Simulation of Electrical Conduction in Cardiac Tissue on High Performance Computers

Abstract

We present computational methods for simulating electrical conduction in cardiac ventricles on parallel machines. We used the network model approach to describe the tissue geometry and biophysically detailed models of ion currents and membrane transporters in cardiac ventricular myocytes. Simulations of biophysically detailed ionic models with anatomically detailed tissue geometries are computationally very expensive. We investigated the use of high performance computers to reduce execution time. Experiments have shown that we can adapt and optimize our existing models of electrical activity in heart tissue for the High Performance Computers. The solution was implemented on the IBM R p690, a shared memory machine and on a cluster of workstations, a distributed memory machine. An efficient algorithm was designed to partition the data and to pass messages between processors using Message Passing Interface (MPI). The algorithm was highly scalable to the problem size as well as the number of processors used, and could easily be ported to other parallel architectures. We were able to achieve speedup of up to 13.5 on the p690 and 27 on the cluster of workstations. Using the method developed, the simulated pattern of electrical activation agreed with the experimental data.

A Survey of Bioinformatics Applications on Parallel Architectures

“…The wellfamous Human Genome Project [10] is a well known example of sequence analysis. Using large amount of high performance computers [11] and various techniques of collecting sequences, the whole human genome was sequenced and stored within a structured database [12].…”