July31T-con

In attendence (add your name if you were there but I didn't note it below): Guido Gerig, Ron Kikinis, Casey Goodlett, Randy Gollub, Ross Whitaker, Dennis Jen, Lauren O'Donnell, Karl Helmer, Vince Magnotta, John Melanakos, CF Westin, Tri Ngo, Doug Markant, Sonia Pujol

1. Data format issues: Synopsis of email discussion ; Some data exploration
   1. Voxel resolution: non-isotropic native resolution vs. upsampling vs. downsampling (to be discussed and decided on common preprocessing)
      1. Native grid size is 144x144x81 [FOV: 240mm] with voxel size 1.667x1.667x1.7
      2. Provided data is 256x256x81 [FOV: 240mm] with voxel size 0.9375x0.9375x1.7
   2. Eddy Current distortion correction: both uncorrected and corrected datasets are available (we need a consensus on which to use. No comparison to determine quality of correction has yet been done.)
2. Input/Output formats
   1. Input/Output image volumes should be in nrrd format using the standard NA-MIC DTI conventions [NAMIC_Wiki:DTI:Nrrd_format]
   2. Streamlines should be in vtk format [legacy or XML vtk format] including the tensor information at each point along the fiber tract
   3. Other output formats (such as statistics spreadsheets, graphs, etc. ) should be documented by the groups

Action Items:

1. All algorithm developers will make sure the wiki page for this project correctly relfects the details of their algorithm and participation.
2. All agorithm developers will run their software on the data as described above by Guido.
3. All agorithm developers will keep laboratory notes on as many details of the processing pipeline as they can and summarize this information in a Lab Report to be shared with the group in Sante Fe. The metrics that were mentioned included: detailed methods, processing steps, time to run each step, user interaction or tweaks that needed to be made, size of resultant files, processor requirements, etc.
4. All algorithm developers will generate final results both by subject and group with output measures that they deem most informative. The summarized information including images of the results and any metrics to be put into a final Powerpoint report to be shared with the group in Sante Fe.
5. The above two items should be a single presentation.
6. Marek and his group will add specific clinically relevant hypotheses for how each tract differs between healthy and matched schizophrenic subjects.
7. Algorithm developers will use the page to post their ideas for statistical approaches to summarize, compare and contrast the data.
8. Sonia, with support of her mentors, will draft an outline to be used to guide our discussion in Sante Fe
9. Randy will extend an invitation to Carlo Piepoli, Susumu Mori, Allen Song and Tim Behrens and amend the wiki to let interested outsiders contact her to inquire about participation. She will ensure that the meeting retains its focus on NAMIC work but welcomes outside expertise and potential collaborations.
10. Agenda for the 2 days has been updated and is open to further discussion from the group.

Miscellaneous ideas:

• Ron/Ross suggested that this Conference be a "dry run" for a future MICCAI session on this same topic.
• Another question to be considered at the meeting is whether this particular dataset is the optimal one to use for this work. Probably we are fine for Sante Fe, but going forward, should we consider a dataset that was collected with a isotropic voxels and other considerations raised during our discussion?

Follow up:

• Marek provided details about the clinical relevance of the 5 chosen tracks to the study of schizophrenia, the information is entered on the ROI Definition page and on these File:Comparison project.ppt slides.

Return to Contrasting Tractography Project Page
Return to SanteFe Tractography Conference Page