Brain entropy mapping in Alzheimer's Disease

摘要。阿尔茨海默病（AD）影响了数千万人，但仍然无法治愈。的 AD的主要危险因素是衰老，由于进行性脑损伤，这需要高脑熵。值得注意的是， 大脑不断消耗大量的能量来维持其功能的完整性，可能会产生一个巨大的 “储备”来抵消高熵。这一储备的功能失常可能预示着疾病的临界点 转化和进步。储备故障可以通过补偿结果来衡量：功能 脑熵（fBEN），这可能揭示了弥合AD之间的知识差距的关键信息 病理学和临床症状：AD病理学早在AD症状之前就开始了并且可能不会导致痴呆， 并且可以提供用于早期疾病检测的脑标记。本项目旨在表征fBEN， 正常老化和AD连续体，并测试倒U形fBEN模型：fBEN随着年龄和AD增加 在正常衰老的病理，但减少在AD连续。我们将使用现有的大型 静息状态fMRI（rsfMRI）数据。我们的团队在2010年开始了基于rsfMRI的fBEN映射，并发布了第一个 开源fBEN映射工具。该工具已被广泛用于许多神经科学和翻译 问题研究为了适应大数据，我们将在这个项目中开发一个更快的版本。目标1： 将使用来自公共数据库的2000多名年轻和老年健康个体的数据计算fBEN， 检查fBEN与年龄，教育和认知功能的关系。我们假设大脑 一个储备相关网络，其fBEN随受教育年限的增加而减少，并与 脑功能，表明该网络中的低fBEN作为脑储备的指标。目标2将表征fBEN 并评估其与AD病理学和临床症状的相关性。我们假设有一个强大的 病理与疾病对fBEN的相互作用：fBEN随着年龄和病理在正常老化和较低 fBEN与更好的认知相关;这些关联将在AD连续体中逆转，因此fBEN将 fBEN水平随病理级别的升高而降低，fBEN水平越低，认知功能越差。目标3将评估可行性 在目标1和2中阐明的区域中基线fBEN的早期疾病检测。因为fBEN可能只会 为了反映AD患者的部分功能异常，我们将联合收割机fBEN与其他影像学生物标志物结合， 模型，以达到更高的预测精度。我们还将建立一个模型，使用fBEN来预测AD 正常衰老的病理学本项目的临床影响包括：1）测试假设的fBEN模型 将有助于描述AD的病理与临床差异; 2）发现疾病相关的fBEN模式可能 提供一个潜在的干预目标; 3）测试预测模型将有助于早期疾病检测; 4）AD 病理预测对于正常衰老和AD管理具有重要的临床价值，因为目前AD病理 测量要么是痛苦的，要么是太昂贵的。这个新项目的可行性和许多新的 我们的大量初步数据初步证实了这些假设。

ABSTRACT. Alzheimer’s Disease (AD) has affected tens of millions of people but remains incurable. The major risk factor for AD is aging, which entails high brain entropy due to the progressive brain damages. Notably, the brain constantly consumes a large amount of energy to maintain its functional integrity, likely creating a big “reserve” to counteract the high entropy. Malfunctions of this reserve may indicate a critical point of disease conversion and progression. Reserve malfunction can be measured by the compensation outcome: functional brain entropy (fBEN), which may reveal critical information for bridging the knowledge gap between AD pathology and clinical symptoms: AD pathology begins long before AD symptoms and may not lead to dementia, and may provide a brain marker for early disease detection. This project is proposed to characterize fBEN in normal aging and the AD continuum and test an inverse U-shape fBEN model: fBEN increases with age and AD pathology in normal aging but decreases in the AD continuum. We will test the model using large existing resting state fMRI (rsfMRI) data. Our group started rsfMRI-based fBEN mapping in 2010 and released the first open-source fBEN mapping tool. The tool has been widely used in many neuroscientific and translational studies. To be scalable for large data, we will develop a further accelerated version in this project. In Aim 1, we will calculate fBEN using data from 2000+ young and old healthy individuals from public databases and examine the associations of fBEN to age, education, and cognitive function. We hypothesize that the brain has a reserve-related network whose fBEN decreases with years of education and is negatively correlated with brain function, suggesting low fBEN in this network as an indicator of brain reserve. Aim 2 will characterize fBEN in AD and assess its associations to AD pathology and clinical symptoms. We hypothesize that there is a strong pathology vs disease interaction on fBEN: fBEN increases with age and pathology in normal aging and lower fBEN correlates with better cognition; those associations will be reversed in the AD continuum so that fBEN will decrease with pathology level and lower fBEN will correlate to worse cognition. Aim 3 will evaluate the feasibility of baseline fBEN in regions elucidated in Aims 1 and 2 for early disease detection. Because fBEN may only reflect some part of the functional abnormalities in AD, we will combine fBEN with other imaging biomarkers in the model in order to achieve higher prediction accuracy. We will also build a model to use fBEN to predict AD pathology in normal aging. The clinical impact of this project includes: 1) testing the hypothetical fBEN model will help delineate the pathology vs clinical discrepancy in AD; 2) finding the disease related fBEN patterns may provide a potential intervention target; 3) testing the prediction model will aid early disease detection; 4) AD pathology prediction is of great clinical value for normal aging and AD management as current AD pathology measurement is either painful or too expensive. The feasibility of this novel project and the many new hypotheses have been initially demonstrated by our extensive preliminary data.