Quantifying Brain Abnormality by Multimodality Neuroimage Analysis,

阿尔茨海默病（AD）仅在美国就影响总共530万人，使其成为第七大死亡原因，并且每年花费约1720亿美元。目前，AD诊断主要基于临床和心理测量评估。然而，只有在尸检报告有进行性痴呆病史的个体在特定脑区存在特征性神经炎性淀粉样斑块和神经原纤维缠结时，诊断才能确定。因此，有一个重要的 对病理学的非侵入性客观诊断和量化以及疾病进展的一般评估的未满足的需求。该项目的目标是开发一种新的神经成像分析框架，该框架将利用来自不同成像模式的互补信息来有效量化疾病引起的病理，从而促进早期检测以进行可能的治疗和预防。实现这一目标需要在神经图像分析技术方面进行重大创新，以检测复杂而微妙的大脑变化模式。因此，该项目的具体目标是（目标1：疾病诊断）开发一种多模态多变量诊断技术，用于准确识别患有以下疾病的个体 有患AD的风险，（目标2：进展监测）设计一种新的多任务内核学习框架，用于预测和量化不同疾病阶段的大脑异常，以及（目标3：评价），以评估开发的方法，使用老年受试者的大型数据库，其在定量AD/MCI患者脑改变模式中的诊断能力，其对处于AD风险中的MCI患者的预测能力，以及它们在疾病进展时量化异常的能力。我们期望，在成功完成该项目后，由此产生的全面、综合和有效的诊断/监测框架将有助于提高MCI/AD以及其他神经系统疾病（包括精神分裂症、自闭症和多发性硬化症）的早期检测成功率。公共卫生相关性声明：在临床病理学出现之前，AD经历前驱期，持续数年至数十年，具有临床上无法检测或不确定的疾病病理或易感性。因此，如果疾病改善治疗是有效的，识别AD风险的个体是至关重要的。由于这个原因，在这个项目中开发的神经图像分析技术与公共卫生显著相关，因为它们将有助于提高患者识别和疾病监测的准确性， 有效治疗。 公共卫生相关性：本项目旨在开发一种基于个体的诊断方法，通过使用多模态成像和非成像数据来早期检测和监测脑部疾病的进展。这是显着不同的传统方法，重点是使用单一的成像模式或多模态数据的简单组合的脑部疾病的组比较。这些组比较方法不能诊断和预测单个患者的脑部疾病，尽管它们可能有助于在组水平上识别疾病对大脑结构和功能的影响。

DESCRIPTION (provided by applicant): Alzheimer's disease (AD) affects a total of 5.3 million individuals in the U.S. alone, making it the 7th leading cause of death and also costing about 172 billion dollars annually. Currently, AD diagnosis is predominantly based on clinical and psychometric assessment. However, diagnosis can only be certain if an autopsy reports the presence of characteristic neuritic ¿-amyloid plaques and neurofibrilatory tangles in specific brain regions in an individual with a history of progressive dementia. Thus, there is a significant unmet need for non-invasive objective diagnosis and quantification of pathologies, as well as general assessment of disease progression. The goal of this project is to develop a novel neuroimaging analysis framework that will harness the complementary information from different imaging modalities for effective quantification of disease -induced pathologies, so as to promote early detection for possible treatment and prophylaxis. Achieving this goal requires significant innovation in neuroimage analysis techniques to detect sophisticated yet subtle brain alteration patterns. Accordingly, the specific aims of this project are (Aim 1: Disease Diagnosis) to develop a multimodality multivariate diagnosis technique for accurate identification of individuals who are at risk for AD, (Aim 2: Progress Monitoring) to design a novel multi-task kernel learning framework for prediction and quantification of brain abnormality at various disease stages, and (Aim 3: Evaluation) to assess the developed methods using a large database of elderly subjects, for their diagnostic power in quantifying brain alteration patterns in AD/MCI patients, their predictive power of MCI patients who are at risk for AD, and also their capability in quantifying abnormalities as the disease progresses. We expect, upon successful completion of this project, that the resulting comprehensive, integrated, and effective diagnosis/monitoring framework will be conducive to improving the success of early detection of MCI/AD, as well as other neurological disorders including schizophrenia, autism, and multiple sclerosis. Public Health Relevance Statement: Prior to the appearance of clinical symptomatology, AD undergoes a prodromal phase, lasting from years to decades, with disease pathology or predisposition that is clinically undetectable or uncertain. Thus, identifying individuals who are t risk for AD is critical if disease-modifying treatments are to be effective. For this reason, the neuroimage analysis techniques developed in this project are significantly relevant to public health in that they will help improve accuracy in patient identification and disease monitoring for effective treatment. PUBLIC HEALTH RELEVANCE: Description of Project This project aims to develop an individual-based diagnosis method for early detection and progression monitoring of brain disease by using multimodality imaging and non-imaging data. This is significantly different from the conventional methods that focus on group comparison of brain disease using a single imaging modality or simple combination of multimodality data. These group comparison methods are not able to diagnose and predict brain disease for an individual patient, although they may help identify the effect of disease on brain structures and functions at a group level.