I'd suggest using a mask. The simplest way is to use a GM threshold
of about 0.1, although there are other ways. Here is another paper of
Ged's that may be of interest:
Gerard R. Ridgway, Rohani Omar, Sébastien Ourselin, Derek L.G. Hill,
Jason D. Warren, Nick C. Fox. Issues with threshold masking in
voxel-based morphometry of atrophied brains. NeuroImage, Volume 44,
Issue 1, 1 January 2009, Pages 99-111
I can't comment on the VBM batch you are using, but in general, I
would suggest using the latest software versions, which are usually
improvements over older versions.
Best regards,
-John
On 2 April 2012 12:37, Valentino Pironti <[log in to unmask]> wrote:
> Dear John,
> Thank you. I am going to read the paper. In the meantime, apology for
> another question that might look naive to experts. I have not used a mask
> for my design. What would you suggest to do in practice? Shall I change any
> option in the VBM batch in SPM5? Would you suggest to use SPM8?
> Thank you in advance for your advice.
> Best wishes,
> Valentino
>
>
>
>
> ----- Reply message -----
> From: "John Ashburner" <[log in to unmask]>
> To: "Valentino Pironti" <[log in to unmask]>
> Cc: <[log in to unmask]>
> Subject: [SPM] VBM with spm5
> Date: Mon, Apr 2, 2012 12:13
>
>
> This paper may help explain what's going on:
>
> Gerard R. Ridgway, Vladimir Litvak, Guillaume Flandin, Karl J.
> Friston, Will D. Penny.
> The problem of low variance voxels in statistical parametric mapping;
> a new hat avoids a ‘haircut’. NeuroImage, Volume 59, Issue 3, 1
> February 2012, Pages 2131-2141
>
> Fred Bookstein's criticisms of VBM (also in NeuroImage) also covered
> this, although it is a bit less straightforward to follow.
>
> For this reason, I'd suggest making sure your GM mask extends far enough.
>
> Best regards,
> -John
>
>
> On 1 April 2012 21:31, Valentino Pironti <[log in to unmask]> wrote:
>> Dear All,
>> I have done some VBM analyses with SPM5. The t test results are funny.
>> Scattered blobs far outside the brain. I am aware it might be common but I
>> checked t1 and segmented images, one by one, and excluded those that looked
>> dodgy. However the problem persists. Any idea why is the problem still
>> there?
>> Thank you in advance for your help.
>> Best
>> Valentino
|