Hi,
> We are trying to compare the default networks using the IC components from two different resting-state series of one single subject.
This is going to be at best descriptive - you won't be able to do a proper statistically valid comparison easily. You should think carefully about what you mean by 'comparing the default networks', e.g. are you interested in areas present in one but absent in another run or do you care about differences in amplitude (in the F-statistic sense)?
> We have tried three different setups:
> - Performing Single ICA in each subject is quite unmeaningful.
Yes, in as much as it would be in task FMRI - you simply do not have sufficient repetitions to perform a valid comparison.
> - Concatenating the series as a single 4D to use together in a Single ICA. Does it make any sense to do things in this way?
Technically possible, though see above, you need to define your question first
> - When trying to make Group-ICA multisession temporal concatenation, it didn't appear any mistake but the mixture modelling to compute z-maps only returned the IC 1 although in melodic_IC.nii.gz appear all the components... Do you know what it is happening? are two series very few for group ICA?
Did melodic complete, i.e. does the log.txt file end with 'finished'? If not, it must have crashed...
hth
Christian
>
> Does anyone have another idea?
>
> Thank you very much in advance.
>
> Cheers
>
> AnaBea
|