1. XIP In-Vivo Imaging Workspace Software SIG February 7, 2007

2. The eXtensible Imaging Platform (XIP) is an open source environment for rapidly developing medical imaging applications from an extensible set of modular elements. Researchers will be able to easily develop and evaluate new approaches to medical imaging problems, and use them in a translational research setting. caGrid makes it possible to develop an XIP architecture that allows users to choose between remotely hosted grid-based components and data sources as well as locally available components and sources. Components may include analytic services, e.g. CAD algorithms, algorithms for quantifying changes in consecutive imaging studies, algorithms associated with a 3-D visualization pipeline etc) Available data sources include NCIA DICOM data repositories Local databases What is ?

3. Why ? One of the goals of the In Vivo Imaging Workspace is to “focus on identifying the ways in which the wealth of information provided by … imaging, performed at academic and other research centers across the country, can be shared, optimized, and most effectively integrated into the ongoing effort to relieve suffering and death from cancer.” To facilitate the increasing use of imaging based end points in clinical trials, the Workspace has identified the need for an easily extensible open source platform to support image analysis and visualization. This platform will make it easier and less expensive to access specific post-processing applications at multiple sites, simplifying clinical trials, and most importantly, increasing the uniformity of imaging and analysis. Imaging applications developed by research groups will more easily be accessible within the clinical operating environment, simplifying workflows and speeding data processing and analysis. Once validated, the software should be readily transitioned into products through streamlined Federal Drug Administration, (FDA), approval processes due to the re-use of already approved libraries and open source development processes.

4. XIP Use Cases Four Use Cases defined by the caBIG IVI Workspace, Software SIG (F. Prior, B. Erickson), in the XIP Requirements Document Imaging as a bio-marker for drug therapy trials with centralized data analysis within a Cooperative Group: NCCTG Imaging as a bio-marker for research and drug therapy trials using distributed analysis within a Cooperative Group: COG, NANT Standardizing and translating emerging imaging methods in a Translational Research Network: NTROI Annotation of tumors as part of curation process for an image archive to be used for algorithm development: NCIA These Use Cases illustrate the spectrum of uses of XIP – the Software SIG will guide the prioritization and roll-out of features to meet the needs of such representative users Clinical Translation 

5. Imaging as a Bio-marker, Classical Phase 2/3 (NCCTG) Use Case: Trial to assess effect on brain tumor of drug with & without radiation Brain MRI at 2 mo. intervals, collected in central review facility Functionality needed in XIP Imaging Application: Query DICOM worklist created for each rater Visualize baseline (pre-gad, post-gad, T2, FLAIR) and current scan Image registration (rigid body) to improve accuracy and reliability Presentation with linked cursors and multi-planar reformats Measurement of tumor size based on margin outlines (T2 and/or FLAIR), both RECIST and volume Controlled order of case presentation to reduce bias Quickly find cases with significant inter-reader differences for adjudication

6. Viewer for assessing temporal change Imaging as a Bio-marker, Classical Phase 2/3 (NCCTG)

7. Use Case: Evaluation of cases of Peripheral Neuroblastic Tumors (PNTs), integrating radiological and pathological images, patient demographics, … Emphasis on MR perfusion analysis as bio-marker for tumor growth rate Functionality needed in XIP Imaging Application: “Virtual Workbench” for PNT research based on original data, derived results, annotations, and mark-ups of PNT data based on the Cooperative Group Grid resources Interactive and integrated radiological and pathological image analysis of PNT’s Publishing of analysis results to Grid Storage systems Querying of grid-based data systems for discovery of outcome correlations Imaging as a Bio-marker in Cooperative Group Trials (COG, NANT)

8. DSC-MR perfusion of intrinsic brainstem (pontine) glioma compared to a bithalamic glioma (left) and methodology Example of the kind of DCE-MR perfusion processing needed from XIP Figure courtesy of Stephan Erberich, USC.

