Multi-dimensional network framework for AD detection and progression

项目总结 阿尔茨海默病(AD)是最常见的痴呆症，对患者、家庭有重大影响 和公共卫生系统。据估计，2018年美国有570万人患有阿尔茨海默氏症1。 在该病临床表现时，已经存在重大的不可逆转的脑损伤， 使在疾病早期阶段诊断AD成为潜在治疗的紧急先决条件 延迟或预防症状2。据估计，早期准确的AD检测可以节省高达7.9万亿美元 医疗和护理费用1。此外，早期发现和进展阿尔茨海默病对于监测 试验性治疗以及通知开发有效的治疗方法。这是一种迫切的临床需要 改善AD的早期检测和进展。随着阿尔茨海默病神经生物学的最新进展，我们对 已经从一个基于临床症状的疾病转变为一个多因素的生物学结构 而且是异质性的，这不能用任何单一的可用的生物标志物来解释。美国国立卫生研究院投入了数十亿美元 在过去的几十年里，为几个中心和针对大量老年人的数据倡议提供资金； 产生了丰富的多模式神经成像、认知、临床、生物谱学和遗传数据。然而， 在实施跨医疗模式聚合数据的创新综合方法方面所做的努力较少 捕捉AD的异质性。为了填补多模式广告数据分析范式的空白， 拟议研究的总体目标是测试和验证多维网络框架 在单个模型中聚合多个医疗设备的数据，以捕捉AD的异质性，并进一步 增强AD检测和进展。我们的核心假设--由先前的证据和初步的 数据-建议的框架将通过提高以下能力来增强AD检测和进展 检测跨多种数据类型的公共信号和互补信号，并通过减少 每种模式在尺度、收集偏差和噪声方面的差异。我们将整合行为、临床、MR 来自ADNI和斯坦福ADRC数据的成像、Aβ和TAU标记物以及神经变性标记物的测试 并对所提出的集成不同数据类型的多维网络框架进行了早期验证 AD(Aim1)的检测以及应用多维网络表征AD进展 不同测量中纵向变化的框架(目标2)。据我们所知，这是第一次研究 在多维网络模型中集成各种AD数据以表征AD以进一步增强AD 检测和进展。如果被证明成功，这项高风险、高回报的提议将对 AD的特征、早期发现和进展具有重大的健康和经济影响。此外， 这项研究的成功完成将为多模式数据的综合分析提供关键工具，并将 帮助转变当前可用的数据集和临床试验的分析范式，这些分析范式主要侧重于 独立分析单一数据类型。

PROJECT SUMMARY Alzheimer's disease (AD) is the most common form of dementia with significant impact on patients, families and the public health system. An estimated 5.7 million Americans have Alzheimer's in 2018 1. At the time of clinical manifestation of the disease, significant irreversible brain damage is already present, rendering the diagnosis of AD at early stages of the disease an urgent prerequisite for potential therapies to delay or prevent symptoms2. It is estimated that early and accurate AD detection could save up to $7.9 trillion in medical and care costs1. Further, early AD detection and progression is crucial for monitoring the effect of experimental treatments as well as for informing developing efficient treatments. It is a pressing clinical need to improve early AD detection and progression. With recent advances in neurobiology of AD, our understanding of the disease has moved from one based on clinical symptoms to a biological construct that is multifactorial and heterogeneous and that cannot be explained by any single available biomarkers. NIH has devoted billions of dollars in the past decades to fund several centers and data initiatives on large cohorts of older adults; resulting in a wealth of multi-modal neuroimaging, cognitive, clinical, biospecimen, and genetic data. However, less effort has been made to implement innovative integrative methods for aggregating data across modalities to capture the heterogeneity of AD. To fill the gap in the analysis paradigm of multi-modal AD data, the overarching goals of the proposed study are to test and validate a multi-dimensional network framework for aggregating data across modalities in a single model to capture the heterogeneity of AD and to further enhance AD detection and progression. Our central hypothesis – backed by previous evidence and preliminary data – is that the proposed framework will enhance AD detection and progression by improving the ability to detect common as well as complementary signals across multiple data types and by reducing the effect of differences in scale, collection bias and noise in each modality. We will integrate behavioral, clinical, MR imaging, Aβ and Tau markers, and neurodegeneration markers from ADNI and Stanford ADRC data to test and validate the proposed multi-dimensional network framework for integration of different data types for early detection of AD (Aim1) as well as to characterize AD progression by applying multi-dimensional network framework to longitudinal changes in various measurements (Aim 2). To our knowledge, this is the first study that integrates various AD data in a multi-dimensional network model to characterize AD to further enhance AD detection and progression. If proven successful, this high-risk high-reward proposal will have a large impact on AD characterization, early detection and progression with significant health and economical impact. Moreover, successful completion of this study will provide critical tools for integrative analysis of multimodal data and will help shift the current analysis paradigm of available datasets and clinical trials which mainly focuses on independent analysis of single data types.