Multiethnic machine learning brain signatures of ADRD

ADRD 的多种族机器学习大脑特征

• 批准号: 10693310
• 负责人: Mohamad Habes
• 金额: $ 70.62万
• 依托单位: UNIVERSITY OF TEXAS HLTH SCIENCE CENTER
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-01 至 2027-06-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10693310
• 关键词: Acceleration; Address; Adult; Affect; Age; Aging; Algorithms; Alzheimer' s Disease; Alzheimer' s disease diagnosis; Alzheimer' s disease pathology; Alzheimer' s disease related dementia; Alzheimer’s disease biomarker; Amyloid beta-42; Atherosclerosis Risk in Communities; Atrophic; Attenuated; Biological; Biological Markers; Black American; Black Populations; Blood; Brain; Brain imaging; CHI3L1 gene; Caring; Cerebrovascular Trauma; Clinical; Cognition; Cohort Studies; Communities; Computer software; Data; Data Set; Dementia; Development; Diabetes Mellitus; Dimensions; Disease; Disparity; Dyslipidemias; Ethnic Origin; Ethnic Population; Exposure to; Framingham Heart Study; Functional Magnetic Resonance Imaging; Funding; Future; Genetic; Glial Fibrillary Acidic Protein; Health Care Costs; Heart; Heterogeneity; Hispanic Americans; Hispanic Populations; Hypertension; Impaired cognition; Incidence; Individual; Insulin Resistance; Intervention; Intervention Trial; Knowledge; Light; Longevity; Longitudinal cohort; Machine Learning; Magnetic Resonance Imaging; Medical; Metabolic syndrome; Microvascular Dysfunction; Morbidity - disease rate; Multi-Ethnic Study of Atherosclerosis; Neuroanatomy; Neurodegenerative Disorders; Not Hispanic or Latino; Obesity; Onset of illness; Participant; Pathogenesis; Pathologic Processes; Pathway interactions; Pattern; Population; Preventive; Race; Research; Rest; Smoking; Stage at Diagnosis; Structure; Training; United States; United States National Institutes of Health; Validation; advanced analytics; aging brain; aging demographic; aging demography; aging population; analytical method; biomarker validation; blood-based biomarker; brain based; cardiovascular health; cardiovascular risk factor; clinical diagnosis; clinically relevant; cognitive function; cohort; comorbidity; dementia risk; disorder risk; ethnic difference; ethnic diversity; ethnic minority; ethnoracial; functional decline; genomic epidemiology; health disparity; high dimensionality; lifestyle data; machine learning algorithm; machine learning method; machine learning model; machine learning prediction; mild cognitive impairment; mortality; multi-ethnic; multimodal neuroimaging; neurofilament; neuroimaging; neuroimaging marker; novel; pre-clinical; predictive signature; public health priorities; racial difference; racial diversity; racial minority; racial population; sex; socioeconomics; tau-1; therapy development; trend; vascular cognitive impairment and dementia; vascular risk factor

