H-Matrix Sparsification Applied to Bioelectromagnetic Analysis of Large Scale Human Models

This paper proposes the use of hierarchical matrices (H-matrices) with adaptive cross approximation for the solution of the very large sparse/dense block linear systems typically arising in hybrid differential/integral formulations applied to bioelectromagnetic problems, which make use of large scale human models. The code implementing this approach is described in detail and the usefulness of the approach is first demonstrated on a benchmark provided with an analytical solution and then applied to a magnetic resonance imaging problem.