I am working on a rat dataset where I am processing both fMRI and DTI to create connectivity matrices and performing graph theory analysis using BCT and NBS. I am trying to perform ACT in rats as a comparison to whole brain tractography (whole brain was successful). I don’t have T1 data but have T2 data and am using the SIGMA Atlas have been working to generate segmentations of GM,WM, and CSF. From there I will concatenate and create the 5tt files necessary for ACT.
I have a quick question that stems from a very old thread on this forum:
and specifically the suggestion listed by @rsmith of:
“Add your ‘cortical GM’ segmentation to the ‘sub-cortical GM’ volume, and use an empty volume for the ‘cortical GM’ component of the 5TT image. This will allow streamlines to propagate into this tissue.”
My questions are:
Is this still recommended?
In practical terms, how does this differ from the next suggestion which simply uses a CSF mask as an exclusion map?
I am trying to perform ACT in rats as a comparison to whole brain tractography (whole brain was successful).
This statement doesn’t quite make sense: one can do whole-brain tractography with or without ACT, and one can do targeted tracking with or without ACT.
Is this still recommended?
I don’t think that I’ve ever actually “recommended” this approach; indeed, if it were my recommendation, I’d have modified the corresponding generator scripts to do this by default. But that is primarily motivated by the empirical tractography behaviour in human data. In conjunction with something like Surface-Enhanced Tractography I’d be more in favour of allowing streamlines to propagate beyond the interface, but in the absence of such I simply don’t trust the resulting trajectories. But that’s based on the very sharp curvature of fibres at that interface in human data. For animal data, I simply don’t have either the prior knowledge or the experience; hence why I’ve made the suggestion that it’s something that people can consider using in such cases. So I can only suggest generating 5TT images using both approaches, performing some tracking, and evaluating the differences for yourself.
In practical terms, how does this differ from the next suggestion which simply uses a CSF mask as an exclusion map?
To be clear, the comparison here is presumably “this” being the use of ACT, with the SIGMA GM segmentation assigned to the second volume of the 5TT image, vs. not using ACT, and instead just using a brain mask and a CSF exclusion mask. In that case, the differences are:
In the former case, the tissue images will be interpolated, whereas in the latter case the images are interpreted as masks;
In the former case, streamlines will be forbidden to terminate within the WM, whereas in the latter case this is acceptable.
In the former case, streamlines will not be allowed to go from WM, into GM, and back into the WM again, whereas in the latter this is acceptable.