Group ICA workflow

[thomas.nickson]

Hi all,

I'm trying to put together a group ICA workflow but I'm not sure on how to go about doing it. The problem is that I want to have two groups, I want to perform the same preprocessing and registration stages on all of the subjects but after that I want to split them in to two groups. As a graphical example, something like in the last diagram here:

http://nipy.sourceforge.net/nipype/users/joinnode_and_itersource.html

but where E is two outputs such that D13 and D14 are the output of E1 and D25 and D26 are the output of E2.

Is this possible? My thoughts were to map over two separate lists at the beginning but I have no idea as to how to keep track of them both as they got through the pipeline.

Thanks,

Tom

[Satrajit Ghosh]

hi tom,

one approach would be to run the first part as iterables, collate across everybody with a joinnode, and then split into two groups, either using a function node or the split interface.

try it on a simple workflow first to understand what the joinnode is doing.

cheers,

satra

--

---
You received this message because you are subscribed to the Google Groups "NiPy Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nipy-user+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Thomas Nickson]

I'm still a little confused as to how this would work.

So the iterable will remain ordered after the join and I can then just split them, assuming I have initially ordered the iterables in to groups I can just use a split node [10,10] to split it in to two groups of 10?

If I use a function node is there a way to do the following. I have an identity node that is iterable. This uses the names of the subjects to grab the data with a datagrabber. Later on down the line can I then look at the identity node that it tells me the subject info if I use a join node? Can I do this so that I just plug the identity node in to the function along with the list that was created by join node? I'm a little confused as to how that would work... after the join node does it still make sense to have the identity interface as an input even though this would be a newly collated list of the inputs...

Thanks,

Tom