[Neurobureau-hubs] Preprocessed data ready to be released ... and a brainstorming phone call ... and an expert discussion

[Cameron Craddock]

Hello!

I would like to schedule a phone call for this week.  The goal of this
phone call is to be a brain storming session to discuss analysis
ideas.  Let me know if you can make it and when.  Would Wednesday work
with everyone?  So far I have scans scheduled until roughly 2100
germany time, so 1530 New York time is the earliest I can make it.

Also I really like  the idea of having an ADHD expert give a short
talk on neuroimaging results from ADHD, as well as some of the
underlying neuro-hypotheses related to ADHD.  I wonder if Pierre's
expert-friend or Daniel could do this?  Or even better maybe MM or X?
This might be a nice thing to record and distribute on the ADHD200 web
site ....  but for now lets try to plan the who/what/when for this.

I have finished preprocessing the data and would like your comments on
what to release.  I want to release tomorrow if I can, so please let
me know of any concerns that you might have before then.

What I have planned to release so far:

The preprocessing script

and for each session/rest/subject

SS T1 in template space (MNI template of 4-18 year olds)
SS T1 grey matter in template space
unfiltered preprocessed EPI in template space (4x4x4 isotropic)
filtered preprocessed EPI in template space
mean EPI in template space (the preprocessed EPI's are demeaned in the
nuisance regression)
motion parameters
Extracted timecourses for Smith's 10 RSNs (results of spatial regression)
FC maps generated with the above timecourses

Above is 80GB, and we have 100 GB available on NITRC, so I probably
will not be able to release much more than this.

The next step that I am planning is a functional parcellation into
ROIs and then release timecourses extracted for these ROIs.

Also, this data was not QC'd by me at any level.  So we will have to
systematically go through and verify that the co-registration worked
and that the motion isn't too bad.  I would like to make this a
communtiy effort, but I am not certain that will yield results that
are quick enough for us.  So ... it would be nice if we could all look
at about 100 brains a piece to evaluate the coregistration and look at
the FC maps to make sure that there isn't anything too fishy with the
data.  Please also let me know if there are other more automated QC
methods that you prefer.  I know that MM has some success with using
falff for QC, I will work on generating those maps.


Cheers,
Cameron