A spiking multi-area network model of macaque visual cortex

The primate visual cortex consists of a set of specialized areas whose inter-connections have been shown to influence its dynamics both in spontaneous and driven conditions. Hitherto, models of this system have either concentrated on local detailed circuits or studied the interplay of areas, each represented by a few dynamical equations. We present a model which bridges this gap between microscopic and macroscopic dynamics by extending a spiking model of a 1mm2 patch of early sensory cortex [1] to all vision-related areas of the macaque cortex. The single-cell dynamics is kept simple in order to bring out the influence of the complex connectivity, which is based on a systematic synthesis of anatomical and electrophysiological findings. The extension to multiple areas allows us to replace random inputs to the network in part by simulated synapses, thereby increasing the self-consistency of the model. Here the immediate aim is not to address network function from a top-down perspective but to explore the relationship between network structure and fundamental multi-scale activity states.Neuron densities and laminar thicknesses are taken from available data sets or determined based on structural regularities across the cortex. The cortico-cortical connectivity is defined combining binary information from a large number of data sets collected in the CoCoMac database [2] with quantitative data from retrograde tracing studies [3], which is completed by exploiting an exponential decay of connection densities over distance [4]. Furthermore, we implement laminar connection patterns [5] and estimate missing data using a sigmoidal relation between the fraction of supragranularly originating projections and architectural types of areas [6]. We perform simulations of the system using NEST and find a broad parameter regime with asynchronous, irregular spiking across populations, characteristic of spontaneous cortical activity. The rich connectivity structure is reflected in a complex pattern of firing rates across areas and populations, where inhibitory neurons show higher activity than excitatory cells despite identical intrinsic dynamics