Computer Aided Early Detection and Diagnosis of Alzheimer's Disease

项目描述（由申请人提供）：本项目旨在培养候选人在神经图像分析方面进行多学科研究的能力，重点是多元神经图像分类及其在阿尔茨海默病（AD）计算机辅助早期检测和诊断中的应用。该研究项目将侧重于开发新的神经图像分类算法，以全自动和集成的方式准确识别和测量大脑异常。候选人将通过整合特征提取，特征选择和分类来开发一个通用的集成神经图像分类框架。在此框架下，本文提出了一种改进的结构和功能图像特征提取的多模态模式分类方法。这些多模态多变量分类方法将被验证并应用于基于神经影像学的早期阿尔茨海默病诊断研究和阿尔茨海默病相关脑异常的纵向测量。作为拟议KOI申请的一部分，候选人寻求医学影像学，神经科学，AD临床诊断和生物统计学方面的教学培训。拟定的培训和研究计划将培养候选人成为一名独立的科学家，使用神经成像和图像分析方法进行阿尔茨海默病的早期检测和诊断。

DESCRIPTION (provided by applicant): The goal of this project is to develop the candidate's ability to perform multidisciplinary research in neuroimage analysis, with emphasis on multivariate neuroimage classification and its application to computer aided early detection and diagnosis of Alzheimer's disease (AD). The research project will focus on the development of novel neuroimage classification algorithms for accurately identifying and measuring brain abnormality in a fully automatic and integrated way. The candidate will develop a general integrated neuroimage classification framework by integrating feature extraction, feature selection, and classification. Within this general framework, a multimodality pattern classification method with improved feature extraction from both structural and functional images will be developed. These multimodality multivariate classification methods will be validated and applied to the neuroimage based studies of early AD diagnosis and longitudinal measurement of brain abnormality related to AD. As part of the proposed KOI application, the candidate seeks didactic training in medical imaging, neuroscience, clinical diagnosis of AD, and biostatistics. The proposed training and research plan will foster the candidate's development into an independent scientist, using neuroimaging and image analysis methods for early detection and diagnosis of Alzheimer's disease. RELEVANCE: The improved neuroimage classification methods will help early detection and diagnosis of Alzheimer's disease. The release of the fully automatic neuroimage classification software will significantly improve the interoperability and adoptability of high dimensional pattern classification algorithms for neuroimage analysis, and result in enhanced dissemination, adoption, and evolution of such tools and resources by the broader neuroimaging research community.