Conformal multi-material mesh generation from noisy labeled volumes: a distance-field approach
People: Francois Malan, M.Sc (LUMC / TUD), Dr Charl P. Botha (TUD)
- Key words: meshing, marching cubes, surface reconstruction, multi-material, distance fields
Description
Biomechanical modelling of individual patients' bodies requires segmentation, and subsequently the creation of finite element meshes. Manual segmentation is so slow that automated segmentation is the only viable option. The problem with automatic segmentation is that it may produce noisy output and/or superfluous detail and false structures.
Modern meshing algorithms are fast and robust but generally require well segmented input since all interfaces (however insignificant) are explicitly attempted to be modelled.
We hope to combine positive aspects of some recent algorithms (see full project description) so that we can generate conformal 3D meshes from noisy labeled multiple material volumes. The distance-field based approach of Meyer et al. is proposed as starting point, but care will have to be taken to avoid the macro-smoothing seen in some existing implementations. The output of this this research will directly be used in an ongoing research project at the LUMC in Leiden.
We suggest that algorithms be developed using Python, VTK and ITK in such a way that it can be integrated in the DeVIDE runtime environment (DRE).
More information
A complete project description can be found here.
If you are interested in doing your M.Sc. project as part of this research or you would just like more information, please contact Francois Malan at <fmalan@medvis.org> or Dr Charl P. Botha at <c.p.botha@tudelft.nl>.