It seems the previous message was sent as an attachment. The same message is below -
Dear Experts,
I am looking to do an ICA that will explore FC differences between a clinical and normal population for different frequency bands (in this case, .01-.03, .03-.05, .05-.07, .07-.09). To this end I would like to filter the pre-processed data at these different frequency bins before applying an ICA.
However, my question is - if the different populations have cortical activation at different frequencies, would concatanation of the two groups eliminate any differences in the beginning? For example, if group 1 has insula activity in the .01-.03 frequency band, but group 2 does not, then wouldn't concatanation using melodic of the two groups eliminate this difference right away? Thus, when I do the dual regression, there would of course be no difference in connectivity because the two groups did not have the insula activation in common for the .01-.03 frequency band.
If I do different ICAs for each group (i.e., melodic for group 1, then melodic for group 2), then apply a dual regression for both groups using the separate melodic.IC.nii.gz outputs (i.e., dual regression using both groups compared against the melodic.IC.nii.gz components for group 1, then a dual regression for both groups compared against the melodic.IC.nii.gz components for group 2), then I would possibly be able to capture this difference in insula function.
Any thoughts on this approach would be much appreciated.
Thank you!!
Jason
|