Hi, Ajay-
1) Create an average B0 image by linearly registering multiple B0's to create an optimized mean image as opposed to just creating a mean (since this does not account for head motion and may make the mean look worse).
-> The choice is yours, but mine own thinking on this kind of point has changed over time. And that is one reason for the change from 1dDW_Grad_o_Mat -> 1dDW_Grad_o_Mat++. Basically, for the quality of fit, I believe you are better off *not* averaging the b=0 volumes together. Those volumes have the highest SNR of the acquired DWI set, and averaging them together effectively reduces their weight/influence on the tensor fit, in the manner used by 3dDWItoDT (if using other software with different fitting, then it might make sense to average together, I s'pose). This point basically propagates to answer other points you raise in this posting.
In terms of performing motion correction across b=0s, I would think that you would actually want to motion correct across *all* the acquired volumes, b=0s and DWIs alike. And to do *that*, you are likely best off putting your data into TORTOISE's DIFFPREP tool for both motion and eddy current distortion, and possibly into their DR-BUDDI tool if you have a dual phase encoded acquisition (e.g., AP-PA acquired data). That is a more rigorous way to handle motion correction, and you benefit by getting other distortion reductions (which exist in all acquired DWI data). NB: TORTOISE 3.0 is now command line processable (THANK YOU, OKAN!), and can be downloaded for free from their website:
[
tortoisedti.nichd.nih.gov]
2) Put the mean image as the first volume for eddy current reference
If you are using TORTOISE, this wouldn't be necessary to do.
Using an older afni package, there was just 1dDW_Grad_O_Mat which would create a mean B0 image (although not linearly registered i'm guessing) when the -proc_dset flag is used and would output the new dataset with the -pref_dset flag. When updating and using 1dDW_Grad_O_Mat++ these flags are no longer available, and so I wanted to see if this is handled elsewhere or if this functionality is not being replaced?
As per noted in the first part, the reason this no longer exists is because I believe it would be a non-preferred step on the way to fitting tensors. Note that 1dDW_Grad_o_Mat *does* still exist in its native form, for backward-compatability reasons, but I don't think it should be used, really-- the basic functionality of the 1dDW*++ function is easier to use and more to my liking. (Of course, that is just a personal opinion.)
The issue is that even though the bvec/bval files are updated by removing other B0 values, I do not see any flags to output a corrected image. It looks like bref_mean_top will handle the text files but I do not see how to output a corrected image (with the mean B0 in place which I can replace with a linearly registered version).
Well, if you're going to be doing motion correction on DWIs, you'd want to update the gradient/matrix table appropriately; for b=0 volumes, this isn't an issue, but, again, I think motion correction on the full dset really makes sense-- TORTOISE handles that part of the processing, as well. NB: I'm not paid by the TORTOISE folks (really), but I would strongly recommend using that for these kinds of processing steps.
Happy to hear your thoughts on that. The above are only recommendations.
--pt