Image-based modeling of functional connectivity in neural networks at single-cell resolution

点击翻译按钮获取中文摘要

Project Summary/Abstract Calcium fluorescence imaging has opened unprecedented opportunities to investigate how neurons are wired in circuits that plastically process information in the brain. Recent advances in microscopy and genetically encoded calcium indicators allow us to record in real time the transient rises of intracellular Ca2+ for a large population of neurons during their electrical activity. However, little is known about mechanisms of information processing in neural circuits at the single neuron level. Even though cutting-edge technologies are capable of optically probing thousands of neurons firing in relation to stimulation or behavior output, we are still unable to track the propagation of the neuron firing events. The key barrier to progress is the lack of computational technologies in image and signal processing for the calcium imaging data. A common but unresolved obstacle to collect calcium activities of neurons from acquired images is deformation of live tissues during imaging. The goal of the project for image processing is to develop an algorithm to automatically extract accurate traces of single-neuron activity from deforming 3D calcium images. A new approach under development generates a dynamic region-of-interest for each jittering and blinking neuron by iteratively learning neuronal identities from local images of firing neurons. As a next step, the goal for signal processing is to develop statistical inference frameworks that can assess the evidence of information flows from external stimuli to sensory neurons, and between interconnected neurons. The responsiveness of neurons upon stimulation will be statistically determined based on an autoregressive hidden Markov model. We will identify causal hierarchy among neuronal activities using Granger-causality inference, in order to reconstruct the functional connectivity networks for large-scale neuronal populations. Subsequent graph theoretical quantification of the connectivity networks at the single-neuron level will enable us to differentiate wiring architectures of neural circuits under different molecular conditions. The long-term career goal of the candidate, Dr. Noh, is to establish an independent research program specialized in image-based stochastic modeling of dynamic nervous systems by translating his expertise in statistics and time series analysis. The training objective of this proposal is to allow Dr. Noh to make a unique contribution to computational methods for complex neuroimaging data and its dynamics, and to train Dr. Noh to gain the ability to conduct hypothesis-driven research for neuroscience by himself. The proposed training is guided by Gaudenz Danuser and Julian Meeks, who are leaders in the fields of computational cell biology and neurobiology, respectively. Being engaged in diverse environment of informatics/experiments and neurobiology, Dr. Noh will immerse himself into neuroscience, acquire experiential learning of neuroimaging experiments, and gain expertise in multidisciplinary team science. The completion of this proposal will enable Dr. Noh not only to establish his groundwork for research in neuroimaging, but also to play leading roles in multidisciplinary research.