ROI registrations for Tractography

Hi all,

I have manually segmented my target region volumes from at atlas that I found and its in Talairach space. However my seed region is in MNI. I have run eddy and bedpostx and so I was wondering if I am to register these volumes to the nodifbrain mask from the BET, or if I am supposed to first register the mask to the standard space. I know that tractography results get stored in diffusion space but I need the output matrices as well so I just want to make sure I get the process right.

I am a bit confused by some of the wording:

Unless you spatially normalized the data somehow, eddy, bedpostx, and BET outputs should all be in the same space.

Ideally, you should get the seed and target region in diffusion space for each subject. Spatially transforming diffusion data is difficult because you need to preserve directionality info. Much easier to just get the masks in native space.

Keep in mind that in this case you might want to scale your matrix edges by the inverse volume of the node volumes (seed and target). That is, nodes that become larger after transformation to diffusion space will tend to have more streamlines running through them than smaller nodes.

Hope this helps,

Sorry about the confusion

the website only details the procedure for the example data so I am trying to follow along with my own. It seems that the website wants the data to go through the two step registration however I did not think that was necessary since my nodif brain mask is already in diffusion space along with my data, however I need to obtain the standard2diff.mat transformation in order to have the rois in the diffusion space. I have read that I can take the inverse from the diff2standard.mat but how do I obtain this one to begin with using FLIRT. I tried already and when the ROIs were overlaid on my mask, it was not a good fit. This may be due to the nodes you mentioned though. I can look up how to scale my matrix if that is the issue. Thank you so much

It would help more to know what website/analysis you are trying to replicate.

What inputs did you use for FLIRT (e.g. what was the static image/reference and what was the moving image?

I am using FSL.

For Flirt I have:
flirt -in OFC_MTLt.nii.gz -ref nodif_brain_mask.nii.gz -out OFC_MTLt2.nii.gz ( input is ROIs, reference is brain mask created from bet/eddy).
This is the output volume

For FLIRT you would want to first calculate a transform between the template space (MNI or tailarch) and your diffusion space. Ideally, that would mean using a template T1 image, and registering to your subject’s T1 image (if using a tool like QSIPrep to preprocess your data, your diffusion and T1 outputs will be aligned). In a pinch, you may be able to align the template T1 to your subject’s fractional anisotropy map. Then you apply that transform to the ROI mask images. After which, you may need to re-binarize the masks using fslmaths.

Thank you Steven!
Im not particularly in a pinch, it’s just that I did not run DTFIT. I can do what you suggested though as it seems like it would be more effective( Im using HPC). The seeds and the targets need to be in the same space, which I also need to align my Talairach target regions to an MNI template since my seeds are MNI.

Rather than warp the Talairach target twice (that is, Talairach → MNI → Native), you should simply warp Talairach → Native.

Thank you for your help, I was able to register my seed volumes following your advice and i ended up resampling. The resampled OFC_MTLt2.nii.gz is still not aligned well so I think Im going to try to align to MNI first then native. This is the link where the atlas came from Scalable Brain Atlas - Neuromorphometrics Inc. manually segmented brain

I used the template glm.nii.gz file. I think I need to do some more conversions.

Thank you again, your feedback was extremely helpful