Hi Francesco,
IMO you don't need a DOF correction in this scenario. But I may be wrong.
However, I would not use the thresholded maps. Instead you may want to use your unthresholded zstats or p-maps (based on nonspatial mixture modelling, as implemented in melodic).
HTH-
Andreas
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Francesco Musso [[log in to unmask]]
Gesendet: Mittwoch, 16. Dezember 2009 17:51
An: [log in to unmask]
Betreff: [FSL] ICA and GLM maps spatial similarity
Dear FSL community,
We currently have a paper under review, its about EEG driven BOLD-modeling
in a simultaneous EEG/fMRI resting state setting.
One of the reviewer asked: "a method should be applied which objectively
demonstrates the similarity between the identified EEG-dependent fMRI
patterns and RSN established in the literature to back up the claim that
such specific RSN are EEG-correlated. After having done this it should be
listed which electric field distribution (per subject) corresponded to which
RSN and to what degree (similarity measure)."
In the paper we used MELODIC to identify fMRI native RSNs and provided a
‘‘functional’’ similarity measure to our EEG derived BOLD maps. We selected
the design.mat and design.con of the EEG-GLM analysis performing a post-hoc
regression analysis on estimated time courses to identify whether or not a
given BOLD IC was related with a specific recurrent ‘‘landscape’’ of the
brain electric field (simple F-test and total model fit).
Apparently the reviewer is more interested in the spatial similarity,
therfore I started to play around with "fslcc".
My intention is to compare at the single subject level the
"thresh_zstatXX.nii" z-maps from FEAT and from MELODIC to obtain the
Pearson's spatial cross-correlation coefficient.
Now my questions:
1) which correction for the spatial degrees of freedom should we apply in
order to obtain the corresponding P value?
2) is this procedure right or do you suggest other similarity masures?
Thank you in advance
Francesco
|