Establishing a Brain Health Index from the Sleep Electroencephalogram

从睡眠脑电图建立大脑健康指数

• 批准号: 10180268
• 负责人: SYDNEY S CASH
• 金额: $ 150.66万
• 依托单位: MASSACHUSETTS GENERAL HOSPITAL
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-05-01 至 2022-09-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10180268
• 关键词: 20 year old; Address; Age; Aging; Alzheimer' s Disease; Amyloid; Amyloid beta-Protein; Archives; Arousal; Biological; Biological Markers; Biological Process; Brain; Brain Diseases; Brain Pathology; Cerebrum; Chronic; Chronology; Claustral structure; Clinical; Clinical Trials; Cognitive; Cognitive aging; Cognitive deficits; Complex; Data; Dementia; Devices; Diabetes Mellitus; Disease; Drainage procedure; Early Diagnosis; Electroencephalogram; Epidemiology; Exhibits; Framingham Heart Study; General Hospitals; Genome; Goals; HIV; HIV Infections; Health; Heart; Hippocampus (Brain); Home; Hypertension; Impaired cognition; Individual; Intervention; Israel; Knowledge; Koreans; Life Expectancy; Link; Lung; Machine Learning; Magnetic Resonance Imaging; Massachusetts; Measurable; Measures; Medial; Medical center; Memory; Modeling; Motivation; Nerve Degeneration; Neurodegenerative Disorders; Neurologic; Neuropsychological Tests; Neuropsychology; Outcome; Pathology; Patient Care; Patients; Pattern; Physiological; Play; Population; Positron-Emission Tomography; Prefrontal Cortex; Process; Property; REM Sleep; Reporting; Research; Risk; Role; Science; Series; Signal Transduction; Sleep; Sleep Apnea Syndromes; Sleep Stages; Sleep disturbances; Smoker; Statistical Models; Structure; System; Test Result; Testing; Thick; Thinness; Time; Variant; Work; age related; aging brain; basal forebrain; base; brain health; cognitive function; cognitive performance; cognitive testing; cohort; deep neural network; epidemiology study; executive function; functional loss; glymphatic system; healthy aging; high risk; indexing; individual patient; large datasets; machine learning algorithm; memory consolidation; mild cognitive impairment; model development; mortality; multi-task learning; neuroimaging; neuropathology; non rapid eye movement; novel; novel therapeutics; predictive modeling; sleep physiology; sleep spindle; specific biomarkers; therapy development

Project Abstract: Establishing a Brain Health Index from the Sleep Electroencephalogram Although cognitive decline is a “normal” part of aging, some individuals clearly age better than others. However, the concept of differential aging has been minimally studied for the brain. Electroencephalogram (EEG) oscillations signals carry rich information regarding brain health and brain aging. Alzheimer's disease (AD) is associated with fragmented sleep and altered sleep oscillations. Clearance of cerebral beta amyloid through the brain's glymphatic drainage system occurs mainly in non-rapid eye movement (NREM) sleep, and depends on EEG slow oscillations. Cortical generators of sleep EEG oscillations overlap with regions of cortical thinning and loss of functional connectivity in AD. Disturbances of NREM disrupt memory consolidation. Finally, deficient REM sleep contributes to dementia. These observations suggest that brain health may be measurable from information contained in the sleep EEG. In preliminary work we have developed EEG-brain age – a machine learning model that predicts a patient's age based on patterns of overnight sleeping EEG oscillations. This allows prediction of age with a precision of +/- 7 years. Our preliminary data suggest diabetes and hypertension, chronic HIV infection, an MCI or AD are reflected in the EEG as excessive brain age, and that excessive brain age is independently associated with reduced life expectancy. Our central hypothesis is that sleep physiology data can provide sensitive and specific biomarkers of brain health. This hypothesis is based on our prior work showing that BAI is elevated in several clinical populations. BAI can be accurately calculated using frontal EEG signals, making it suitable for implementation on at-home EEG devices. The rationale for the proposed research is that validating sleep EEG-derived biomarkers as measures of brain health at the level of individual patients would lay the ground for use in clinical trials and patient care. We plan to accomplish the central objective by pursuing two complementary aims. In Aim 1, we will take a hypothesis-driven approach, and test for associations of specific sleep features with specific cognitive deficits and specific structural pathology. In Aim 2, we will take ad data-driven approach, and develop optimized biomarkers of brain health using a novel form of machine learning known as multitask learning, which combine multiple features of sleep – including conventional features, as well as data-driven features directly learned from the data – to predict or “explain” variation in cognitive performance and in structural brain MRI measures that are indicative of brain health or disease. The project will take advantage of a large and diverse set of sleep data (>33,000 patients), as well as thousands of brain MRI and cognitive testing results. At the conclusion of this study, we expect to have a better understanding of the role sleep oscillations play in brain health, and clinically useful brain health biomarkers. These outcomes will aid development of interventions to promote brain health.

项目摘要：从睡眠脑电建立脑健康指数 虽然认知能力下降是衰老的“正常”部分，但一些人显然比其他人更容易变老。 然而，对大脑的差异衰老概念的研究还很少。 脑电(EEG)振荡信号携带着丰富的有关脑健康和脑老化的信息。 阿尔茨海默病(AD)与零碎的睡眠和改变的睡眠振荡有关。净空 大脑β淀粉样蛋白通过大脑的淋巴引流系统主要发生在非快眼 运动(NREM)睡眠，并依赖于脑电的缓慢振荡。睡眠脑电的大脑皮层发生器 在AD中，振荡与皮质变薄和功能连接丧失的区域重叠。扰乱 NREM破坏了记忆的巩固。最后，缺乏快速眼动睡眠会导致痴呆症。这些观察结果 这表明，大脑健康状况可能可以从睡眠脑电中包含的信息来衡量。 在前期工作中，我们开发了EEG-Brain AGE-一种机器学习模型，可以预测患者的 基于夜间睡眠脑电振荡模式的年龄。这使得对年龄的预测可以达到 +/-7年。我们的初步数据表明，糖尿病和高血压、慢性艾滋病毒感染、MCI或AD 在脑电中反映为大脑年龄过大，而大脑年龄过大独立地与 减少了预期寿命。 我们的中心假设是，睡眠生理学数据可以为大脑提供敏感和特异的生物标志物 健康。这一假设是基于我们之前的工作，表明BAI在几个临床人群中都是升高的。 利用额部脑电信号可以准确计算出BAI，因此适合在家中实施 脑电波设备。这项拟议研究的基本原理是，将睡眠脑电衍生的生物标志物确认为 在个体患者水平上衡量大脑健康状况将为临床试验和 病人护理。我们计划通过追求两个相辅相成的目标来实现这一中心目标。在目标1中，我们 将采取假设驱动的方法，并测试特定睡眠特征与特定睡眠特征之间的关联 认知缺陷和特定的结构病理学。在目标2中，我们将采取广告数据驱动的方法，并开发 使用一种名为多任务学习的新机器学习形式来优化大脑健康的生物标记物， 它们结合了睡眠的多个特征--包括常规特征以及数据驱动特征 直接从数据中学习--预测或“解释”认知表现和大脑结构的变化 核磁共振测量表明大脑健康或疾病。该项目将利用一个大型和 不同的睡眠数据集(33,000名患者)，以及数千个大脑核磁共振和认知测试结果。 在这项研究的结论中，我们希望能更好地理解睡眠振荡在 脑健康，以及临床上有用的脑健康生物标志物。这些成果将有助于 促进大脑健康的干预措施。