BME PhD Proposal Presentation
Time: 10:30-11:30 AM
Location / Meeting Link: CODA C1215 - https://gatech.zoom.us/j/97180260397?pwd=YkhDRHBTRnVXU3JxWVg4VTcwOHBnZz09
Christopher J. Rozell, PhD (advisor); Nabil Imam, PhD; Chethan Pandarinath, PhD; Garrett B. Stanley, PhD; Patricio A. Vela, PhD
Title: Towards Optogenetic Feedback Control of Neural Population Dynamics
As the importance of causal inference becomes increasingly apparent in neuroscience, so does the need for technology enabling precise manipulation of neural activity. Feedback control is a prime example—widely used throughout the engineering disciplines, it has had a significant impact via techniques such as the voltage clamp and the dynamic clamp on cellular neuroscience. However, it has yet to be widely applied at the mesoscale of circuits and networks despite recent improvements in interfacing technology, such as optogenetics. Challenges to such applications include the complexity of implementing fast closed-loop experiments, the need to adapt the mature methods of control theory to the idiosyncratic constraints of systems neuroscience experiments, and the lack of established technical guidelines for applying feedback control to address complex scientific questions. In this work I propose to begin to address these challenges in three aims. In Aim 1, I develop a simulation framework for easily prototyping closed-loop optogenetic control (CLOC) experiments in silico, thus enabling CLOC experiment design and method development without the costs of in-vivo experiments or up-front investments in hardware-software systems. In Aim 2, I will translate sophisticated model-based feedback control algorithms to the experimental setting of multi-input CLOC—the simultaneous use of multiple light sources and/or opsins—and test the virtues of these algorithms compared to the limited, simpler ones previously used. Finally, in Aim 3, I will explore how control quality varies with experimental parameters in a promising future application of CLOC—controlling the latent dynamics of neural population activity—and test the hypothesis that per-neuron actuation will not be needed. I will do this in silico with recurrent spiking neural networks trained using biologically plausible methods and differing degrees of brain-like structure. This work will thus equip the systems neuroscience community to more fully take advantage of CLOC with an accessible testing and development environment, multi-input actuation, and a point of reference for designing experiments capable of answering complex scientific questions.