AAL template extends beyond MNI cortical surface about antsr HOT 7 OPEN

Can you post the precise set of commands used to produce these results?

from antsr.

After downloading as above, I did:

plot(mymni[[1]], mymni[[2]], alpha=.8)

I also note that if I threshold the MNI at values 1 and higher (which surely includes some CSF), and plot the AAL mask on that, it still doesn't match up perfectly:

> mniMask2<-thresholdImage(mymni[[1]],1,Inf)
> plot(mniMask2, mymni[[2]], alpha=.8)

which results in this image:
antsMNIextremeMask_with_AAL_overlay.pdf

If I do a 3-category kmeans segmentation, combine presumed gray and white matter portions of this, and then plot the AAL on top, it shows even less of a match to the surface:

> mni_kmeans3_seg<-kmeansSegmentation(mymni[[1]], 3)
> mni_kmeans3_GW<-thresholdImage(mni_kmeans3_seg$segmentation,2,3)
> plot(mni_kmeans3_GW, mymni[[2]], alpha=.8)

Which results in this:
antsMNIkmeans3Mask_with_AAL_overlay.pdf

I realize things won't be exact, but I just want to be sure I'm not misunderstanding something here.

My thought is the following: The original segmentation of the MNI was a best-guess to begin with, of course. Since my ultimate goal is to compare populations with respect to size of different AAL areas, then as long as I register the MNI atlas to each individual within both populations the same way, the differences between these populations in average AAL area sizes will likely be reasonable, even if the registration to any given individual is imperfect, and even if the AAL itself doesn't exactly match the MNI atlas used. This assumes either random errors or the same biases in the same ways in both populations. Given that the study I'm comparing my present work to essentially ignored these issues, this seems reasonable to me - but I'd be interested in any thoughts.

-Tom

from antsr.

yes - AAL labels are very crude, from the start. undergrads paid to roughly segment structures - done something like 20 years ago.

from antsr.

OK. Any thoughts about why the AAL doesn't match the MNI particularly closely (the areas were drawn on a version of the MNI, so it seems odd to me they wouldn't match more closely)? Also: are there other segmentation akin to the AAL that anyone might recommend in place of them?

-Tom

from antsr.

they are intended to be masked by tissue after the fact

from antsr.

That makes sense. Is there a published reference by any chance that makes this clear, that I can include in my manuscript?

-Tom

from antsr.

My go-to for all things MNI is the LEAD-DBS page

https://www.lead-dbs.org/helpsupport/knowledge-base/atlasesresources/cortical-atlas-parcellations-mni-space/

from antsr.