Summary: This work is driven by: A) recent theoretical developments in the area of nonlinear complex networks. It has been
documented that disparate networks of interacting elements obey similar organizational principles. For example, in
metabolic networks, it is shown that, despite a wide variation in constituents and pathways, they share topological scaling
properties that are similar to the organization of complex non-biological systems, such as the Internet.
The approach is also driven by: B) the large body of electrophysiological evidence indicating that adequate understanding of
the dynamics of information processing in the brain must take into account interactions between neurons. Therefore, we
combine A and B to examine brain information processing from the viewpoint of a fully connected dynamical system. fMRI
data is used to this end.
Standard analysis of network properties of brain fMRI involves two steps: (I) regions involved in a given task are identified
by relating individual voxel activity to the timings of task presentation. (II) The inter-relationship between the identified
regions are then analyzed based on their cross-covariance matrix. Here we extend functional connectivity analysis of fMRI
by studying connectivity across all brain voxels.
We demonstrate how studying the full connectivity of functional brain states one extracts different information than
traditional stimulus-driven methods. The networks identified by the methods parallel to brain areas consistent with the
tasks investigated, and seem to provide additional information regarding sensory-motor transformation integrated by
connectivity through regions involved in attention.
The extracted networks have topological properties shared with many biological and technological communication
networks, as they show a Scale-free, Non-hierarchical, Small-world organization. It is the first time these properties have
been characterized in brain activity networks.