Hey guys !
Is there a problem with registering a partial brain coverage functional images to a full view anatomical ?
Because the result in SPM coregistration is bad and crooked .
by the way the dataset I’m working on is “ds000113c” available on openfMRI .
For what it is worth, I have worked on a dataset (ICBM from the 1000 functional connectome project) where many functional runs have partial brain coverage, with the field of view cropped both ventrally and dorsally. I was using minc tools instead of SPM, but the registration often failed, and I have kept using that dataset to test my pipelines for that reason. May be worth checking how boundary based registration would work on that.
2 Likes
thanks for the help !
Have you tried attempting to register to a skull stripped version of the anat? You could keep all your analyses in SPM this way. You can generate this by creating an SPM batch that runs SPM: Segment on the whole brain anatomical dataset, and then using ImCalc, with the dependencies set to Bias Corrected, c1, c2, c3. These correspond to GM/WM/CSF masks (range 0 to 1).
For the ImCalc expression, use i1.*((i2+i3+i4)>0.2). This is assuming that the first image in the dependency is the Bias Corrected whole brain, don’t have SPM handy. The .2 just means that after adding up the GM/WM/CSF masks, use anything that is above a value of .2. Usually ends up with just a nice, skull stripped brain. You can then use that skull stripped anat as your coregistration target.
BBR is also likley to work, and FSL wants you to skulls strip anyway. I find SPM’s segment produces more reliable skull stripped brains, but BET is also an option.
Good luck!
1 Like