@auto_tlrc vs. @SSwarper anatomy contrast

Hi all,

Our group has been using @auto_tlrc for aligning anatomical human T1 data to the MNI152 template. I recently attended an AFNI workshop and worked through warping/skull-stripping using @SSwarper and it sounds like this is the preferred option if you have the computing power to do it.

I’ve been running @SSwarper on some of our subjects to compare warping between our old @auto_tlrc method and this new nonlinear warping and wanted to run a side-by-side comparison by others (attached image).

On the left is the non-linear warping and skull-stripping done via @SSwarper (anatQQ dataset) and on the right is skull-stripped data warped via @auto_tlrc. The two don’t seem dramatically different but in the @SSwarper (left) image I feel as though there is some blurring between the sulci and gyri - the contrast just doesn’t seem as clear as it is in the linearly warped data on the right. Wouldn’t this negatively impact alignment and registration of functional data, by making it more difficult to find edges and by reducing contrast?

I want to be sold on @SSwarper but am not sure what I’m missing here.


Hi, Walker-

Thanks for looking at this so closely.

One difference is that the anatQQ volume has gone through some preliminary “polishing” as part of the @SSwarper processing. By default, the input anatomical is put through 3dUnifize and 3danisosmooth. So, there is some adjustment of brightness and blurring due to the combination of those.

If you don’t want one or the other of those steps done, modern versions of @SSwarper do have -unifize_off and -aniso_off flags. I haven’t looked into all permutations of leaving those on/off. Those are typically on to try to reduce the effect of brightness inhomogeneity and/or bright spots. Note that in the end, the specific look of the anatomical probably matters a bit less than the quality of alignment. However, if tissues boundaries are being blurred, then that indeed would be a negative effect of the initial polishing before the warping begins.

When you overlay the results on the standard space brain, which looks better, in terms of sulcal+gyral patterns matching?