Hi Volker -
On 25 October 2010 14:06, Volker Baur <[log in to unmask]> wrote:
> Dear forum members,
>
> I am currently exploring fMRI resting-state data with MELODIC ICA. I've set the number of components to 9, highpass filter: off, highpass filter cutoff: 100, the remaining was left default.
> After completing a few test runs, I have unfortunately not obtained "reasonable" results. Just was not able to determine well-known components like the default-mode network, both on the single-subject level (which is the primary goal) and on the group level (10 subjects) using multi-session temporal concatenation ICA. It looks like there are many artefactual components, even though MCFLIRT was applied. Are there preferable settings in MELODIC for resting-state data? Might there be a systematic error?
>
Is this single-subject Melodic or group? How much data do you have?
You're usually best to let Melodic estimate the number of components
itselt. 9 is very low.
Eugene
> Would be great if anyone could help!
> Kind regards,
> Volker
>
>
|