Collaboration:GT-UCI

Georgia Tech (Core 1) & UC-Irvine (Core 3) Collaboration

Logistics

Dates: December 17, 2005
Place: Georgia Tech
Attendees:

1. Allen Tannenbaum (Georgia Tech,Core 1)
2. Delphine Nain (Georgia Tech,Core 1)
3. John Melonakos (Georgia Tech,Core 1)
4. Ramsey Al-Hakim (Georgia Tech,Core 1)
5. Tauseef ur Rehman (Georgia Tech,Core 1)
6. Shawn Lankton (Georgia Tech,Core 1)
7. Eric Pichon (Georgia Tech, Core 1)
8. Jim Fallon (UCI,Core 3)

Agenda

1. All: Discussion with Jim Fallon on neuroanatomy, in particular to get more deep insight on the striatum and DLPFC/DMPFC structures and the rules. Write biological motivation section of the DLPFC SPIE 2006 paper.
2. Delphine and Ramsey show Jim how to use Slicer for manual segmentations and obtain expert ground truth segmentation of a few slices of the Putamen, DMPFC and DLPFC structures on 5 cases. This will be used for the validation part of the DLPFC 2006 paper and future papers.
3. Tauseef shows Jim the DMPFC/DLPFC segmenter Matlab program, in order to validate the current segmentations and get input from the prototype. If time permits, produce more segmentations with Jim's help.
4. John shows Jim the Putamen segmenter program, in order to validate the current segmentations and get input from the prototype. If time permits, produce more segmentations with Jim's help.

Workshop Results

Specific Accomplishments

1. Allen trained Jim on 3D Revolver Software
2. Tauseef and Ramsey demonstrated the Matlab version of the DLPFC/DMPFC rule based segmentation software.
3. Delphine demonstrated the 3D Slicer viewer of raw data overlaid with labelmaps.
4. Jim presented lecture on the "Prefrontal Circuits".
5. Jim manually segmented the DLPFC for 5 cases, will be used for validation.
6. Jim manually segmented the Pulvinar, and Medio-Dorsal (MD) Part of the Thalamus for future work.

Important Discussions

• Common Knowledge When Segmenting Multiple Areas: For the Slicer RuleBacsedSegmentationModule, the user should have the option to segment single structures or combinations of structures. In the case of segmenting a combination of structures, the Module will have to be able to merge common information in an intelligent manner. For example, when segmenting the DLPFC and the DMPFC, it is important to recognize the common boundary shared by the two regions, instead of computing that boundary independently for each structure.
• Bayesian Classification Options: When performing Bayesian Classification, we should include a slide bar to allow the user to adjust the kmeans calulated means and standard deviations of the classes.
• DLPFC Buffer Zone: There is a likely zone of interest surrounding the DLPFC which is a buffer between the DLPFC and surrounding regions. In the DLPFC SubModule, provide the user with the option to create this buffer region as a percentage of areas on each side of the DLPFC cortical surface rhombus.
• General Zooming Options: Provide the user with the ability to zoom arbitrarily when performing the segmentations and do not base region of calculation on the level of zoom.
• Local Probabilistic Morphological Operations: Provide the ability to dilate/erode areas of the resulting segmentation in order to manually tweak results to better fit the data.
• Striatal Bridges (in the Interior Capsule): Provide the option to segment the gray-matter striatal bridges from the interior capsule when segmenting striatal areas.

Action Items

• GT will implement Fallon's Rules to Rule Base Segment all the areas of the prefrontal cortex, the pulvinar, and the MD.
• Jim will continue to provide expert manual segmentations for these areas.

Pictures

Neuroanatomy can be fun! Lots of drawing, a little bit of connecting the dots, some head scratching, but in the end, we all learned the ABC of neuroanatomy :)

To see full size, click on the bottom right icon below each picture.