Generation and uncertainty quantification of patient-specific purkinje network models

Abstract

Cardiac Purkinje networks are a fundamental part of the conduction system and are known to initiate a variety of cardiac arrhythmias. However, patient-specific modeling of Purkinje networks remains challenging due to the high morphological complexity and the lack of non-invasive imaging techniques to identify these structures. This work aims to present a novel method called Shocker based on optimization principles for the generation and uncertainty quantification of patient-specific Purkinje networks that combine geometric and electrical accuracy in branch size, bifurcation angles, and Purkinje-Ventricular-Junction activation. Several Purkinje network models are generated over four different biventricular meshes with increasing complexity to reach this goal. They are used to evaluate the performance of our model in a range of different scenarios. Additional Purkinje-tissue coupled monodomain simulations are executed to evaluate the generated networks in a more realistic scenario using the most recent Purkinje/ventricular human cellular models, physiological values for the Purkinje-Ventricular-Junction characteristic delay, and a high-performance GPU solver. The results demonstrate that the new method can generate patient-specific Purkinje networks with controlled morphological metrics, local activation times at the Purkinje-Ventricular-Junctions, estimated points given by the patient electro-anatomical map and electrocardiogram. In addition, the generation of multiple Purkinje network models that can reproduce the same patient-specific data is an important tool for quantifying uncertainties associated with the computational modeling of this important conduction system of the human heart