Computational study of correlated neuronal activity in a model neural network

Synchronized oscillation is a commonly observed phenomenon in many parts of the brain. Among various types of these oscillations, gamma frequency rhythms in the cortex are thought to play an important role in the sensory information processing. To study how this neural oscillatory activity affects the information processing in the brain, a large neural network model of the visual cortex, built from Hodgkin-Huxley type excitatory and inhibitory neurons, was used. The spontaneous gamma oscillations in response to feedforward input spikes were generated by the isotropic local lateral connections between excitatory and inhibitory cells. The response function of a cortical neural network was considerably modified by the spontaneous oscillations. The response function of a population became more nearly linear in the presence of the oscillations, unlike a step-like single neuron response function. As a result, the sensitivity to weak inputs was increased and the encoding capability of the network was enhanced. The excitatory-excitatory cell couplings in the network generated spatially traveling waves of fluctuating voltage. Both the strength of the gamma oscillations and the spread of synchronized cortical activities were proportional to the excitatory-excitatory coupling strength. Moreover, the cortical network selectively amplified feedforward inputs according to their strengths, because the excitatory-excitatory coupling strength was dependent on the feedforward input strength. The relative weight of the thalamic feedforward and the recurrent cortical inputs to the visual cortex response was also modulated by the spontaneous cortical oscillations. The spontaneous local gamma oscillation increases the weight of cortical lateral inputs and decreases that of thalamic inputs to network outputs. Thus the spontaneous gamma oscillations can dynamically modulate the properties of the neural network response by controlling the relative weights of the feedforward and recurrent inputs to the network. This is a general mechanism by which neural oscillatory activity may control the information flow in the nervous system.