I’m using a dataset which has released the preprocessed DWI data. The preprocess they’ve done includes denoising and eddy-current correction, and since they’ve acquired using two b-values they have released two preprocessed diffusion scans per subject, one for each b-value. I’m trying to use them for multi-tissue Fixel-based Analysis.
I merged both scans using mrcat. I first thought they’ve done the motion correction and both scans are registered to each other, but skimming through the images using mrview showed the opposite.
Now I want to do the subject motion correction but not eddy current correction (due to computational loads on the system). Are there any options to only do the motion correction? If not, can I simply run dwipreproc to do so? I’m not sure whether running this command on an eddy-current corrected scan may damage the data or not.
I think that if you only want to align the two scans (both have been already preprocessed independently) the best way to do that is corregister (rigid only) one B0 of one adquisition to another B0 of the other acquisition and the apply the resulting transformation to one of the 4D volumes.
The problem is that even within each acquisition the motion correction has not been done and I need to rigidly transform all the volumes to the base b0.
The only way that comes to my mind is to split the 4D images to b0 and gradient-applied scans, then rigidly transform the latter to the former.
That seems very unusual to me that one would even be able to estimate the eddy currents in order to correct them, without also estimating and correcting subject motion. It would need to be done based on a prediction of eddy currents from the diffusion gradient table only, which is going to be intrinsically limited in accuracy.
Personally I would err toward just running the data through
dwipreproc. Even if eddy current correction “has been done”, I would prioritise having the confidence that such corrections have in fact been performed, as have other corrections, over any potential loss in precision from eddy current correction potentially having been performed twice.