Hi Emmaly,

I believe those "bilateral arches" you are observing could also simply be a natural byproduct of aCompCor. Similar negative seed-to-voxel correlations at the center of white matter areas can easily appear for seeds that have relatively high global correlation (ie. seeds that tend to show more positive than negative associations with the rest of the brain) as a consequence of spillage into white matter areas from nearby voxels (due to a combination of spatial smoothing and inaccuracies in subject-to-subject coregistration) combined with zero-average properties of aCompCor. In particular, the correlation with an arbitrary seed is going to be approximately zero when averaging across all white matter voxels, so if boundary white matter voxels tend to have positive connectivity due to spillage, interior white matter voxels will tend to have negative connectivity due to centering (hence those arches at the center of white matter areas)

That said, it is a bit strange that those features would reach significance in your results while other features of amygdala connectivity (like anticorrelations with frontoparietal network areas) would not (or appear only weakly), so I would probably recommend taking a second look at your data to make sure that all possible noise sources have been properly addressed. That typically involves: 1) in the denoising tab looking at the histogram of voxel-to-voxel correlation values across all subjects to see if there are any indications of incomplete denoising (e.g. not properly centered distributions, or very wide or dissimilar distributions across subjects); and 2) looking at measures of subject motion, like the QA_InvalidScans and QA_MaxMotion measures created during ART preprocessing, and seeing whether your connectivity patterns are modulated by amount of subject motion (e.g. create two roughly balanced subject groups, High- and Low- motion, and see whether there are significant differences in connectivity between the groups). Last, you probably also want to make sure that you amygdala ROIs are properly aligned with each subject anatomy (e.g. in Setup.ROIs select the amygdala seed and use some of the 'ROI tools. check-registration' options). If all of these look ok, then there is probably no reason to be concerned with your current results. 

Hope this helps
Alfonso


Originally posted by Emmaly Owens: