Hi Shary,
sry, I meant Voxel-based morphometry--the methods.
Neuroimage. 2000 Jun;11(6 Pt 1):805-21. Review.
Thanks for pointing this out :-)
Because the images enode probabilities you should use an absolute
threshold to discard low probabilities.
Thanks, Stefan
> I usually do Relative masking (0.8) to analyze VBM images (gray
> segmented, normalized, modulated, smoothed imaged) using PET models.
> I looked (searched in PDF) at John's 2001 publication (COMMENTS AND
> CONTROVERSIES Why Voxel-Based Morphometry Should Be Used) , and
> could not see anything about thresholding. Could you give me the
> title of the paper and I also could let me know whether you mean that
> I should set the relative masking to 0.05 when it comes to VBM.
>
> Thanks,
> Shary
>
>
>
> At 28/02/2006 Tuesday 07:18 AM, you wrote:
>> Hi Susie,
>>
>> I had a look at why you find these rather odd smoothness estimates.
>>
>> I assume that you were doing a VBM study?
>>
>> The smoothness estimator fails for very low probability voxels, i.e.,
>> voxels where the estimated probability for being grey matter is
>> extremely small. These are usually 'edge of the brain' voxels.
>>
>> In SPM, the 2nd-level model interface was designed with having
>> functional images in mind. When you're doing VBM you have to modify
>> one of the default masking options: you should change the explicit,
>> absolute threshold to something a little bit greater than 0, say 0.05.
>>
>> Also, John recommends the 0.05 threshold in his 2001 Neuroimage VBM
>> paper, because low-intensity voxels might not follow a normal
>> distribution.
>>
>> All the best, Stefan
>>> Hi Will
>>>
>>>
>>> >> load SPM
>>> >> SPM.xVol
>>>
>>> ans =
>>>
>>> XYZ: [3x692811 double]
>>> M: [4x4 double]
>>> iM: [4x4 double]
>>> DIM: [3x1 double]
>>> FWHM: [1.0882e-04 2.7407e-06 4.7969e-07]
>>> R: [1 2.2577e+08 6.7308e+15 4.6652e+21]
>>> S: 692811
>>> VRpv: [1x1 struct]
>>>
>>> I guess that means spm has estimated the smoothness wrongly! Is it
>>> something I am doing wrong, or something that can be sorted in spm?
>>>
>>> Thanks for all your help.
>>>
>>> Susie
>>>
>>>
>
>
>
>
|