Massively Parallel Simulation of Cardiac Electrical Wave Propagation on Blue Gene

Heart rhythm disorders are a leading contributor to morbidity and mortality in the industrialized world. Treatment and prevention of cardiac rhythm disorders remains difficult because the electrical signal that controls the heart rhythm is determined by complex, multi-scale biological processes. Data-driven computer simulation is a promising tool for facilitating a better understanding of cardiac electrical properties. Conducting detailed, large-scale simulations of the cardiac electrical activity presents several challenges: the heart has a complicated 3D geometry, conduction of the electrical wave is anisotropic, and cardiac tissue is made up of cells with heterogeneous electrical properties. Our group has developed a software platform for conducting massively parallel simulations of wave propagation in cardiac tissue. The code is being used in an on-going study to connect drug-induced modifications of molecular properties of heart cells to changes in tissue properties that might lead to a rhythm disorder. The platform uses a finite difference method for modeling the propagation of electrical waves in cardiac tissue using the cable equation with homogeneous Neumann boundary conditions. We use a novel algorithm which is based on the phase field method for handling the boundary conditions in complicated geometries. To map grid points within the spatial domain to compute nodes, an optimization process is used to balance the trade-offs between load balancing and increased overhead for communication. The performance of the code has been studied by simulating wave propagation in an anatomically realistic model of the left and right ventricles of a rabbit heart on Blue Gene partitions of up to 4,096 processors. Based on these results, the Blue Gene architecture seems particularly suited for cardiac simulation, and offers a promising platform for rapidly exploring cardiac electrical wave dynamics in large spatial domains.