Dear Ozkan,
you can find many answers on your questions in the CAT12 manual and in the online help. Anyway, see below for my answers...
On Fri, 2 Mar 2018 22:27:39 +0000, Ozkan C <[log in to unmask]> wrote:
>Dear SPMusers,
>I am confused with some points on using SPM12-CAT12(DARTEL) for structural MRI data preprocessing (I have Parkinson's disease and control groups).
> -How can I check the performance of each preprocessed smwp1 data (Is it logical to check values on report pdfs and say higher results of resolution/noise/bias/weighted average are well processed)?
Try the tool for checking sample homogeneity where you can also use the quality measures that are displayed in the pdf. In the online help there is also a more detailed explanation of these quality measures.
>
> -I used two sample t test. For masking, I only use implicit masking:yes (leave threshold masking none as default). Would using threshold absolute value 0.2 increase my classification result? Should i take it into account? (Since I have few clusters detected in the mask, i prefer not to take it).
If you are not using an absolute threshold you might also find effects in white matter regions although you are analyzing gray matter data. If you are sure that your finding are exclusively in GM this might be fine. However, I always recommend to use an absolute threshold to ensure that analysis focuses on the tissue type your are intention.
>I use age and TIV as two nuisance variables by taking them under covariates in the factorial design spec. Should I use TIV as global value as follows:
>o Global Calculation → User <-X → Define here the TIV values
>o Global Normalization
> • Overall grand mean scaling → Yes (and value 50)
>o Normalization → User → Proportional
>
This is explained in the manual. If you keep the default value of "50" you will get into trouble if you define an absolute threshold.
> -In result, I check t-contrast (FWE: none, p<0.001, k_threshold=0) to see the differences between two groups . I have few clusters and hence a few number of voxels (Around 6 clusters and 300 voxels). If I use FWE p<0.05, then I have no clusters detected. Is it normal in PD data (I have 40PD and 40 HC) or am i doing something wrong?
Your sample size is fine and usually there should be differences in the putamen and the globus pallidum in PD patients. However, maybe other factors cause variance in your data. You can try to add sex as nuisance parameter. I would also display the ResMS image (residual mean squares, which are the remaining residuals after fitting the model) and check that no large values are visible in the putamen and pallidum. I would also check that no age differences exist between your groups. If you then add age as nuisance parameter you not only remove variance explained by age but also by the group differences. Same holds true for TIV as nuisance parameter.
Best,
Christian
>
>I would appreciate your help a lot.
>Thank you in advance.
>Best
>--
>Ozkan
|