The grand mean scaling is done before ICA+FIX.
Peace,
Matt.
On 7/20/16, 4:40 PM, "FSL - FMRIB's Software Library on behalf of Benjamin
Ely" <[log in to unmask] on behalf of [log in to unmask]> wrote:
>Following up on this thread - any advice on how to quantify the variance
>removed from a particular region? One simple approach would be running
>something like:
>
>fslstats DATA_noFIX.nii -k MASK.nii -S
>
>and comparing to
>
>fslstats DATA_FIX.nii -k MASK.nii -S
>
>But I don't think this will work for the HCP datasets, since the
>whole-brain scaling to a mean of 10,000 +/- 3100 is apparently done after
>ICA-FIX for cleaned datasets. When I tried this in a test subject, it
>showed an increase in SD within my masked region following ICA-FIX (from
>1959 to 2210), which doesn't seem all that plausible.
>
>Maybe instead there's a way to make use of the individual component maps?
>
>Many thanks,
>-Ely
|