Using Point Clouds for Material Properties Smoothing in Low-Frequency Numerical Dosimetry Simulations

Numerical dosimetry simulations are used to determine the electric field induced in the human body by low-frequency external magnetic fields. These simulations require a discretization of the human body to resolve the different tissues. Human models discretized with structured meshes are commonly prone to staircase errors at tissue interfaces caused by large gradients in the material properties. This paper presents an approach using point clouds to create mixed material fillings from existing voxel models to reduce the error in numerical simulations. A simplified head model for which analytical solutions exist can be used to quantify these errors. The error is reduced to a few percent if the structured mesh allows mixed material fillings.