Implementation of a fast simulation framework for the computation of vulnerability to atrial arrhythmias on a GPU
Atrial fibrillation (AF) is a common cardiac pathology affecting 10% of the elderly population. AF is characterized by an abnormal rapid and irregular activation. Sustained AF contributes to congestive heart failure, ventricular arrhythmia, cardiac mortality and thromboembolic stroke. Additionally, AF modifies the atrial electrical properties (electrical remodeling) promoting the occurrence and maintenance of AF. One curative therapy option for AF patient is radio frequency ablation (RFA), which aims to destroy pathological triggers or block unwanted excitation pathways.
Until today, the choice of ablation lines and the success of RFA is solely based on the experience of the medical doctor and on empirical studies. In this work, we aim to develop a framework that can quantify, which regions of the atrium are more vulnerable to ectopic beats. Therefore, additional stimuli in different locations in the atrium should be evaluated for the initiation of persistent reentries. Those stimulation locations that lead to persistent reentries should be marked in the anatomical model of the atrium with the color red, the others with green. The color red denotes regions that we expect to be necessary to target during RFA for an improved therapy outcome.
As the simulation framework needs to calculate excitation propagation in the atrium over the range of several seconds for many stimulation locations, standard reaction-diffusion-based models are not applicable. Therefore, a fast-marching level set method considering rule-based cell properties and possibilities to include inhomogeneities and anisotropies needs to be implemented on a GPU.