Two-Photon Imaging of Brain Regions in Fissures and Learning Manifolds from Neural Dynamics

Abstract

Progress in systems neuroscience requires effective tools and techniques for probing neural circuits, and for analyzing the resulting data in ways that drive theoretical insight. This thesis consists of three parts, aimed broadly toward furthering the measurement and analysis of neural circuits. In the first part, we present methods for two-photon imaging of brain regions situated in deep fissures, enabling the use of cellular resolution optical tools for probing areas such as the medial prefrontal cortex (mPFC) and medial entorhinal cortex (MEC). We demonstrate recordings of population activity in the mPFC and grid cells in the MEC in behaving mice. In the second part, we present an optical approach for measuring dopaminergic input to the mPFC with high spatiotemporal resolution, which has not been feasible using traditional methods. We demonstrate recordings of mPFC dopamine signals in behaving mice, and present preliminary evidence for fine-scale heterogeneity across individual dopaminergic axons. In the third part, we present a new unsupervised learning algorithm for inferring underlying, nonlinear structure in neuronal population activity. We use this algorithm to characterize the geometric properties of hippocampal activity and their relationship to behavior. And, we propose a conceptual model explaining how neural coding and trial-to-trial variability both arise from movement along a low dimensional, nonlinear activity manifold, driven by internal cognitive processes.