Hi all,
I am a beginner in fMRI and am trying to analyze data with FSL. However, I encountered questions for some basic issues in data analysis (I'm sorry if someone posted these before~).
1. Slice timing correction. I read in the FSL lectures that the recommend solution is to use temporal derivative in GLM model in FEAT ( http://fsl.fmrib.ox.ac.uk/fslcourse/lectures/feat1_part1.pdf ). When setting an EV in GLM, it is optional to "Add temporal derivative". There is also an option in FEAT which is called "Slice timing correction" and requires the slice order to be specified. Does this mean that I only need to check or specify one of the 2 options?
2. I want to find the brain regions that is correlated to my features with GLM, both negative and positive. For negative correlations, I read in previous posts that I need to set corresponding coefficients to -1 in Contrast. Do I need additionally to change the Z threshold in Post-Stats to a negative one (e.g. -2.3)? Is there any other settings I need to change?
3. About GLM EVs:
(1) Orthogonalization and de-meaning of EVs. I read in some papers that the authors orthogonalize EVs, which is not recommended in FSL. So when do I need this option? Also, do I need to manually de-mean each of my custom EVs in advance?
(2) In FEAT Stats, there is an option "Add additional confound EVs". Is there any difference when I use this option compared to add confound EVs through GLM interface?
(3) I'm doing analyses regarding simultaneous EEG-fMRI. Several kinds of EEG features are extracted and are to be used as EVs in GLM. I found the results are very different when only one of them is used in GLM as compared to all kinds are used in GLM. I am not sure if there is something wrong with my analysis, but I would like to know what is the principles when there are more than one EVs of interest?
Best,
Chao
|