CRCNS: Bayesian Modeling of Interacting Time Series to Discover Cortical Networks Associated with Auditory Processing Dysfunction

尽管接受了正常的听力学评估，一些听者仍然向临床医生抱怨他们很难听到，特别是在嘈杂或拥挤的环境中。然而，对大脑如何处理声音（以及它如何出错）的系统调查还很缺乏。这个资助项目的目标是应用新颖的统计方法来研究大脑在复杂情况下处理声音时的活动模式，比如在一个多说话的环境中听声音。在几种不同类型的听觉刺激下，将收集各种各样的行为和电生理反应。脑干反应的行为数据和生理测量将用于表征个体在单耳和双耳通路中的听力健康，为他们在类似听觉任务中的皮质磁和脑电图（M-EEG）反应提供补充信息。听觉注意网络连接也将被分析，以解释听觉功能障碍方面的神经基础，如无法在说话者之间保持或切换注意力。使用计算驱动的统计方法，将学习基于高维时间序列的灵活图形模型表示，以便基于收集的M-EEG数据表征听觉注意网络。具体来说，将解决两个计算目标：1)构建贝叶斯模型来表征不同空间分辨率下的动态皮层相互作用；2)建立模型来推断不同标准皮层节奏带的连接结构。该研究项目利用了两位研究者的互补专业知识，将听觉行为和系统神经科学结合起来，采用灵活和可扩展的统计时间序列建模方法。在大数据分析和系统神经科学中，时间结构经常被忽略，而这项资助的研究将通过使用高维时间连续神经数据集直接解决这一缺点。通过这种方法发现的皮层网络将使神经科学家更好地理解听觉注意网络在不同任务类型和听力能力个体差异中的内在变异性。

Despite receiving normal audiological assessments, some listeners still complain to clinicians that they struggle to hear, particularly in noisy or crowded environments. However, systematic investigations into how the brain processes sound (and how it can go wrong) are lacking. The goal of this funded project is to apply novel statistical approaches to study patterns of activity in the brain while it processes sound in complex situations like when listening in a multi-talker environment.A wide variety of behavioral and electrophysiological responses will be collected under several different types of auditory stimulation. Behavioral data and physiological measures of brainstem response will be used to characterize individuals' hearing health in both monaural and binaural pathways, providing complementary information to their cortical magneto- and electroencephalography (M-EEG) responses during similar auditory tasks. Auditory attentional network connectivity will also be analyzed, to account for the neural underpinnings of aspects of auditory dysfunction such as the inability to maintain or switch attention between speakers. Using computationally-driven statistical approaches, flexible graphical model-based representations of high-dimensional time series will be learned, in order to characterize the auditory attentional network based on collected M-EEG data. Specifically, two computational aims will be tackled: 1) Construct Bayesian models to characterize dynamical cortical interactions at different spatial resolutions and 2) Develop models that infer connectivity structure at different canonical cortical rhythmic bands. This research program leverages the complementary expertise of the two investigators, bringing together auditory behavioral and systems neuroscience, with flexible and scalable statistical time series modeling approaches. Temporal structure is often ignored in big data analyses as well as in systems neuroscience, and this funded research will directly address this shortcoming by using a high-dimensional set of temporally continuous neural data. The cortical network discovered by this approach will enable neuroscientists to better understand the variability inherent in the auditory attentional network across both task types and individual differences in listening abilities.