Decoding retinal signals with denoising natural image priors
Abstract/Contents
- Abstract
- The retina transforms and compresses visual information as it encodes incident patterns of light into the spike trains of retinal ganglion cells. Understanding the nature of these signals and the cells that carry them is fundamental both to understanding the visual system, and to the development of retinal prosthetic devices that restore vision. This thesis first explores the content and meaning of the retinal code, using a novel Bayesian maximum a posteriori method for reconstructing (decoding) natural images from the recorded spike trains of large populations of retinal ganglion cells. This method achieves state-of-the-art performance for reconstructing statically-presented natural images, and generalizes straightforwardly to reconstructing natural movies with emulated fixational drift eye movements, while providing an interpretable framework for understanding retinal coding. Application of the method to reconstructing natural movies demonstrates that fixational drift eye movements improve the fidelity of the retinal signal, even if the eye movements are unknown a priori and must inferred from the spike trains. Spike timing precision is found to be particularly important in the presence of eye movements, and stimulus-induced correlated firing between nearby cells is shown to contribute significantly to the content of the retinal code. Separately, this thesis develops a novel optimization-based technique to decompose the extracellularly-recorded spiking waveforms of retinal ganglion cells into distinct contributions from the somatic, dendritic, and axonal cellular compartments. This simple, biophysically-motivated representation effectively extracts physiological properties of retinal ganglion cells from their electrically-recorded waveforms, and correlates strongly with the morphology, receptive field location and structure, and functional cell type of retinal ganglion cells. This technique enables substantial advances in inferring the receptive field locations and the functional cell types of retinal ganglion cells from recorded spiking waveforms alone, addressing challenges in the calibration and operation of an epi-retinal prosthetic device.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2024; ©2024 |
| Publication date | 2024; 2024 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Wu, Eric Gene |
|---|---|
| Degree supervisor | Chichilnisky, E.J |
| Thesis advisor | Chichilnisky, E.J |
| Thesis advisor | Mitra, Subhasish |
| Thesis advisor | Wetzstein, Gordon |
| Degree committee member | Mitra, Subhasish |
| Degree committee member | Wetzstein, Gordon |
| Associated with | Stanford University, School of Engineering |
| Associated with | Stanford University, Department of Electrical Engineering |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Eric Gene Wu. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis Ph.D. Stanford University 2024. |
| Location | https://purl.stanford.edu/yr893dw9922 |
Access conditions
- Copyright
- © 2024 by Eric Gene Wu
- License
- This work is licensed under a Creative Commons Attribution 3.0 Unported license (CC BY).
Also listed in
Loading usage metrics...