Conductivity and mesh refinement in finite element computations of electrocardiographic fields

Computational simulations are a valuable research tool in biomedical applications because they allow researchers to test a variety of hypotheses and designs, some of which are impractical or impossible to test experimentally. Such simulations have been used to investigate implantable cardioverter defibrillator (ICD) design, electrical impedance tomography and the basic nature of the electric fields generated by the heart (electrocardiographic fields). An essential component of simulation is to determine the level of detail required for the specific application, which results in the need for data on the effects of potential model features.; This dissertation describes three simulation studies that systematically examined the effects of several model parameters and features of simulated electrocardiographic fields. The first study evaluated the effects of inhomogeneous electric conductivity in the finite element model. The second study extended the first by evaluating the specific role of anisotropy in model conductivity. The final study compared two approaches for reducing the discretization error in the simulations by refining either the finite element mesh structure ( h-adaptation) or basis functions (p-adaptation) in local regions.; The results of the first study suggest that the effects of specific inhomogeneities vary widely depending on the distribution of epicardial potentials and their gradients present at any one time during the QRS complex. In general, however, the inhomogeneities with the larger volumes (the lungs, anisotropic skeletal muscle and subcutaneous fat) had the larger effects on the results (>11% average relative error, RE). The exception to this general trend was the bones of the torso, the removal of which resulted in a 7.2% RE. The results also indicate that no single inhomogeneity is an appropriate substitute for the fully inhomogeneous anisotropic model as no model with just one inhomogeneous region came within 10% RE of the full model. The results of the second study demonstrate that even when only changing the implementation of the skeletal muscle anisotropy, it is common for the results to differ by more than 10% RE. These effects also varied as a function of the distribution of epicardial potentials and gradients, as in the first study. In the third study, adaptive p-refinement proved to be the more efficient approach for reducing discretization error in the cases simulated. In the simulation of the field generated by the heart in a fully inhomogeneous torso model, the adaptive p approach reduced the global energy error by 98% compared to a reduction of 64% for the adaptive h approach and it did so using 13% fewer degrees of freedom. When an ICD was modeled for the source of the electric field, the adaptive p approach reduced the error by 97% using 9% fewer degrees of freedom than the h approach, which reduced the error by 54%.