Neurodegenerative diseases and the role of green space: A deep learning assessment

项目摘要 阿尔茨海默氏病和相关痴呆症（ADRD）有明确的危险因素，如体力活动（PA）， 抑郁症和高血压（HTN）。这些风险因素不成比例地影响少数民族人口，但 造成种族健康差异的机制尚未得到充分了解。在这方面，地理因素可能是关键，因为巴勒斯坦权力机构， 抑郁症和HTN的强烈影响的地理曝光，包括绿色空间。然而，绿色空间是 通常用问卷调查来衡量，问卷调查有很大的误差，或者用卫星指数来衡量，卫星指数是不具体的， 不提供关于植被类型的信息（例如，树与草），也不是植被是否在街道上可见 水平因此，没有研究量化了绿色空间对ADRD种族差异的贡献。虽然新奇 谷歌街景（GSV）成像等技术是很有前途的数据来源，可用于捕捉 绿色空间，管理、处理和分析高维数据， 这是一个挑战，特别是在将这些数据与来自大型前瞻性队列的现有数据联系起来时。最后，我们需要了解 绿色空间在其他潜在相关的地理暴露的背景下，或城市麻烦-生活的整体- 课程地理暴露（绿色空间、空气污染物、噪音、建筑环境和社会环境）-- 估计哪些因素驱动健康。该提案将通过使用GSV成像来评估效果， PA、抑郁症和HTN的绿色空间，以及随后的ADRD风险， 动脉粥样硬化（梅萨）-一项为期10年的纵向研究，包括6,814名无临床心血管疾病的男性和女性， 基线来自4个人种/种族组（非西班牙裔白色、非洲裔美国人、中国人和西班牙裔）。目标1将量化 绿色空间的特定方面（如树木，草，灌木，植物）对ADRD和认知能力下降的影响，并评估 这些关联是否因种族/民族而异。目标2将确定绿色空间对 通过PA、抑郁和HTN介导的ADRD。目标3将量化与ADRD的不良关联， 使用非目标性数据驱动方法结合降维技术来评估认知下降， 他们是否因种族/民族而异。这项研究计划是由一个培训计划，建立在 申请人的流行病学和生物统计学背景，包括（1）实施深度学习的新培训 分析高分辨率地理数据的算法，（2）认知功能流行病学，以及（3）开发和完善 数据驱动的方法，以执行麻烦知情的流行病学研究。这些组合计划将成功地 准备申请人的独立研究生涯，重点是确定可修改的地理决定因素， ADRD在不同人群中使用创新的地理环境措施。

PROJECT SUMMARY Alzheimer’s disease and related dementias (ADRD) have well-established risk factors such as physical activity (PA), depression, and hypertension (HTN). These risk factors disproportionately affect racial minority populations, but the mechanisms underlying racial health disparities are not well understood. In this, geographic factors could be key, as PA, depression and HTN are strongly affected by geographic exposures, including green space. However, green space is typically measured with questionnaires, which have substantial error, or satellite-based indexes that are nonspecific and provide no information on the type of vegetation (e.g., tree vs. grass), nor whether the vegetation is within view at the street level. As a result, no study has quantified the contribution of green space to racial disparities in ADRD. And while novel technologies such as Google Street Views (GSV) imaging are promising data sources for capturing unique measures of green space, managing, processing, and analyzing high-dimensional data present significant logistical and analytical challenges, especially when linking these data to existing data from large prospective cohorts. Finally, we need to understand green space in the context of other potentially correlated geographic exposures, or the urban exposome—the totality of life- course geographic exposures (the set of green space, air pollutants, noise, built environment, and social environment)—to estimate which factors drive health. This proposal will address these challenges by using GSV imaging to assess the effect of green space on PA, depression, and HTN, as well as subsequent ADRD risk within the Multi-Ethnic Study of Atherosclerosis (MESA)—a 10-year longitudinal study of 6,814 men and women without clinical cardiovascular disease at baseline from 4 racial/ethnic groups (Non-Hispanic White, African-American, Chinese, and Hispanic). Aim 1 will quantify the effect of specific aspects of green space (e.g. trees, grass, shrubs, plants) on ADRD and cognitive decline and evaluate whether these associations differ according to race/ethnicity. Aim 2 will determine the indirect effect of green space on ADRD that is mediated through PA, depression, and HTN. Aim 3 will quantify exposome associations with ADRD and cognitive decline using untargeted data-driven approaches in conjunction with dimension reduction techniques and evaluate whether they differ according to race/ethnicity. This research plan is complemented by a training plan that builds on the applicant’s background in epidemiology and biostatistics and includes new training in (1) implementing deep learning algorithms to analyze high-resolution geographic data, (2) cognitive function epidemiology, and (3) developing and refining data-driven approaches to perform exposome-informed epidemiological studies. These combined plans will successfully prepare the applicant for an independent research career focused on identifying modifiable geographic determinants of ADRD in diverse populations using innovative measures of geographic context.