A theory of orientation selectivity emerging from randomly sampling the visual field

Abstract

Neurons in rodent primary visual cortex are simultaneously tuned to several stimulus features, including orientation and spatial frequency of moving gratings used in experiments. Light-induced signals emitted by retinal ganglion cells (RGC) are relayed to the primary visual cortex (V1) via cells in the dorsal lateral geniculate nucleus (dLGN). However, there is currently no agreement on which thalamocortical transformation leads to the neuronal tuning curves observed in experiments. Here, we outline a model that explains the emergence of feature-specific neural responses as the result of a two-step integration process: First, the compound input to cortical neurons comes from a set of retinal sensors randomly placed in the receptive field. Second, the cortical responses to the combined input are shaped by the rectification caused by the spike threshold of the neurons. We performed numerical simulations of a thalamocortical network stimulated by moving gratings and found that simultaneous tuning to orientation and spatial frequency results from this spatio-temporal integration process. We also show how this tuning is related to the complex structure of the receptive fields that reflect the input. We conclude that different types of feature selectivity arise naturally from random thalamocortical projections. Moreover, we describe in detail the underlying neural mechanism.