Dear Oh Dong Hoon,
this question pops us periodically in the list. The simple answer is:
use a two-sample t test. As usual, the larger the control group, the
better. To determine the size of the control group, you need estimates
of the precision of the data you are using. I can personally say
nothing about the precision of SPECT data.
The more complex answer is that distributional assumptions play a much
larger role in a t test of this kind than is usually the case. In
particular, using a large control group does not protect against
skewness, kurtosis etc. as is usually the case.
I do not think that significance values in such a situation can be
trusted too much. However, I would agree that standardization might
contribute to a more sensitive assessment of images than when done on
the raw images. For this however you do not need to compute
significance levels.
Roberto Viviani
Department of Psychiatry III
University of Ulm, Germany
Quoting Oh Dong Hoon <[log in to unmask]>:
> Dear SPMers.
>
> How can I analysis individual patient's SPECT (or PET) image by using SPM?
> I want to know how to make individual analysis.
>
> In comparison study (analysis),
> we are able to compare the difference between the experimental group and
> control group using [compare-population: 1scan/subject] option ;
> 'two sample t-test' or 'AnCova' in SPM software.
>
> However, in clinical (or other experiemtal) situations, if I would
> like to compare
> individual patients' SPM results with visual interpretation of brain SPECT
> images by experiecned radiologist, how can I do that?
>
> I hypothesized that individual SPM analyses of SPET/PET images could be
> more sensitive in searching rCBF changes than visual interpretation.
>
> In my case, is it suitable to use 'two sample t-test' or 'AnCova' in SPM
> software? (comapring 'single subject image' with control group is possible
> option?)
>
> If a control group are needed, how many subjects are appropriate?
> (for examle, 10 age & sex-matched healthy control subjects is suitable?)
>
> If not, how can I make individual analysis of SPECT(PET) image?
>
> Please let me know how to do that in detail or introduce helpful references
>
> Thanks in advance
> Best regards
>
|