>Date: Tue, 11 Dec 2001 17:41:14 -0400
>To: Sam Reyes <[log in to unmask]>
>From: Badreddine Bencherif <bench@[162.129.121.62]>
>Subject: Re: resels and global normalization ?
>Cc:
>Bcc:
>X-Attachments:
>
>Sam
>Thank you for your prompt reply.
>There is a large variability in my PET neuroreceptor data (binding
>potential) between subject (mean range 0.7 to 1.4). This is the
>reason why I use AnCova when I want for example to assess the effect
>of age on binding after discounting inter subject differences in
>global.
>The question is how using AnCova versus Non global nomalization
>translates into computation of strikingly different smoothness
>(calculated using the variance-covariance matrix of the first
>partial spatial derivatives and used for correction for multiple
>comparisons) by SPM99.
>The same data (Gaussian smooth gaussian 14x14y14z , same threshold
>etc ...) was used for both Ancova and no global normalization.
>
>Thanks
>Didine
>
>>Badreddine Bencherif wrote:
>>
>>>Dear SPM ers:
>>>Using SPM99 I used no global normalization (ngn) and then ancova (anc
>>>for global differences nuisance) on the same set of data to compare
>>>the results.
>>>PET data = [(n=14 subjects) voxel size 2,2,2, smoothing Gaussian
>>>14,14,14 mm. same grand mean scaling and absolute threshold]
>>>
>>>I found
>>>1) mask produced by the 2 analysis was similar (visually)
>>>2) RPV.img resels images were dissimilar. far more contrast and
>>>variation using anc than ngn.
>>>3) Striking differences in number of resels.
>>>ngn smoothness fwhm 43.9, 47.6, 47.3 (mm) 1 resel = 12372.74
>>>voxels, S=7.1 resels.
>>>anc smoothness fwhm 15.7 17.0 17.3 (mm) 1 resel = 577.34 voxels,
>>>S=151.6 resels.
>>>S is 102089 for both data.
>>>
>>>Can someone explain these large differences.
>>
>>Hi,
>>
>>What are the values you are seeing for a typical voxel in the
>>RPM.img file for the ngn analysis? My guess is that it is much
>>higher than anywhere in the anc RPM.img file. If this is true, it
>>is likely that there are large variability in gCBF or tracer dose
>>across scans. This variability in gCBF is being modeled in the anc
>>study. Based on the fact that a substatianl amount of the
>>variability in between scans seems to be accounted for by changes
>>in gCBF or tracer dose, I would model gCBF and adjust for it. The
>>only reason I wouldn't is if you have quantitative blood flow data
>>and you expect gCBF to change with the task.
>>
>>Hope it helps,
>>---sam
--
Badreddine Bencherif,MD
Department of Radiology
Division of Nuclear Medicine
Johns Hopkins University School of Medicine
601 North Caroline Street / JHOC 4230
Baltimore, MD 21287-0855 USA
Voice: (410) 614-2787 Pager: (410) 283-2050
Fax: (410) 614-1977
|