Functions and Origins of Nonlinear Processing in the Retina
by Steffen Nitsche née Krüppel
Date of Examination:2024-01-19
Date of issue:2024-11-29
Advisor:Prof. Dr. Tim Gollisch
Referee:Prof. Dr. Tim Gollisch
Referee:Prof. Dr. Alexander Gail
Referee:Prof. Dr. Thomas Euler
Persistent Address:
http://dx.doi.org/10.53846/goediss-10899
Files in this item
Name:Functions and Origins of Nonlinear Processin...pdf
Size:77.0Mb
Format:PDF
This file will be freely accessible after 2025-01-18.
Abstract
English
Millions of people worldwide suffer from terminal blindness, but state-of-the-art vision restoration devices can only produce a sensation fundamentally different from natural seeing. We lack the sufficient understanding of the retina – a thin sheet of neurons much more complex than a camera’s chip – to be able to replace it. Ganglion cells, the retina’s output neurons, perform complex computations on the incoming visual image, with nonlinear calculations, particularly in the spatial domain, lying at the heart of their processing power. Spatially nonlinear processing refers to the observation that many ganglion cells sum the luminance in their receptive fields (essentially the part of the visual image that an individual ganglion cell “sees”) in a nonlinear fashion with bright and dark regions not canceling. In this thesis, I have compiled our current knowledge of spatially nonlinear processing by retinal ganglion cells. It is involved in various aspects of encoding the visual environment, like motion sensitivity, and endows the retina with a wider range of functionalities compared to linear receptive fields. But many functions are still poorly understood, especially in natural viewing conditions. Spatial nonlinearity is thought to arise from the nonlinear transmission of signals from bipolar cells. Bipolar cells provide the link between the light-sensitive photoreceptors and the ganglion cells and can manifest as an orderly mosaic of functional subunits in a ganglion cell’s receptive field. Few methods are available to infer the subunits and thus bipolar cells connecting to a ganglion cell being measured from, and their utility is limited by their high data demands and intensive post-processing. To shed more light on ganglion cell computations during natural viewing, I investigated their encoding strategies during saccades – rapid eye movements that shift the gaze. These trigger a diverse set of responses in the different types of ganglion cells in the primate retina, recorded with multi-electrode arrays. Spatially nonlinear integration of luminance can explain many of these responses, but it is not sufficient to generate a particular sensitivity to changes of the image across the saccade common for Off ganglion cells (preferring light decrements) but not On cells (preferring light increments). Intricate interactions between functional subunits likely generate this type of sensitivity. I also developed a novel method to infer the subunits in a ganglion cell’s receptive field from its recorded responses that draws from concepts of super-resolution microscopy and tomography. In comprehensive simulations, this method can quickly and reliably identify the nonlinear inputs to model ganglion cells, and experiments suggest a similar efficacy. In contrast to previous methods, its speed might reveal subunits early on during an experiment leaving ample time to further study the uncovered retinal circuitry and exactly how it generates the complex computations observed, e.g., during saccades. Combined, these findings further our understanding of the retina’s function and bring us closer to proper vision restoration.
Keywords: retina; ganglion cells; primate; saccades; subunits; nonlinear; tomography; super-resolution; receptive field
