Dynamic Model of the dLGN Push-Pull Circuitry - Foundations on Natural and Artificial Computation - page 128

Dynamic Model of the dLGN Push-Pull
Circuitry

Ruben Ferreiroa 1 , Eduardo Sanchez 1 ,andLuisMart ınez 2

1 Grupo de Sistemas Intelixentes (GSI)
Departamento de Electronica e Computacion,
Universidade de Santiago de Compostela,
15782 Santiago de Compostela, Spain
[e-mail] , [e-mail]
2 Visual Neuroscience
Instituto de Neurociencias de Alicante,
03550 Alicante, Spain
[e-mail]
http://www-gsi.usc.es/index.html

Abstract. In the present work we propose a dynamic model of the lat-
eral geniculate nucleus (dLGN) that allows the implementation of dif-
ferent configurations of the push-pull circuitry in order to study the
spatio-temporal filtering being carried out. It is widely accepted that
each relay neuron receives only one input from a single retinal ganglion
cell, which leads to interpret that the thalamus preserves the retinal
structure of the receptive field and thus works as a simple relay station.
We believe that this assumption is not fully valid and the thalamus could
perform a more relevant processing of information through its complex
push-pull circuitry. To test this hypothesis, a computational model was
developed with a wiring configuration (convergence/divergence) between
the retina and the dLGN based on experimental evidences, and a realistic
description of the ON and OFF channels of dLGN. We found that this
configuration may help improve the contrast of a stimulus by increasing
its synaptic weight on higher frequencies.

1 Introduction

Visual information is processed through multiple pathways that originate in
different types of retinal ganglion cells and remain relatively well segregated
through the dorsal lateral geniculate nucleus (dLGN) until reaching the visual
cortex (Cleland and Lee 1985; Cleland et al. 1971a; Hamos et al. 1987; Mas-
tronarde 1992; Sincich et al. 2007; Usrey et al.1999). In the cat's visual system,
there are two main pathways (X and Y) that differ in their response proper-
ties and axonal projections. Whereas most X retinal afferents project to a single
dLGN layer (layer A), Y retinal afferents can diverge into two dLGN layers (layer
A, C). In our study we will focus on X-type afferents located in the fovea, re-
sponsible for processing the details of the scene, and the DLGN's layer A, which
is composed by two types of neurons: relay cells and interneurons. In addition