Dwi2response for a specific fa value

Hi All,

I was trying to use dwi2response function to generate response function output for a specific FA value.

How can I generate response function for certain grey matter tissue?

Kind regards.

Do you mean that you want to generate a response corresponding to the signal that you would expect assuming a tensor model with a given mean diffusivity and FA? There’s no functionality to do that in MRtrix3, it’s expected you’d always measure it directly from the data.

I do have Matlab code to do this though, I should push it up to GitHub at some point…

I’m not sure whether you’re asking about the isotropic response function needed for a multi-tissue CSD, or whether you genuinely want a full anisotropic response suitable for e.g. tracking in GM? If the former, then dwi2response with the tournier, manual or msmt_5tt algorithms should do the trick (dhollander is also an option if you want all responses), but each will require different inputs.

If the latter however, then I’m not sure what the best approach would be. You’d need to be able to select voxels with a single coherently-oriented population of GM fibres - I don’t think you’d ever get that, GM is simply not organised like that. You could always try dwi2response using the tournier or manual algorithms, which will give you an answer, but I’m not convinced it’ll necessarily be the right thing one - this will depend on exactly what you’re trying to achieve…

Well, this one not really of course. :wink:

This is a common confusion (about terminology mostly, I reckon) these days… If you’re referring to tracking in “the region that people typically refer to as gray matter, e.g. the cortex”, then there’s not per se an issue going with a 3-tissue CSD output with an anisotropic response and FOD for WM, and isotropic compartments for GM and CSF. If responses are chosen well, you’ll see an anisotropic WM FOD in the “region of the GM”. This one can definitely be tracked, and depending on your spatial resolution will give you a result mostly related to axonal contributions to the signal. The isotropic GM response and the resulting GM compartment ends up modelling what I would professionally refer to as “a lot of crap somewhere in the middle of the diffusivity range, that is preferably quite isotropic in nature”. Read: contributions from certain spaces related to other cells. This doesn’t have to per se relate to what you may think of as “GM” specifically, and happens right in the middle of certain bits of WM as well (with perfectly clear and consistent explanations to go with it). And the CSF response/compartment is simply free water of course.

But, @isAarya, your initial question is indeed a bit confusing. Can you elaborate what you specifically mean by




I was wondering if I can use a specific FA value to obtain streamlines/tractography for specific part of tissue such as either GM or WM.

I believe dwi2response function helps to obtain DWI signal information from a single-coherently-oriented fibre bundle as mentioned in explanation of this function.

I use the default dwi2response tournier -options with my input image.

This response function is later send to dwi2fod csd -options to carry out spherical deconvolution and thus estimate fibre orientation distribution. Which is further sent to tckgen to get the streamlines.

But with this setting I don’t see that all the GM regions that I am interested in are having any corresponding streamlines; hence the question if I can use a specific FA value can help me to achieve this.

At this stage I am just looking at only isotropic response, I believe anisotropic response maybe more difficult to achieve.

Currently I am investigating if multi-tissued csd will help me achieve this.

I’m still not 100% sure about your goal, but it sounds like this should be the solution. Note that this wouldn’t affect where you get your WM response from; it just adds a response for GM/CSF. This will result in your WM FODs being cleaned up in regions such as the cortex, where partial voluming with GM tissue happens.

If you haven’t already, definitely give the original multi-tissue CSD paper a read, and in particular observe the results to see if that’s the difference in performance/outcome your’re looking for!