Using Artificial Intelligence to Identify Accelerated Brain Aging in World Trade Center Responders

项目摘要/摘要 在9/11世界贸易中心(WTC)现场参与救援和恢复工作的男男女女 是否在中年时出现认知障碍，通常比基于年龄的认知障碍早几十年 检测到。到目前为止，认知功能障碍和损害的最一致的危险因素之一 人口包括长期暴露在WTC灾难现场和创伤后应激障碍症状 (创伤后应激障碍)。我们的初步分析发现，与痴呆症患者相比，应答者的皮质厚度降低 给认知未受损的WTC应答者。虽然这项工作对增进我们对 对于WTC应答者的认知损害，皮质厚度减少到什么程度仍不清楚 表明了一种已知的疾病，到目前为止还没有研究能够可靠地量化 核磁共振上明显的模式符合人群正常标准。我们的团队最近发现了一种高度敏感的生物标志物 对于功能性的“脑年龄”，我们早在40年代末就已经显示出发现恶化的迹象，S。 这一指标被称为大脑“网络稳定性”，可在多个大规模的静息状态功能中重复使用 磁共振成像数据收集并与逐渐的认知能力下降相关。两者之间的区别是 根据MRI数据预测的个人年龄(“大脑年龄”)与他们的实际年龄相比，提供了一个衡量标准 加速大脑老化。因此，关键的下一步是确定WTC响应者的大脑年龄，两者都是 结构上(皮质厚度)和功能上(网络稳定性)，这可能与观察到的WTC损伤有关 中年时的认知障碍。在目前的工作中，我们建议完成一个二次数据分析 大规模脑MRI训练数据集(UK Biobank，N=19831)，用于训练神经生物学的深度学习模型 衰老的特征及其可能的机制。然后，我们将比较在WTC看到的神经生物学特征 这些签名的回应者。在目标1中，我们使用和测量WTC应答者的大脑加速老化 在没有创伤后应激障碍的情况下，使用将大脑老化与人群正常以及阿尔茨海默氏症的蛋白质组标志进行比较的方法 疾病和相关痴呆，包括β-淀粉样蛋白和tau。在目标2中，我们利用以前的方法 人工智能在神经成像数据方面的发展，以开发针对关键机制的神经生物学分类器 与WTC的相关性：颗粒物、糖皮质激素、炎症、焦虑、抑郁和创伤后应激障碍，以确定 人工智能是否将WTC大脑归类为与其中一个或多个特定的神经生物学特征匹配 机械装置。这项研究响应了一项呼吁对WTC患者进行与衰老相关的研究建议。 (RFA-OH-21-004)，并将提高我们对现有 WTC应答者的神经影像研究。为了成功地预防ADRD，可靠的措施是 中年时大脑加速老化所需的亚临床变化。这项研究旨在实现 一种新的脑年龄测量方法，优化为对中年神经变化敏感，并将其与人工智能相结合 确定暴露可能影响WTC应答者的机制。

PROJECT SUMMARY/ABSTRACT The men and women who worked in rescue and recovery operations at the 9/11 World Trade Center (WTC) site are developing cognitive impairment at mid-life, decades before age-based cognitive impairment is usually detected. To date, one of the most consistent risk factors for cognitive dysfunction and impairment in this population include long-term exposures to the WTC disaster sites and symptoms of posttraumatic stress disorder (PTSD). Our preliminary analyses identified reduced cortical thickness in responders with dementia compared to cognitively unimpaired WTC responders. While this work has been valuable in advancing our understanding of cognitive impairment in WTC responders, it remains unknown to what extent reduced cortical thickness is indicative of a known disorder, and no studies to date have been able to reliably quantify the extent to which patterns evident on MRI match population norms. Our team has recently identified a highly sensitive biomarker for functional “brain age,” which we have shown to detect the first signs of deterioration as early as the late 40’s. Known as brain “network stability,” this measure replicates across multiple large-scale resting-state functional magnetic resonance imaging datasets and correlates with gradual cognitive decline. The difference between an individual’s predicted age based on MRI data (“brain age”) versus their chronological age provides a metric for accelerated brain aging. Therefore, a critical next step is to characterize WTC responders’ brain ages, both structurally (cortical thickness) and functionally (network stability), which may relate WTC trauma to observed cognitive impairment at mid-life. In the present work, we propose to complete secondary data analyses of a large-scale brain MRI training data set (UK Biobank, N=19,831) to train a deep learning model for neurobiological signatures of aging and its potential mechanisms. We will then compare neurobiological features seen in WTC responders to these signatures. In Aim 1, we measure accelerated brain aging for WTC responders with and without PTSD, using comparing brain aging to population norms, as well as to proteomic markers of Alzheimer’s Disease and related dementias, including β-amyloid and tau. In Aim 2, we leverage our previous methods development in AI of neuroimaging data to develop neurobiological classifiers specific to key mechanisms of relevance to WTC: particulates, glucocorticoids, inflammation, anxiety, depression, and PTSD, to determine whether AI classifies WTC brains as matching neurobiological signatures specific to one or more of these mechanisms. This study responds to a call for aging-related research proposals in WTC-affected individuals (RFA-OH-21-004) and will improve our understanding of accelerated neurobiological aging in an existing neuroimaging study of WTC responders. For the prevention of ADRD to be successful, reliable measures are needed for subclinical changes in accelerated brain aging that occur in midlife. This study seeks to implement a novel measure of brain age optimized to be sensitive to midlife neurological changes and combines it with AI to identify the mechanisms through which exposures may have affected WTC responders.