An evaluation of three approaches to tetrahedral mesh generation for deformable registration of brain MR images

In this paper we evaluate three conceptually different approaches to mesh generation for deformable Finite Element Method (FEM) registration of Magnetic Resonance (MR) images of brain volume. Precise approximation of brain volume segmentations and good shape of the mesh tetrahedra are the main requirements imposed by the application. Our contributions are (1) application-motivated comparison and analysis of practical mesh generation implementations and (2) open source implementation of a mesh generation algorithm which we show delivers mesh quality comparable with the best commercial software products available. The preliminary results indicate, that our implementation provides a solid foundation for further development of application-specific mesh generation tools