Modularity and saccade influence in the cortical vision network: an fMRI / graph theory approach

Considerable regional evidence has accumulated for dorsal-ventral modularity in the visual system, but it is not clear how these modules function at the whole brain network level, or how these networks are influenced by naturally occurring saccades. We addressed these questions using graph theory analysis of fMRI data collected during a task where participants had to remember, then discriminate between two different object features. Seventeen participants judged whether a remembered object changed shape or orientation with or without an intervening saccade. BOLD activation from 50 cortical nodes was used to identify local and global network properties, which indicated greater interconnectivity and efficiency of information transfer during saccades. A network modularity analysis revealed three sub-networks during fixation: a bilateral dorsal sub-network linking areas involved in visuospatial processing and two lateralized ventral sub-networks linking areas involved in object feature processing. Importantly, when horizontal saccades across the remembered object required visual comparisons between hemifields, the two lateralized ventral sub-networks became functionally integrated into a single bilateral sub-network. Comparisons of betweenness centrality between conditions identified several significant hub regions in occipital, parietal, and frontal cortex involved in linking distant network nodes during saccades. These results provide support of a ventral and dorsal stream distinction in human perception and show how hemispheric sub-networks are modified to functionally integrate information across saccades.