I got multishell data (b=0,1000,3000) and testing global tractography using dhollander response functions. All RF look the same, an sphere (I thought I would get a pancake shaped RF). Wondering whether this is ok or not.
I used dwi2fod msmt_csd to get FODs and display them as RGB image, it seems the unsupervised dhollander algorithm detects some gray matter as white matter, any idea how to improve it ?
Those are probably completely fine, but the spheres you’re seeing is just the b=0 part of the response (which is indeed by definition isotropic). Use the right (and left) arrow buttons on your keyboard to see the other shell’s responses for each tissue response! We should probably make this feature even more clear, because you’re not the only one asking this question every once in a while. Note that for the GM and CSF responses, you’ll also only find spheres for the other b-values as well, but as you press the right arrow, you’ll see the sphere becoming smaller, proportionally to the signal decrease for those responses. For the WM response, you should see 2 anisotropic responses for the b=1000 and b=3000 shells, the former being smoother and the latter being a sharper/flatter disk.
Slightly careful with the wording here: it’s actually the (unsupervised) MSMT-CSD algorithm that “detects” this, given the responses from the (unsupervised) dhollander algorithm. What gray matter bits are you exactly referring to, that appear as white matter? Within the cortex, it’s normal/expected to see mixtures of WM-like and GM-like tissue compartments: the WM-like contributions are mostly due to axons, the GM-like contributions are mostly due to excess extra-axonal contributions, e.g. from neuronal cell bodies and potentially other (extra) glial cells.
Were you, in the previous question, referring maybe to the bits on the far outside of that tractogram visualisation? If so, those probably aren’t part of the brain (or you should be lucky enough to pick up the outer layer of the cortex that actually has axons parallel to the surface; but I don’t think so). These may be bits due to motion, or even other structures that induce anisotropic diffusion. They may appear “WM-like” in signal, but they’re indeed not WM. In this case, slightly eroding the brain mask would do the job.