Identification of time-varying multivariate physiological systems and applications to cerebrovascular regulatory mechanisms and dynamic brain functional connectivity from multimodal measurements

提出的研究旨在开发最先进的方法来分析时变多元系统，这特别适合于生理系统和功能神经成像。虽然设想的方法贡献具有普遍适用性，但我们将把它们应用于模拟系统和区域脑血流动力学以及时变脑功能连接。具体目标是：(i)开发一种方法框架，用于对可能表现出非线性行为的时变多元系统进行建模。在这个目标中，我们将研究递归和线性参数变化估计的结合来提高性能，以及遗传算法/正则化方法来选择模型顺序，特别是在非线性系统的情况下。（ii）通过收集和分析多模态神经成像数据（同时收集EEG-fMRI和TCD-fNIRS数据），获得改进的区域fMRI HRF定量描述。在这一目标中，我们还将生成真实的生物物理模拟数据，将神经激活获得的经验hrf与不同的脑电图信号特征联系起来，旨在更好地理解潜在的神经血管耦合机制。（iii）研究血管迷走神经性晕厥患者的时变血流动力学，设计预测晕厥发作的算法，从长远来看，这些算法将应用于可穿戴设备收集的数据。（iv）从多模态神经成像测量中量化时变的功能性脑连接。在这个目标中，我们的目标是通过同时使用EEG-fMRI数据来量化神经和生理因素在时变静息状态fMRI功能连接出现中的贡献。我们还致力于利用时变连通性作为无创脑刺激（如经颅直流电刺激）的标志，利用EEG/MEG数据促进运动学习，以及利用长时间EEG数据设计更准确的癫痫发作检测/预测算法。这项提议的研究为使用可穿戴设备的实时监测应用带来了巨大的希望，这有可能改变现代医学。它还有望获得基于时变功能连通性的更敏感的生物标志物，以及使用后者在闭环环境（生物反馈）中识别非侵入性方案的治疗靶点。

The proposed research aims to develop state-of-the art methods for the analysis of time-varying multivariate systems, which are particularly well-suited for physiological systems and functional neuroimaging. While the envisioned methodological contributions are of general applicability, we will apply them to modeling systemic and regional cerebral hemodynamics as well as time-varying brain functional connectivity. The specific objectives are to: (i) develop a methodological framework for modeling time-varying multivariate systems that may exhibit nonlinear behavior. Within this objective, we will investigate the combination of recursive and linear parameter-varying estimation to improve performance, as well as genetic algorithm/regularization methods to select model order, particularly in the case of nonlinear systems. (ii) obtain improved quantitative descriptions of the regional fMRI HRF, by collecting and analyzing multimodal neuroimaging data (simultaneous EEG-fMRI and TCD-fNIRS data). Within this objective, we will also generate realistic biophysical simulation data to link the empirical HRFs obtained by neural activation and different EEG signal features, aiming to obtain better understanding of the underlying neurovascular coupling mechanisms (iii) investigate time-varying hemodynamics in vasovagal syncope patients and design algorithms to predict the occurrence of fainting episodes that, in the longer term, will be applied to data collected from wearable devices. (iv) quantify time-varying functional brain connectivity from multimodal neuroimaging measurements. Within this objective, we aim to quantify the contribution of neural vs. physiological factors in the emergence of time-varying resting-state fMRI functional connectivity by using simultaneous EEG-fMRI data. We also aim to pursue the use of time-varying connectivity as a marker for noninvasive brain stimulation (e.g. transcranial direct current stimulation) to promote motor learning using EEG/MEG data, as well as to design more accurate detection/prediction algorithms for epileptic seizures using long-duration EEG data. The proposed research yields great promise for real-time monitoring applications using (among others) wearable devices, which have the potential to transform modern medicine. It also yields promise for obtaining more sensitive biomarkers based on time-varying functional connectivity, as well as using the latter for identifying therapeutic targets for noninvasive protocols in a closed-loop context (biofeedback).