TR&D 3: Advanced Statistical Methods for Functional MRI

3.第三章 功能磁共振成像的先进统计方法。 主要研究者： James J. Pekar博士，放射学副教授 布莱恩·S Caffo博士，生物统计学教授 总结 大脑作为分布式神经网络的进化集合体的生物学描述是 将功能连接的影像学测量应用于临床研究的意义。我们的协作 项目使用血氧水平依赖的功能性MRI（BOLD fMRI）来评估大脑的变化， 自闭症、多动症、阿尔茨海默病、多发性硬化症、精神分裂症、原发性进行性痴呆、 失语症和亨廷顿氏病，寻求开发基于非侵入性成像的生物标志物， 揭示疾病机制，改善诊断和预后，并评估治疗干预措施。他们的 研究受到BOLD fMRI采集的灵敏度和特异性的限制。这个项目的首要目标是 TR&D将与我们的合作者合作，以增强其功能连接的敏感性和特异性 通过开发新的经验贝叶斯分析方法来衡量， 这些转换极大地改善了fMRI数据的获取和可用性，即 同步多层（SMS）MRI，以及大型公共数据集的可用性。因此，我们有 制定了三个具体目标：1。开发自回归建模的时不变方法，以及 优化他们的短信功能磁共振成像数据。2.开发时间不变的方法来进行滋扰回归，以及 优化他们的短信功能磁共振成像数据。3.设计，实施和评估经验贝叶斯方法， 将来自大型公共数据库的信息与从单个受试者/小样本获得的数据相结合 问题研究

TR&D 3. Advanced Statistical Methods for Functional MRI. Principle Investigators: James J. Pekar, PhD., Associate Professor of Radiology Brian S. Caffo, Ph.D., Professor of Biostatistics SUMMARY The biological description of the brain as an evolved ensemble of distributed neural networks underlies the significance of applying imaging measures of functional connectivity to clinical research. Our collaborative projects use blood oxygenation level dependent functional MRI (BOLD fMRI) to assess changes in brain networks in autism, ADHD, Alzheimer's disease, multiple sclerosis, schizophrenia, primary progressive aphasia, and Huntington's disease, seeking to develop noninvasive imaging-based biomarkers in order to reveal disease mechanisms, improve diagnosis and prognosis, and assess therapeutic interventions. Their studies are limited by the sensitivity and specificity of BOLD fMRI acquisitions. The overarching goal of this TR&D is to work with our collaborators to enhance the sensitivity and specificity of their functional connectivity measures by developing novel empirical Bayesian analysis approaches that exploit two ongoing transformations that are dramatically improving the acquisition and availability of fMRI data, namely simultaneous multi-slice (SMS) MRI, and the availability of large public datasets. Accordingly, we have developed three specific aims: 1. To develop time-invariant approaches to autoregressive modeling, and optimize them for SMS fMRI data. 2. To develop time-invariant approaches to nuisance regression, and optimize them for SMS fMRI data. 3. To design, implement, and assess empirical Bayesian methods for combining information from large public databases with data obtained from single subject/small sample studies.