Network Analyses of Multisensory Processing

Paul J Laurienti, Christina E Hugenschmidt, Joseph A Maldjian, Benjamin Wagner, Satoru Hayasaka
Poster
Time: 2009-06-30  09:00 AM – 10:30 AM
Last modified: 2009-06-04

Abstract


The brain is a complex network and multisensory processes occur within the framework of this highly integrated system. While most traditional studies have evaluated unisensory and multisensory brain regions in isolation, a more comprehensive understanding of multisensory processes will be achieved through studies of the brain as a complex network. Since the discovery of small-world 1 and scale-free2 networks studies of complex networks have emerged in virtually every scientific discipline. Small-world networks have the unique property that all elements (vertices or nodes) in the network are within a few links (edges) of each other but maintain high clustering between neighboring elements 1. Scale-free networks have nodes that have an extraordinarily high number of connections (hubs) and a distribution of connections that follows a power law 2. The brain is known to exhibits regional specificity and distributed processing. In the past several years investigators have begun to evaluate the structural and functional connectivity of the brain using network tools and have repeatedly demonstrated that the brain is a small-world network with specific regions serving as the network hubs. Most studies have evaluated the brain at rest and no studies to date have examined network properties under various multisensory conditions.

The current study was designed to evaluate network properties in the human brain under various multisensory conditions. Functional MRI (fMRI) data were collected from 14 subjects while they viewed a fixation point, watched a silent movie, and watched a movie with a sound track. Functional imaging was acquired based on the blood oxygenation dependent level (BOLD) signal using echo-planar imaging (EPI). Networks were generated by performing cross-correlation analyses (functional connectivity) of the temporal signal between each voxel across the entire brain image. This analysis generated a correlation matrix that contained 400 million cells for each subject. A threshold was applied to each matrix based on the correlation coefficient to generate binary adjacency matrix. Global properties such as clustering coefficient, path length, and degree distributions were compared across stimulus conditions. In addition, the location of connector hubs and community structure were assessed.

The global network metrics did not reveal differences between sensory conditions indicating that the network did not exhibit overall organizational changes across the stimulation conditions. However, changes in the local network properties were consistently revealed. Brain hub structure exhibited spatial changes with hubs in the default-mode network at rest. The hubs were significantly increased in visual cortex during the silent movie and in visual and auditory cortices during the multisensory movie. The community structure analysis revealed interesting findings in sensory cortex. During viewing of the silent movie the somatosensory and auditory cortices were clustered into a single neighborhood likely due to suppressive effects from visual attention. During the multisensory movie the community structure changed and auditory cortex became a highly organized and localized neighborhood. The data demonstrated that network analyses provide a unique and powerful method for evaluating multisensory processes. Identification of network neighborhoods that dynamically change with sensory conditions reveals the functional organization associated with multisensory processing.


References

1. Watts, D. J. & Strogatz, S. H. Collective dynamics of 'small-world' networks. Nature 393, 440-2 (1998).
2. Barabasi, A. L. & Albert, R. Emergence of scaling in random networks. Science 286, 509-12 (1999).

Conference System by Open Conference Systems & MohSho Interactive Multimedia