Hi CONN users!

Some of my functional images from some subjects are cut-off at the bottom of the cerebellum (because the scann didn't take the whole brain, including the whole cerebellum), so I was wondering whether conn toolbox includes those voxels from those cut-off images which show 0 activation in order to perform the connectivity analysis. 

Do those images influence my final analysis, or conn just ignore those missing voxels? 

Thank you in advance!

Hi Albert,

Generally the analyses will include all voxels within your analysis-mask, even if that mask include voxels which were not scanned for some subjects (those will be treated as 0-connectivity values). More specifically, CONN does not include those voxels when computing ROI-level timeseries (e.g. if you have an ROI that encompass some portion of the cerebellum which was scanned, and some which was not, CONN will extract the average BOLD signal only within those cerebellar voxels that were scanned), but it does include those voxels (all voxels within the analysis mask defined in Setup.Options) when computing voxel-level statistics (e.g. in your second-level SBC analyses,  statistics for an individual cerebellar voxel will be computed using the actual seed connectivity for those subjects where that individual voxel was scanned, and it will include 0-connectivity values for those subjects where that individual voxel was not szanned), so this means that you may see effects in those "partially-scanned" areas which may be caused just by the differences in acquisition parameters across subjects. 

In general, both anatomical as well as scanning differences across subjects can result in artifactual connectivity differences (through many different causal pathways), so it is not a bad idea to try to discern and potentially control for those anatomical/scanning differences explicitly in your second-level analyses if possible, particularly if you find results in these same areas (e.g. if you are trying to compare connectivity between controls and patients and you find a significant difference in cerebellar connectivity, then you want to find out if those "cropped-cerebellum" scanning issues appear selectively across controls vs. patients, and if so, quantify those effects and enter them into your second-level analysis as additional control covariates to make sure that the between-group differences in connectivity that you are observing are not caused by between-group differences in scanner-coverage)

Hope this helps
Alfonso

Originally posted by Albert Bellmunt :