I currently have alligned and upscaled some rois from the HCP atlas to my subjects’ T2w space. However, I will need the rois to do a diffusion analysis and therefore I will need to register the ROIs with the DWI scan, which does not have surface files.
How can I do that?
I was thinking about creating surface files from the DWI scan but it seems unpractical.
If you are using FSL probtrackx then you can use structural space ROIs I think? Either way you need the volumetric transform from the subject’s T2 image to it’s diffusion space.
Thank you for the answer.
Yes, I will use FSL probtrackx. Shouldn’t I register the GIFTI labels to the diffusion space? And if I have to, how can i apply a warp to a surface file?
the masks can also be provided in a different space by selecting Seed space is not Diffusion on the GUI Data tab or using the --xfm and/or --invxfm.
So you shouldn’t transform the surface ROIs you should leave them in structural space and provide the diffusion to structural transformation, or structural to diffusion. Not sure which way round they ask for - best to check with FSL. Maybe this registration is what is causing you problems - you can check how good the ROIs are in diffusion space by mapping them first into the structural volume using wb_command -surface-to-volume-mapping and then propagate them to diffusion using your volumetric BBR diffusion to T2 registration. TBH this proved highly non trivial for us during dHCP processing so it’s highly worth carefully checking (or double checking the released files to see if it is provided for you)
I left the ROIs in the structural space and I ran probtrack with both the direction of transformation and without them but it returned the same newmat:indexException error for always the same ROIs while others work.
I will try wb_command -surface-to-volume-mapping to check what could be wrong with those problematic areas.
I have a question regarding the BBR diffusion to T2 registration:
In the released files, within xfm folder a trasformation between dwi to T2w and viceversa.
I assume those are the BBR transforms you are referring to given the fact FLIRT does the BBR. Is it right?
I think checking the ROIs is the way to go right now. I will follow you suggestion.
I think so but maybe start a new post on that in order to raise support from the people who worked on them