Sang Min Han

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2024-26

May 1, 2024

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-26.pdf

We present a high-throughput, scalable, and biophysically informed effective connectivity mapping method for a population of neurons via two photon holography optogenetics. Specifically, we derive a simple neural circuit dynamics model from basic biophysical principles and introduce a fast algorithm that best estimates the connectivity among the neurons in a network given the observed neural population activity. The algorithm leverages the state-of-the-art two photon holography optogenetic technique and calcium trace imaging as a proxy for neuron membrane potential to optimally estimate the connectivity of neurons residing inside a three-dimensional volume with dimensions spanning hundreds of microns. Using 3D-SHOT, a two photon holography optogenetics technique capable of stimulating custom ensembles of neurons with cellular resolution and millisecond-order time precision, combined with GCaMP6, a contemporary genetically encoded calcium indicator, imaging, we observe the activity of both stimulated and neighboring neurons inside the neocortex of both awake and anesthetized mice in vivo. With these modern technologies at hand, we use the derived deterministic and linear autoregressive model of an arbitrary order with a feed-forward optogenetic stimuli input to describe the population-level time evolution of neural activities in a network. We utilize the ideas in causal inference, compressed sensing, and parallel computing to efficiently estimate the aforementioned model parameters, which directly translate to the connectivity matrices that characterize the effective interactions among the observed neurons. Furthermore, interference framework in experimental design and network control algorithm based on a graph-theoretic centrality measure are applied to provide a higher fidelity summary statistics of the connections among a subset of selected neurons and to artificially drive the network to specific brain states, respectively. With the estimated biophysical model describing the partial dynamics of neuronal interactions, inferences regarding both the spatial and temporal signatures of a local region of the brain can be made.

Advisors: Chunlei Liu

---

BibTeX citation:

@phdthesis{Han:EECS-2024-26,
    Author= {Han, Sang Min},
    Title= {Neural Circuit Dynamics Estimation and Control},
    School= {EECS Department, University of California, Berkeley},
    Year= {2024},
    Month= {May},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-26.html},
    Number= {UCB/EECS-2024-26},
    Abstract= {We present a high-throughput, scalable, and biophysically informed effective connectivity mapping method for a population of neurons via two photon holography optogenetics. Specifically, we derive a simple neural circuit dynamics model from basic biophysical principles and introduce a fast algorithm that best estimates the connectivity among the neurons in a network given the observed neural population activity. The algorithm leverages the state-of-the-art two photon holography optogenetic technique and calcium trace imaging as a proxy for neuron membrane potential to optimally estimate the connectivity of neurons residing inside a three-dimensional volume with dimensions spanning hundreds of microns. Using 3D-SHOT, a two photon holography optogenetics technique capable of stimulating custom ensembles of neurons with cellular resolution and millisecond-order time precision, combined with GCaMP6, a contemporary genetically encoded calcium indicator, imaging, we observe the activity of both stimulated and neighboring neurons inside the neocortex of both awake and anesthetized mice in vivo. With these modern technologies at hand, we use the derived deterministic and linear autoregressive model of an arbitrary order with a feed-forward optogenetic stimuli input to describe the population-level time evolution of neural activities in a network. We utilize the ideas in causal inference, compressed sensing, and parallel computing to efficiently estimate the aforementioned model parameters, which directly translate to the connectivity matrices that characterize the effective interactions among the observed neurons. Furthermore, interference framework in experimental design and network control algorithm based on a graph-theoretic centrality measure are applied to provide a higher fidelity summary statistics of the connections among a subset of selected neurons and to artificially drive the network to specific brain states, respectively. With the estimated biophysical model describing the partial dynamics of neuronal interactions, inferences regarding both the spatial and temporal signatures of a local region of the brain can be made.},
}

EndNote citation:

%0 Thesis
%A Han, Sang Min 
%T Neural Circuit Dynamics Estimation and Control
%I EECS Department, University of California, Berkeley
%D 2024
%8 May 1
%@ UCB/EECS-2024-26
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-26.html
%F Han:EECS-2024-26