Computer simulation of action potential propagation on cardiac tissues: An efficient and scalable paralell approach

Abstract

Publisher Summary This chapter discusses the computer simulation of action potential propagation on cardiac tissues. The chapter discusses a complete parallel computing implementation for the electrical activity simulation in a two-dimensional monodomain cardiac tissue, using a cost-effective beowulf cluster. High performance computing techniques have reduced the simulation time by a factor of nearly the number of processors, reaching a 94% of efficiency when using 32 processes. Besides, the application of stable numerical ODE integration methods has allowed the use of larger time steps, reducing the whole simulation time. Moreover, the use of a grid infrastructure has allowed the simultaneous execution of a huge amount of different parametric simulations, thus increasing the global research productivity. The coefficient matrix associated to the linear equation system is symmetric, positive definite, and shows a very regular pattern, with a high level of sparsity as it is a pentadiagonal matrix. The computation of the ionic currents across the membrane of each cell requires solving, for instance, 14 differential equations in the Luo-Rudy Phase II model: 6 for updating the so-called “gate-variables” and 8 for updating the ionic concentrations. It has been estimated that the ionic model used is responsible for up to 60% of the total execution time. The purpose of all these biomedical simulations is to investigate cardiac electrical behavior. Thus, it is necessary to record intermediate simulation data when they are calculated. These data may include membrane potential, main ionic currents and any of the cell characteristics, such as ionic concentrations and intermediate currents.