While performing some SBC analyses I realized that the default analysis mask in CONN (i.e., MNI152 2mm isotropic) did not include some PFC regions (e.g., forntal poles) and was therefore too restrictive for my data. My data was preprocessed using fMRIPrep and is in MNI152 space at 2mm isotropic resolution. I found via some old posts that it is not possible in CONN to use a different analysis mask for each subject based on searching the forum and playing around with the GUI. To ensure I wasn't removing any brain regions across my sample due to my analysis mask, I took the union of all of my participants BOLD volumes. However, this came with a tradeoff. The tradeoff is that although susceptibility distortion correction was performed on all of my data, the effectiveness varies and some subjects still have some shearing near the occipital and prefrontal cortices as well as some superior regions (e.g., pre and post central gyrus). Throughout my analyses, I have had a mix of some significant clusters containing mostly brain matter and some containing mostly non-brain matter. Early on I decided that I was going to ignore any significant clusters that contain more than 50% non-brain matter as determined by the label "... covering 0% of not-labeled with center at ..." and visual inspection of the center coordinates on an MNI152 atlas. However, this is concerning if I want to use the cluster size p-values to do my statistical inferences.
This leads me to two questions:
- Is there currently a way to use individual analysis masks for
each subject? If there isn't, are there recommended best practices
to handle the masking issue I explained above?
- When we get significant clusters that contain non-brain matter,
is there a recommended way to adjust our cluster size p-values to
assess if the cluster is still significant after removal of
non-brain voxels?
Any help with these two would be greatly appreciated! Also I am using CONN22v2407.