Digital twinning of the human ventricular activation sequence to Clinical 12-lead ECGs and magnetic resonance imaging using realistic Purkinje networks for in silico clinical trials

Cardiac in silico clinical trials can virtually assess the safety and efficacy of therapies using human-based modelling and simulation. These technologies can provide mechanistic explanations for clinically observed pathological behaviour. Designing virtual cohorts for in silico trials requires exploiting clinical data to capture the physiological variability in the human population. The clinical characterisation of ventricular activation and the Purkinje network is challenging, especially non-invasively. Our study aims to present a novel digital twinning pipeline that can efficiently generate and integrate Purkinje networks into human multiscale biventricular models based on subject-specific clinical 12-lead electrocardiogram and magnetic resonance recordings. Essential novel features of the pipeline are the human-based Purkinje network generation method, personalisation considering ECG R wave progression as well as QRS morphology, and translation from reduced-order Eikonal models to equivalent biophysically-detailed monodomain ones. We demonstrate ECG simulations in line with clinical data with clinical image-based multiscale models with Purkinje in four control subjects and two hypertrophic cardiomyopathy patients (simulated and clinical QRS complexes with Pearson's correlation coefficients > 0.7). Our methods also considered possible differences in the density of Purkinje myocardial junctions in the Eikonal-based inference as regional conduction velocities. These differences translated into regional coupling effects between Purkinje and myocardial models in the monodomain formulation. In summary, we demonstrate a digital twin pipeline enabling simulations yielding clinically consistent ECGs with clinical CMR image-based biventricular multiscale models, including personalised Purkinje in healthy and cardiac disease conditions.

Graph-based homogenisation for modelling cardiac fibrosis

Fibrosis, the excess of extracellular matrix, can affect, and even block, propagation of action potential in cardiac tissue. This can result in deleterious effects on heart function, but the nature and severity of these effects depend strongly on the localisation of fibrosis and its by-products in cardiac tissue, such as collagen scar formation. Computer simulation is an important means of understanding the complex effects of fibrosis on activation patterns in the heart, but concerns of computational cost place restrictions on the spatial resolution of these simulations. In this work, we present a novel numerical homogenisation technique that uses both Eikonal and graph approaches to allow fine-scale heterogeneities in conductivity to be incorporated into a coarser mesh. Homogenisation achieves this by deriving effective conductivity tensors so that a coarser mesh can then be used for numerical simulation. By taking a graph-based approach, our homogenisation technique functions naturally on irregular grids and does not rely upon any assumptions of periodicity, even implicitly. We present results of action potential propagation through fibrotic tissue in two dimensions that show the graph-based homogenisation technique is an accurate and effective way to capture fine-scale domain information on coarser meshes in the context of sharp-fronted travelling waves of activation. As test problems, we consider excitation propagation in tissue with diffuse fibrosis and through a tunnel-like structure designed to test homogenisation, interaction of an excitation wave with a scar region, and functional re-entry.

Novel aggregated multiposition noncontact mapping of atrial tachycardia in humans: From computational modeling to clinical validation

BACKGROUND

A novel aggregated multiposition noncontact mapping (AMP-NCM) algorithm is proposed to diagnose cardiac arrhythmias.

OBJECTIVE

The purpose of this study was to computationally determine an accuracy threshold and to compare the accuracy and clinical utility of AMP-NCM to gold standard contact mapping.

METHODS

In a cellular automata model, the number of catheter positions and chamber coverage were varied to establish accuracy requirements for clinically relevant AMP-NCM. This guided the clinical study protocol. In a prospective cohort of patients with atrial tachycardia (AT), noncontact mapping (NCM) recordings from a single position (SP) and multiple positions were compared to contact mapping with a high-density multipolar catheter using morphology and timing differences of reconstructed signals. Identification of AT mechanisms and ablation targets using both AMP-NCM and contact mapping were randomly evaluated by 5 blinded reviewers.