9. Imaging as a Bio-marker in Cooperative Group Trials (COG, NANT) Figure courtesy of Stephan Erberich, USC. Example of the kind of Support for Distributed Data Storage Needed from XIP

10. Standardizing Imaging in a Translational Research Network (NTROI) Use Case Standardization and validation of optical imaging methods for breast cancer screening and therapy monitoring Starting point is collaborative validation of physiological measures of unique instruments at each university End point is trial using common, validated instrument Functionality Needed in XIP Imaging Application Multi-modal registration, visual fusion, tumor segmentation Measurement of tumor volume, optically-derived physiological parameters

11. GUI for Optical/MR Visualization and Analysis Standardizing Imaging in a Translational Research Network (NTROI)

12. Computations for Optical/MR Visualization and Analysis Standardizing Imaging in a Translational Research Network (NTROI)

13. Annotation of tumors for an image archive (NCIA) Use Case Remote and on-site expert 3D segmentation and rigorously defined annotation of tumors in National Cancer Image Archive Annotated images used both to develop and test CAD and other algorithms by academia and industry Functionality Needed in XIP Imaging Application Advanced thin-client 3D visualization and navigation tools for remote curation by domain experts Mark-up, segmentation, annotation and measurement of tumor volume using a variety of 2D and 3D metrics, using rigorously defined vocabularies

14. Annotation of tumors for an image archive (NCIA) Figure courtesy of John Freymann, NCI.

15. What is Included in XIP Rapid Application Development Tools and Libraries (RAD) Development and application build environment Extensive and extensible set of libraries for imaging and visualization Uses Open Inventor framework Includes code generating wizards to create new objects and wrap existing libraries XIP Workstation (WS) A reference implementation of a medical imaging workstation developed using XIP RAD Integrated via middleware into caGRID Optimized to support basic cancer research use cases Includes two key components: XIP Application – a use case specific “plug-in” application integrated via the DICOM WG-23 Interface XIP Host – the hosting environment that provides XIP Applications with access to services such as data stores, remote processing, etc.

16. XIP Framework & Architecture XIP Application (Can be replaced with any WG23-compatible Host) XIP Host Adapter API XIP LIBITKVTK... Host-Specific Plugin Libraries Remote Processing caGrid Analytical Svc. Data Access caGrid Data Svc. Annotation and Markup AIM MetaData... XIP Modules Host Independent WG23 XIP Development Tools XIP Host WG23 (Enables rapid development of applications)

17. Categories of Users XIP Host WG-23 API (Socket) WG-23 API (Plug) XIP Application CustomXIPClassesStandard XIP Library Classes XIP Application Developer XIP Library Developer XIP Host Developer XIP Application User IVI Middleware caGRID Interface DICOM/IHE Intefaces

18. Application Builder Open Inventor Scene Graphs Engines enable the creation of processing pipelines Nodes support the concept of scene graphs, which are hierarchical structures of objects describing what needs to be visualized in 2D/3D Manipulators handle input devices, measurements and coordinate transforms in response to user interaction

19. Library Builder Wrapped ITK functions include Region Growing, Neighborhood, Isolated, Confidence, Watershed, Thresholding, Edge Detection, Laplacian, Gaussian, … Support for ITK Data Meshes and Vector Fields Automatic Wrapper generation for 2D/3D libraries/toolkits such as ITK and VTK Example: ITK for image processing, segmentation, registration User can review parsed results and choose to support only the desired data types, hide some methods, exclude some classes, etc.

