Comparison of fMRI Analytic tools

fMRI 分析工具比较

• 批准号: 6663895
• 负责人: TERRENCE R OAKES
• 金额: $ 20.37万
• 依托单位: UNIVERSITY OF WISCONSIN-MADISON
• 依托单位国家: 美国
• 财政年份: 2002
• 资助国家: 美国
• 起止时间: 2002-09-25 至 2005-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6663895
• 关键词: adult human (21+); bioimaging /biomedical imaging; brain imaging /visualization /scanning; clinical research; computer data analysis; computer human interaction; computer program /software; computer system design /evaluation; functional magnetic resonance imaging; human subject; image processing; mathematical model

DESCRIPTION (provided by applicant): There are a variety of software packages available for analysis of functional MRI (fMRI) data. Widespread use of a thoroughly tested package is advantageous not only because of the ready availability to research labs without extensive computer programming resources, but also is a vehicle to facilitate comparisons and reproducibility among different research sites. We have experience with the following packages in our lab: AFNI, BrainVoyager, MEDx, SPM99, and VoxBo. Each of these packages is used by several laboratories around the world, and is generally accepted by the research community. Furthermore, each of these packages has a similar general approach to fMRI analysis, consisting of pre-processing (motion correction, coregistration to a template), parametric voxelwise comparison with a time-varying activation paradigm, definition of a statistical model, and interrogation and display of statistically meaningful results. However, the implementation of these basic steps varies considerably from one package to another, and many users find it difficult to select the best package for a given application based on criteria other than simply having started out using one package and becoming familiar with it. Each package has a different approach to preprocessing, specifying and estimating a statistical model, corrections for multiple comparisons, and to interrogation of intermediate and final results. Other considerations include the ease of learning and using a package, the availability and complexity of a Graphical User Interface (GUI), the ability to employ scripts to batch-process large analysis efforts, and the computational time required to perform the various analytic steps. Of course, the most important criteria is the accuracy, sensitivity, and specificity of the results with respect to a range of BOLD signal changes; while each of the above packages has an acceptable level of accuracy and sensitivity, how these levels compare across the packages has not been thoroughly investigated. We propose to perform a series of well-characterized fMRI scans, representative of research efforts in a variety of neuroimaging applications. We will use these scans to investigate the utility and effectiveness of five common fMRI analysis packages. We will also construct several series of synthetic data with simulated activations of varying magnitude and extent, which we will use to quantify the accuracy of motion correction and spatial normalization, and the ability of each package to detect small activations. Given the large amount of energy expended on neuroimaging experiments and the importance of knowledge to be gained, it is vital for researchers to select the best tools to reduce and analyze these complex data.

描述（由申请人提供）：有多种软件包可用于分析功能MRI （fMRI）数据。广泛使用经过彻底测试的软件包是有利的，不仅因为无需大量计算机编程资源就可以随时获得研究实验室，而且还可以促进不同研究地点之间的比较和可重复性。我们在实验室中有以下包的经验：AFNI， BrainVoyager, MEDx， SPM99和VoxBo。这些软件包中的每一个都被世界各地的几个实验室使用，并被研究界普遍接受。此外，这些软件包中的每一个都有类似于fMRI分析的通用方法，包括预处理（运动校正，对模板的共配准），与时变激活范式的参数体素比较，统计模型的定义，以及统计有意义的结果的询问和显示。然而，这些基本步骤的实现因包而异，许多用户发现很难根据标准为给定的应用程序选择最佳包，而不是简单地开始使用一个包并熟悉它。每个包都有不同的方法来进行预处理、指定和估计统计模型、多次比较的修正以及中间和最终结果的询问。其他考虑因素包括学习和使用包的容易程度、图形用户界面（GUI）的可用性和复杂性、使用脚本批量处理大型分析工作的能力，以及执行各种分析步骤所需的计算时间。当然，最重要的标准是相对于一系列BOLD信号变化的结果的准确性、灵敏度和特异性；虽然上述每个软件包都具有可接受的准确性和灵敏度水平，但这些水平如何在软件包之间进行比较还没有得到彻底的研究。我们建议进行一系列具有良好特征的功能磁共振成像扫描，代表各种神经成像应用的研究努力。我们将使用这些扫描来研究五种常见的fMRI分析包的效用和有效性。我们还将构建一系列具有不同大小和程度的模拟激活的合成数据，我们将使用这些数据来量化运动校正和空间归一化的准确性，以及每个包检测小激活的能力。考虑到神经影像学实验所耗费的大量精力和所获得知识的重要性，研究人员选择最好的工具来减少和分析这些复杂的数据是至关重要的。