| This dissertation developed several major utilities for generating anatomically accurate, computationally efficient, 3D, finite element meshes from 2D medical images.; These enhancements included innovative algorithms for creating multiple-material surface meshes from sequential, 2D, segmented medical images. Current surface mesh generation methods can only reconstruct single-material surface meshes in an integrated manner. Multiple-material reconstructions are iterative and disjoint. Continuity between materials (or tissues) does not exist consistently. On the contrary, our multiple-material marching-cubes algorithm reconstructs multiple-material surfaces within one sweep of the staked segmented images. The continuity and integrity of the surfaces are ensured with this robust algorithm. Surface mesh adjustment algorithms, such as smoothing and simplification, were also revised to adapt to the multiple material nature.; An integrated surface mesh is suitable for visualization and virtual endoscopy simulations. However, a volume mesh is necessary for nonlinear simulations and numerical analyses. A biomedical engineering model can have numerous complex geometries that are difficult for the user to refine appropriately while maintaining computational efficiency. An automatic detection system was developed that identifies and refines regions requiring geometry-dictated refinement. It could automatically create efficient volume meshes of complicated biomedical models from only a few initial large building blocks.; No volume mesh generation system has demonstrated the ability to handle all situations without creating some poor quality or even invalid elements. A local optimization scheme was developed to improve problematic regions in existing meshes, independent of the mesh creation process. Elements below a user-specified acceptable threshold are removed along with a small buffer zone of neighboring elements. Then a search is performed for the best possible element re-connection. At times this search can be exhaustive, hence the name “Last Resort”.; The validity of these algorithms was demonstrated via numerous models created from 2D medical images. These examples included the detailed modeling of breast tissues and mammary glands for an ongoing NIH breast cancer research effort; a human torso undergoing an elecctromagnetic hyperthermia treatment for a cancerous prostate tumor; and an intricate reconstruction of the Visible Male torso in the kidney region. Each example produced valid 3D finite element meshes suitable for numerical computations. |
