Dear Darren,
Thanks very much for the reply.
> Ah a topic dear to me heart. We have taken the approach of stating the
> issue of Talairach space thusly:
>
> the images were normalized using the template image supplied with SPM-96
> (99b). This template conforms to the space defined by the International
> Consortium for Brain Mapping project (ICBM) (NIH P-20 grant), and closely
> approximates the space described in the atlas of Talairach and Tournoux.
>
> The description, however, belies the difficulty in figuring out where one
> is in the atlas and in performing meta-analyses.
I guess this is the problem. How in practice do you work out where your
activations are in terms of Brodmann's areas? Do you modify the coordinates
in any formal way to work out where you are in the Talairach atlas? Do
you use the Talairach atlas? I guess most of us do...
> Please note that the transform of Andreas' has an error in it which he
> corrected and wrote me later.
>
> The revised transform is:
>
> X' = 0.88 - 0.8 <---- here is the error
> Y' = 0.97Y - 3.32
> Z' = 0.05Y + 0.88Z - 0.44
I will add the correction to my web page. This error will only have a slight
effect on the transform though (of 1.6 mm in X).
> You may want to check your transform against it.
I have tried to do this by taking landmarks on the MNI brain. I didn't
think to put this on the web page. The
landmarks I used were from the left side of the
MNI brain: the highest point (in Z), the leftmost point, the most
anterior and the most posterior point:
High Left Front Back
X 13 69 9 11
Y -30 -35 73 -107
Z 78 7 -1 -2
I ran Andreas' transform (with the correction, just now), and my transform,
on these points, and they generated estimated Talairach points thus:
Andreas':
10.6400 59.9200 7.1200 8.8800
-32.4200 -37.2700 67.4900 -107.1100
66.7000 3.9700 2.3300 -7.5500
mni2tal:
12.8700 68.3100 8.9100 10.8900
-25.4771 -33.5857 70.6795 -103.7443
73.1247 8.1287 -4.3780 3.5094
These transformed points are:
9.7187 10.0629 7.6403 11.7337
mm different between the two transforms, which I think is significant.
I thought that the estimates from my transform were closer to the atlas;
which was partly why I posted it. I believe Andreas has also
had a look, and found my transform to work reasonably well. He
might want to comment?
I am not suggesting the transform is anything other than a temporary solution
to a rather painful problem, but as I've said in the web page, I
think some sort of non-linear (not affine) transform is necessary
because of the different brain shapes.
> I agree it would help if
> the SPM documentation was more precise about the issue of canonical
> Talairach space vs. all other spaces which approximate it.
Yes, I agree with your agreement!
Thanks a lot,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|