RESULTS

AMP-NCM accuracy was asymptotic at 60 catheter positions in computational modeling. Twenty patients (age 65 ± 12 years; 19 male) with 26 ATs (5 focal, 21 reentrant) were studied. Morphologic correlation of signals derived from AMP-NCM was significantly better than those from SP-NCM compared to contact signals (median 0.93 vs 0.76; P <.001). AMP-NCM generated maps more rapidly than contact mapping (3 ± 1 minutes vs 13 ± 6 minutes; P <.001) and correctly diagnosed AT mechanisms in 25 of 26 maps (96%). Overall, 80% of arrhythmia mechanisms were correctly identified using AMP-NCM by blinded reviewers.

CONCLUSION

Once 60 catheter positions were achieved, AMP-NCM successfully diagnosed mechanisms of AT and identified treatment sites equal to gold standard contact mapping in 3 minutes of procedural time.

Deep learning formulation of electrocardiographic imaging integrating image and signal information with data-driven regularization

AIMS

Electrocardiographic imaging (ECGI) is a promising tool to map the electrical activity of the heart non-invasively using body surface potentials (BSP). However, it is still challenging due to the mathematically ill-posed nature of the inverse problem to solve. Novel approaches leveraging progress in artificial intelligence could alleviate these difficulties.

METHODS AND RESULTS

We propose a deep learning (DL) formulation of ECGI in order to learn the statistical relation between BSP and cardiac activation. The presented method is based on Conditional Variational AutoEncoders using deep generative neural networks. To quantify the accuracy of this method, we simulated activation maps and BSP data on six cardiac anatomies.We evaluated our model by training it on five different cardiac anatomies (5000 activation maps) and by testing it on a new patient anatomy over 200 activation maps. Due to the probabilistic property of our method, we predicted 10 distinct activation maps for each BSP data. The proposed method is able to generate volumetric activation maps with a good accuracy on the simulated data: the mean absolute error is 9.40 ms with 2.16 ms standard deviation on this testing set.

CONCLUSION

The proposed formulation of ECGI enables to naturally include imaging information in the estimation of cardiac electrical activity from BSP. It naturally takes into account all the spatio-temporal correlations present in the data. We believe these features can help improve ECGI results.

Inference of ventricular activation properties from non-invasive electrocardiography

The realisation of precision cardiology requires novel techniques for the non-invasive characterisation of individual patients’ cardiac function to inform therapeutic and diagnostic decision-making. Both electrocardiography and imaging are used for the clinical diagnosis of cardiac disease. The integration of multi-modal datasets through advanced computational methods could enable the development of the cardiac ‘digital twin’, a comprehensive virtual tool that mechanistically reveals a patient's heart condition from clinical data and simulates treatment outcomes. The adoption of cardiac digital twins requires the non-invasive efficient personalisation of the electrophysiological properties in cardiac models. This study develops new computational techniques to estimate key ventricular activation properties for individual subjects by exploiting the synergy between non-invasive electrocardiography, cardiac magnetic resonance (CMR) imaging and modelling and simulation. More precisely, we present an efficient sequential Monte Carlo approximate Bayesian computation-based inference method, integrated with Eikonal simulations and torso-biventricular models constructed based on clinical CMR imaging. The method also includes a novel strategy to treat combined continuous (conduction speeds) and discrete (earliest activation sites) parameter spaces and an efficient dynamic time warping-based ECG comparison algorithm. We demonstrate results from our inference method on a cohort of twenty virtual subjects with cardiac ventricular myocardial-mass volumes ranging from 74 cm³ to 171 cm³ and considering low versus high resolution for the endocardial discretisation (which determines possible locations of the earliest activation sites). Results show that our method can successfully infer the ventricular activation properties in sinus rhythm from non-invasive epicardial activation time maps and ECG recordings, achieving higher accuracy for the endocardial speed and sheet (transmural) speed than for the fibre or sheet-normal directed speeds.

•

Estimation of the ventricular speeds and earliest activation sites from ECG and CMR.

•

Evaluation with twenty virtual subjects shows the effect of anatomical variability.

•

Bayesian-inspired simultaneous estimation of continuous and discrete parameters.

•

Efficient dynamic time warping-based comparison of electrocardiograms (ECG).

•

Changing fibre and sheet-normal speed does not affect healthy activation sequence.

