New Tools for Image-Based Mesh Generation of 3D Imaging Data

Abstract

There has been increasing interest in the generation of models for computational modeling from imaging modalities such as MRI and CT.

Although a wide range of mesh generation techniques are available, these on the whole have not been developed for meshing from 3D imaging data. The paper will discuss new tools specific to image-based meshing, and their interface with commercial FEA and CFD packages.

Automated mesh generation techniques can generate models with millions of nodes. Reducing the size of the model can have a dramatic impact on computation time and memory and CPU (Central Processing Unit) requirements. A proprietary technique allowing the setting of different density zones throughout the model was developed, reducing of the number of elements required to capture a given geometry, while increasing the density around areas of greater interest.

Micro-architectures can be generated to conform to existing domains. To control the mechanical properties of the structure, a re-iso-surfacing technique combined with a bisection algorithm is shown to allow density variations and porosity control. The concept of a relative density map, to represent the desired relative densities in the micro-architecture, is introduced. Both 2D and 3D examples of functionally and arbitrary graded structures are given.

Finally, a new homogenization algorithm has been implemented by using the meshing techniques described above and parallel processing strategies innovatively to compute orthotropic mechanical properties from higher resolution scans, enhancing the value of micro level information, and enabling it to be used for macro models on desktop computers.

The ability to automatically convert any 3D image dataset into high quality meshes is becoming the new modus operandi for anatomical analysis.

New tools for image-based modeling have been demonstrated, improving the ease of generating meshes for computational mechanics and opening up areas of research that would not be possible otherwise.