High-Quality Multi-Tissue Mesh Generation for Finite Element Analysis

Mesh generation on 3D segmented images is a fundamental step for the construction of realistic biomechanical models. Mesh elements with low or large dihedral angles are undesirable, since they are known to underpin the speed and accuracy of the subsequent finite element analysis. In this paper, we present an algorithm for meshing 3D multi-label images. A notable feature of our method is its ability to produce tetrahedra with very good dihedral angles respecting, at the same time, the interfaces created by two or more adjoining tissues. Our method employs a Delaunay refinement scheme orchestrated by special point rejection strategies which remove poorly shaped elements without deteriorating the representation of the objects’ anatomical boundaries. Experimental evaluation on CT and MRI atlases have shown that our algorithm produces watertight meshes consisting of elements of very good quality (all the dihedral angles were between 19 and 150 degrees) which makes our method suitable for finite element simulations.