Dear all,

I'm running resting state analysis on Conn of one patient vs multiple controls.

I have preprocessed my images using spm, and add them as inputs in my conn analysis. I don't use the GUI, and use a script instead, which mainly uses the default settings.

My patient has extended lesions in the visual areas, and I have a small ROI in the visual cortex, corresponding to some spared cortex in my patient. I'm trying to look at the difference of correlation of this ROI (seed-based-connectivity) between my patient and my controls (1 vs 30+).

My issues are that:

1/ I don't find differences in lesioned visual cortex, while it's where I should find the biggest differences since in my patient there is no signal correlation with a given ROI / voxel?

2/ I have important correlations in these lesioned regions in my controls when I look at them only, and no correlation when I look at my single patient only (either when I look at it through Conn GUI, or through SPM UI). I think this could be a lack of power, but I still find some differences between my 2 groups in other regions.

3/ When I look at the Beta maps (that are entered in the second model in Conn) of the SBC of the visual ROI, in my patient I have high Z-values (= correlation) of the visual seed with almost the whole brain, including the visual cortex, while in my controls I have an even bigger correlation (Z-values) in the visual cortex, but really low Z-values with the rest of the brain, resulting in a brain splitted in 2 parts in each control individually (visual cortex with high correlation, and rest of the brain with low correlation).

--> I don't understand why I have high Z-values in my single patient in the whole brain, while when I look at spmT maps through conn UI or SPM UI, I don't find any significant cluster? Is there some kind of normalization by the global correlation value of the brain?

--> I also don't understand why I can't find differences of correlation between my single patient and group in the lesioned regions, while there is no healthy cortex, or even CSF since there is a lot of atrophy. Is it just a matter of statistical power since it's one vs multiple subjects, or is it something else that I'm missing?

I have tried to make my second-level models through SPM with the BETA maps from Conn, using ANOVA or T-test models, and I always find the same differences in the same regions, but never differences in these lesioned areas, even if I lower the voxelwise threshold...

So please, do you have any idea about why does this happen, what could be the issue, and how to fix it?

Thanks all,

Fabien