Patient-Specific Identification of Atrial Flutter Vulnerability-A Computational Approach to Reveal Latent Reentry Pathways

Atypical atrial flutter (AFlut) is a reentrant arrhythmia which patients frequently develop after ablation for atrial fibrillation (AF). Indeed, substrate modifications during AF ablation can increase the likelihood to develop AFlut and it is clinically not feasible to reliably and sensitively test if a patient is vulnerable to AFlut. Here, we present a novel method based on personalized computational models to identify pathways along which AFlut can be sustained in an individual patient. We build a personalized model of atrial excitation propagation considering the anatomy as well as the spatial distribution of anisotropic conduction velocity and repolarization characteristics based on a combination of a priori knowledge on the population level and information derived from measurements performed in the individual patient. The fast marching scheme is employed to compute activation times for stimuli from all parts of the atria. Potential flutter pathways are then identified by tracing loops from wave front collision sites and constricting them using a geometric snake approach under consideration of the heterogeneous wavelength condition. In this way, all pathways along which AFlut can be sustained are identified. Flutter pathways can be instantiated by using an eikonal-diffusion phase extrapolation approach and a dynamic multifront fast marching simulation. In these dynamic simulations, the initial pattern eventually turns into the one driven by the dominant pathway, which is the only pathway that can be observed clinically. We assessed the sensitivity of the flutter pathway maps with respect to conduction velocity and its anisotropy. Moreover, we demonstrate the application of tailored models considering disease-specific repolarization properties (healthy, AF-remodeled, potassium channel mutations) as well as applicabiltiy on a clinical dataset. Finally, we tested how AFlut vulnerability of these substrates is modulated by exemplary antiarrhythmic drugs (amiodarone, dronedarone). Our novel method allows to assess the vulnerability of an individual patient to develop AFlut based on the personal anatomical, electrophysiological, and pharmacological characteristics. In contrast to clinical electrophysiological studies, our computational approach provides the means to identify all possible AFlut pathways and not just the currently dominant one. This allows to consider all relevant AFlut pathways when tailoring clinical ablation therapy in order to reduce the development and recurrence of AFlut.

Spatio-temporal Analysis of Multichannel Atrial Electrograms Based on a Concept of Active Areas

Atrial tachycardia and atrial flutter are frequent arrhythmia that occur spontaneously and after ablation of atrial fibrillation. Depolarization waves that differ significantly from sinus rhythm propagate across the atria with high frequency (typically 140 to 220 beats per minute). A detailed and personalized analysis of the spread of depolarization is imperative for a successful ablation therapy. Thus, catheters with several electrodes are employed to measure multichannel electrograms inside the atria. Here we propose a new concept for spatio-temporal analysis of multichannel electrograms during atrial tachycardia and atrial flutter. It is based on the calculation of simultaneously active areas. The method allows to identify atrial tachycardia and to automatically distinguish between subtypes of focal activity, micro-reentry and macro-reentry.

Strategies of data layout and cache writing for input-output optimization in high performance scientific computing: Applications to the forward electrocardiographic problem

Input-output (I/O) optimization at the low-level design of data layout on disk drastically impacts the efficiency of high performance computing (HPC) applications. However, such a low-level optimization is in general challenging, especially when using popular scientific file formats designed with an emphasis on portability and flexibility. To reconcile these two aspects, we present a novel low-level data layout for HPC applications, fully independent of the number of dimensions in the dataset. The new data layout improves reading and writing efficiency in large HPC applications using many processors, and in particular during parallel post-processing. Furthermore, its combination with a cached write mode, in order to aggregate multiple writes into larger ones, substantially decreased the writing times of the proposed strategy. When applied to our simulation framework for the forward calculation of the human electrocardiogram, the combined strategy resulted in drastic improvements in I/O performance, of up to 40% in writing and 93–98% in reading for post-processing tasks. Given the generality of the proposed strategies and scientific file formats used, our results may represent significant improvements in I/O performance of HPC applications across multiple disciplines, reducing execution and post-processing times and leading to a more efficient use of HPC resource envelopes.

