Solving the cardiac bidomain equations using graphics processing units

The computational modeling of the heart has been shown to be a very useful tool. The models, which become more realistic each day, provide a better understanding of the complex biophysical processes related to the electrical activity in the heart, e.g., in the case of cardiac arrhythmias. However, the increasing complexity of the models challenges high performance computing in many aspects. This work presents a cardiac simulator based on the bidomain equations that exploits the new parallel architecture of graphics processing units (GPUs). The initial results are promising. The use of the CPU accelerates the cardiac simulator by about 6 times compared to the best performance obtained in a general-purpose processor (CPU). In addition, the CPU implementation was compared to an efficient parallel implementation developed for cluster computing. A single desktop computer equipped with a CPU is shown to be 1.4 times faster than the parallel implementation of the bidomain equations running on a cluster composed of 16 processing cores. (C) 2012 Elsevier B.V. All rights reserved.