1 Introduction

Visual attention, attracted to the most salient regions in visual space, lies at the core of our understanding of the visual world. Computational saliency maps have long been utilized in this regard to figure out where attention is directed. In this article, we analyze electrophysiological data from the retina and the primary visual cortex V1 in mice to uncover how this works.

2 Methods

Our investigation involves the use of electrophysiological data from mouse retina and V1. On one hand, we've got the retinal ganglion cells (RGC) response data which is obtained from a deep model trained on this response to natural images. On the other hand, we have the cortical response data, which is measured experimentally.

The Mouseland dataset was utilized for the V1 response data, providing invaluable insights from multiple simultaneous recordings of V1 cell responses. The data from the retina was obtained via recordings from 60 biological (mouse) RGCs, providing crucial insights into retinal activity.

3 Joint analysis of cortical (V1) response and visual attention

To understand the effect of visual saliency on neuronal firing rate, we performed statistical analyses using the Kolmogorov-Smirnov (KS) test. This test enabled us to compare two distribution functions to see whether there were significant differences in the neuronal responses at salient versus non-salient image regions.

4 Joint analysis of retinal response and visual attention

Further, we conducted an analysis of retinal response and visual attention using firing rate ratios and correlation coefficients. We looked at two firing rate ratios, which allowed us to compare the firing rates at the most salient image regions to a baseline firing rate.

After analysis, we found that in the primary visual cortex (V1), a subset of around 10% of the neurons respond differently to salient versus non-salient visual regions, whereas visual attention information was not traced in retinal response. It appears that the retina remains naive concerning visual attention, and it is the cortical response that gets modulated to interpret visual attention information.

This information is crucial for the design of improved visual prosthesis systems that create artificial visual percepts for visually impaired individuals through electronic implants on either the retina or the cortex.

In conclusion, our study establishes a valuable premise for further experimental animal studies that can be designed to delve deeper into the biological basis of visual attention that we have traced in this study.