Visual Stimuli for Selective Modulation of Neural Pathways in the Retina

University Of Washington, Seattle WA

Investigators

Abstract

PROJECT SUMMARY / ABSTRACT The primate retina’s output is split into 15-20 Retinal Ganglion Cell (RGC) types, each processing and transmitting different aspects of the visual world. Different RGC types are thought to have distinct roles in visual processing and behavior, and type selective impairment is implicated in optic neuropathies and developmental disorders. The response properties of numerically dominant RGC types have been described using predominantly linear models, and a large body of psychophysical work uses these models to propose visual stimuli that can selectively drive a given RGC type. However, these stimuli face three important limitations: 1. Recent studies challenge the accuracy of key model assumptions, 2. They lack direct neural population testing and the underlying models focus on single cells, and 3. They are predominantly simple artificial stimuli and the underlying models fail to capture responses to naturalistic stimuli. A lack of validated, ethologically relevant RGC type selective stimuli limits the ability to answer key questions in visual neuroscience such as whether Midget RGCs dominate ventral stream processing and Parasol RGCs dominate dorsal stream processing. The central aim of this proposal is to acquire the necessary training in sensory neuroscience, psychophysics, and computational neuroscience to directly test existing approaches, refine and evaluate new models of RGC responses, and leverage new models to generate effective, ethologically relevant RGC type selective stimuli. Preliminary data from in vitro population recordings of peripheral Macaque retina suggest that stimuli based on assumed contrast gain differences between Parasol and Midget RGC types fail to achieve their proposed selectivity. I will confirm and extend this to other stimuli based on assumed response property differences between RGC types, which will aid interpretation of behavioral results in healthy and diseased subjects tested with these stimuli. Further, I will refine and evaluate newer mechanistic and deep neural network models of RGC responses that can capture nonlinear circuit components driving responses to naturalistic stimuli. I will leverage these models for precise manipulation of RGC population outputs by systematically generating stimuli that minimize modulation of one RGC type population while maximizing modulation of others. This will provide validated tools for studying the downstream contributions of the numerically dominant RGC types in naturalistic contexts with applications in basic research and the clinic. Through this training and research plan, I will develop the necessary expertise and skills for contributing substantially to sensory neurophysiology and pathology as a physician scientist.