1 Challenge the future Conformal multi-material mesh generation from labelled medical volumes
Conformal multi-material mesh generation from labelled medical volumes (Dec 2012)
Jul 15, 2015
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1 Challenge the future
Conformal multi-material mesh generation from labelled medical volumes
2 Challenge the future
Introduction
• Generation of volume meshes for FEA
• Particular use case: hip prostheses analysis
• Typical pipeline:
Segmentation from patient’s CT-scan (a) to labelled volume image (b). Volume Meshing (c) of the image and FEA for stress-strain results (d,[Dick2011]).
3 Challenge the future
Introduction
• Mesh requirements:
• precise meshes
• segmentation-conform
• minimal mesh element number feature-adaptive
4 Challenge the future
Related Work
Weighted Delaunay Tetrahedralization refinement [Boltcheva2009]
Dynamic Particle System Meshing [Meyer2007]
Multi-labelled volumes meshes with particle systems [Meyer2008]
5 Challenge the future
Challenges
• long computation time • oversampling of edges and corners • no sharp-feature recreation ε-sample requirement
wrong topology, bad
reconstruction
too many samples
6 Challenge the future
Contribution
• Application of Integer Medial Axis (IMA) as fast, discrete
medial axis scheme
• proposal of local surface triangulation scheme for volume
images
7 Challenge the future
Integer Medial Axis - Analysis
8 Challenge the future
Integer Medial Axis - Idea
BioMesh3D – Centres of
maximal spheres
IMA – shortest path in feature
transform
9 Challenge the future
Integer Medial Axis – Results
Runtime
dataset BioMesh3D DeVIDE FE-Mesher
artificial 26 min 0.1 sec
Tooth 1h 41 min 1 sec
real femur 14h 11 min 2 sec
10 Challenge the future
Integer Medial Axis – Results
Quality
Tooth # Tetra 142795 DeVIDE FE-Mesher
Max. Min. Avg. Variance # bad
Tetra
%
bad
Aspect
Ratio
119.69 1.01 1.94 1.00 9282 6.50
Radius
Ratio
105.43 1.00 1.69 0.83 6736 4.72
Volume 272.34 0.0 2.96 21.74 0 0.0
Tooth # Tetra 118110 Simpleware FE+
Max. Min. Avg. Variance # bad
Tetra
%
bad
Aspect
Ratio
44.59 1.02 1.54 0.16 816 0.69
Radius
Ratio
921.55 1.00 1.37 7.92 997 0.84
Volume 46.68 0.0 3.41 14.24 0 0.0
14 Challenge the future
Integer Medial Axis – Results
Precision
16 Challenge the future
Minimal Sample for accurate Meshing Concept
• ε-sampling:
• ensures topologic conformity
• applies to dense and sparse
samples
• Loss of sharp features
• only applies for 3D meshes
without additional information
• our idea:
• mesh surface locally
• take surface mesh to
generate volume mesh
0,, xxBESx
17 Challenge the future
Minimal Sample for accurate Meshing Concept
1. Get TBN-Matrix per sample
vertex
2. Get Neighbourhood per
vertex
3. re-project points in
tangent plane
4. mesh via Local Delaunay
Triangulation tangent
plane neighbourhood
5. use established
connections in 3D
18 Challenge the future
Minimal Sample for accurate Meshing Results
VTK CGAL – no constraint CGAL – Convex Hull constraint
formation of holes due unsuitable Neighbourhood determination
19 Challenge the future
Conclusion and Future Work
• Improved runtime behaviour due to Medial Axis Transform
Algorithm change
• Local Triangulation in tangent space not ε-sample bound, but
dependent on Neighbourhood operation
• k-Nearest Neighbour not suitable for non-uniformal, sparse
samples
• In future: usage of natural neighbours for neighbourhood
determination
Related Documents
