Kitware, Inc. Carrboro NC, USA.
Neuro Image Research and Analysis Laboratories, Department of Psychiatry, University of North Carolina Chapel Hill, NC, USA.
Front Neuroinform. 2013 Dec 30;7:46. doi: 10.3389/fninf.2013.00046. eCollection 2013.
Image processing is an important quantitative technique for neuroscience researchers, but difficult for those who lack experience in the field. In this paper we present a web-based platform that allows an expert to create a brain image processing pipeline, enabling execution of that pipeline even by those biomedical researchers with limited image processing knowledge. These tools are implemented as a plugin for Midas, an open-source toolkit for creating web based scientific data storage and processing platforms. Using this plugin, an image processing expert can construct a pipeline, create a web-based User Interface, manage jobs, and visualize intermediate results. Pipelines are executed on a grid computing platform using BatchMake and HTCondor. This represents a new capability for biomedical researchers and offers an innovative platform for scientific collaboration. Current tools work well, but can be inaccessible for those lacking image processing expertise. Using this plugin, researchers in collaboration with image processing experts can create workflows with reasonable default settings and streamlined user interfaces, and data can be processed easily from a lab environment without the need for a powerful desktop computer. This platform allows simplified troubleshooting, centralized maintenance, and easy data sharing with collaborators. These capabilities enable reproducible science by sharing datasets and processing pipelines between collaborators. In this paper, we present a description of this innovative Midas plugin, along with results obtained from building and executing several ITK based image processing workflows for diffusion weighted MRI (DW MRI) of rodent brain images, as well as recommendations for building automated image processing pipelines. Although the particular image processing pipelines developed were focused on rodent brain MRI, the presented plugin can be used to support any executable or script-based pipeline.
图像处理是神经科学研究人员的一项重要定量技术，但对于缺乏该领域经验的人来说却很困难。在本文中，我们提出了一个基于网络的平台，允许专家创建一个脑图像处理管道，使即使是那些图像处理知识有限的生物医学研究人员也能够执行该管道。这些工具作为 Midas 的插件实现，Midas 是一个用于创建基于网络的科学数据存储和处理平台的开源工具包。使用这个插件，图像处理专家可以构建一个管道，创建一个基于网络的用户界面，管理作业，并可视化中间结果。管道使用 BatchMake 和 HTCondor 在网格计算平台上执行。这为生物医学研究人员提供了一个新的能力，并为科学合作提供了一个创新的平台。目前的工具运行良好，但对于缺乏图像处理专业知识的人来说可能无法访问。使用这个插件，研究人员可以与图像处理专家合作，创建具有合理默认设置和简化用户界面的工作流程，并且可以轻松地从实验室环境中处理数据，而无需强大的桌面计算机。这个平台允许简化故障排除，集中维护，以及与合作者轻松共享数据。这些功能通过在合作者之间共享数据集和处理管道，实现了可重复的科学。在本文中，我们介绍了这个创新的 Midas 插件的描述，以及在构建和执行几个基于 ITK 的扩散加权磁共振成像（DW MRI）啮齿动物脑图像处理工作流程方面获得的结果，以及构建自动化图像处理管道的建议。虽然开发的特定图像处理管道专注于啮齿动物脑 MRI，但提出的插件可用于支持任何可执行或基于脚本的管道。
