STATISTICAL METHODS TO MODEL SPATIOTEMPORAL BRAIN FUNCTIONAL PATHWAYS IN DEMENTIA

描述（由申请人提供）：该候选人是华盛顿大学的助理教授，她的研究重点是开发用于处理缺失数据和高维数据的新型统计方法，重点是开发痴呆症，神经科学和遗传学的预测工具。她的长期职业目标是成为一名独立的转化生物统计学家，在痴呆症研究中分析功能磁共振成像（fMRI）数据的统计方法方面拥有额外的专业知识。这项资助的总体目标是拓宽候选人在痴呆症和功能磁共振成像统计技术方面的知识，运用她的高维数据、多元数据和推理专业知识，进一步发展统计方法，解释功能磁共振成像数据的时空特性。虽然功能连接磁共振成像（fcMRI）目前是一种研究工具，其在痴呆症中的临床应用尚未建立，但功能连接可能为进一步了解阿尔茨海默病（AD）中大脑网络异常的发展提供见解。虽然在时间属性和空间属性建模方面取得了一些进展，但这些方法通常没有被使用，并且对功能连通性的群体差异的重视程度不够。该基金的第一个研究目标是开发功能连通性的推理统计方法，该方法结合fcMRI数据的空间和时间特性，并可以估计群体效应。这些方法包括边缘模型广义估计方程和边缘过渡模型，以及广义潜变量建模和向量自回归模型，以进一步解释相关区域数据。这项研究将确定阿尔茨海默病与健康人群之间的区域间联系和大脑网络的差异。我们还将从第一个目标进一步发展统计方法来分析fcMRI数据，通过发展协方差结构和应用单指数方法来模拟纵向数据的额外可变性。这些纵向研究的统计技术将用于改进对阿尔茨海默病进展的评估。本研究将通过应用新颖的统计方法确定fcMRI作为AD临床工具的实际效用。

DESCRIPTION (provided by applicant): The candidate is an Assistant Professor at Washington University with her research focus on the development of novel statistical methods for handling missing data and high-dimensional data, with an emphasis on developing prediction tools in dementia, neuroscience, and genetics. Her long-term career goal is to become an independent translational biostatistician with additional expertise in statistical methods for analysis of functional magnetic resonance imaging (fMRI) data in dementia studies. An overall objective of this grant is to broaden the candidate's knowledge of statistical techniques for dementia and fMRI, applying her high- dimensional data, multivariate data, and inferential expertise to further develop statistical methods that account for spatial and temporal properties of fMRI data. While functional connectivity magnetic resonance imaging (fcMRI) is currently a research tool and its clinical utility in dementia is yet to be established, functional connectivity may provide insight to further understand how abnormalities in brain networks develop in Alzheimer's disease (AD). Although some progress has been made with modeling temporal properties and spatial properties, these methods are not conventionally utilized and insufficient emphasis has been placed on group differences in functional connectivity. The first research goal of this grant is to develop inferential statistical methods of functional connectivity that incorporate spatial and temporal properties of fcMRI data and can estimate group effects. These methods include marginal model generalized estimating equations and marginalized transition models, and generalized latent variable modeling and vector autoregressive models to further account for correlated regional data. This research will determine how the inter-regional associations and brain networks differ between AD and healthy cohorts. We will also further develop statistical methods from the first goal for the analysis of fcMRI data that model the extra variability of longitudinal data by developing covariance structures and applying single-index methods. These statistical techniques for longitudinal studies will be intended to improve the assessment of the progression of AD. This research will determine the practical utility of fcMRI as a clinical tool for AD through the application of novel statistical methods. PUBLIC HEALTH RELEVANCE: This research is linked to studies in the area of AD and fMRI that will be of great value to evaluate brain networks that will aid in: 1) evaluating and detecting early-stage AD; and, 2) assessing the progression of AD. fcMRI also has the potential to investigate the pathobiology of AD through the analysis of brain networks and their changes over time. The candidate will provide recommendations for optimal statistical techniques in these areas.