Deformation axes after permutation, longitudinal study, 2 weeks left to submit dissertation

[Yuval A.]

Dear ExploreDTI community,

I'm having some ongoing trouble with deciphering the SM/EPI/EC correction pipeline, and could some advice.

First - in the mat file calculation step, I had to permute the gradient components to Y X Z, and flip them to -x y z (based on the acquisition parameters and the radiologists' instructions).

The phase encoding direction in my scan is anterior>posterior.

Therefore - should I still define the deformation axes as [1 0 0], or should I change them to [0 1 0]?

Second - I'm performing a longitudinal analysis in my study. Each participant has 4 scans, which I rigidly co-registered to their respective T1w images, and masked the transformed data with the brain mask I BETed earlier.

After the SM/EPI/EC correction, I exported the FA images for each scan, co-registered them to the first scan, and created subject-wise mean FA images.

I did this to get rid of whatever intra-subject variability there is, before delving into network atlas analysis.

So, the second SM/EPI/EC correction was also rigid, and involved a single dataset. I chose 'FA' as the image type for co-registration.

Many of the images came out misshapen- either elongated, or missing significant portions of the ventral forebrain or the brainstem.

Considering the ELASTIX transforms the atlas to the individual images, rather than the other way around - it this step actually necessary? If so, what other options can I tweak to get this right?

I'm aware of the 'hack' suggesting co-registering each of the longitudinal scan to one of the single subject-specific T1 images, but I'd rather avoid the consequent frowns from my current thesis committee and future peer-reviewers.

Thanks in advance,

Yuval.