help
489 Subscribers 
help > Group ICA w/ multiple Sessions?
Showing 1-6 of 6 posts
Jan 5, 2017 12:01 AM | Shady El Damaty - Georgetown University
Group ICA w/ multiple Sessions?
Will the CONN automatically concatenate multiple sessions together
to estimate group ICA? That is if I have 2 task sessions and
1 rest session, will ICA components be estimated from the
concatenated task and rest sessions or are they estimated
separately for each task?
Jan 30, 2017 08:01 PM | Shady El Damaty - Georgetown University
RE: Group ICA w/ multiple Sessions?
bumpity bump bump - any thoughts? thought id ask the question
first before spending time loading a very large dataset.
Feb 8, 2017 07:02 PM | Alfonso Nieto-Castanon - Boston University
RE: Group ICA w/ multiple Sessions?
Hi Shady,
Typically yes, it will concatenate all of your data. More precisely, you can actually control exactly which portion of your data your want to include in your ICA analyses by unchecking the 'all conditions' checkbox and selecting a subset of conditions there instead (e.g. select your 'rest' condition if that includes only the rest session and you do not want to include the 2 task sessions in your ICA analyses). Note that when selecting a subset of conditions, backpropagated subject-level maps will be also computed only for those same conditions. If, for example, you want to use only your 'rest' data to obtain ICA networks, but then you would like to look at ICA-network connectivity during your 'task' sessions, then you would select only 'rest' when computing the ICA analyses, and then you would import the group-ICA networks as new ROIs and perform any desired additional analyses on those (e.g. seed-to-voxel analyses during 'task' sessions)
Hope this helps
Alfonso
Originally posted by Shady El Damaty:
Typically yes, it will concatenate all of your data. More precisely, you can actually control exactly which portion of your data your want to include in your ICA analyses by unchecking the 'all conditions' checkbox and selecting a subset of conditions there instead (e.g. select your 'rest' condition if that includes only the rest session and you do not want to include the 2 task sessions in your ICA analyses). Note that when selecting a subset of conditions, backpropagated subject-level maps will be also computed only for those same conditions. If, for example, you want to use only your 'rest' data to obtain ICA networks, but then you would like to look at ICA-network connectivity during your 'task' sessions, then you would select only 'rest' when computing the ICA analyses, and then you would import the group-ICA networks as new ROIs and perform any desired additional analyses on those (e.g. seed-to-voxel analyses during 'task' sessions)
Hope this helps
Alfonso
Originally posted by Shady El Damaty:
Will the CONN automatically concatenate multiple
sessions together to estimate group ICA? That is if I have 2
task sessions and 1 rest session, will ICA components be estimated
from the concatenated task and rest sessions or are they estimated
separately for each task?
Feb 9, 2017 12:02 AM | Shady El Damaty - Georgetown University
RE: Group ICA w/ multiple Sessions?
Thanks for the exquisitely detailed reply as always Alfonso! I'll
give it a try :)
Feb 14, 2017 09:02 AM | Alice Yo
RE: Group ICA w/ multiple Sessions?
Hi
I also want to include only some of my conditions in the ICA analysis, but I could not find the "all conditions" checkbox in the GUI. It doesn't appear in the "first-level analyses, ICA networks" screen.
Any suggestions how to fix this?
Thanks for your help
Alice
I also want to include only some of my conditions in the ICA analysis, but I could not find the "all conditions" checkbox in the GUI. It doesn't appear in the "first-level analyses, ICA networks" screen.
Any suggestions how to fix this?
Thanks for your help
Alice
Feb 15, 2017 08:02 AM | Alice Yo
RE: Group ICA w/ multiple Sessions?
Sorry, got it- it is the checkbox that appears after you press the
"done" button in the ICA first level processing.