Setting acquisition parameters for DWI Preprocessing and CSD

Hi MRTrix experts,

I have some DWI studies acquired with a b value of 1500 s/mm2 and 32 directions (because our license does not allow a greater number of directions unless you acquire 32 directions twice to get 64).

Is a b value=1500 and 32 directions enough to do the preprocessing, the Constrained Spherical Deconvolution and get good results?

Thanks a lot and best regards,


Hi José,

Similar questions have been asked a few times in the past (see e.g. this post or this post). Basically, CSD is certainly feasible and will run, but the quality of results may be questionable. It depends to a large extent on the SNR, and also what you plan to do with the results exactly. Assuming sufficient SNR, I’d expect you would get decent results here, though you could obviously get better with a slightly higher b-value and more directions, and even simply repeating the acquisition to boost SNR will already achieve a lot (32 directions is enough for a decent lmax=6 fit) – just make sure you don’t average the images if you do that.

Hope this helps,

Thanks you very much for the information



Just a brief follow-up:

Acquiring 32 directions twice is not equivalent to acquiring 64 well-distributed directions. While the “amount of data” is the same, the angular resolution is not. E.g. Imagine acquiring 1 direction 64 times then trying to do CSD.

For b=1500 it probably wouldn’t make much difference as there’s unlikely to be much angular contrast in the signal beyond l=6 which is captured by >= 28 directions (article), but it’s worth understanding the difference nonetheless.