Hi Jonathan,
>I'm confused as to why you say most of the masking happens during>normalization? Can you explain what you mean?
Thanks for spotting that. That was of course not really true in the way I've wrote it. I actually had the bounding box in mind, which cuts out parts of the cerebellum during the normalisation, when the default options are used. The masking of non-brain voxel happens indeed during the estimation of the model.
Karsten
--
-------------------------------------------------------------
Karsten Specht, PhD
Department of Biological and Medical Psychology
Bergen fMRI group
University of Bergen
Jonas Lies vei 91
5009 Bergen
Norway
Tel.: +47-555-86279
Fax: +47-555-89872
[log in to unmask]
http://fmri.uib.no/
-----Ursprüngliche Nachricht-----
Von: Jonathan Peelle
Gesendet: 12.aug.2010 16:20:42
An: [log in to unmask]
Betreff: Re: [SPM] explicit masking in a whole-brain analysis?
>Just adding my 2 cents in:
>
>Karsten,
>I'm confused as to why you say most of the masking happens during
>normalization? Can you explain what you mean?
>
>Asaf,
>During 1st level model estimation SPM calculates a mask (saved as
>mask.img in the stats directory); this is generally a good brain mask,
>and limits the number of voxels that are estimated when fitting the
>model. So, I don't think you would gain much by using an explicit
>whole brain mask. It probably wouldn't hurt, although you would want
>to make sure that, however you calculated your brain mask, you weren't
>cutting out any voxels of BOLD signal.
>
>Best regards,
>Jonathan
>
>
>On Thu, Aug 12, 2010 at 9:04 AM, Karsten Specht wrote:
>> Hi Asaf,
>>
>>>Suppose I don't want to choose or to leave out certain areas in the analysis (i.e., i want to
>>>perform a whole brain analysis), is there any point in using this option? >I read in the manual (section 8.8) that it speeds up the estimation process - Will it be wise
>>>to use my SPGR image of the whole brain, have I got anything to lose if I do this?
>> yes, of course, you can use your own brainmask, but note that most of the masking already happens during the normalisation, not during the estimation of your model.
|