Context integration in visual processing: a computational model of center-surround suppression in the visual system

Abstract

A dysfunction of GABAergic neurotransmission is hypothe-sized to be an important factor in the pathophysiology ofschizophrenia [1], depression and anxiety disorders. Find-ings of decreased center-surround suppression (CSS, i.e. themutual inhibition of a focalvisual stimulus and its sur-rounding) have been interpreted in terms of GABAergicdysfunction [2]. Consistently, strongly decreased CSS isreported in schizophrenic patients [3]. However, the under-lying mechanisms of this decrease remain unclear.MethodsWe developed a biologically inspired neural networkmodel of the human primary visual system, consisting oftwo retinas, two layers of dorso-lateral geniculatenucleus and section B and C of layer 4 of the primaryvisual cortex. The model was built using GENESIS(www.genesis-sim.org) and comprises about 150,000neurons. Three different kinds of neurons were incorpo-rated, namely thalamic relay neurons, and excitatory andinhibitory cortical neurons. The processing of input inthe retina was simply modeled as random-spike units,where the average firing rate corresponds to the inputintensity. The neurons consisted of a soma, an axon andtwo or four dendrites for relay or inhibitory and excita-tory neurons, respectively. We used different ionic chan-nels to reproduce the desired firing behaviour. Synaptictransmission was implemented using the built-in excita-tory and inhibitory synaptic channels.ResultsWe designed a protocol to explore the center-surroundsuppression capabilities of our model. The retina wasstimulated with contrast stimuli similar to those used in[2]. We compared activity in central neurons of layer 4Band 4C in response to two stimuli with a focal circle of40% Michelson contrast, where the target stimulus hada high-contrast surrounding while the reference had auniform surrounding. We found a decrease in activity of27.73% in layer 4B and one of 31.41% layer 4C whencomparing responses to the target with responses to thereference.A plausible neural network model of the human pri-mary visual system which is consistent with psychophy-sical effects previously reported in humans [2] is astarting point to further investigate the role of GABAer-gic inhibition in visual context integration. The capabil-ities of the model to explain both individual variance ofCSS strength and CSS in pathological conditions ofGABAergic neurotransmission remains to be tested. Weplan to do so on the basis of psychophysical CSS data inhumans and correspondingin vivo measurements ofGABA and Glutamate concentrations in primary visualcortex using Proton Magnetic Resonance Spectroscopy(1H-MRS).