Accelerating Cardiac Bidomain Simulations Using Graphics Processing Units

Abstract

Anatomically realistic and biophysically detailed multiscale computer models of the heart are playing an increasingly important role in advancing our understanding of integrated cardiac function in health and disease. Such detailed simulations, however, are computationally vastly demanding, which is a limiting factor for a wider adoption of in-silico modeling. While current trends in high-performance computing (HPC) hardware promise to alleviate this problem, exploiting the potential of such architectures remains challenging since strongly scalable algorithms are necessitated to reduce execution times. Alternatively, acceleration technologies such as graphics processing units (GPUs) are being considered. While the potential of GPUs has been demonstrated in various applications, benefits in the context of bidomain simulations where large sparse linear systems have to be solved in parallel with advanced numerical techniques are less clear. In this study, the feasibility of multi-GPU bidomain simulations is demonstrated by running strong scalability benchmarks using a state-of-the-art model of rabbit ventricles. The model is spatially discretized using the finite element methods (FEM) on fully unstructured grids. The GPU code is directly derived from a large pre-existing code, the Cardiac Arrhythmia Research Package (CARP), with very minor perturbation of the code base. Overall, bidomain simulations were sped up by a factor of 11.8 to 16.3 in benchmarks running on 6–20 GPUs compared to the same number of CPU cores. To match the fastest GPU simulation which engaged 20GPUs, 476 CPU cores were required on a national supercomputing facility.

A balanced accumulation scheme for parallel PDE solvers

“…To validate the performance of the standard and optimized accumulation algorithms we solve an elliptic PDE, discretized with linear tetrahedral finite elements, that describes the electric potential on the rabbit heart [13,14] shown in Fig. 5, with a parallel conjugate gradient algorithm with algebraic multigrid preconditioner (PCG-AMG).…”

“…5. For more details on the original problem setting and its solution we refer the reader to [13,14].…”

Parallel ADI Smoothers for Multigrid

“…Applications of this approach in biomedical simulations can be found in [4], [5]. We only consider parallel storage schemes for vectors and matrices that are data consistent on each processor, i.e., if any element is on more than one processor, the value is the same on all processors.…”

GPU accelerated solver for nonlinear reaction–diffusion systems. Application to the electrophysiology problem

“…However, with unstructured three dimensional (3D) bidomain simulations, the number of iterations required for convergence became prohibitive. In a more recent work Neic et al [26] showed that 25 processors were equivalent to a single GPU when computing the bidomain equations. This new capability to solve the governing equations on a relatively small GPU cluster makes it possible to one day introduce simulation using patient specific computer models into a clinical workflow.…”