Strange dwi2response results

Hi everyone. I am getting rather strange results for dwi2response (msmt_5tt). I also tried the dhollander method, but the results look similar.
The white matter response is completely spherical. It should look like a donut, right? The gm response looks like a white matter fODF. See attached images. Any suggestions there?





The tournier white matter response looks fine:


Hi Peter,

Could you do me a favour and post the actual contents of the response function files?

In both dwi2response msmt_5tt and dwi2response dhollander, the GM and CSF responses are intrinsically calculated as isotropic functions (they are simply mean values scaled by 1/sqrt(4*pi) to account for the SH basis), so can’t possibly be anisotropic. I suspect what’s happening here is that shview is in fact taking the response function values across shells, and plotting them as though they were response function coefficients with increasing l.

For the WM response: For your first ‘shell’ (b=0), the WM response is isotropic, so this might be what you’re seeing. See if the arrow keys allow you to navigate between shells.

shview was supposed to have been brought up-to-date to reflect multi-shell processing here; the issue of column vector transposing may have been missed as part of that… @ThijsDhollander?

Woops, yep, that wasn’t resolved yet.

@pneher : this is definitely just a bug in shview. We can double-check your responses, as Rob suggested; but I’ve actually checked multi-shell responses for those tissue types myself, and can 100% confirm that what you see, is simply shview treating the values of the GM and CSF responses as if they were SH coefficients of a single shell. The funny thing is: I can even reproduce those erroneous shapes. :smile: So, apart from the WM one (where you’re just seeing the b=0 isotropic shell in this case, use the arrows as Rob suggested to see the other shells), I can even reassure you that your responses are probably of good quality.

But in any way, shview is of not much use for the GM and CSF responses (even if it didn’t make that mistake), because it’ll normalise/scale each shell when you show it… so all you’d see if it were working right, would be just a nice sphere each time. If you really want to assess your responses, or more specifically the decay curves across b-values, just open them as text files, or import the file or numbers directly into Matlab, or even Excel.

That makes sense. WM looks fine on the other shells. I will have a look at the gm/csf responses but I suppose they are fine too. Thank you guys!

1 Like