Particle::todo

2/18/2011

Sungkyu and Bea working on using a proper entropy for matching normals
- fitting great circle to the normals at each vertex, using both residuals and location on the circle as two entropy terms
- Sungkyu will send his code to Bea for initial experiments
Some problems with mesh to volume (scan conversions) in Christine's ventricles and the liver shapes

New member: Lucile
Minutes from the namic meeting are on the NAMIC wiki
Utah's IPMI submission
- Precompute and store the inter-vertex geodesic distances
- Includes normal directions
- Stanford meshing library - Bea will investigate
Next steps
- Not user-friendly software yet, Manasi
  - We should be able to apply this to the mouse cortex (Joohwi)
- Move to NITRC SVN - Clement created a project, we should start populating the documentation
- GenerateCLP version
- Quality control in slicer3 - Bea & Lucile
- Landmarks - still unclear who's doing this

Joohwi - mouse cortex has flip problems even with directionality features - he'll try taking out the problematic subject and increasing the weight of directionality feature
Lei - had problems with one of the subjects, hopefully ok now
Bea - atlas-based initialization
- inflate meshes, scanconvert, areg, atlaswerks, parcellate
- directionality feature - demo data so we can start thinking about writing a paper
Clement - sulcal depth attribute with gambit - running tests

New time and place - fridays 1pm in SN 284
Joohwi - mouse cortex has flip problems. Procrustes crashes
- Bea will try getting shapeworks to run with attributes so we can test its procrustes implementation
- Joohwi will try MeshMath procrustes to circumvent the need for procrustes inside particles
Bea has been working on Rohit's mandibles
- Very jagged surfaces, so curvature computation is problematic - using MeshInflate

Two new members: Greg (BME - heart imaging) and Lei (Steve's new RA - liver)
Mandibles - Correspondence with curvature only (no xyz) failed miserably
- we should talk to Marc about the heat flow if we can't get rohit's tool to work
- bea will try rohit's heat flow
Joohwi wrote his own version of normal directions projection
email shapeworks tutorial to greg, lei - DONE
email bea wiki address - DONE

Joohwi will start coming to our meetings
Xiaoxiao had trouble getting the general attribute version working - you have to "hold min variance" in the old version; you also have to set the "min variance" to a value on the same order of magnitude as the largest eigenvalue (printed on the console as 0: ... 1: ... 2: ...)
Beatriz has very nice features to stop the flipping (dir-x). The initialization is still a problem because the 6 extreme points can start out at very different locations on different surfaces in the population. Her next plan is to segment the mandible using an atlas, so she can initialize using the segmentation.
There is a concern that 2 surfaces are not enough for the entropy method to make sense - so she's using 2 additional surfaces that are just 1-dilated versions of the two surfaces.

Xiaoxiao's lungs look well with "simple" bronchi but when the structures get too close, there are problems. She will try the kick-out-particles-with-bad-normals approach from Bea.
Bea will email Josh&Manasi to discuss what to do with the kick-out code - include it in the GUI, maybe? It's in MeshMath for now.
MeshMath also has C and S computation.
Scaling the meshes shouldn't matter, because the attribute weights are taking care of the scale.

Xiaoxiao's meshes a little bit better with subdivision, but she didn't try the projection to surface yet
Clement's meshes look much better - he has 5-25 tiny triangles (area<0.0001) but otherwise
Beatriz has been working on kicking out bad particles (inconsistent normals)
Soft tissue project (face) is going well
ShapeWorksGroom might have a bug - the last shape seems to be treated differently (Bea is looking into this).
Rohit: antialiasing seems to be working very well

Rohit's going to try using his own distance maps (danielsson distance map from binary segmentation) and the attribute volumes created from preprocessing.
Bea's new project: lip distortion before/after brace removal. Just minimal entropy. Particles are working great, and so is TPS!
Ipek: Project particles to the original surface.
No meeting next week (good friday)

TPS works now
Bea's meshes look better than before, but she still has particle-flip problems. She'll look into throwing out "bad" particles when doing TPS.
Xiaoxiao's data: looks good but the flat surfaces get un-flat in the postprocessing. She'll try to subdivide her big triangles and see if it helps.
Rohit pointed out that one should be careful when using TPS because it can cause inverted triangles, etc, because it's not necessarily diffeomorphic.

Clement will try out: surface reconstruction w/ thin plate splines from Rohit's tool
Bea: Particle Initialization
What does "ratio" do? look up in shapeworks manual
Everyone: Write more stuff in the wiki
Bass compilation still not working (argh) - Clement said 3.10 itk, 5.2 vtk, GenerateCLP /opt/local/GenerateCLP
Rohit: antialiasing