Title : A fitting pipeline to reproduce the dynamic of high complexity electrophysiology models
Abstract:
In the last 30 years, people have devoted great efforts to generate models that reproduce the electrophysiological complexity of patient hearts. These models are referred to as digital twins and require a large number of parameters to be calibrated through experimental data and careful analysis. The choice of observables to perform the fit is crucial in determining if the model will provide useful information. In the literature, it is common to find articles in which the cardiac action potential (AP) morphology or a single restitution curve serves as the only data source for the fit. This problem leads to a situation in which the model reproduces a limited set of features and is susceptible to overfitting, which means that there is no way to ensure that the model behaves appropriately in a physiological and dynamical way. In this work, we present a new fitting pipeline that incorporates AP duration, conduction velocity, and activation time restitution features in single-cell and tissue. These observables are commonly measured in experimental and clinical setups and provide information at a tissue level; thus avoiding the need for slow and convoluted experiments or sampling them from other systems. Furthermore, the model we fit is a phenomenological model with a low number of parameters; hence, we are able to create a one-to-one map between the observables and most of the parameters. Our results show that the pipeline is able to reproduce the restitution and dynamical complexity of realistic human models in single-cell and tissue while increasing the computational speed of the simulations and avoiding redundant parameter fits. Moreover, we show that our method can be applied to ventricle and atria systems, corroborating the universality of our new fitting paradigm. Finally, we discuss the clinical applicability of the pipeline and suggest optimizations to it.
