reversal potential of −40 mV, which made propagation through affected regions impossible. Thus, our findings suggest GtACR1-based optogenetic defibrillation of the human heart may be feasible with ≈2–3 orders of magnitude less energy than ChR2.

Combining mechanistic modeling and machine learning, a multivariable sudden cardiac death predictor outperforms clinical metrics., Cardiac sarcoidosis (CS), an inflammatory disease characterized by formation of granulomas in the heart, is associated with high risk of sudden cardiac death (SCD) from ventricular arrhythmias. Current “one-size-fits-all” guidelines for SCD risk assessment in CS result in insufficient appropriate primary prevention. Here, we present a two-step precision risk prediction technology for patients with CS. First, a patient’s arrhythmogenic propensity arising from heterogeneous CS-induced ventricular remodeling is assessed using a novel personalized magnetic-resonance imaging and positron-emission tomography fusion mechanistic model. The resulting simulations of arrhythmogenesis are fed, together with a set of imaging and clinical biomarkers, into a supervised classifier. In a retrospective study of 45 patients, the technology achieved testing results of 60% sensitivity [95% confidence interval (CI): 57-63%], 72% specificity [95% CI: 70-74%], and 0.754 area under the receiver operating characteristic curve [95% CI: 0.710-0.797]. It outperformed clinical metrics, highlighting its potential to transform CS risk stratification.