A model for neuronal processes underlying colour vision

Abstract

To provide a physiologically plausible computational model for the development of chromatically selective cells in macaque primary visual cortex. The model comprises of cones, horizontal, bipolar, ganglion, lateral geniculate nucleus and primary visual cortex input layer cells. Each cell was represented by a differential equation and all equations were solved simultaneously to provide time courses of membrane potential and, for spiking cells, impulse rate.

Chapter 3 describes the construction of ganglion cell receptive fields. Negative feedback from horizontal cells to cones was built into the model and two notable results emerge. First, the full model can be reduced to a ratio-of-Gaussians model which corresponds more closely with the anatomy and has a temporal component. Second, the ratio-of-Gaussians model shows how centre and surround radii can be calculated from the radii of the optical point spread function, horizontal cell receptive field, and ganglion cell dendritic field.

Chapter 4 concentrates on the development of cortical circuits before the onset of vision. Stimulation of the model in this first development phase depends on spontaneous waves of retinal activity. I assumed that the geniculocortical synapse was plastic and Hebbian: positive correlations of lateral geniculate nucleus and cortical responses strengthened the intervening synapse, and negative correlations weakened it. This led to stronger cortical responses, orientation selectivity, and mapping of orientation preference across the visual field.

Chapter 5 is devoted to the second phase of development, in which stimuli were scenes from the natural world. The development here was again determined by a Hebbian mechanism, and two types of cortical cell emerged. The first, luminance cells, were best driven by the stimulus component varying in luminance. Luminance/colour cells, by contrast, produced a relatively strong response to equiluminant stimuli.