Eikonal Model Personalisation Using Invasive Data to Predict Cardiac Resynchronisation Therapy Electrophysiological Response

Abstract

In this manuscript, we personalise an Eikonal model of cardiac wave front propagation using data acquired during an invasive electrophysiological study. To this end, we use a genetic algorithm to determine the parameters that provide the best fit between simulated and recorded activation maps during sinus rhythm. We propose a way to parameterise the Eikonal simulations that take into account the Purkinje network and the septomarginal trabecula influences while keeping the computational cost low. We then re-use these parameters to predict the cardiac resynchronisation therapy electrophysiological response by adapting the simulation initialisation to the pacing locations. We experiment different divisions of the myocardium on which the propagation velocities have to be optimised. We conclude that separating both ventricles and both endocardia seems to provide a reasonable personalisation framework in terms of accuracy and predictive power.

Extrapolation of Ventricular Activation Times From Sparse Electroanatomical Data Using Graph Convolutional Neural Networks

“…This could explain at least in part the higher estimation errors on the epicardium. Similar differences were observed by the challenge participants Cedilnik et al, who personalized an Eikonal model to both endocardial and epicardial measurements (Cedilnik and Sermesant, 2019). The authors reported a mean RMSE between 9 and 17 ms depending on the used personalization strategy.…”

“…We observed no significant improvement with increased number of endocardial samples. To measure the overall fit of the prediction to the ground truth measurements (endocardial and epicardial), we follow the approach by Cedilnik and Sermesant ( 2019 ) and compute the average over case-wise root median squared errors (RMSE) on the training and test set, respectively. Errors between 5.6 and 7.8 ms were measured for the network for different sampling ratios, with 6.6–9.3 ms for the personalized computational model (see Table 3 ).…”

“…Errors between 5.6 and 7.8 ms were measured for the network for different sampling ratios, with 6.6–9.3 ms for the personalized computational model (see Table 3 ). Cedilnik et al reported on the training set a mean RMSE between 9 and 17 ms depending on the personalization scheme of an Eikonal model to both endocardial and epicardial measurements (Cedilnik and Sermesant, 2019 ). Furthermore, when comparing qualitatively the epicardial ground truth to the solutions provided by the two methods (see Figure 9 ), the two methods provided very similar wave propagations.…”

Automated Localization of Focal Ventricular Tachycardia From Simulated Implanted Device Electrograms: A Combined Physics–AI Approach

“…One limitation of the majority of these studies is the dependence on monodomain formulations to represent electric sources in the form of transmembrane voltages. These models are time-consuming, and thus to achieve clinical translation, fast reaction-eikonal (RE) simulations ( Neic et al, 2017 ; Cedilnik et al, 2019 ) have been the preferred choice. More recently, realistic simulations of full extracellular potentials at specific locations (e.g., ECG electrodes) have been obtained from the combination of lead field (LF) methods ( Potse, 2018 ) with fast RE models ( Gillette et al, 2021 ), achieving accuracy comparable to pseudodomain or bidomain formulations, but within a fraction of the computational time.…”

Cardiac Activation Maps Reconstruction: A Comparative Study Between Data-Driven and Physics-Based Methods

“…For example, in Cedilnik and Sermesant (2019), authors suggest a model personalization based on eikonal equation to compute activation times. In Van Dam et al (2009), authors suggest to estimate activation times directly from BSPs using the equivalent double layer source model.…”