NCS-FO: Extracting Functional Cortical Network Dynamics at High Spatiotemporal Resolution

该项目由理解神经和认知系统的综合策略（NSF-NCS）资助，这是一个由计算机与信息科学与工程（CISE）、教育与人力资源（EHR）、工程（ENG）和社会、行为和经济科学（SBE）联合支持的多学科项目。神经科学家已经非常成功地了解了许多大脑区域的功能，通过对它们进行孤立的研究，并描述了它们在行为中的个别作用。然而，近年来越来越多的证据表明，复杂的大脑功能来自于表现出网络活动的多个大脑区域的共同激活。这些网络迅速组织起来，使大脑能够适应环境的变化，从而产生强健的行为。因此，破译这些网络动态背后的神经机制对于理解大脑如何进行注意力、决策和学习等认知过程至关重要。近年来，无创神经成像技术的进步在很大程度上解决了研究人类动态神经网络的实验挑战，并在无数的临床和实验条件下提供了丰富的神经数据。然而，这些数据的高维性以及这些网络的复杂性在数据分析、建模和统计推断方面造成了各种瓶颈。为了利用丰富的非侵入性神经数据提供的独特机会，该项目(1)开发一种统一的方法来推断这些皮质网络的动态和统计特征，以一种计算效率高的方式；(2)将这种方法应用于行为人类受试者的脑磁图（MEG）数据，以解决有关听觉处理的几个基本问题。这项工作在前所未有的时空分辨率下为大脑网络的动态组织带来了新的见解，从而可以影响脑机接口和神经形态工程领域的技术。它还允许创建工程解决方案，用于早期检测和监测涉及听觉感知和注意力的认知障碍。该项目的成果将以可公开访问的数据分析工具箱的形式传播给更广泛的科学界，并附有教程和网络研讨会。该研究计划由K-12、本科和研究生各级的教育活动补充，包括讲习班、本科项目和课程开发，重点是妇女和代表性不足的少数民族的参与。现有的提取皮质功能网络动态的范式面临着挑战，包括由于滑动窗口处理导致的时间分辨率损失，由于非侵入性记录的限制导致的空间分辨率损失，以及由于网络本质上是非线性的，大量使用线性估计技术导致的统计偏差。该项目通过使用现代信号处理技术，以统计稳健的方式将高时间分辨率非侵入性记录与高空间分辨率相结合，为解决这些挑战提供了统一的研究计划。这种方法将特别应用于从行为的人类受试者获得的MEG数据，并将用于破译适应性听觉处理的神经机制。

This project is funded by Integrative Strategies for Understanding Neural and Cognitive Systems (NSF-NCS), a multidisciplinary program jointly supported by the Directorates for Computer and Information Science and Engineering (CISE), Education and Human Resources (EHR), Engineering (ENG), and Social, Behavioral, and Economic Sciences (SBE). Neuroscientists have been remarkably successful in understanding the function of numerous brain regions by studying them in isolation and characterizing their individual roles in behavior. Growing evidence in recent years, however, suggests that sophisticated brain function emerges from the co-activation of multiple brain regions that exhibit networked activity. These networks organize rapidly in order to allow the brain to adapt to changes in the environment, resulting in robust behavior. Deciphering the neural mechanisms underlying these network dynamics is therefore crucial in understanding how the brain carries out cognitive processes such as attention, decision-making and learning. Recent technological advances in noninvasive neuroimaging have largely addressed the experimental challenges in studying these dynamic networks in humans and have provided abundant neural data under countless clinical and experimental conditions. However, the sheer high-dimensionality of these data together with the complexity of these networks has created various bottlenecks in data analysis, modeling, and statistical inference. In order to exploit the unique window of opportunity provided by the abundance of noninvasive neural data, this project is (1) developing a unified methodology for inferring the dynamics and statistical characteristics of these cortical networks, in a computationally efficient fashion, and (2) applying this methodology to magnetoencephalography (MEG) data from behaving human subjects to address several fundamental questions about auditory processing. This work brings new insight as to the dynamic organization of brain networks at unprecedented spatiotemporal resolutions, and can thereby affect technology in the areas of brain-computer interfacing and neuromorphic engineering. It also allows for the creation of engineering solutions for early detection and monitoring of cognitive disorders involving auditory perception and attention. The outcome of this project will be disseminated to the broader scientific community in the form of publicly accessible data analysis toolboxes accompanied with tutorials and webinars. The research plan is complemented by educational activities at the K-12, undergraduate, and graduate levels, including workshops, undergraduate projects, and course development, with an emphasis on the involvement of women and underrepresented minorities.The existing paradigm for extracting cortical functional network dynamics faces challenges, including loss of temporal resolution due to the common sliding window processing, loss of spatial resolution due to the constraints of noninvasive recording, and statistical bias due to the heavy usage of linear estimation techniques given that network properties are intrinsically non-linear. This project provides a unified research plan for addressing these challenges, by combining high temporal resolution non-invasive recordings with high spatial resolution in a statistically robust way, using modern signal processing techniques. This methodology will specifically be applied to MEG data acquired from behaving human subjects, and will be used to decipher the neural mechanisms of adaptive auditory processing.