Oct 08 2009 meeting: schedule for Oct 9

With Ron:

• what caused the hole in Valentino data? Ron: the impression is that this is a cyst

• check the quality of N3 bias correction. Ron: recovered bias field makes sense.

• learn how to manually segment sub-cortical structures -- need 3-class segmentation to do this, not ready yet

With Sandy:

• discuss EM segmenter -- talk with Chris, go through all recommendations for Ginger to take home

• workflow -- how much benefit we can expect if we make it more complex, is it worth it (e.g., do EM segment on each subject, then re-run atlas construction without bias field, considering there is improvement)

• can we think of quantitative criterion for bias correction with N3? -- we could script and run parameter exploration. Sandy: can possibly check the result by running classification, compute gaussians for each class, and calculate classification error as a metric.

• check the quality of the smoothed data. Sandy: run EM on the non-smoothed data, because boundaries may have been affected, and MF should take care of Gaussian noise.

With Chris:

• information about the second timepoint data / data transfer issue. Data transfered. Verified image quality for the second time point -- not better than the first one, based on Calvin images.

• confirm there are no intellectual property issues for the workflow tools originating from VT -- can we put everything in NA-MIC sandbox to maximize reuse later? Confirmed

With all:

• look at the current results of the initial registration (up to affine) -- Registraton ok, but Tommy is not a good template candidate (motion and susceptibility artifacts; head shape not normal, possibly to premature birth). Switch to Gucci.

• build up the list of non-slicer3 tools / scripts, discuss sharing of code and data (data -- after the publication(s) are out)

• discuss protocols on documentation;

• conform proposed work flow;

• discuss time-line / winter NAMIC meeting/ potential publications. Confirmed meet Chris and Ginger at SLC. Try to arrange phone call with imaging people before that.

Some other points:

• consider XNAT for storing raw data