PhD Theme

Tetrahedral mesh generation from discrete volume data for Electromagnetic and Thermal Simulations.

Context

In his daily environment, the man is voluntarily or involuntarily exposed to Radio-Frequency Electromagnetic Fields. The prediction of inferred temperature rise by this radiation, by means of the calculation on a meshed anatomical model, depends a lot from quality of this model. The current methods of tetrahedral mesh generation use intermediate surface representations on the anatomical data.We show that it is not necessary to have a surface representation to generate a volume mesh. We propose a tetrahedral mesh generation based on the Centroidal Voronoi Diagram and their dual, the Delaunay Triangulation. This approach, contrary to surrounding areas of the literature, treats directly a volume of segmented voxels, obtained by MRI or by tomography X. To obtain meshes, either uniform but adapted at the anatomical complexity, we proposed a new method of capture of this complexity by means of medial axis approximation. A comparison with three other methods of tetrahedral mesh generation shows that our approach builds tetrahedra of better geometrical quality for various criteria. Of this quality ensues a better precision of the temperature inferred by the electromagnetic radiation, calculated by finite element methods, as well as a calculation time reduction. These results show the potential of our discrete approach of type "Voronoï-Delaunay" for tetrahedral mesh generation.

Download full PDF