Abstract

A freely moving animal must detect independently moving objects, such as predators or prey, against a backdrop of self-generated retinal image motion (i.e., optic flow) and must do so reliably in complex visual environments. This thesis investigates how the mouse retina resolves this challenge, focusing on the spatial and temporal circuit mechanisms that enable robust object motion detection during active vision. We reveal how an evolutionarily conserved inhibitory interneuron, the TH2 amacrine cell, constructs a geometrically structured inhibitory surround for downstream object-motion-sensitive (OMS) ganglion cells, output neurons of the eye. Using electron microscopy, calcium imaging, and computational modeling, we show that the architecture of individual TH2 cells gives rise to a daisy-shaped population-level filter. This filter selectively cancels coherent global motion while remaining insensitive to spatially uncorrelated background motion, resolving a fundamental trade-off between self-motion cancellation and sensitivity to independently moving objects. We then dissect the temporal mechanisms by which the same circuit segments object motion from self-generated optic flow. This computation arises from precise interactions of excitatory and inhibitory inputs differentially tuned to speed and coincidence. The inhibitory TH2 pathway acts as a robust global-motion silencer with specific speed preferences, while the excitatory drive is differentially tuned to coincidence. At the output, W3/UHD ganglion cells integrate these inputs, with center excitation and surround inhibition canceling only when object and self-generated motion overlap in speed and time, ensuring robust object motion detection both during sustained optic flow and within the brief stationary pauses that punctuate natural locomotion. Together, these findings reveal how the retina detects independently moving objects against a backdrop of self-generated motion during active vision. We discover motion-coherence selectivity implemented by a population-level geometric filter of TH2 amacrine cells and temporal segmentation of object motion by precise interactions between excitatory and inhibitory inputs to OMS ganglion cells. Together, these spatial and temporal mechanisms ensure that freely moving animals can reliably detect objects (incl. predators and prey) across a broad range of object densities during active vision.

Committee Chair

Daniel Kerschensteiner

Committee Members

Josh Morgan, Baranidharan Raman; Martha Bagnall; Philip Williams; Timothy Holy

Degree

Doctor of Philosophy (PhD)

Author's Department

Biomedical Engineering

Author's School

McKelvey School of Engineering

Document Type

Dissertation

Date of Award

8-1-2026

Language

English (en)

Available for download on Sunday, July 30, 2028

Included in

Ophthalmology Commons

Share

COinS