Novel Bayesian linear dynamical systems-based methods for discovering human brain circuit dynamics in health and disease

项目摘要/摘要 对人脑如何产生认知的理解最终取决于精确的量化 基于上下文的分布式动态功能网络(DFN)的特征 大脑区域。由于缺乏理论框架，在实现这一目标方面取得的进展有限 描述DFN的特征和测试它们的适当计算方法。设计和验证 因此，研究人脑中DFN的计算方法具有重要意义。第一大调 这项提议的目标是通过开发新的算法来解决人脑研究中的一个关键需求 识别DFN并描述分布的大脑区域之间的动态网络交互。至 为了实现这一目标，我们将在贝叶斯框架内开发和验证新的计算方法 具有矢量自回归模型(VAR)和因子的切换线性动力系统(BSD) 分析(FA)，克服了现有方法的主要局限性，用于研究 人脑。该提案的第二个主要目标是使用BSD来调查潜在的DFN 健康成人和帕金森氏病患者的认知功能。严重认知 功能障碍是帕金森病患者最具破坏性的行为后果之一，但人们对此知之甚少 在这种衰弱障碍中，神经认知系统功能障碍的时间特性。计算性的 我们计划开发、验证和应用的算法将使我们能够严格研究大脑动力学 支持关键的认知功能，并显著提高我们对动态过程的理解 潜在的人脑功能和功能障碍。我们提议的研究还将首次调查 在模拟的、活体的啮齿类动物的光遗传功能磁共振成像中的DFN，以及使用最先进的人类数据(亚 第二)由美国国立卫生研究院资助的斯坦福阿尔茨海默病产生的高时间分辨率fMRI数据 研究中心(ADRC)，重点介绍我们建议的方法的关键翻译应用。 我们提出的研究将为研究人类动态功能网络提供新的工具 大脑，及其在人类连接组计划(HCP)和神经学研究中的创新应用 更广泛的障碍和临床神经科学。拟议的研究与以下任务高度相关 大脑倡议(RFA-EB-15-006)，该倡议呼吁开发和传播创新 用于探测人脑功能和功能障碍的计算工具。我们的计算工具将被广泛使用 传播以促进对人脑功能动力学方面的研究。

Project Summary/Abstract Understanding how the human brain produces cognition ultimately depends on precise quantitative characterization of context-dependent dynamic functional networks (DFN) that transiently link distributed brain regions. Progress in achieving this goal has been limited due to a lack of theoretical frameworks for characterizing DFNs and appropriate computational methods to test them. Devising and validating computational methods for investigating DFNs in the human brain is thus of great significance. The first major goal of this proposal is to address a critical need in human brain research by developing novel algorithms for identifying DFNs and characterizing dynamic network interactions between distributed brain regions. To achieve this goal, we will develop and validate novel computational methods within the framework of Bayesian switching linear dynamical systems (BSDS) with vector autoregressive models (VAR) and factor analysis (FA) that overcome major limitations of existing methods for investigating dynamic interactions in the human brain. The second major goal of this proposal is to use BSDS to investigate DFNs underlying cognitive function in healthy adults, and in patients with Parkinson's disease (PD). Severe cognitive impairment is one of the most devastating behavioral outcomes in patients with PD, yet little is known about the temporal properties of dysfunctional neurocognitive systems in this debilitating disorder. The computational algorithms we propose to develop, validate, and apply will allow us to rigorously investigate brain dynamics that support critical cognitive functions and significantly advance our understanding of dynamic processes underlying human brain function and dysfunction. Our proposed studies will also, for the first time, investigate DFNs in simulated, rodent in vivo optogenetic fMRI, as well as human data using state-of-the-art (sub- second) high-temporal resolution fMRI data generated by the NIH-funded Stanford Alzheimer's Disease Research Center (ADRC), highlighting critical translational applications of our proposed methods. Our proposed studies will provide novel tools for investigating dynamic functional networks in the human brain, with innovative applications to the Human Connectome Project (HCP) and the study of neurological disorders and clinical neuroscience more broadly. The proposed studies are highly relevant to the mission of the BRAIN Initiative (RFA-EB-15-006), which calls for the development and dissemination of innovative computational tools for probing human brain function and dysfunction. Our computational tools will be widely disseminated to facilitate research into the dynamical aspects of human brain function.