Intrinsic Modeling and Tracking of Neuroanatomy in Alzheimer's Disease

描述（由申请人提供）：项目摘要/摘要神经影像学在阿尔茨海默病（AD）的早期诊断中发挥着越来越重要的作用。来自大规模、多地点研究的数据的可用性，例如阿尔茨海默病神经影像学倡议（ADNI）和显性遗传阿尔茨海默病网络（DIAN），为提高我们对这种复杂疾病的理解提供了前所未有的机会。另一方面，这些不断增长的大规模、高维成像数据迫切需要开发和验证鲁棒的自动制图工具。为了成为AD脑成像研究的独立研究者，候选人在此K01申请中提出接受多模态图像分析，AD临床诊断，MR成像技术和生物统计学方面的培训。这些培训活动将极大地增强候选人在神经影像分析方面的背景，并为他成为计算机辅助AD早期诊断领域的领先研究人员的长期目标奠定坚实的基础。在研究计划中，候选人将开发和验证一套新的工具，用于在AD的发展和进展过程中使用解剖结构的内在几何结构进行神经解剖学映射。与传统的方法，在一个典型的欧几里德空间，如Talairach图集，大脑对齐，候选人建模的解剖结构固有的特征函数的拉普拉斯-贝尔特拉米（LB）算子和他们的Reeb图。这种光谱方法是不变的自然姿态变化，由于病理和疾病进展的几何变形的鲁棒性，并导致新的表面重建，建模和映射的方法。具体目标是：1.继续开发一个内在的框架映射的皮层下结构的基础上的LB特征函数。2.开发并验证一个自动化系统，用于皮质表面提取、主要脑沟识别和映射。3.开发和验证新算法，用于多模态融合与皮质映射。新算法将使用ADNI和DIAN的数据进行认知测量验证，并与现有方法在AD早期诊断中的辨别能力进行比较。本项目开发的软件工具将公开分发。

DESCRIPTION (provided by applicant): Project Summary/Abstract Neuroimaging plays an increasingly important role in the early diagnosis of Alzheimer's disease (AD). The availability of data from large scale, multi-site studies, such as the Alzheimer's Disease Neuroimaging Initiative (ADNI) and Dominantly Inherited Alzheimer's Network (DIAN), provide unprecedented opportunities of improving our understanding of this complicated disease. On the other hand, these large scale, high dimensional imaging data of ever growing size call for the urgent needs of developing and validating robust and automated mapping tools. To become in independent investigator of brain imaging research in AD, the candidate proposes in this K01 application to receive training in multimodal image analysis, clinical diagnosis of AD, MR imaging techniques, and biostatistics. These training activities will greatly augment the candidate's background in neuroimage analysis and establish a solid foundation for his long term goal of being a leading researcher in computer-aided early diagnosis of AD. In the research plan, the candidate will develop and validate a suite of novel tools for the mapping of neuroanatomy during the development and progression of AD using intrinsic geometry of the anatomical structure. In contrast to conventional approaches that align brains in a canonical Euclidean space such as the Talairach atlas, the candidate models the anatomy intrinsically with the eigenfunctions of the Laplace-Beltrami (LB) operator and their Reeb graphs. This spectral approach is invariant to natural pose variations, robust to geometric deformations due to pathology and disease progression, and leads to novel methods for surface reconstruction, modeling, and mapping. The specific aims are: 1. Validate and continue to develop an intrinsic framework for the mapping of sub-cortical structures based on the LB eigenfunctions. 2. Develop and validate an automated system for cortical surface extraction, major sulci identification, and mapping. 3. Develop and validate novel algorithms for multimodal fusion with cortical mapping. The new algorithms will be validated with cognitive measures using data from ADNI and DIAN, and compared with existing methods in terms of the discrimination power in the early diagnosis of AD. The software tools developed in this project will be distributed publicly.