Simple Model for Encoding Natural Images by Retinal Ganglion Cells with Nonlinear Spatial Integration

Abstract

A central goal in sensory neuroscience is to understand the neuronal signal processing involved in the encoding of natural stimuli. A critical step towards this goal is the development of successful computational models of this encoding. For ganglion cells in the vertebrate retina, the development of satisfactory models for responses to natural visual scenes is an ongoing challenge. Standard models typically apply linear integration of visual stimuli over space, yet many ganglion cells are known to show nonlinear spatial integration in natural stimulus contexts. We here study the encoding of natural images by retinal ganglion cells, using multielectrode-array recordings from isolated salamander retinas. We assess how responses to natural and blurred images depend on first- and second-order statistics of spatial patterns inside the receptive field. This leads us to a simple extension of current standard ganglion cell models, which are based on linear spatial integration. We show that taking not only the weighted average of light intensity inside the receptive field into account but also its variance over space yields substantially improved response predictions of responses to novel images. Finally, we demonstrate how this model framework can be used to assess the spatial scale of nonlinear spatial integration. Our results underscore the importance of nonlinear spatial stimulus integration in the retina in responses to natural images. Furthermore, the introduced model framework provides a simple, yet powerful extension of standard models and may serve as a benchmark for the development of more detailed models of the nonlinear structure of receptive fields.

Representing the dynamics of high-dimensional data with non-redundant wavelets

“…To demonstrate the advantage of WCMI, we applied it to spike responses of a population of cells under the stimulation of natural images. This dataset contains neural spikes of 80 retinal ganglion cells in response to 300 different natural images, 29 where each image was presented for 200 ms then removed with a gray background of 800 ms to account for delayed responses before the next image. Since each cell has a unique response pattern for each individual stimulus image, we decoded different image identities from the whole population of neural spikes.…”

“… 48 This was to test neural encoding of natural images, as described previously. 29 , 31 , 61 The stimulus set contains 300 grayscale natural images using a McGill calibrated color image database. 62 Each image was displayed for 200 ms following an 800 ms interval with a gray background as one trial.…”

“…We used the first 300 ms spiking data for our analysis as this time window was enough to capture the spiking activity of image stimulation. 29 , 31 Then each trial per image per cell is a time series of 300 points. Wavelet analysis was applied to each trial to get a set of 300 coefficients.…”

“…In the data, the RFs of cells were given as a 2D Gaussian function, where the centers and covariances were given in the unit of physical distance, μm. 29 To compute the distance between cells, the Euclidean distance between the centers of Gaussian RFs was computed for pairwise cells. The process was repeated for 50 groups of images, where each group has 4 random images out of all the 300 images.…”

Decoding Pixel-Level Image Features From Two-Photon Calcium Signals of Macaque Visual Cortex

“…In the research on biologic visual mechanism, it is common to see that natural scenes or artificial stimuli have been used as the visual input (Liu et al, 2017;Liu and Gollisch, 2021). However, in most of these works, researchers always investigated visual mechanism through these two types of stimuli separately.…”