Center for Reproducible Neuroimaging Computation (CRNC)

 在过去的二十年中，出现了一个巨大的技术，计算和社会基础设施，并改变了如何收集信息和收集科学各个方面的知识。在医学界，响应众多NIH数据共享倡议和任务，以及通过基层的努力，社区已经成功地积累了广泛的共享数据。在科学过程中，方法应该是可重复的。在科学经济学中，数据和方法应该最大限度地重复使用，以最大限度地提高数据采集投资的科学回报。数据和分析处理方法的重用已经成为一个焦点，在一个日益关注的可复制性和权力，今天的许多科学研究。这种可重复性问题的严重性表明，我们如何从越来越多的公共和私人神经成像知识库中生成和报告知识，可能是一种范式转变。这些因素阻碍了科学发现，并最终损害了所有利益相关者，包括研究人员本身，他们的同行和同事，他们的机构和资助机构。我们提出的BTRC资源，可再现神经成像计算中心（CRNC），旨在实现神经成像研究方式的转变。通过开发支持基础研究和临床活动的一套全面的数据管理，分析和利用框架的技术，我们的总体目标是提高神经影像学科学的可重复性，并扩大我们在神经影像学研究中的国家投资的价值。重现性至关重要，因为目前的文献包含大量错误的结论（由于效力有限、假阳性、出版偏倚和偶尔的错误）。在神经影像学研究中，很难区分假阳性和真阳性结果，因为数据很难汇总，准确的方法也很难复制。为了以可重复的方式推进分析和出版领域，整个中心将有以下目标：A）交付一个可复制的分析系统，该系统由包括数据和软件发现的组件组成（TR&D 1），实施标准化的工作流程描述，开发机器可读的标记，并存储这些工作流程的结果（TR&D 2）开发执行选项，促进在多种计算环境中的操作，并减少规模和可靠性的障碍（TR&D 3）; B）与协作者和服务用户的社区合作，部署，通过从软件开发人员到应用科学家的各种用例测试和验证可重现的分析系统，这些用例支持原始数据以及衍生结果的存档和重复使用，以促进多个不同应用领域的可重复临床研究（及其出版物）;以及C）向社区提供培训和教育，以促进神经影像学研究中可重复框架的持续使用和开发。

 DESCRIPTION (provided by applicant): Over the last two decades a vast technological, computational and societal infrastructure has emerged and transformed how information is collected and knowledge is gathered in all facets of science. Within the medical community, in response to numerous NIH data sharing initiatives and mandates, as well as through grassroots efforts, the community has succeeded in accumulating an extensive array of shared data. In the scientific process, methods should be reproducible. In the economics of science, data and methods should be maximally reusable in order to maximize the scientific return on the data acquisition investment. Data and analytic processing methods reuse have become a focal point in a growing concern about the replicability and power of many of today's scientific studies. The magnitude of this reproducibility issue indicates that a paradigm shift may be in order as to how we generate and report knowledge from our mounting public and private neuroimaging repositories. These factors impede scientific discovery and is ultimately a disservice to all the stakeholders, including the investigators themselves, their peers and colleagues, their institution, and their funding agencies. Our proposed BTRC resource, the Center for Reproducible Neuroimaging Computation (CRNC), seeks to implement a shift in the way neuroimaging research is performed. Through the development of technology that supports a comprehensive set of data management, analysis and utilization frameworks in support of both basic research and clinical activities, our overarching goal is to improve the reproducibility of neuroimaging science and extend the value of our national investment in neuroimaging research. Reproducibility is critical because the current literature contains large numbers of erroneous conclusions (due to limited power, false positive, publication bias and occasionally mistakes). Given a neuroimaging study, it is exceedingly difficult to discern between false positive and true positive findings as data is hard to aggregate, and exact methods are hard to replicate. In order to advance the field in terms of analysis and publication in a way that embraces reproducibility, the overall Center will have the following aims: A) Deliver a reproducible analysis system comprised of components that include data and software discovery (TR&D 1), implementation of standardized workflow description and development of machine-readable markup and storage of the results of these workflows (TR&D 2) and development of execution options that facilitates operation in multiple computational environments and reduces barriers to scale and reliability (TR&D 3); B) Working with a community of collaborator and service users, deploy, test and validate the reproducible analysis system with a wide variety of use cases ranging from software developers to applied scientists that support the archiving and reuse of raw data and the archival and reuse of derived results to promote reproducible clinical research (and its publication) in multiple different application areas; and C) Provide training and education to the community to foster continued use and development of the reproducible framework in neuroimaging research.