Abstract

The observation that there are relatively consistent distributed patterns of activity during rest has lead to the suggestion that it might be possible to characterize network dynamics without needing an explicit task to drive brain activity. This possibility has been explored in studies of resting-state networks (RSNs). Resting state dynamics is mainly characterized by the resting functional connectivity matrix that reflects long-range correlations between different brain areas even under no task, no stimulation conditions. Nevertheless, on the other hand, many neurophysiological studies demonstrate that at the single brain area level the activity is largely uncorrelated. Traditional network models ignore this fact, so in this talk I will discuss the effect of constraining the emergence of resting long-range correlations by maintaining the local level uncorrelated. This can achieved by requiring that the local dynamics of each single area is balanced, i.e. the amount of local excitatory and inhibitory currents are equilibrated. We show that under this constraint, the dynamical working region required for the generation of resting state activity is much broader and thus robust if the network is accordingly balanced. We demonstrate this effect by large-scale simulation based on DTI /DSI extracted structural anatomical matrices and detailed spiking and synaptic models emulating empirical obtained resting functional BOLD correlations. Further more, using advanced mean-field techniques we are able to show the same effect analytically.