Hi folks,
I have a data set at b=3000. Does MRtrix3 have a strategy for dealing with the free water confound in this case?
Many thanks,
Tom.
Hi folks,
I have a data set at b=3000. Does MRtrix3 have a strategy for dealing with the free water confound in this case?
Many thanks,
Tom.
Hi Tom,
As mentioned in a few other threads, use of multi-tissue CSD with a CSF tissue component performs something akin to what is typically referred to as “free water elimination” in DWI. It’s not precisely equivalent however, so I would suggest reading the multi-tissue CSD manuscript, and other threads on this forum where the concept is discussed (should be found relatively easily using the search function).
Cheers
Rob
Thank you Rob. I may be misunderstanding something.
I thought multi-tissue CSD could only be performed for multi-shell data. I only have the b zero image and one shell at b=3000.
Or is it still valid only with the following caveat?
“SSST-CSD has a tendency to overestimate the WM volume in voxels containing GM and/or CSF”.
In the case of b=0 and b=3000, it is still possible to use multi-tissue CSD; the caveat is that you are restricted to two tissues only. Therefore a common approach with “single-shell” data is to use a WM response and a CSF response. This provides something akin to “free-water elimination”, in that signal arising from fluid is attributed to the CSF component by the tissue decomposition. So while this approach may not give the separation between WM and GM that the “typical” multi-shell multi-tissue approach with three tissue types provides, and therefore the GM signal is mostly attributed to the WM resulting in noisy FODs, what it does provide bears some resemblance to your original request.
Eventually the “Single-shell 3-tissue (SS3T)” method will enable separation of WM, GM and CSF using only b=0 volumes and one b!=0 DWI shell; unfortunately this is not yet available in MRtrix3.
Hi @tom,
See these topics and their replies for some more useful information on this topic:
Don’t hesitate to click all the links to other posts and sources in these discussions as well, they’ll lead you further down the multi-tissue rabbit hole.
As @rsmith suggested, another recommended read would be the original multi-tissue CSD paper; to get an understanding in how we generally model the signal in a multi-tissue CSF framework.
Performing just single-shell single-tissue CSD will indeed leave you with the “caveat” of overestimating the WM compartment, because basically all signal will end up in that WM (FOD) compartment. So anything that still generates signal at b=3000, which includes grey matter, and potentially some other types of (non-axon) cells as well, will end up in your FOD, leading not only to overestimation of the “apparent fibre density”, but also distortion of your FOD by spurious peaks and other noisy structures, further limiting your abilities for tractography or even separating fibre density contributions in specific fixels.
Hope this all makes some sense… if not, definitely go clicking and reading some stuff.
Cheers,
Thijs