Print

Print


Hi Leonardo,

When performing group-ICA (which you have effectively done), melodic
uses an optimisation techinique called MIGP which has the effect that
the component time courses are no longer generated.

To generate component-level time courses for your sessions, the
easiest approach is probably via dual regression.

Cheers,

Paul

On Wed, 5 Jun 2019 at 00:29, Leonardo Tozzi <[log in to unmask]> wrote:
>
> Dear Experts,
>
>
>
> I am running MELODIC on multiband resting state data for some subjects. For each subject, I have 4 runs (TR=0.71, scan duration=5 min, 426 volumes).
>
> After running the ICA concatenating the 4 sessions (I use a text file with the 4 files as inputs) like this:
>
>
>
> melodic -i REST_ICA_files_${sub}.txt -o ${sub}.ica --report --Oall
>
>
>
> MELODIC seems to run as expected. However, when I want to look at the components using FSLeyes like this:
>
>
>
> fsleyes --scene melodic -ad ${sub}.ica &
>
>
>
> I have noticed that something strange happens: when I look at the timeseries all components for all subjects seem to be fluctuating a lot in the last part of the scan, but are almost “flat” for the first 2/3rds
>
> At first I thought it could be motion or fatigue in the last resting state session, but I find it strange that it’s consistent for all components and all subjects. Also, relative mean motion for the 4 scans seems comparable. I am wondering if there is something wrong with the visualization or if really something is completely different in the fourth scan, which would be strange, because it’s collected back to back with the third.
>
> Here are screenshots of my FSLeyes window for some ICs on Dopbox, but all of them look like this (there are always ~180 components).
>
>
>
> https://www.dropbox.com/sh/ver43b0bbhknz06/AADfjlWvo04H4vwqC4xmjMOUa?dl=0
>
>
>
> Has anyone encountered something similar or has any suggestions on how to proceed? My intention was to run MELODIC to then pick noise components to train FIX on my dataset.
> Thank you,
>
>
>
>
>
>
>
> Leonardo Tozzi, MD, PhD
>
> Williams PanLab | Postdoctoral Researcher
>
> Stanford University | 401 Quarry Rd
>
> [log in to unmask] | (650) 5615738
>
>
>
>
> ________________________________
>
> To unsubscribe from the FSL list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

########################################################################

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1