Which version of SPM8 are you using? Some random numbers are used to
initialise various parameters, and there was an update in revision
3687 to include "randn('state',0)", in addition to the existing
"rand('state',0)". If you update to the latest version, the same
random numbers should be used to initialise each time.
Note also that if you run MATLAB on multiple cores, you can not expect
exactly identical answers each time. For example, try:
x=(randn([1000000,1],'single'));
s1=sum(x);
s2=sum(x);
s1-s2
I don't think this is what you are experiencing though.
Best regards,
-John
On 4 April 2011 10:45, Janin, Hugo <[log in to unmask]> wrote:
> Dear all,
>
>
> I am processing anatomic (T1-weighted) images for volumetric study and I have a problem with New Segment. When I run it several times on the same image (copy/paste before running NS), the segmented images are always different. These differences are invisible to the naked eye but when I calculate volumes (of GM, WM and CSF) with the get_totals script, there are sometimes big volume differences (10 mL) between 2 processes of the same image. This is a problem because I try to find small volume changes between different images of a same patient.
>
> Thanks in advance for your help,
>
> Hugo
>
