 |
 |
Dear all, Just a comment after Matthew's answer. On the Talairach webpage they suggest to use the icbm2tal transform: http://www.talairach.org/manual.html (et the end) Nevertheless, I checked the BioImage suite and it works very well with the talairach atlas. It would would be unfortunate if they are suggesting the wrong tool. The BioImage suite also offers the matlab script for the non-linear conversion. Best regards. Dorian 2009/6/1 Matthew Brett <[log in to unmask]>: > Hi, > > So - I wrote mni2tal and... > >> I have used two functions to convert SPM coordinates for Talairach usage. >> >> 1. mni2tal.m (from Matthew Brett) >> >> 2. icbm_spm2tal.m. >> >> The two give quite different results (sometimes differing some mm). >> Should I take in consideration only the second one (said to be more >> precise)? Or is there another tool even more precise? > > I think the answer is fairly simple. > > Case 1) If you want to match your MNI normalized brain, to a template > of normal brains that has been Talairach transformed (the rather odd > piecemeal linear transform used by Talairach), then the icbm_spm2tal > transform is the way to go. > > Case 2) If you want to match to the Talairach atlas brain, the > icbm_spm2tal transform will give you some wrong results, and I would > use another transform. If you look here: > > https://cirl.berkeley.edu/mb312/icbm_transform/ > > you will see a set of png images showing the relationship between the > MNI template, the downloaded Talairach daemon brain, the MNI template > transformed with the icbm_spm2tal transform, and the MNI brain > transformed with the mni2tal transform. You'll see that the > icbm_spm2tal transform tips up the MNI atlas, making the match in the > orbitofrontal and lateral temporal regions worse. The reason for > this is probably that the Talairach transform does not do a good job > of matching to the very odd shape of the Talairach atlas brain. > > So, personally, I would definitely use the icbm_spm2tal transform in > case 1), but not case 2). If you want a case 2) transform, I believe > the transform implemented in BioimageSuite is particularly good, but > mni2tal is still better than nothing. > > Best, > > Matthew >