20. Host Functions Provides the infrastructure in which an XIP Application runs Authenticates user Manages installation, launching, and termination of XIP Applications Provides data and services to XIP Applications Accepts status information and results back from XIP Applications Deals with auditing and controls access to services and data Isolates the XIP application from the nature of databases, archives, networks, and possibly image data formats Manages caGRID interactions and security Manages access to DICOM networks, objects, and services Maps images and associated meta-data from various sources between their native form and a common form useable by the XIP application Handles workflow issues General Purpose Worklist support, following IHE pattern

21. DICOM WG23 Plug-in Framework WG-23 addresses clinical integration and vendor inter-operability by defining standardized “plugs” and “sockets” (APIs) caBIG XIP addresses an open-architecture, integrated development environment for rapid prototyping & collaboration based on WG 23 APIs. Unix, Mac, PCInternet ServerCommercial Vendor #2 … Commercial Vendor #1  Clinical   Prototype & Collaboration  XIP developed Application Standard API

22. Who is Contributing to ? The caBIG In Vivo Imaging Workspace, Software SIG Released the XIP RFP Provides primary feedback to the XIP development team Washington University in St. Louis, Electronic Radiology Lab Main coordinating site Actively involved in caBIG, DICOM, IHE, and serves as imaging core for several clinical trials Siemens Corporate Research (SCR) Contributing a suite of tools – ivRAD – that will form the basis for XIP Experienced in moving ideas from prototypes to commercial reality DICOM WG-23 Standardizing the interfaces between a hosting system (e.g. workstation) and hosted post-processing applications (a.k.a. “Plug-ins”) Representation from both vendors and user communities ITK/VTK community Providing image processing and visualization libraries with the assistance of Kitware

23. What does ivRAD bring to the table? Application development framework following the Open Inventor API Scene graphs coupled to processing pipelines and manipulators Well established, well documented, open source, free Easy to extend create custom Open Inventor-style objects create Open Inventor-style wrappers around existing libraries Extensions to support medical visualization and image processing Import of medical image data sets (e.g. DICOM, raw) Manipulation of medical image data (e.g. registration, fusion, segmentation) Multi-dimensional visualization (e.g. cine, MPR, MIP, Shaded-surface display, Volume Rendering) Tools for wrapping 3rd party class libraries into Open Inventor objects Used to incorporate the open source ITK (Insight ToolKit) into ivRAD Hides peculiarities of the underlying host system from the application Allows the same application to run stand-alone on MS Windows, or within various versions of Siemens’ medical workstations Since the interfaces that the application sees remain constant, the application is unaware of the differences in the underlying system

24. ivRAD to 1.Strip out Siemens-proprietary classes Siemens-proprietary visualization and processing functions Functions for Siemens-specific data sources Siemens ‘look-and-feel’ 2.Change copyright notices to support an open source license 3.Replace Siemens-proprietary classes with open-source classes Continued use of ITK (Insight ToolKit) Continued use of open source DICOM toolkits ( e.g. DCMTK or DCM4CHE ) Add visualization via VTK and other open-source classes 4.Modify the host – application interaction per the DICOM WG-23 APIs 5.Add support for new functionality caGRID via the IVI Middleware software Annotations and markup via the AIM project Additional platforms (e.g. Linux, MAC) Other data formats and functions requested in the RFP XIP ‘look-and-feel’

25. Open Inventor Open Inventor ® is an object-oriented 3D toolkit offering a comprehensive solution to interactive graphics programming problems. URL: http://oss.sgi.com/projects/inventor/ It presents a programming model based on the Model/View/Controller design pattern and the concept of Pipelines. Open Inventor: is built on top of OpenGL® defines a standard file format for 3D data interchange introduces a simple event model for 3D interaction provides animation objects called Engines provides high performance object picking is window system and platform independent is a cross-platform 3D graphics development system encourages programmers to create new customized objects

26. Open Inventor modules in XIP C++ modules represent Engines, Nodes and Manipulators Engines enable the creation of processing pipelines Nodes support the concept of scene graphs, which are hierarchical structures of objects describing what needs to be visualized in 2D/3D Manipulators handle input devices, measurements and coordinate transforms in response to user interaction

