Dear Sean,
I have now dilated the mask with a 9 mm sphere kernel (3 voxels). It looks like there is not more room around the data (see picture attached). Indeed, running ICA with this mask gives me 47 components. The components now do extend a bit outside the brain boundaries, but I guess this is tolerable then? I have also tried with a dilation of 6 and 3. With 6, I still get a comparable number of components (48). With 3, again only one component is estimated. So dilating by at least 6 mm could be the fix I was looking for.
To answer your other question whether I had that one component if I forced the number of dimensions the answer is no. That component really seems to come up only if I let the ICA free to decide the number of components. As you say, it's probably the piece of data cut off by the mask, that logically has a very distinct spatial distribution and hence becomes a super component for the estimation procedure... this is an interesting outcome.
Thank you for your help.
Leonardo Tozzi
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
|