CAREER: Control-Aware System Identification of Heterogenous Multiscale Brain Network Dynamics

该项目旨在建立人类大脑的数学模型，并为其在临床治疗中的应用奠定基础。仅在美国，每年用于治疗神经和精神疾病的费用就超过1万亿美元。大量关于神经科学数学建模的文献对临床治疗几乎没有影响，因为大多数模型要么依赖于简化假设，要么使用模糊了与潜在生物学联系的机器学习方法。这项研究将为大脑开发一种新的数学模型，这种模型既具有生物学意义，又可解释，而不依赖于简化的假设。这些模型将植根于结合数据驱动和非线性动力系统方法的工程方法。该研究与针对UCR，当地内陆帝国社区以及全球高中生，本科生和研究生的多样化和坚实的教育活动紧密结合在一起。该项目将开发数据驱动的人脑模型，严格结合生物神经网络的三个关键但经常被忽视的方面：(i)跨越多个尺度，（ii）异质性，（iii）对神经调节的反应。第一个推力将通过数据驱动的神经动力学在不同空间尺度（神经元、神经种群和大脑区域）之间的前馈和反馈相互作用的建模来实现，这超越了微观动力学的简单宏观读数，增强了我们对宏观动力学如何从更小尺度产生并反馈到更小尺度的理解。第二个重点将发展结构异质性的大脑模型，该模型将大脑异质性的非线性、维度和皮层和皮层下区域的神经编码数据结合起来。第三个重点将产生一个数据驱动的、样本高效的框架，用于在全脑网络水平上对深部脑刺激的影响进行建模，从而超越基于电磁扩散第一原理建模的局部预测。预期的结果是潜在的变革性模型和建模技术，为神经科学界提供可靠的和临床可翻译的神经调节设计工具，同时也显著增加了我们对人脑多尺度、异构和输入驱动动力学的理解。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Faculty Early Career Development (CAREER) project aims to develop mathematical models of the human brain, with an eye on laying the foundation for their implementation in clinical treatments. Treatments of neurological and psychiatric disorders impose over $1 trillion annually on the US population alone. A large body of literature on mathematical modeling in neuroscience had little impact on clinical treatments because most models either rely on simplifying assumptions or use machine learning methods that obscure the link to the underlying biology. This research will develop a new category of mathematical models for the brain that are at once biologically meaningful and interpretable, without relying on simplifying assumptions. These models will be rooted in engineering approaches that combine data-driven and nonlinear dynamical systems methods. The research is tightly integrated with a diverse and solid body of educational activities targeted towards high school, undergraduate, and graduate students at UCR, the local Inland Empire community, and across the globe.This project will develop data-driven models of the human brain that rigorously incorporate three critical but often ignored aspects of biological neural networks: (i) spanning across multiple scales, (ii) heterogeneity, and (iii) response to neuromodulation. The first thrust will be achieved through data-driven modeling of feedforward and feedback interactions between neural dynamics at different spatial scales (neurons, neural populations, and brain regions), that move beyond simple macroscopic readouts of microscopic dynamics and enhance our understanding of how macroscopic dynamics emerge from, and feed back into the smaller scales. The second thrust will develop structurally heterogeneous brain models that incorporate data of brain heterogeneities in nonlinearity, dimensionality, and neural code across cortical and subcortical regions. The third thrust will generate a data-driven, sample-efficient framework for modeling the effects of deep brain stimulation at the level of the whole-brain network, thus moving beyond local predictions based on first principle modeling of electromagnetic diffusion. The expected outcomes are potentially transformative models and modeling techniques that provide the neuroscience community with solid and clinically translatable tools for the design of neuromodulation, while also significantly increasing our understanding of the multiscale, heterogeneous, and input-driven dynamics of the human brain.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.