1. NTR Case Study Project Archival storage of non-DICOM and non-image data from the NTR in the National Cancer Imaging Archive Siemens Corporate Research Bob Schwanke, John Pearson, Fabian Moerchen, Philipp Hussels March, 2009

2. NTR (Network for Translational Research) NTR is the largest imaging translational research group funded by the NCI CIP (20 universities & companies) NTR validates clinical utility of new emerging imaging methods to be used in multi-center clinical trials NTR now entering the 2 nd 5-year grant period Years 1-5 focused on optical imaging (NTR-OI) Years 6-10 generalize this beyond optical But translation imaging theme remains The new QIN grant from NCI CIP is organized like NTR (QIN – Quantitative Imaging Network, PAR 08-225)

3. Management of Experimental Data by the NTR-OI with SciPort Current experimental data management and collaboration system (SciPort) used by the Network for Translational Research, Optical Imaging (NTR-OI) Diverse categories of data. Diverse file types. Much of it non-standard & constantly evolving with the imaging innovations

4. Purpose of this “NTR Case Study” Project This project is to “build two-way bridges” between the informatics efforts of NTR and caBIG NTR gains access to caBIG tools (as usual) XIP, AIM, AVT, caGRID,… caBIG gains access to NTR data (as usual) E.g., Optical/MR fused data for monitoring cancer therapy But caBIG also adapts useful tools developed by NTR Imaging data management (SciPort); Image fusion/analysis (OMIRAD) First Step Publish NTR data directly to NCIA from SciPort This enables the NTR to share data beyond their consortium, and provides archival storage Utilize the CTP “ZipObject” import scheme of NCIA

5. SciPort Background Introduction

6. SciPort workflow (1/5) Define document templates User specifies names and types of data stored in a document. Documents can store text, numbers, dates, lists, tables with user defined constraints. User specifies types and number of additional files that can be attached.

7. SciPort workflow (2/5) Define a folder hierarchy Folders store several documents created with the same template. Folders can be assigned to a specific group of users.

8. SciPort workflow (3/5) Create documents Documents can be created and edited with input masks generated for each template. Files can be uploaded and attached to documents. DICOM images can be anonymized and previews are generated. Ontologies guide user in using common vocabulary. Each user can tag documents with personal keywords.

9. SciPort workflow (5/5) Searching documents The hierarchy of folders can be browsed for documents. Documents can be search specifying with keywords and fields in document template. Search history Keyword search Boolean operators Common fields Hierarchy search Schema based searchAdd fields for predicatesRemove fields from predicates Schema field predicates

10. SciPort workflow (5/5) Collaboration Templates and documents can be shared via a central server. Documents from other servers can be searched and downloaded. Central Server Local Server Structured Documents and Schemas Published Documents Linked Data Files Upon Request Local Server User

11. SciPort Technology SQL/XML data storage Efficient and scalable storage in IBM DB2 relational database with pureXML technology for XML indexing. Highly customizable modeling of scientific data with arbitrarily nested structures. Web 2.0 application Highly interactive AJAX web based user interface. Based on scalable web technologies (Tomcat, JSP, web services) Medical imaging support DICOM images are automatically anonymized and previews are generated. HIPAA: anonymization, encrypted passwords and secure protocols.

12. SciPort – NCIA bridge (focus of current work)

13. NTR-NCIA Bridge (1/3) SciPort document and attached files are stored in NCIA using CTP Generate manifest.xml file with meta-data. Generate dummy DICOM image with patient information. Add non-DICOM and non-image data to ZIP object. Send to Clinical Trial Processor (CTP). Store in NCIA database. DB2 MySQL CTP

14. NTR-NCIA Bridge (2/3) NTROI SciPort Database NCIA Data Model Mapping Metadata Publishing Image Data Publishing Image Data Publishing NTROI –NCIA Bridge

15. NTR-NCIA Bridge (3/3) ZIP Object: Create empty ZIP Object. Add SciPort XML Document to ZIP Object. Add non-DICOM image attachments to ZIP Object. Generate valid DICOM identifiers and write to manifest.xml file: StudyInstanceUID, SeriesInstanceUID, SOPInstanceUID, PatientID Add manifest file to ZIP Object. Send ZIP Object to CTP using HTTP Post Method. Wrapper DICOM Object: Create DICOM Object with IDs matching the manifest.xml file Add tags required by DICOM Standard and CTP/NCIA, e.g. “Modality” Send DICOM Oject to CTP using HTTP Post Method.

16. Example Define document template in SciPort –Breast cancer data from BLI at UCI –Category for patient information –Category with table Lesions –MR scan as DICOM data –Non-DICOM optical data

17. Example Data entry SciPort generates forms according to the customized data model. Table rows can be added for lesions.

18. Example Attach DICOM and Non-DICOM images Files can be uploaded as ZIP files Previews are generated.

19. Example Publish experimental data from SciPort to CTP/NCIA –Log into SciPort. –Open document in SciPort. –Select Publish to NCIA. –Data is sent to CTP.

20. Data is received, checked, and saved to NCIA by CTP Login to CTP. Select Status. DICOM files are imported/exported. Example

21. Curate incoming data in NCIA to make it visible –Login to NCIA and select the QC Tool –Select uploaded image. –Change QC status to Visible and Submit.

22. Example Search data in NCIA –Select OT (Other) as the modality and press Submit.

23. Example Download data from NCIA –Put into basket and select View my basket. –Select download all items.

24. Example ZIP file contains the wrapper DICOM image and all attached data

25. Next steps Introduction

26. Summary to this Point Problem: –Translational Researchers need to share non-standard data from emerging imaging systems, e.g., optical. –But NCIA is tailored to DICOM images with annotations. Solution: –Develop metadata models of all data (DICOM and other) with SciPort –Upload data into NCIA as ZIP objects. Observations –Our generic approach can serve as a model for other translational research networks with non conventional data formats. –SciPort was used as it was already in use by the NTROI and is very general, so convenient for doing a proof-of- concept study –But SciPort is not open-source at present… Conclusion

27. Next steps 1.Create open-source bi-directional NCIA bridge –SciPort is not currently open source, so can’t be use “as is” in caBIG –SciPort-NCIA-bridge covers only upload, not download –Create an open source version of the tools demonstrated thus far that would enable using the NCIA as a platform for sharing non-DICOM data: Parse, package, and batch upload typical NTR data to NCIA. Batch-download and unpack data for researchers. 2.Extend project to the new NTR partners –New NTR has replaced old NTROI, only Stanford remains. –Many new partners could benefit from the technology. –Fred Prior leads informatics working group of new NTR 3.Demonstrate large-scale feasibility by upload UCI data to NCIA –non-DICOM optical data on breast cancer. –Data should be parsed, unified and uploaded to the NCIA. Conclusion