Fast and robust deep learning tools for analysis of neuroimaging data of Alzheimer's disease

快速、强大的深度学习工具，用于分析阿尔茨海默病的神经影像数据

• 批准号: 10573337
• 负责人: Yong Fan
• 金额: $ 69.99万
• 依托单位: UNIVERSITY OF PENNSYLVANIA
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-03-15 至 2026-02-28
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10573337
• 关键词: Adopted; Aging; Algorithms; Alzheimer' s Disease; Alzheimer' s disease diagnosis; Alzheimer' s disease model; Anatomy; Brain; Brain imaging; Brain scan; Clinical; Cognitive; Computer Assisted; Computer software; Consumption; Data; Data Analyses; Data Set; Dementia; Development; Disease Progression; Ensure; Environment; Event; High Performance Computing; Hour; Image; Image Analysis; Imaging technology; Impaired cognition; Individual; Intervention; Knowledge; Learning; Machine Learning; Magnetic Resonance Imaging; Maps; Measures; Medicine; Methods; Modeling; Neurodegenerative Disorders; Pattern Recognition; Pennsylvania; Performance; Persons; Prognosis; Research; Research Personnel; Resources; Risk; Scanning; Source Code; Speed; Structure; Surface; Techniques; Time; Translations; Universities; brain based; brain morphology; brain tissue; cluster computing; convolutional neural network; cost; deep learning; deep learning algorithm; disease prognosis; graph neural network; image processing; image registration; imaging Segmentation; improved; large scale data; learning strategy; mild cognitive impairment; neuroimaging; non-invasive imaging; novel; open source; portability; pre-clinical; predictive modeling; prognostic model; programs; reconstruction; segmentation algorithm; supervised learning; tool; user-friendly; web app

Alzheimer's disease (AD) is the most prevalent neurodegenerative disorder. Interventions at the preclinical and prodromal stages are appealing targets for slowing or halting disease progression. It is desired to achieve accurate prognosis of AD dementia and cognitive decline for people with mild cognitive impairment who have increased risk to develop AD. In order to achieve fast and accurate prognosis of AD dementia based on neuroimaging data, we will develop and validate novel deep learning techniques. Particularly, we will develop unsupervised deep learning methods for segmenting brain images and reconstructing cortical surfaces from structural magnetic resonance imaging data. These fast and accurate image processing methods will be used in conjunction with advanced deep learning methods to build prognosis models of AD dementia and cognitive decline in a time-to-event analysis framework using large-scale imaging datasets. Finally, we will develop and disseminate a user friendly, open source, modular, and extensible software package to improve prognosis of AD dementia. Source code, standalone programs, and web-application interfaces of all the algorithms will be made available on GitHub and NITRC. Our tools will enable real-time neuroimaging data analysis and can find applications in diverse fields, including quantifying brain changes associated with aging and development.

阿尔茨海默病（Alzheimer's disease，AD）是最常见的神经退行性疾病。临床前干预 和前驱期是减缓或停止疾病进展的有吸引力的目标。则期望 实现AD痴呆和轻度认知障碍患者认知功能下降的准确预后 他们患AD的风险增加。为了实现AD痴呆的快速准确预后 基于神经成像数据，我们将开发和验证新的深度学习技术。特别是，我们 将开发无监督的深度学习方法，用于分割大脑图像和重建大脑皮层。 表面结构磁共振成像数据。这些快速准确的图像处理 方法将与先进的深度学习方法结合使用，以建立AD的预后模型。 使用大规模成像数据集，在事件发生时间分析框架中研究痴呆和认知能力下降。 最后，我们将开发和传播一个用户友好的，开放源代码的，模块化的，可扩展的软件 改善AD痴呆症预后。源代码、独立程序和Web应用程序 所有算法的接口都将在GitHub和NITRC上提供。我们的工具将使实时 神经成像数据分析，可以在不同领域找到应用，包括量化大脑变化 与衰老和发育有关。