From Terabytes of Pixels to Intuitive Brain Networks (Administrative Supplement)

 描述(由申请人提供)：在连接论时代思潮中，显微镜图像以前所未有的速度被收集。收集的数据量如此之大，对于研究人员来说，提取嵌入其中的神经网络的组织信息是一个挑战。开发为研究人员提供感兴趣连接的直观视觉表示的软件，将极大地帮助他们生成关于神经网络功能重要性的可测试假说--首先是创造哺乳动物连接体的动力。开发这样一个可视化系统的首要和最初的挑战是彻底和可靠地量化海量图像数据的艰巨任务。在没有大量计算辅助的情况下实现这一点是困难的。尽管存在用于配准图像并自动重建用于数据分析的神经元过程和轴突路径的算法，但由于它们的处理时间较长，因此对于连接大数据来说效率不高。利用南加州大学鼠标连接组项目(MCP)的数据，我们开发了连接透镜的测试版，这是一种创新的信息管道，用于高效和方便地扭曲、分割和量化连接数据。我们已经成功地将Connection Lens应用于有限的显微镜图像数据集，并建议将其功能扩展到处理我们的整个档案。此外，利用我们的Connection Lens量化数据，我们建议开发一个补充的可视化Web应用程序。它被称为投影透镜，它将以连接图、邻接矩阵、网络图和平面图的形式呈现用户指定的感兴趣的连接的可发布可视化。与路线图类似，该程序将配备为显示两个感兴趣区域之间的所有可能路线，并说明故障节点将如何影响网络中的整体信息流。这些功能将使研究人员能够快速浏览、理解和发布有关功能不同的神经网络的基本发现，从而最大限度地利用嵌套在TB级显微镜扫描中的连接学数据。投影透镜的基于网络的界面将使世界各地的科学家能够轻松访问。此外，为连接/投影透镜(C/PL)框架开发的代码将在网上免费发布，并通过开放源码许可证发布，使其他实验室能够量化和可视化其数据。

 DESCRIPTION (provided by applicant): In the midst of a connectomics zeitgeist, microscopy images are collected at an unprecedented rate. The amount of data collected is so overwhelming that it is a challenge for researchers to extract the organizational information of the neural networks embedded within. Developing software that provides researchers intuitive visual representations of their connections of interest would greatly aid them to generate testable hypotheses regarding the functional significance of neural networks -- the impetus for creating the mammalian connectome in the first place. The primary and initial challenge toward developing such a visualization system is the daunting task of thoroughly and reliably quantifying the enormous amount of image data. Achieving this without significant computational aid is intractable. Although algorithms for registering images and automatically reconstructing neuronal processes and axonal pathways for analysis of data exist, they are not efficient for connectivity Big Data given their protracted processing times. Utilizing the data fro the Mouse Connectome Project (MCP) at USC, we developed a beta version of Connection Lens, an innovative informatics pipeline for efficiently and expediently warping, segmenting, and quantifying connectivity data. We have successfully applied Connection Lens toward a limited set of our microscopy image data and we propose to extend its functionality to process our entire archive. Furthermore, leveraging our Connection Lens quantified data we propose to develop a complementary visualization web application. Called Projection Lens, it will render publishable visualizations of user specified connections of interest as connectivity maps, adjacency matrices, network graphs, and flatmaps. Similar to a roadmap, the program will be equipped to show all possible routes between two regions of interest and illustrate how a dysfunctioning node will affect overall information flow within the network. These features will empower researchers to quickly browse, comprehend, and publish fundamental findings regarding functionally distinct neural networks, thereby maximizing the utility of connectomics data nested in terabytes of microscopy scans. The web-based interface of Projection Lens will grant easy access to scientists world-wide. In addition, the code developed for the Connection/Projection Lens (C/PL) framework will be published freely online, and released via an open source license enabling other laboratories to quantify and visualize their data.