面向阿尔茨海默病的脑功能网络及分类研究

Alzheimer's disease (AD) is one of the major causes of disability in late-life. Beyond the impact on those directly affected, the increases of AD patients entail significant society-wide psychological, social and economic burdens. A desirable purpose of AD research is to develop means of making early and accurate diagnoses of individuals who either have or are at increased risk of developing AD. This will allow for timely interventions that may be effective in managing or treating AD. However, currently there are no accurate, effective, non-invasive, economic, and secure biomarkers for early diagnosis of AD. Recent studies have demonstrated that the abnormal functional connectivity based on resting state functional magnetic resonance imaging (fMRI) is prior to structural atrophies during the progression of AD. Resting state fMRI is non-invasive, convenient and secure, which is promising to be a new tool to investigate AD biomarkers.The first aim of this study is to construct robust, adaptive, data-driven, and high-dimensional functional connectivity maps on an individual basis using multicenter resting state fMRI dataset. To this end, the dynamic network pattern tracking disease progression will be investigated, and the biomarkers on network level will be unravelled. The second aim of the study is to take the advantage of the complementarities and diversity of different modalities, features and classifiers, and to study a meta-learning strategy for fusing multiple AD classifiers to improve the performance of early diagnosis. In conclusion, we plan to develop a promising network biomarker and methods for accurate diagnosis of AD in its early stages in order to help prevent the development of this syndrome and/or slow its progression.

阿尔茨海默病（AD）是威胁老年人身心健康和生活质量的严重疾病之一。AD的早期诊断和及时干预可以很大程度上延缓疾病的发展。然而，目前为止，尚缺乏准确有效的非侵入、经济安全的生物标志进行AD的早期识别。研究表明在AD的发展过程中基于静息态功能磁共振（fMRI）的脑网络异常早于结构萎缩，且静息态fMRI具有非侵入、简单方便、容易被患者接受等特点,有望成为AD早期识别的新工具。本项目拟基于多中心大样本数据集的静息态fMRI数据，在个体上构建鲁棒的、自适应的、数据驱动的高分辨率功能连接图，探索疾病发展过程中的网络模式的动态变化，明确其脑网络水平生物标志。同时，利用多模态、多特征、多分类器的互补性和多样性，研究基于元学习的疾病预测框架，以提高AD早期识别的正确率。本项目的研究可以从一定程度上为理解AD发展过程中脑网络层次的异常提供新的科学证据，同时也有可能为AD的早期诊断提供有效的检测方法。

伴随着我国人口老龄化，阿尔茨海默病（AD）逐渐成为家庭和社会沉重的负担。AD的早期识别、早期干预对延缓疾病发展有重要作用。本项目基于多中心的静息态fMRI数据，发展脑功能连接建立方法，研究脑功能网络在正常老人、轻度认知障碍 (MCI) 及AD发展过程中的异常表征及其遗传基础，AD的早期预测等内容。我们建立了个体的脑功能连接密度图，首次在多中心数据集发现AD患者中功能连接在楔前叶/后扣带显著降低，在内侧前额叶显著升高，证明AD患者中可能在默认网络中存在功能代偿作用。功能异常区域与MMSE评分显著相关。将此指标用于识别MCI人群，识别率高达81%。进一步地，我们研究了典型功能网络内和网络间的连接异常模式。横断面的研究表明，在AD阶段，默认网络、感觉运动、注意、警觉、执行控制等功能网络内和网络间普遍出现失连接。在MCI阶段，感觉运动和注意网络连接最先受损。纵向研究表明，默认网络内的功能连接变化程度和认知水平显著相关，且默认网络和警觉网络连接变化水平在疾病早期更为明显。此外，我们基于双生子的研究策略考察了功能连接的遗传基础，发现功能连接与遗传相关之间显著相关，相关系数为0.344。表明功能连接可部分由遗传相关解释，功能网络表征可作为AD等疾病遗传学研究的内表型。上述研究成果系统研究了全脑网络、子网络及中心节点在AD发展中的异常变化模式，发展了AD早期多中心稳定的网络水平生物特征，有望服务于临床的早期识别和早期干预中。