Developing a Univariate Neurodegeneration Imaging Biomarker with Optimal Transportation

项目总结/摘要 我们将开发和应用一种新的单变量神经退行性变成像生物标志物， 从阿尔茨海默病神经成像倡议（ADNI）数据集获得的共振图像（MRI）， 亚利桑那州APOE症状前人群的特征。我们最近的工作表明， 基于Wasserstein距离的脑成像指数优于其他几个单变量脑成像指数 在区分阿尔茨海默病（AD）患者与认知未受损（CU）受试者中， 断层脑MR和氟脱氧葡萄糖正电子发射断层扫描（FDG-PET）图像。在当前 项目，我们将继续开发新的结构磁共振成像分析方法的基础上谐波地图和 变分原理具体来说，我们将开发4D谐波映射算法来计算正则成像 并进一步计算基于4D Wasserstein距离单变量 纵向神经退行性变指标与有效的变分框架。拟设系统可 生成简单、客观和可靠的神经退行性变成像生物标志物，以量化渐进性 与AD相关的症状前解剖学变化，并提供简洁和信息丰富的单变量结果 随机临床试验（RCT）。为了研究我们的方法的可靠性和实用性，我们 将研究从ADNI和亚利桑那州APOE症状前队列中获得的大脑结构MRI扫描 科目我们试图（1）将计算机神经变性成像指数与纵向 ADNI和独立的亚利桑那州APOE队列的认知轨迹;（2）评估其能力， 通过区分β-淀粉样蛋白阳性轻度认知障碍（MCI）/CU受试者和 ADNI队列中β-淀粉样蛋白阴性MCI/CU受试者;（3）研究其预测 ADNI和年轻的CU受试者进展到遗忘型MCI临床阶段的速度 亚利桑那州APOE队列的症状前个体;（4）验证其促进 通过减少所需RCT样本量来评估AD治疗。我们将进行正面交锋 提出的单变量神经退行性疾病生物标志物与其他最先进的生物标志物之间的比较 单变量结构MRI指数与这些任务。我们还将开发和免费传播我们的软件 研究社区的工具。

PROJECT SUMMARY / ABSTRACT We will develop and apply a novel univariate neurodegeneration imaging biomarker to brain magnetic resonance images (MRI) obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset and the well-characterized Arizona APOE cohort of presymptomatic individuals. Our recent work has shown that Wasserstein distance-based brain imaging indices outperformed several other univariate brain imaging indices in discriminating Alzheimer’s disease (AD) patients from cognitively unimpaired (CU) subjects with cross- sectional brain MR and fluorodeoxyglucose positron emission tomography (FDG-PET) images. In the current project, we will continue developing novel structural MRI analysis methods based on harmonic maps and the variational principle. Specifically, we will develop 4D harmonic map algorithms to compute canonical imaging spaces of longitudinal brain images and further compute 4D Wasserstein distance-based univariate longitudinal neurodegeneration indices with an efficient variational framework. The proposed system will generate simple, objective, and reliable neurodegeneration imaging biomarkers to quantify progressive presymptomatic anatomical changes related to AD and provide concise and informative univariate outcome measures for randomized clinical trials (RCT). To investigate the reliability and practicality of our method, we will study brain structural MRI scans obtained from the ADNI and the Arizona APOE cohort of presymptomatic subjects. We seek to (1) correlate the computed neurodegeneration imaging indices with longitudinal cognitive trajectories in both ADNI and the independent Arizona APOE cohorts; (2) assess its ability to identify early AD by distinguishing beta-amyloid-positive mild cognitive impairment (MCI)/CU subjects from beta-amyloid-negative MCI/CU subjects in the ADNI cohort; (3) investigate its potential to predict progression rate to the clinical stage of amnestic MCI on CU subjects of the ADNI and the younger presymptomatic individuals of the Arizona APOE cohort; and (4) validate its potential to facilitate the evaluation of AD treatments by reducing the required RCT sample sizes. We will conduct head-to-head comparisons between the proposed univariate neurodegeneration biomarker and other state-of-the-art univariate structural MRI indices with these tasks. We will also develop and freely disseminate our software tools to the research community.