PROJECT SUMMARY / ABSTRACT The underlying pathology of Alzheimer's disease and related dementias (ADRDs) accumulates gradually over decades, making the identification of non-invasive, sensitive biomarkers in the preclinical stage a critical public health priority. Harnessing advanced analytic methods, our team and others have established neuroimaging signatures of advanced brain aging (Spatial Pattern of Atrophy Recognition of Brain Aging, SPARE-BA) and functional decline (fSPARE-BA), and ADRDs (SPARE-AD and SPARE-Small vessel disease), which predict incident cognitive decline. Unfortunately, most research to date has been conducted in predominantly non- Hispanic white populations, which limits the ability to generalize results to the diverse ethnoracial makeup of the United States' growing aging demographic. If current trends continue, machine learning models will primarily be trained in ethnically imbalanced datasets, leading to biases that may affect clinical relevance. Thus, the primary aims of the current proposal are to: leverage an ethnically diverse neuroimaging consortium to build new machine learning models trained by data from ethnically well-balanced populations, derive sensitive and specific neuroimaging signatures of brain aging and ADRD, and evaluate whether they can be practical non-invasive biomarkers of incident cognitive decline, mild cognitive impairment (MCI), and dementia across ethnoracial groups. We propose to leverage the rich clinical and neuroimaging (structural MRI and resting-state functional MRI) data within the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium, including the Atherosclerosis Risk in Communities Study (ARIC), the Cardiovascular Health Study (CHS), the Genetics of Brain Structure and Function Study (GOBS), the Framingham Heart Study (FHS), the Vascular Contributions to Cognitive Impairment and Dementia consortium (MARK-VCID) and the Multi-Ethnic Study of Atherosclerosis (MESA). We will leverage a collaborative research framework across existing longitudinal cohorts to address unanswered questions contributing to disparities in ADRD burden. Machine learning algorithms will be applied to brain imaging data of over 7,200 non-Hispanic Whites, 1,400 Blacks, and 1,425 Hispanics to address our Specific Aims: 1) Generate and evaluate clinical utility of machine learning-based signatures of brain aging and ADRD for each race/ethnic group and uncover multidimensional heterogeneity in aging across groups; 2) Examine associations of vascular risk factors with the derived machine learning-based brain signatures of ADRD by race/ethnicity, and 3) Explore blood-based biomarker predictors of these machine learning-based brain signatures by ethnoracial group to elucidate underlying biological mechanisms. Further, we will share our robust machine learning models together with implementation software with the scientific community. This project will develop and validate neuroimaging markers with robust predictive utility for incident cognitive decline and to identify underlying pathophysiologic pathways, expanding opportunities for novel intervention development across diverse ethnoracial cohorts. ii

项目摘要/摘要 阿尔茨海默病和相关痴呆(ADRD)的基本病理逐渐积累 几十年，使临床前阶段非侵入性、敏感生物标志物的识别成为关键 健康优先。利用先进的分析方法，我们的团队和其他人建立了神经成像 高级脑老化的特征(脑老化的萎缩识别的空间模式，Spare-BA)和 功能下降(fSPARE-BA)和ADRDS(备用AD和备用小血管疾病)，它们可以预测 突发事件认知功能下降。不幸的是，到目前为止，大多数研究都是在主要非 西班牙裔白人人口，这限制了将结果推广到不同民族构成的能力 美国日益老龄化的人口结构。如果目前的趋势继续下去，机器学习模型将主要是 在种族不平衡的数据集上进行培训，导致可能影响临床相关性的偏差。因此，主要的 当前提案的目标是：利用一个不同种族的神经成像联盟来制造新的机器 通过来自种族平衡的人群的数据训练的学习模型，得出敏感和具体的 脑老化和ADRD的神经影像特征，并评估它们是否可以实用于非侵入性 跨民族的认知功能衰退、轻度认知障碍(MCI)和痴呆的生物标志物 组。我们建议利用丰富的临床和神经成像(结构MRI和静息状态功能 MRI)数据在基因组流行病学(CHARE)联盟心脏和衰老研究队列中， 包括社区动脉粥样硬化风险研究(ARIC)、心血管健康研究(CHS)、 大脑结构和功能遗传学研究(GOBS)，Framingham心脏研究(FHS)，血管 对认知障碍和痴呆症联盟的贡献(MARK-VCID)和多种族研究 动脉粥样硬化(MESA)。我们将利用跨现有纵向研究的协作研究框架 解决造成ADRD负担差异的悬而未决的问题。机器学习 算法将应用于超过7200名非西班牙裔白人、1400名黑人和1425名黑人的脑成像数据 拉美裔解决我们的具体目标：1)生成和评估基于机器学习的临床效用 每个种族/民族的脑老化和ADRD的特征，并揭示了 跨组老化；2)检查血管危险因素与基于机器学习的派生的相关性 按种族/民族分类的ADRD的脑信号，以及3)探索这些机器的基于血液的生物标志物预测因素 以学习为基础的脑信号，由人脑小组阐明潜在的生物学机制。此外，我们 将与科学界共享我们健壮的机器学习模型和实现软件 社区。该项目将开发和验证具有强大的事件预测实用功能的神经成像标记 认知衰退和识别潜在的病理生理途径，扩大小说的机会 在不同的民族队列中的干预发展。 II