Has anyone on the run group-wise analysis on the HCP resting state data,
and if so what tools did you use? I am having memory issues when running
more than 10 subjects and I was wondering if anyone has a way of getting
around the large memory requirements when concatenating in time.
Thanks!
Thanks Matthew, I'll check out the latest version of FSL. Is there a paper
the discusses the MIGP method?
On Thu, Jan 29, 2015 at 7:24 PM, Micah Chambers micahc...@gmail.com wrote:
Well the resting state data provided through the HCP around 1GB each so
I'm guessing the memory requirement is a
If you update to the latest version of FSL the MIGP algorithm may be in there and is designed to help with this. The main limitation with melodic has been that it was single threaded. I don’t know if that has been addressed yet.
Peace,
Matt.
From: Micah Chambers micahc...@gmail.com
Well the resting state data provided through the HCP around 1GB each so I'm
guessing the memory requirement is a good bit more than 8GB now, unless the
PCA is performed per subject. Has anyone done Group ICA on datasets this
size?
-Micah
On Thu, Jan 29, 2015 at 6:40 PM, Glasser, Matthew
Yeah, I sort of need to do whole-brain.
-Micah
On Thu, Jan 29, 2015 at 7:33 PM, Glasser, Matthew glass...@wusm.wustl.edu
wrote:
Finally I’ll note that MIGP and lots of RAM are not needed when using
parcellated datasets (though if you want ICA, you probably aren’t planning
to run it on
Finally I’ll note that MIGP and lots of RAM are not needed when using parcellated datasets (though if you want ICA, you probably aren’t planning to run it on parcellated data).
Peace,
Matt.
From: Glasser, Matt Glasser glass...@wusm.wustl.edu
Date: Thursday, January 29, 2015 at 9:32 PM
http://www.sciencedirect.com/science/article/pii/S105381191400634X
We’re working on some improvements for it when used for functional connectivity (i.e. all to all correlations), but it works fine as is for ICA.
Peace,
Matt.
From: Micah Chambers micahc...@gmail.com
Date: Thursday,
Thanks Matt
That's correct that the --migp option (see a recent post from me to the FSL
list about this new option) inside melodic is still single-threaded.
However the journal paper (linked below by Matt) includes simple matlab code
for doing MIGP and that can easily be parallelised as well