question about preprocessing in spm5, analyze in resting state toolkit, & conducting paired sample t-test in spm5

Submitted by brodhinooch on
hello

i am have some questions relating to my resting state fMRI analysis. i have 6 subject that were scanned pre medication and post medication. these images were preprocessed in spm5 which included realignment, normalization, & smoothing. Next, I conducted functional connectivity resting state analysis with the toolkit on this website which resulted in fcMAPs for each subject pre & post for a total of 12 maps. Next, I used spm5 to conduct a paired samples t-test on the pre & post maps which included 6 pre fcMAPs & 6 post fcMAPs. 2 separate contrasts were created which a) compared the pre fcMAPs-post fcMAPs and b) compared the post fcMAPs-pre fcMAPs. 

The question I have resulted from these contrast. Controlling for none at the p = .0001 level the glass brain showed activations which were outside the brain for each of the contrast. This does not seem possible since each participant was normalized in preprocessing. Control for FWE at the p = .05 level resulted in no significant activations.

Could someone please explain to me why I had significant activations on the glass brain when i did not control for FWE that were outside the normalized brain and did I do something wrong in the analysis. Is the resting state toolkit fcMAPs not compatible with the spm5 normalized brain. All I can think of is the images were not normalized correctly, but I am sure normalization is correctly performed in spm5, but that is all i can think of. Thank you in advance for all your suggestions on this matter.

Jef

YAN Chao-Gan

Wed, 09/23/2009 - 16:25

Hi!
    1. What's your seed regions?
    2. Did you use FCmaps or zFCmaps entering 2nd analysis? Usually we use the zFCmaps after Fisher's r-to-z transformation.
    3. The difference outside brain is difficult to interpret. It may caused by the data processing, or by the signal it self (if a special seed ROI was chosen.) We usually use a mask to remove the outside brain regions, since they are not region of interests.
    If you use REST V1.2 (REST V1.1 is for SPM2) or above release, then there is no problem to preprocess data in SPM5 and calculate FC in REST, then come back to SPM5 for statistical analysis. Basically, this is our daily procedure.
    Best wishes!
Forums