Hello,
From a 2nd-level generalised psychophysiological interaction (PPI) analysis, I would like to extract the value of significant PPI activations (within specific clusters of interest), from the group t-contrast of interest. The t-contrast results from a mixed ANCOVA using a flexible factorial design, with a 3-level between-subject factor (controls, patient-group A, patient-group B), and a 2-level within-subject factor (time 1 and time 2), and covariate (memory score at time 1 and time 2 for each subject). For a specific t-contrast [-1 1 1 -1 0 0 1], I would like to retrieve the value of specific significant clusters to be able to use it in further types of correlations analysis, such as with structural data. Aka, for each patient, I am interested in the contrast estimate within a certain cluster that is specific to the group t-contrast.
I have tried different approaches but they all give me different results. I have tried:
1) Manual parameter extraction via Marsbar
http://andysbrainblog.blogspot.com/2012/11/parameter-extraction-with-marsbar.html
2) Using ROI_Extract_Data.m
http://andysbrainblog.blogspot.com/2014/07/quick-and-efficient-roi-analysis-using.html
3) ROI Analysis with SPM12
After loading the group-level flexible factorial t-contrast within a mask/sphere, I extracted whitened and filtered y within a cluster of interest.
4) adjusted responses via SPM 12
- upload an spm.mat with the contrast of interest, go to -->plot --> fitted responses
- Select contrast of interest --> predicted or adjusted: adjusted response
- plot against: Scan or time
- then in matlab see the "y" variable.
-> This method yields 62 values (one for each subject times two, due to the repeated timepoints), which are mirrored and centred around 0 (e.g., 1.71 and -1.71 for subject 1; 0.14 and -0.14 for subject 2, 0.25 and -0.25 for subject 3 etc. until subject 31. I am not sure why this is doing this/ should this method be used only at the 1st-level analysis?
The example documentations show methods for one-sample or two-sample t-tests but I am not sure how to reliable retrieve the contrast estimates from a mixed ANCOVA with repeated measures.
My main questions are:
1. Is there another/a better method to extract the PPI contrast estimate for my repeated measures, flexible factorial design? If not, which of the ones listed above should I use since I am getting different values for each method.
2. For each method, I get two values per subject, but are these values actually representing the connectivity value between a brain region and my seed region used in PPI analysis that correlates with memory scores in that cluster at time 1 and time 2?
I hope my questions make sense; I would appreciate any help on this!
Thank you!
Kind regards,
Marine
|