Hi all,

I asked this question about the failure of DARTEL normalization after CONN denoising a few days ago.  (That question contains detailed information about my preprocessing pipeline and what I have done to my data).  After DARTEL failed I ran regular SPM normalization and after it looked pretty good I tried all my basic analyses.  I started to hypothesize that denoising had somehow altered the image (changed the field of view? put it in a different space?) to take it out of concordance with what other software expects when some idiosyncratic results visualization code started failing in ways that it never had before. It would really be laborious to describe how my code is failing and I'm not sure how illuminating it would be. But here are two simple observations:

1) When I open up my normalized functional image now along with SPM's canonical T1 and EPI in FSLeyes I get this warning: ""Displaying images with different orientations/fields of view!" That does not happen with my old normalized data that were never denoised.  

2) Way downstream, after I have run CAN lab's robust regression, the resulting beta map always used to have dimensions [90, 109, 91] and now has dimensions [79, 95, 79].  This is on the same data, and it's a map produced by the same scripts.


Admittedly, two things are different now, not one.  I ran denoising and then, because DARTEL normalization failed, ran regular SPM normalization.  Also, there is a piece of evidence that the problem is actually my new plain vanilla normalization and not denoising: when I open up my failed, DARTEL normalized denoised files, I don't get that warning in FSLeyes. It just seems so bizarre that running SPM's normalization procedure would cause the normalized image to not correspond to SPM's own canonical brains.  Also, if denoising did this, then it explains why DARTEL normalization failed.  

If this sounds like something recognizable about what the denoising steps would do and not just some feverish conjecture on my part, if anyone has any tips on how to revert this change I would appreciate it.  On the other hand, if this is something SPM normalization has done, I realize it's not really a question for this forum to solve.

Thanks in advance!