C-PAC: A configurable, compute-optimized, cloud-enabled neuroimaging analysis software for reproducible translational and comparative

摘要 大脑计划旨在利用复杂的神经调节，电生理记录， 和宏观神经成像技术在人类和非人类动物模型，以开发一个 对人脑功能的多层次理解。然而，组织、处理和 分析通过这些努力产生的神经成像数据并不像连贯的和容易的那样广泛可用， 使用软件包。差距对于非人类数据尤其明显（即，猴子，啮齿动物），因为大多数 现有的处理和分析软件包是专门为人体成像而设计的。方法 已经提出用于解决处理非人类数据所固有的挑战（例如，大脑 提取、组织分割、空间标准化、脑分割、时间去噪）;迄今为止，这些 还没有容易地集成到一个易于使用的，强大的，和可重复的分析包。同样地， 许多为神经成像分析开发的复杂机器学习和建模方法， 大多数研究人员无法使用，因为它们尚未集成到易于使用的管道软件中。作为 因此，翻译和比较神经成像研究人员将神经信息学管道拼凑在一起， 使用不同软件包和内部代码的各种组合。 我们建议扩展可配置管道用于连接组分析（C-PAC）开源 软件，为功能和结构MRI数据提供强大和可再现的管道。我们将整合 各种不同的图像处理和分析方法，用于处理非人类的挑战， 成像数据，整合到一个开源、可配置、易于使用的端到端分析管道包中， 可以在本地或通过云访问。最终产品将不仅提高质量，透明度， 非人类翻译和比较成像的可重复性，而且还使科学研究的新途径成为可能。 通过我们包含尚未应用于非人类成像数据的方法的查询（例如，梯度- 基于皮质包裹方法，超对齐）。拟议工作的具体目标包括： 集成针对BRAIN Initiative数据优化的神经成像处理和分析方法，2）实施 对人类和非人类群体进行比较研究的战略，以及3）将C-PAC扩展到 包括用于识别大脑功能机制的尖端分析策略。所有的发展将 使用GitHub和其他协作工具“公开”进行，以最大限度地参与C-PAC 项目将举办年度黑客松，与来自BRAIN Initiative奖项和其他奖项的调查人员合作。 神经信息学开发项目将其工具与C-PAC集成。将举行实践培训， 培训调查人员如何最佳使用新开发的工具。

ABSTRACT The BRAIN Initiative is designed to leverage sophisticated neuromodulation, electrophysiological recording, and macroscale neuroimaging techniques in human and non-human animal models in order to develop a multilevel understanding of human brain function. However, the necessary tools for organizing, processing and analyzing neuroimaging data generated through these efforts are not widely available as coherent and easy-to- use software packages. Gaps are particularly apparent for nonhuman data (i.e., monkey, rodent), as most of the existing processing and analytic software packages are specifically designed for human imaging. Methods have been proposed for addressing the challenges inherent to the processing of nonhuman data (e.g., brain extraction, tissue segmentation, spatial normalization, brain parcellation, temporal denoising); to date, these have not been readily integrated into an easy-to-use, robust, and reproducible analysis package. Similarly, many of the sophisticated machine learning and modeling methods developed for neuroimaging analyses are inaccessible to most researchers because they have not been integrated into easy-to-use pipeline software. As a result, translational and comparative neuroimaging researchers patch together neuroinformatics pipelines that use various combinations of disparate software packages and in-house code. We propose to extend the Configurable Pipeline for the Analysis of Connectomes (C-PAC) open-source software to provide robust and reproducible pipelines for functional and structural MRI data. We will integrate the various disparate image processing and analysis methods used to handle the challenges of nonhuman imaging data, into a single, open source, configurable, easy-to-use end-to-end analysis pipeline package that is accessible locally or via the cloud. The end product will not only improve the quality, transparency and reproducibility of nonhuman translational and comparative imaging, but also enable new avenues of scientific inquiry through our inclusion of methods that are yet to be applied to nonhuman imaging data (e.g., gradient- based cortical parcellation methods, hyperalignment). Specific aims of the proposed work include to: 1) Integrate neuroimaging processing and analysis methods optimized for BRAIN Initiative data, 2) Implement strategies for carrying out comparative studies of human and non-human populations, and 3) Extend C-PAC to include cutting-edge analytical strategies for identifying mechanisms of brain function. All development will occur “in the open” using GitHub and other collaborative tools to maximally involve participation in the C-PAC project. Annual hackathons will be held to collaborate with investigators from BRAIN Initiative awards and other neuroinformatics development projects to integrate their tools with C-PAC. Hands-on training will be held to train investigators on optimal use of the newly developed tools.