Discovering brain state dependent dynamics in large scale perceptual ensembles.

点击翻译按钮获取中文摘要

Project Summary Active perception, the ability to seek out behaviorally relevant information, is guided by both cognitive and motor behaviors and is influenced by fluctuations in endogenous brain state. It is a result of the concerted activity of ensembles of neurons in the sensory hierarchy. These ensembles interact flexibly and dynamically as the organism transitions between various behavioral and brain states. However, the state-dependent information processing principles that underlie the activity of such ensembles are largely unknown. A rich body of theoretical and recent experimental work has shown that dependencies, such as co-variability, within an ensemble strongly influence their information carrying capacity and hence their functional efficacy. Further, theoretical investigation of these dependencies has revealed that their influence is determined by the extent of their alignment with the information coding dimension of an ensemble. The source of dependencies – shared vs. local – has been identified as a key determinant of this alignment. A critical step toward determining this source is to characterize the joint spiking activity (beyond pairwise correlations) of populations in these ensembles. This is, however, a challenging task owing to the varied non-linear nature of neuronal interactions, and the fact that neural populations are sparsely sampled by current recording techniques. Tackling this problem requires a highly interdisciplinary approach spanning advanced techniques in systems and computational neuroscience. Based on our preliminary data and prior studies, our broad hypothesis is that the computations of active perception are cortical layer-specific and that they are mediated by ensembles of neural sub-populations defined by their layer identity and cell-class. We propose to answer several key questions regarding this hypothesis: how does active perception modulate information flow in laminar circuits, both during attention (Aim 1A) and saccadic eye movement (Aim 1B)? How are the laminar circuits of active perception modulated by internal brain state fluctuations such as those during cued attention (Aim 2A) and spontaneous vision (Aim 2B)? We will achieve these aims using laminar high-density recordings in the visual cortex of non-human primates, while animals are engaged in either task-based or spontaneous active perception. Using a novel combination of dynamic Bayesian networks among (DBN) and partial information decomposition (PID) we will infer distinct categories of dependencies cortical network components. Our proposal will achieve the first systematic characterization of modulation of information flow (beyond pairwise correlations) by active perception processes in the context of laminar cortical circuits. Our proposal is the first study to investigate how internal brain state fluctuations shape the ensemble level causal motifs of active perception. The results of these investigations will significantly advance our understanding of information flow structure in a canonical cortical circuit and provide a broad framework for investigating such structures in brain-wide circuits.