There are two problems with what you are trying to do.  First, you are extracting the surface with a marching cubes algorithm and then using a mesh fairing algorithm.  The mesh fairing algorithm, or any laplacian based smoothing algorithm, does not guarantee valid surface elements.  Often you will get elements that cross each other and this will crash Tetgen, Cubit, or pretty much any other program you are trying to run.  You can try reducing the amount of smoothing, or take it out all together if you want to guarantee that the surface elements will work with Tetgen.  If you take out the smoothing you will see that the resulting mesh is somewhat stair stepped and the elements often have poor aspect ratios.  You will be able to make a mesh from this, but it may not be the best quality.

Second, it looks like you have at least two materials in your segmentation.  I don't have your data, but there is most likely a place where these materials share a boundary.  If so, you need the materials to share nodes at these boundaries.  This pipeline doesn't handle these types of interfaces.

Long story short, it looks like Biomesh3D would be exactly what you are looking for.  

Good luck,

Darrell

On Jul 25, 2011, at 6:08 AM, Jeffery, Nathan wrote:

Hi All, 
I am still struggling with my goal of creating a mesh suitable for the likes of Abaqus from segmented nrrd files produced by seg3D. I am using the example tooth-seg-unorient.nrrd file which can be found under the SCIrun\BioMesh3D_Examples\tooth folder. 

SCIrun crashes everytime I attempt to Tetgen the surface of the different parts. I have attached my network. Any advice would be greatly appreciated.

cheers
nathan

<meshing-network.srn>