Functional ROIs as seeds? resampling and grey matter to white matter registration

I am new to diffusion imaging and have a few basic questions

  1. How can I use the fmri ROI coordinates to get the FA and other values form specific tracts? I am not sure how that can be implemented in the diffusion NATIVE space of the participant? The fMRI ROI coordinates are in a resampled MNI space. I also have the ROI masks (6mm), but again in the resampled space. How do I register? or do I need to register? and how do I deproject/move into the diffusion space from the GM space? Any documentation/guidance in this direction is a bliss… Thank you !
  2. I plan to use SS3D_CSD now as my data is restricted (single shell,b=700 + b0 (x10)). here, my doubt is regarding the stage at which I implement the fROIs. I believe first I will do the tractography and then use the fROI masks to select only those tracts that pass through it? I need a direction here.
    Thank you much!
1 Like


Regarding 1, yes, you will need to register your fMRI and diffusion data. There are multiple ways to do this. I’d register the first volume or an average across (a subset of) volumes of the fMRI data to the average b0. This gives you transformations that you can apply to your ROIs and/or the images to perform your analysis in either space. If you need nonlinear transformations, have a look here. Otherwise there is more information for instance here or here.


Hi @mvm,

Nice to hear. It’ll be interesting to see how well it performs at such a low b-value of b=700. Consider showing some of the output as feedback over here. I can then also advise you if / how well it worked. I’ve had lots of feedback, both on- as well as offline about performance on low b-values: looking good so far, but I’m happy to keep on confirming this on a variety of data.

Follow this pipeline all the way to the end, including tractography (at the end); but make sure you generate a good (large) number of streamlines, and choose an appropriate -cutoff value. Experiment a bit with the latter to get it right for your data. Finally, use tckedit and provide it for each tract you’re after e.g. with 2 separate -inclusion regions, located where you expect the tract to end up (at both ends). If your fMRI results are a bit messy or all over the place, you might want to manually draw 2 regions, guided by the fMRI results. This might further help in getting alignment right, even after registration, as fMRI results are often quite blurred due to various smoothing operations in the analysis. Once you get decent looking results, I always tend to re-run tckgen itself again, with the -inclusion ROIs and parameters I’ve figured out through the process of the initial tckgen and tckedit. In this way, you can assure that the final tract itself has enough streamlines or is “dense” enough for your purposes.