27. Flexible Application Deployment Standalone Application Web-based Application Commercial Workstation Development IDE

28. XIP modules extend Open Inventor to facilitate medical imaging application development: Database access (read/write) DICOM query/retrieve Image/Volume types Lookup tables Transfer function editor MPR intersection lines/manipulators camera control (pan/zoom/rotate) 2D Image display ROIs, Annotations, Measurements … Modules for DICOM loading and 2D/3D Display XIP Features Modules for DICOM loading and 2D/3D Display

29. Modules for Fused Volume Rendering XIP Features Modules for Fused Volume Rendering Features: Volumes are stored separately and fused at rendering time, not as a preprocessing step Support for unlimited number of fused large volumes Each volume has independent control of: Transformation (rot, trans, scale, shear … ) Color/opacity Transfer function Crop box Cut-planes Rendering mode (VRT, MIP, MinIP, DRR, SSD) Voxel Resolution Sampling rate Performance 20 frames/sec during interactivity 1 sec for final diagnostic-quality update

30.  Visual creation and configuration of distributed services Thin Client and Smart Client configurations are supported Support for caGRID remote grid computing services XIP modules for DICOM Query, Sorting, remote data transmission XIP can serialize the entire state to a file, thereby facilitating support for client/server state management and recovery as well as workflow management. Modules for Client/Server Remote Visualization XIP Features Modules for Client/Server Remote Visualization

31. Fused MPRs (screen shot of live demo)

32. Fused Volume Rendering (screen shot of live demo)

33. Animal Imaging Prototype (screen shot of live demo)

35. Lymphnode Segmentation (screen shot of live demo)

36. ROI Tools, Measurement and Filtering (screen shot of live demo)

37. 4D Beating Heart (screen shot of live demo)

38. ITK Demos: Level Set Networks (screen shot of live demo)

39. XIP Development Plan Phase 1 – Planning and RSNA/caBIG demonstrations (3 months) Planning done in parallel with RSNA/caBIG demonstration creation RSNA/caBIG demo done utilizing Siemens-internal proprietary development tools Source not released until phase 2. RSNA/caBIG demo serves as discussion point for generating use cases and requirements for ongoing development. Phase 2 – Conversion of base Siemens SW to open source (4 months) Largely utilizes the existing code pool, stripping out proprietary references and preparing the code for open source distribution Final output will be similar to, though not exactly the same as the RSNA/caBIG demonstration. May include some new additions, but may not fulfill all of the requirements listed in the SOW. Intermediate releases of documentation and code with partial feature sets during the course of conversion. Becomes the first prototype implementation Phase 3 – Addition of new features to XIP, creation of sample reference applications (3 months) Drafts of all documents available with all implemented features available at the end of this phase Intermediate releases available throughout the phase to foster discussion, review, and use. Prototype code with all implemented features at the end of this phase Phase 4 – Finalization of the year’s work products (2 months) No new features added after the start of this phase Bug fixes and documentation corrections as needed Final testing, review, and approvals

40. First Prototype Release – Early May Focused on the needs of the AIM Project Targeted toward continued internal development, not necessarily end users Will be available open source on the NCI’s gForge site for developers and the curious What is included: Loading and display of stacks of 2D DICOM images From the caGRID via the IVI Middleware From DICOM image archives From the local file system Processing by Open Inventor-wrapped ITK objects Graphics provided by stock Open Inventor objects Graphical medical image markup Simple measurement functions What is not included: The GUI-based XIP Application Builder (not till summer) The wrapper-generating toolkit for existing class libraries 3D and 4D Rendering and VTK integration Non-DICOM image formats Workflow, Security, Audit, etc.

41. XIP Development Team Washington University in St. Louis Lawrence Tarbox Jaroslaw Krych David Maffitt Steve Moore Fred Prior Others Other Consultants Siteman Cancer Center Kitware Collaborative Projects IVI Middleware Annotation and Image Markup NCIA MIRC Siemens Corporate Research Gianluca Paladini Thomas Moeller Daphne Yu Klaus Engel John Pearson Others NCI caBIG IVI David Kupferschmid Booz Allen Hamilton Paul Mulhern

42. An Open Platform for Cancer Research