Personalized Modeling of Cardiac Electrophysiology Using Shape-Based Prediction of Fiber Orientation

摘要

Fibers play an important role in electrophysiological (EP) simulations as they determine the shape and directions of the electrical waves traveling throughout the myocardium. Due to the limited unavailability of in vivo images of the fiber structure, computational modeling of electrophysiology has been performed thus far mostly using the well-known rule-based Streeter model. The aim of this paper is to present an EP simulation study based on a statistics-based fiber model. With this approach, the missing subject-specific fiber model is predicted directly from the available shape information based on a predictive model constructed from a training sample of ex vivo DTI images. Experiments are carried out based on a database of canine datasets (including normal and abnormal cases), by considering the DTI-, the Streeter-, and the statistics-based fiber models. The results show that the shape-based predicted fiber models improve significantly the estimation accuracy of the electrical activation times and patterns, from average errors of about 10% to 1%.