Dynamic Spatiotemporal Restructuring of Visual Receptive Fields through Selective Attention: A Model

Cells in the lateral geniculate nucleus (LGN) strongly change their behaviour covarying with different EEG states. During sleep and drowsiness (synchronised alpha, delta-wave EEG) short transient responses prevail whereas during a desynchronised ‘alert’ EEG (beta-waves) long-lasting tonic responses are observed. We propose that this is part of a mechanism used to restructure the spatial and temporal characteristics of the receptive fields in LGN and cortex reflecting changing states of selective attention. To this end we present a model of the primary visual pathway using integrate-and-fire neurons to simulate the afferent signal flow (retina, LGN, V1). The model also implements excitatory topographically arranged lateral intracortical and corticofugal connections which act as a positive feedback and trigger spatial winner-takes-all (WTA) mechanisms enhanced by lateral inhibition at both levels. Furthermore, the LGN membrane characteristic can switch from phasic (hyperpolarised) low-threshold Ca²⁺ bursting mode to tonic (depolarised) signal-transmission mode. Switching is triggered by feedback and amplified by intracellular intrinsic positive-feedback mechanisms in the model LGN. All positive-feedback mechanisms are subject to damping such that they remain ineffective below a certain threshold. Salient stimuli which ‘attract attention’ will push the system above threshold and a self-amplifying process is started which sharpens the cortical receptive fields spatially (by spatial WTA) and drives the winners in the LGN into signal transmission mode (by intrinsic intracellular mechanisms). These results predicted by the model are in accordance with LGN cell behaviour. In addition, the model predicts that cortical receptive fields should be wider during synchronised EEG than during desynchronised EEG.