John McGonigle1, 2, Majid Mirmehdi2, Laurence Reed1, Andrea Malizia3
1Neuropsychopharmacology Unit, Imperial College London, London, United Kingdom; 2Computer Science, University of Bristol, Bristol, United Kingdom; 3Psychopharmacology Unit, University of Bristol, Bristol, United Kingdom
When examining functional connectivity in the brain it is common to compare the synchrony of the mean time courses of spatially separated regions of interest and model these as edges between nodes in a graph. However, in creating a node, due to the commutative nature of the averaging, the quality of the time course can be driven by the number of voxels in the region in the native space of the subject. We explore this issue using real and simulated data and find that differences in apparent connectivity between groups with systematically different structure and volume may be artefactual.