Dear Rob
In addition here is further evidence that you should not rely upon the
VBM cluster results reported by the SPM application without adjustment.
However it is possible to report on cluster inference when the cluster
size is adjusted to reflect the underlying deviation from uniform
smoothness. Also there are important constraints to be observed such as
the df and uncorrected threshold level.
Moorhead TWJ, Job DE, Spencer MD, Whalley HC, Johnstone EC, Lawrie SM.
Empirical comparison of maximal voxel and non-isotropic adjusted
cluster extent results in a voxel-based morphometry study of comorbid
learning disability with schizophrenia. Neuroimage. 2005 Nov
15;28(3):544-52.
Bill Moorhead
Div of Psychiatry
Univ of Edinburgh
Quoting "(Mr.) Shahryar (Shary) Rafi-Tari" <[log in to unmask]>:
> Dear Rob, Sartu and other colleagues with similar questions:
>
>
> Just to add a bit more info in support of what Satoru is saying:
>
> Please read the following and take a look a the
> formulas used for multiple comparisons. You will
> see that only if you use set level for inference,
> the use of extend threshold means somthing:
>
> Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J.-B., Frith,
> C. D., and Frackowiak, R. S. J. 1995. Statistical parametric maps
> in functional imaging: A general linear approach. Hum. Brain
> Mapp. 2: 189–210
>
>
> As far as inference in VBM is concerned, again in
> support to Satoru's comments, the following reference is what you
> should read:
>
> Ashburner J, Friston KJ.
> Voxel-based morphometry--the methods.
> Neuroimage. 2000 Jun;11(6 Pt 1):805-21. Review.
>
> Indeed, you should look at corrected voxel level p-value.
>
> I have been through this before and hope this is helpful.
>
> Cheers,
> Shary
>
>
>
>
> At 27/01/2006 Friday 02:56 PM, Satoru Hayasaka wrote:
>>
>> Dear Rob and VBMers,
>>
>> The extent threshold in SPM does not affect the cluster extent
>> p-values whatsoever. As you said, it just “screens out” small
>> clusters. If you are certain that even small clusters (say <25
>> voxels) indicate local gray matter change, then probably it’s not a
>> good idea to use the extent threshold. Otherwise it’s a good way to
>> get rid of salt & pepper (i.e., tiny clusters) from your glass brain.
>>
>> Just a footnote. You shouldn’t use cluster extent p-values in your
>> VBM analysis. This is because the smoothness is not uniform in VBM
>> image data. In other words, there are smooth (high FWHM) areas and
>> rough (low FWHM) areas within the same brain. Large clusters tend to
>> occur smooth areas, and this could result in increased false
>> positives in such areas.
>>
>> Hope this helps.
>> -Satoru
>>
>>
>> Satoru Hayasaka PhD ----------
>> Assistant Professor, Public Health Sciences & Radiology
>> Wake Forest University School of Medicine
>> (ph) +1-336-716-8504 / (fax) +1-336-716-2870
>> (email) shayasak _at_ wfubmc _dot_ edu
>>
>> From: SPM (Statistical Parametric Mapping)
>> [mailto:[log in to unmask]] On Behalf Of Robert (Rob) Kevin McClure
>> Sent: Friday, January 27, 2006 1:01 PM
>> To: [log in to unmask]
>> Subject: [SPM] VBM and cluster extent
>>
>> Dear John and SPMers:
>>
>> I have noticed that some investigators doing VBM use an extent
>> threshold of 25, 50 or 100 voxels in order to "screen out" very
>> small volume differences and "sensitize" their analysis to larger
>> volume differences.
>>
>> Is this an appropriate use of the extent threshold for VBM ? What
>> effect (if any) would the use of the extent threshold have on
>> corrected or uncorrected p-values?
>>
>> thanks much,
>>
>> Rob McClure
>>
>>
>> Robert K. McClure M.D.
>> Assistant Professor
>> Department of Psychiatry
>> UNC School of Medicine
>> Room 247, Medical School Wing C
>> Chapel Hill, NC, 27510-7160, USA
>> E-mail: [log in to unmask]
>> Office phone: 919-843-6629
>> Fax: 919-966-8004
>> Pager: 919-216-2058
>
|