Dear Ralph and Randy
We experienced similar problems.
Under SPM96 the approach suggested by Randy was the one we choose with
interesting results (at least on the 10 scans population we choose to test
this approach, the accuracy was constantly < 1.5 cm using temporal poles,
frontal horns and red nucleus as landmark). In fact rather than using EPI
images, we took a T2 scan for each subject, that was previously normalize.
So that each subject mean EPI was normalize on its respective normalized
T2.
In fact there is one very problematic region using this approach which is
the cerebellum on its medial part. In SPM 96 we had to use the maximum
amount of basis function to have a good result here (to be precise, we used
the windows version of SPM96, but the algorithm should have been the same).
For unknown reasons, we haven't found so good results using SPM99b (not
already tried SPM99). May be John Ashburner could find a reason for this.
Our idea is that the major difference we have in this region is the
intensity of the venal blood. I feel like we are facing most problem with
EPI heavily darken in this regions.
Notice that this approach permit a perfect matching of EPI on its anatomy.
Although we haven't test this in a large population, for 2 AVM (arterio
venous malformations), the abnormalities match with a 1.5-3 mm accuracy
(EPI was interpolated to 1.5 mm), and all the activation were perfectly
along the sulcus (supposed to be the central one).
As we are moving to another study, we should try to use high definition EPI
rather than classical T2 (RARE, we do not have 3D T2).
This sounds a bit like a cooking recipe no ?
Thanks for sharing your experience with us about this topic.
Sincerely
Jack
________________________________________________________________
| Jack Foucher Universite Louis Pasteur |
| Institut de Physique Biologique UPRES-A 7004 du CNRS |
| 4 rue Kirschleger Tel: 33 (0)3 88 77 89 90 |
| 67085 STRASBOURG Fax: 33 (0)3 88 37 14 97 |
| France |
| Faster E-mail: [log in to unmask] |
| Other [log in to unmask] |
|_______________________________________________________________ |
-----Message d'origine-----
De: R. Axel Mueller [SMTP:[log in to unmask]]
Date: mercredi 16 fevrier 2000 19:07
A: Benson,Randall R.
Cc: spm.mail
Objet: Re: Talairach
Dear Randy: I have had similar experience with using the T2 EPI template.
Two things at least seem to be important. First, when EPI volumes are
located off the center of the field of view (i.e. differently from the
template) the talairaching can go seriously awry. Second, reducing the
number of basis functions often helps to prevent strange warps. The problem
with warping one of your own brains to the template and then using this in
turn as a template is that you may introduce a systematic inaccuracy, which
is then applied to all of your subjects. This is hard to rule out unless
you have very good anatomical information in your EPIs (typically not the
case).
Axel
>Users of spm:
>
>Was wondering if there are any new methods for converting
>"Talairach" coordinates obtained using spm96 spatial normalization with
MNI
>single brain template to the canonical Talairach's atlas coordinates. I
used
>the transform from this template to the template from spm95 (PET) that
>Andreas posted but it is not perfect since the template is not matched to
>Talairach. Since the Talairach Daemon contains a scanned version of the
>atlas has anybody warped this to the MNI template?
>
>Also, what has been the experience of people warping their T2 EPI
functional
>data to the spm templates? My success is about 60% with the remainder not
>being close at all. I've taken to warping one subject brain to the
template
>and using this as my new template.
>
>Appreciate any help.
>
>Randy Benson
***************************************
Ralph-Axel Mueller, PhD
Dept. of Cognitive Science
University of California, San Diego
Tel.: (858)-551-7926
Fax: (858)-551-7931
E-mail: [log in to unmask]
Mailing address:
R.-A. Mueller
Children's Hospital Research Center
8110 La Jolla Shores Dr. #200
San Diego, CA 92037
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|