I am seeing some sort of artefact on an acquisition that I can’t seem to resolve. It clearly appears when a multiband factor of 3 is used (although it is there but not very noticeable when a factor of 2 is used, and no artefact when a lower parallel imaging factor is used). I have tried different head coils but the result is always the same.
It is very clear on the noise map and also appears when averaging the shell (b=1000).
I have tried different dwidenoise options and different mtnormalise options but the result is always the same. Does anyone have any insight as to what could be causing this artefact?
It is not uncommon to see a discontinuity in the noise level map at the multiband boundary (e.g. between slices 20 and 21 and between slices 40 and 41 with MB=3 and 60 slices). This discontinuity arises from the way in which signal and noise propagate through the multiband unfolding. In fact, there should be a hard edge, but the patch-based approach in MP-PCA blurs this out across the slice direction.
The fact that you also notice this in the average b=1000 shell points to signal leakage. As you already observed, you can reduce this with lower acceleration factors (the multiband factor and also in-plane acceleration). I would try to reduce it to the point where you don’t see an edge in the shell-averaged data, but not worry too much about the noise level maps.
As this is related to the acquisition and the MR reconstruction, it is not something you can suppress with