Dear CONN experts,
I am doing a ROI to ROI connectivity analysis in CONN (v2020b).
Along with the "anatomical" AAL parcellation, I would like to use other "connectivity-based" brain parcellations (Schaefer and Brainnetome atlases, in particular).
However, I could only find versions of them in the "FSL-MNI" space (https://github.com/ThomasYeoLab/CBIG/tre...) or the "HCP-MNI" space (https://atlas.brainnetome.org/), which, as far as I know, are slightly different from the version of the MNI space implemented in SPM12 (and CONN).

I am aware that differences are very minor, but I was thinking if this discrepancy may be relevant for my purpose (i.e. to obtain full ROI-to-ROI connectivity matrices with different atlases). If yes, are you aware of versions of these atlases in the SPM-MNI space (or of any other meaningful method to overcome this issue - e.g. registering the atlas to SPM-MNI?)?

Also, does the imported atlas ROIs have to be at 1mm-isotropic resolution (like the CONN built-in atlases) or are other resolutions (e.g. 2mm or 1.25mm isotropic) equally acceptable?

Thanks in advance for your help,
Giuseppe

Dear Giuseppe,

In practice those differences are often considered sufficiently small not to be worrisome, perhaps mainly because there are many other potential sources of differences (e.g. depending on details of acquisition, details of your sample, etc.) that can be reasonably expected to produce similar or larger differences between the results of normalization in your sample from the results of normalization in the data used to generate the atlas. That said, one possible/reasonable approach to try to minimize those differences would be to normalize in SPM a reference anatomical image in the atlas space (e.g. MNI152 NLIN 6th generation for FSL?) and then apply the same normalization transformation to your atlas file (e.g. using nearest-neighbor resampling to avoid breaking the structure of the data labels) in order to bring it to the same MNI-space as your data (e.g from FSL-MNI to SPM-MNI).

And regarding your question about resolution, atlases and ROI files can be saved in any resolution, as long as they are properly co-registered to the functional data (this information is stored in the nifti header files, encoding the transformation between voxel coordinates and "world" coordinates) the details of voxel-size, image bounding box, etc. of your files can be arbitrary.

Hope this helps
Alfonso

Originally posted by Giuseppe Pontillo: