Model-based decoding of time-varying visual information during saccadic eye movements using population-level information

Abstract

Our visual system scans the environment by directed ballistic movements of the eyes, known as saccades, to bring the locations of interest to the fovea for further visual processing. Accurately characterizing visual responses in the perisaccadic period is an important step toward understanding how the visual world is represented during saccades, and therefore how the brain maintains the stability of visual perception across saccades. Here we use our recently developed probabilistic model in the generalized linear model framework, which uses both simultaneously recorded spiking responses and the local field potential, to extract time-varying visual information represented by individual neurons in the middle temporal cortex during saccades. Our results indicate that the optimal model-based decoding exploiting the response correlation structure between neurons, or the network state of the neuronal population, or both, extracts significantly more information about the visual scene than decoding based solely on the single neuron spiking activity. This model-based approach reveals the role of several external and internal covariates in the extrastriate coding of visual stimulus during eye movements. Moreover, the modelbased decoding provides a general framework for testing the contribution of specific perisaccadic changes in neuronal responses to our perception of the visual scene during eye movements.