Accuracy-efficiency comparison of finite-difference time-domain and adaptive integral method based simulators for bioelectromagnetics
MetadataShow full item record
A detailed study of the performance of finite-difference time-domain (FDTD) and adaptive integral method (AIM) based simulators is presented for bioelectromagnetic (BIOEM) analysis in the UHF band. The comparison is complicated because modern simulators based on these methods can routinely perform high-fidelity BIOEM analysis with hundreds of millions of degrees of freedom. In this thesis, an empirical approach is adopted to investigate the accuracy-efficiency tradeoffs of an FDTD and an AIM based simulator. Specifically, comprehensive numerical experiments are performed using several benchmark multi-layered spherical phantoms. Scattering from these phantoms are computed by using increasingly finer resolution meshes and the results are compared to analytical solutions to investigate the accuracy as well as computational costs of the different methods. The results from the benchmark problems show that both FDTD and AIM based simulators achieve similar error levels for staircased voxel meshes but FDTD based simulation is less expensive, especially when the memory requirement and preprocessing cost are considered. The results also show that although both simulators can reduce errors by refining voxel meshes, AIM based simulators can significantly reduce errors by using CAD meshes instead of voxel ones without significant cost increase.