A work flow to build and validate patient specific left atrium electrophysiology models from catheter measurements

Biophysical models of the atrium provide a physically constrained framework for describing the current state of an atrium and allow predictions of how that atrium will respond to therapy. We propose a work flow to simulate patient specific electrophysiological heterogeneity from clinical data and validate the resulting biophysical models. In 7 patients, we recorded the atrial anatomy with an electroanatomical mapping system (St Jude Velocity); we then applied an S1-S2 electrical stimulation protocol from the coronary sinus (CS) and the high right atrium (HRA) whilst recording the activation patterns using a PentaRay catheter with 10 bipolar electrodes at 12 ± 2 sites across the atrium. Using only the activation times measured with a PentaRay catheter and caused by a stimulus applied in the CS with a remote catheter we fitted the four parameters for a modified Mitchell-Schaeffer model and the tissue conductivity to the recorded local conduction velocity restitution curve and estimated local effective refractory period. Model parameters were then interpolated across each atrium. The fitted model recapitulated the S1-S2 activation times for CS pacing giving a correlation ranging between 0.81 and 0.98. The model was validated by comparing simulated activations times with the independently recorded HRA pacing S1-S2 activation times, giving a correlation ranging between 0.65 and 0.96. The resulting work flow provides the first validated cohort of models that capture clinically measured patient specific electrophysiological heterogeneity.

A conduction velocity adapted eikonal model for electrophysiology problems with re-excitability evaluation

Computational models of heart electrophysiology achieved a considerable interest in the medical community as they represent a novel framework for the study of the mechanisms underpinning heart pathologies. The high demand of computational resources and the long computational time required to evaluate the model solution hamper the use of detailed computational models in clinical applications. In this paper, we present a multi-front eikonal algorithm that adapts the conduction velocity (CV) to the activation frequency of the tissue substrate. We then couple the eikonal new algorithm with the Mitchell-Schaeffer (MS) ionic model to determine the tissue electrical state. Compared to the standard eikonal model, this model introduces three novelties: first, it evaluates the local value of the transmembrane potential and of the ionic variable solving an ionic model; second, it computes the action potential duration (APD) and the diastolic interval (DI) from the solution of the MS model and uses them to determine if the tissue is locally re-excitable; third, it adapts the CV to the underpinning electrophysiological state through an analytical expression of the CV restitution and the computed local DI. We conduct series of simulations on a 3D tissue slab and on a realistic heart geometry and compare the solutions with those obtained solving the monodomain equation. Our results show that the new model is significantly more accurate than the standard eikonal model. The proposed model enables the numerical simulation of the heart electrophysiology on a clinical time scale and thus constitutes a viable model candidate for computer-guided radio-frequency ablation.

Parametrization of activation based cardiac electrophysiology models using bidomain model simulations

Eikonal models are useful to compute approximate solutions of cardiac excitation propagation in a computationally efficient way. In this work the underlying conduction velocities for different cell types were computed solving the classical bidomain model equations for planar wavefront propagation. It was further investigated how changes in the conductivity tensors within the bidomain model analytically correspond to changes in the conduction velocity. The error in the presence of local front curvature for the derived eikonal model parametrization were analyzed. The conduction velocity simulated based on the bidomain model was overestimated by a maximum of 10%.

Influence of Modeling Errors on the Initial Estimate for Nonlinear Myocardial Activation Times Imaging Calculated With Fastest Route Algorithm

Noninvasive reconstruction of cardiac electrical activity has a great potential to support clinical decision making, planning, and treatment. Recently, significant progress has been made in the estimation of the cardiac activation from body surface potential maps (BSPMs) using boundary element method (BEM) with the equivalent double layer (EDL) as a source model. In this formulation, noninvasive assessment of activation times results in a nonlinear optimization problem with an initial estimate calculated with the fastest route algorithm (FRA). Each FRA-simulated activation sequence is converted into the ECG. The best initialization is determined by the sequence providing the highest correlation between predicted and measured potentials. We quantitatively assess the effects of the forward modeling errors on the FRA-based initialization. We present three simulation setups to investigate the effects of volume conductor model simplifications, neglecting the cardiac anisotropy and geometrical errors on the localization of ectopic beats starting on the ventricular surface. For the analysis, 12-lead ECG and 99 electrodes BSPM system were used. The areas in the heart exposing the largest localization errors were volume conductor model and electrode configuration specific with an average error <10 mm. The results show the robustness of the FRA-based initialization with respect to the considered modeling errors.

