Pathology Image Informatics Platform for visualization, analysis and management

 描述(申请人提供)：随着整个幻灯片数字扫描仪的出现，组织病理切片可以数字化为非常高分辨率的数字图像，实现了一种新的“大数据”流，在大小和复杂性上可能与“组学数据”相媲美。就像分析高通量的基因和表达数据一样，复杂的图像分析工具和数据管道的应用可以将数字病理(DP)档案中通常被动的数据转化为强大的来源：(A)丰富的癌症生物学定量见解和(B)精确医学的配套诊断决策支持工具。数字病理学支持的伴随诊断测试可以以类似于分子诊断测试的方式预测癌症风险和侵袭性。然而，在广泛采用DP之前，需要对DP成像的临床解释(DPI)和伴随的决策支持工具进行广泛的评估。由于缺乏易于查看、管理和定量分析DPI的公开可用的开放获取图像信息学平台，DPI被癌症社区(临床和研究)更广泛地接受。虽然存在一些用于查看和分析DPI数据的商业平台，但这些平台都不是免费的。面向放射学(例如XNAT)和计算生物学社区的开源图像查看/管理平台通常不利于处理DPI数据集所遇到的非常大的文件大小。这份多PI U24提案旨在扩展现有的、可免费获得的病理图像查看器(Sedeen Image Viewer)，以创建一个用于管理、注释、共享和定量分析DPI数据的病理信息平台(PIIP)。Sedeen被设计为DPI的通用平台(通过解决几种专有扫描仪格式和“大数据”挑战)，以提供(1)可靠和有用的图像注释工具，以及(2)DPI数据的图像配准和分析。此外，Sedeen已经成为一个用于裁剪大型DPI的应用程序，以便可以将它们输入到MatLab或ImageJ等程序中。Sedeen已经向公众免费开放了三年，拥有来自20多个国家的160多个独立用户。在最初的基础上 对于Sedeen及其现有用户群的成功，我们的目的是在癌症研究社区和临床试验工作中大量增加DPI和算法的传播，并为采用一套合理和标准化的DP操作约定做出贡献。这个独特的项目将允许具有不同需求和技术背景的终端用户无缝地(A)存档和管理、(B)共享和(C)可视化从不同站点、格式和平台获得的DPI数据。PIIP将为第三方算法(核分割、颜色归一化、生物标记物量化、放射学-病理学融合)提供统一的用户界面，并将允许对来自多个来源站点的数据进行算法评估。通过与专业协会的合作，我们设想PIIP的用户群将扩大到包括肿瘤学、病理学、放射学和制药界。

 DESCRIPTION (provided by applicant): With the advent of whole slide digital scanners, histopathology slides can be digitized into very high-resolution digital images, realizing a new "big data" stream that can potentially rival "omics data" in size and complexity. Just as with the analysis of high-throughput genetic and expression data, the application of sophisticated image analytic tools and data pipelines can render the often passive data of digital pathology (DP) archives into a powerful source for: (a) rich quantitative insights into cancer biology and (b) companion diagnostic decision support tools for precision medicine. Digital pathology enabled companion diagnostic tests could yield predictions of cancer risk and aggressiveness in a manner similar to molecular diagnostic tests. However, prior to widespread clinical adoption of DP, extensive evaluation of clinical interpretation of DP imaging (DPI) and accompanying decision support tools needs to be undertaken. Wider acceptance of DPI by the cancer community (clinical and research) is hampered by lack of a publicly available, open access image informatics platform for easily viewing, managing, and quantitatively analyzing DPIs. While some commercial platforms exist for viewing and analyzing DPI data, none of these platforms are freely available. Open source image viewing/management platforms that cater to the radiology (e.g. XNAT) and computational biology communities are typically not conducive to handling very large file sizes as encountered with DPI datasets. This multi-PI U24 proposal seeks to expand on an existing, freely available pathology image viewer (Sedeen Image Viewer) to create a pathology informatics platform (PIIP) for managing, annotating, sharing, and quantitatively analyzing DPI data. Sedeen was designed as a universal platform for DPI (by addressing several proprietary scanner formats and "big data" challenges), to provide (1) reliable and useful image annotation tools, and (2) for image registration and analysis of DPI data. Additionally, Sedeen has become an application for cropping large DPIs so that they can be input into programs such as Matlab or ImageJ. Sedeen has been freely available to the public for three years, with over 160 unique users from over 20 countries. Building on the initial successes of Sedeen and its existing user base, our intent is to massively increase dissemination of DPI and algorithms in the cancer research community and clinical trial efforts, as well as to contribute towards the adoption of a rational and standardized set of DP operational conventions. This unique project will allow end users with different needs and technical backgrounds to seamlessly (a) archive and manage, (b) share, and (c) visualize their DPI data, acquired from different sites, formats, and platforms. The PIIP will provide a unified user interface for third party algorithms (nuclear segmentation, color normalization, biomarker quantification, radiology-pathology fusion) and will allow for algorithmic evaluation upon data arising from a plurality of source sites. By partnering with professional societies, we envision that the PIIP user base will expand to include the oncology, pathology, radiology, and pharmaceutical communities.