Print

Print


Hello Lester and Ged,
 
Thank you for the detailed information from you guys. I will try all the options and hope I won't
have any problem then..Else I will call contact you again.
 
Thank you.
 
Sincerely,
--
Abhinay D. Joshi
UT Southwestern
E06.122 MC 9061
Dallas, TX 75390
P -214-648-9265

Biomedical Engineer
Avid Radiopharmaceuticals Inc
3624 Market Street
5th floor
Philadelphia, PA 19104

 
On 1/15/07, Ged Ridgway <[log in to unmask]> wrote:
Hi Abhinay,

> 1) How to get good FLAIR templates in SPM

I think the easiest way around this would be to use SPM5's unified
segmentation, which just uses tissue probability maps, and doesn't
assume any particular intensity relationship between these and the images.

You may still have problems if e.g. the gray/white contrast is
inadequate (I don't have much experience with FLAIR, but some of the
images I've seen have rather unclear GM/WM boundaries), but then I
don't think there's much you could do about that within SPM.

> 2) how to measure hippocampus volume by using SPM?

Well, SPM is designed *not* to do this kind of volumetry. The idea
with SPM is that you analyse e.g. the whole GM across subjects, and if
they have significant volume differences in the hippocampus then
hopefully this gets picked up in the resulting statistical maps.

You could define a rough hippocampal ROI in MNI space and use SPM
(with small volume correction for FWE p-values), but this will still
amount to looking for significant voxel-wise hippocampal differences,
rather than measuring volume.

To measure volume, you really need to segment the hc on each scan. You
could try using an automatic segmentation propagation method for this
(e.g. the SPM extension AAL -- http://www.cyceron.fr/freeware/ or
perhaps something like FreeSurfer http://surfer.nmr.mgh.harvard.edu/
which I've heard has good seg prop parcellation as well as cortical
analysis).

However, you might find that these aren't accurate enough, in which
case I think your only option is to manually (or "semi-automatically")
segment the hc. You could try e.g. itk-SNAP for this
http://www.itksnap.org/

Best,
Ged.