MRI: Development of an Instrument for a Comprehensive Study of Alzheimer's Disease: Multimodal Imaging, Visualization, Machine Learning and Therapeutic Brain Stimulation

该项目开发了一种工具，不仅在多种记录方式方面，而且在结构、功能和代谢脑数据的收集和管理方面，建立了一个综合的基础设施。作为一个应用，它利用了阿尔茨海默病(AD)的综合研究，特别是神经心理测试、遗传学和人口统计因素，所有这些都与具有共同评估和标准化测量的数据库相链接，从而建立了一个最适合多站点研究和跨站点数据合并的环境。预计这一工具将为寻求机器/深度学习的黄金标准(即稳定性、稀疏性、可解释性、准确性以及处理临床研究固有的缺失数据和对抗多重共线性的能力--这是纵向研究中重复测量设计的内在特征)创造一个适当的环境。这项工作集成了神经成像网络服务接口、多模式神经成像平台和计算平台。该仪器的创建是为了通过稀疏但高度可解释的决策树和过程来克服集成方法(如ANN(人工神经网络))的不透明度。通过了解哪些重要特征是导致给定分类和/或预测结果的关键，诊断或预后的审议过程可能会得到加强。该设计的模块化结构允许扩展到涉及其他神经疾病的其他应用领域，其中大脑成像是任何主题危重护理的一部分。通过网络接口将该仪器用作网络物理系统，可使研究界能够在多模式和计算有效的神经成像平台上应用新的数据挖掘概念或执行新的分类和预测算法，从而为多站点研究以及广泛的数据共享打开领域。这种复杂的仪器允许培养将工程和计算知识结合到生物科学、计算机和医学领域的应届毕业生。它通过唤起人们对这种疾病的新理解，极大地促进了科学和国民健康。根据阿尔茨海默氏症协会2017年报告的数据，这种疾病影响了10%的65岁以上人口(不断增长的成本估计超过2590亿美元)。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project develops an instrument to establish an integrated infrastructure not only in terms of multiple recording modalities, but also in terms of collection and management of structural, functional, and metabolic brain data. As an application it utilizes the comprehensive study of Alzheimer's Disease (AD),notably neuropsychological testing, genetics, and demographic factors, all linked to a database with common evaluations and standardized measures, thus setting up an environment most suitable for multi-site studies and the merging of data across sites. This instrument is expected to create an appropriate environment for seeking the gold standards of machine/deep learning (i.e., stability, sparsity, interpretability, accuracy, and ability to handle missing data inherent to clinical studies and to confront multicollinearity--an intrinsic characteristic from the repeated measures design in longitudinal studies). This work integrates a Neuroimaging Web-Services Interface, Multimodal Neuroimaging Platform, and a Computational Platform. The instrument is created to overcome the opacity of ensemble methods such as ANNs (Artificial Neural Networks) through sparse, and yet highly interpretable, decision trees and processes. The deliberation process for a diagnosis or prognosis is likely to be enhanced by learning what significant features are essential to lead to a given classification and/or prediction outcome. The modular structure of the design allows extensions to other application domains that involve other neurological disorders where brain imaging is part of any subject critical care. Utilizing the instrument through the web interface as a cyber-physical system may enable the research community to apply new data mining concepts or to execute novel classification and prediction algorithms on a multimodal and computationally-effective neuroimaging platform, thus opening the field for multi-site studies, as well as extensive data sharing. This sophisticated instrument allows for training new graduates who combine engineering and computing know-how into the fields of bioscience, computing, and medicine. It significantly promotes science and national health by eliciting new understanding of a disease that according to the data reported in 2017 by the Alzheimer's Association, affects 10% of the population over 65 (with growing costs estimated above $259 billion).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.