Events Calendar

Title: “Machine Learning Framework for Analyzing Shifting Neuronal Network Dynamics in Calcium Recordings”

Advisory Committee: John A. White, PhD – BME (Co-Advisor) Mark Kramer, PhD – Mathematics & Statistics (Co-Advisor) Kamal Sen, PhD – BME (Chair) Laura Lewis, PhD – BME

Abstract: Calcium imaging allows recording from 100s of neurons in a single wide field of view. The responses of a large number of cells, along with possible fluorescent cell-type tagging and varying experimental contexts, give rise to extremely high dimensionality data sets. Wide-spread analyses employ traditional statistics that aim to summarize a large volume of stochastic individual neural responses into single quantitative metrics, discounting the temporal dynamics of individual and coupled responses. In contrast, dimensionality reduction models provide an avenue to conduct nuanced analysis of a data set, while also considering all the dimensions. We aim to adapt a series of easily interpretable dimensionality reduction methods to analyze shifts in increasingly complex neuronal network dynamics that arise as a function of different experimental contexts. We then plan to harness the lower dimensional representation provided by these models to build Deep Neural Networks to predict the experimental context of the recordings being analyzed. After developing this framework, we will apply it to study the neuronal network dynamics of two novel contexts: 1) the primary somatosensory cortex (S1) under increasing concentrations of anesthesia, and 2) the hippocampus during optogenetic stimulation of memory encoding ensembles of neurons. We have successfully adapted and begun characterizing a series of dimensionality reduction methods and have found them to extract underlying structure from calcium recordings. We also find that internal structure collapses and re-expands as a function of entering and existing anesthesia.