A self-taught artificial agent for multi-physics computational model personalization

Personalization is the process of fitting a model to patient data, a critical step towards application of multi-physics computational models in clinical practice. Designing robust personalization algorithms is often a tedious, time-consuming, model- and data-specific process. We propose to use artificial intelligence concepts to learn this task, inspired by how human experts manually perform it. The problem is reformulated in terms of reinforcement learning. In an off-line phase, Vito, our self-taught artificial agent, learns a representative decision process model through exploration of the computational model: it learns how the model behaves under change of parameters. The agent then automatically learns an optimal strategy for on-line personalization. The algorithm is model-independent; applying it to a new model requires only adjusting few hyper-parameters of the agent and defining the observations to match. The full knowledge of the model itself is not required. Vito was tested in a synthetic scenario, showing that it could learn how to optimize cost functions generically. Then Vito was applied to the inverse problem of cardiac electrophysiology and the personalization of a whole-body circulation model. The obtained results suggested that Vito could achieve equivalent, if not better goodness of fit than standard methods, while being more robust (up to 11% higher success rates) and with faster (up to seven times) convergence rate. Our artificial intelligence approach could thus make personalization algorithms generalizable and self-adaptable to any patient and any model.

In Silico Organ Modelling in Predicting Efficacy and Safety of New Medicines

The development of new medicines faces important challenges due to difficulties in the assessment of their efficacy and their safety in the targeted human population. In silico approaches through the use of mathematical modelling and computer simulations are increasingly being used to overcome some of the limitations of current experimental methods used in the development of new medicines. This chapter describes state-of-the-art in silico approaches for the evaluation of the safety and efficacy of medicines targeting important causes of mortality such as cardiovascular disease. Firstly, we describe the in silico multi-scale mathematical models and simulation techniques required to describe drug-induced effects on physiological systems such as the heart from the subcellular to the whole organ level. Then we illustrate the power of in silico approaches used to augment experimental and clinical investigations, by providing the framework to unravel multi-scale mechanisms underlying variability in the response to medicines and to focus on effects in human rather than animal models. We devote the last part of the chapter to discussing the process of validation of in silico models and simulations, which is key in building up their credibility.

Effect of statistically derived fiber models on the estimation of cardiac electrical activation

Myocardial fiber orientation plays a critical role in the electrical activation and subsequent contraction of the heart. To increase the clinical potential of electrophysiological (EP) simulation for the study of cardiac phenomena and the planning of interventions, accurate personalization of the fibers is a necessary yet challenging task. Due to the difficulties associated with the in vivo imaging of cardiac fiber structure, researchers have developed alternative techniques to personalize fibers. Thus far, cardiac simulation was performed mainly based on rule-based fiber models. More recently, there has been a significant interest in data-driven and statistically derived fiber models. In particular, our predictive method in [1] allows us to estimate the unknown subject-specific fiber orientation based on the more easily available shape information. The aim of this work is to estimate the effect of using such statistical predictive models for the estimation of cardiac electrical activation times and patterns. To this end, we perform EP simulations based on a database of ten canine ex vivo diffusion tensor imaging (DTI) datasets that include normal and failing cases. To assess the strength of the fiber models under varying conditions, we consider both sinus rhythm and biventricular pacing simulations. The results show that 1) the statistically derived fibers improve the estimation of the local activation times by an average of 53.7% over traditional rule-based models, and that 2) the obtained electrical activations are consistently similar to those of the DTI-based fibers.

Data-driven estimation of cardiac electrical diffusivity from 12-lead ECG signals

Diagnosis and treatment of dilated cardiomyopathy (DCM) is challenging due to a large variety of causes and disease stages. Computational models of cardiac electrophysiology (EP) can be used to improve the assessment and prognosis of DCM, plan therapies and predict their outcome, but require personalization. In this work, we present a data-driven approach to estimate the electrical diffusivity parameter of an EP model from standard 12-lead electrocardiograms (ECG). An efficient forward model based on a mono-domain, phenomenological Lattice-Boltzmann model of cardiac EP, and a boundary element-based mapping of potentials to the body surface is employed. The electrical diffusivity of myocardium, left ventricle and right ventricle endocardium is then estimated using polynomial regression which takes as input the QRS duration and electrical axis. After validating the forward model, we computed 9500 EP simulations on 19 different DCM patients in just under three seconds each to learn the regression model. Using this database, we quantify the intrinsic uncertainty of electrical diffusion for given ECG features and show in a leave-one-patient-out cross-validation that the regression method is able to predict myocardium diffusion within the uncertainty range. Finally, our approach is tested on the 19 cases using their clinical ECG. 84% of them could be personalized using our method, yielding mean prediction errors of 18.7ms for the QRS duration and 6.5° for the electrical axis, both values being within clinical acceptability. By providing an estimate of diffusion parameters from readily available clinical data, our data-driven approach could therefore constitute a first calibration step toward a more complete personalization of cardiac EP.

Statistical personalization of ventricular fiber orientation using shape predictors

This paper presents a predictive framework for the statistical personalization of ventricular fibers. To this end, the relationship between subject-specific geometry of the left (LV) and right ventricles (RV) and fiber orientation is learned statistically from a training sample of ex vivo diffusion tensor imaging datasets. More specifically, the axes in the shape space which correlate most with the myocardial fiber orientations are extracted and used for prediction in new subjects. With this approach and unlike existing fiber models, inter-subject variability is taken into account to generate latent shape predictors that are statistically optimal to estimate fiber orientation at each individual myocardial location. The proposed predictive model was applied to the task of personalizing fibers in 10 canine subjects. The results indicate that the ventricular shapes are good predictors of fiber orientation, with an improvement of 11.4% in accuracy over the average fiber model.

Computational methods to reduce uncertainty in the estimation of cardiac conduction properties from electroanatomical recordings

Cardiac imaging is routinely used to evaluate cardiac tissue properties prior to therapy. By integrating the structural information with electrophysiological data from e.g. electroanatomical mapping systems, knowledge of the properties of the cardiac tissue can be further refined. However, as in other clinical modalities, electrophysiological data are often sparse and noisy, and this results in high levels of uncertainty in the estimated quantities. In this study, we develop a methodology based on Bayesian inference, coupled with a computationally efficient model of electrical propagation to achieve two main aims: (1) to quantify values and associated uncertainty for different tissue conduction properties inferred from electroanatomical data, and (2) to design strategies to optimize the location and number of measurements required to maximize information and reduce uncertainty. The methodology is validated in an in silico study performed using simulated data obtained from a human image-based ventricular model, including realistic fibre orientation and a transmural scar. We demonstrate that the method provides a simultaneous description of clinically-relevant electrophysiological conduction properties and their associated uncertainty for various levels of noise. By using the developed methodology to investigate how the uncertainty decreases in response to added measurements, we then derive an a priori index for placing electrophysiological measurements in order to optimize the information content of the collected data. Results show that the derived index has a clear benefit in minimizing the uncertainty of inferred conduction properties compared to a random distribution of measurements, reducing the number of required measurements by over 50% in several of the investigated settings. This suggests that the methodology presented in this work provides an important step towards improving the quality of the spatiotemporal information obtained using electroanatomical mapping.

Bridging experiments, models and simulations: an integrative approach to validation in computational cardiac electrophysiology

Computational models in physiology often integrate functional and structural information from a large range of spatiotemporal scales from the ionic to the whole organ level. Their sophistication raises both expectations and skepticism concerning how computational methods can improve our understanding of living organisms and also how they can reduce, replace, and refine animal experiments. A fundamental requirement to fulfill these expectations and achieve the full potential of computational physiology is a clear understanding of what models represent and how they can be validated. The present study aims at informing strategies for validation by elucidating the complex interrelations among experiments, models, and simulations in cardiac electrophysiology. We describe the processes, data, and knowledge involved in the construction of whole ventricular multiscale models of cardiac electrophysiology. Our analysis reveals that models, simulations, and experiments are intertwined, in an assemblage that is a system itself, namely the model-simulation-experiment (MSE) system. We argue that validation is part of the whole MSE system and is contingent upon 1) understanding and coping with sources of biovariability; 2) testing and developing robust techniques and tools as a prerequisite to conducting physiological investigations; 3) defining and adopting standards to facilitate the interoperability of experiments, models, and simulations; 4) and understanding physiological validation as an iterative process that contributes to defining the specific aspects of cardiac electrophysiology the MSE system targets, rather than being only an external test, and that this is driven by advances in experimental and computational methods and the combination of both.