1. Sharing Medical Images Your Way
The RSNA Image Share Network
Wyatt M. Tellis, Ph.D.
Informaticist
Laboratory for Radiological Informatics
Department of Radiology
& Biomedical Imaging

2. Project Overview
In Sept. 2009, RSNA was awarded two-year $4.7 million contract by the National Institute of Biomedical Imaging and Bioengineering (NIBIB)
Funding extended for four additional years
Explore use of open standards (IHE) for sharing images and reports
Phase One: Provide for patient control by enabling access via personal health record (PHR) accounts
Phase Two: Expand network to support direct transfer between sites for clinical and research images

3. Imaging and Information Sharing
Medical imaging is fasting growing physician service in US health system
Medicare data show annual utilization increases of about 9%, three times the rate of other physician services
Driven by growth of advanced imaging techniques such as MRI, CT, PET, Nuclear, and Cardio.
Value of medical imaging in providing accurate, noninvasive diagnosis is unquestioned
Costs of “overutilization” and excessive radiation exposure are growing concerns
Inaccessibility of prior studies is responsible for a significant number of duplicate studies

4. Current State of Image Sharing
Part of the health record most frequently provided to patients & physicians in digital form
No more film
Nearly all images are now in a standard digital format (DICOM)
Frequently distributed on portable media (CD, DVD)

5. Limitations of Physical Media
Lost or misplaced disks
Damaged or unreadable disks
Scratches
Disk wasn’t “closed” by burning software
Non-DICOM disks
A few vendors store images in proprietary format only viewable by their systems
Locked down PCs prevent viewing of images on disk
Wastes physician and patient time and is responsible for a significant number of duplicate studies
There must be a better way

6. To The Cloud… Electronic exchange means images everywhere at anytime
Growing market with a lot of companies, but no standards
Tower of babel problem: Numerous vendors with proprietary solutions that don’t talk to each other

7. Project Goals
Demonstrate a method for sharing images and reports that improves upon portable media
Simulate “CD over the Wire”
Utilize open standards to create image exchange network
Demonstrate patient control of image exchange via personal health record (PHR) accounts
Deliver open source reference implementation of systems developed for project

8. Participants Five initial sites: Phase Two Sites:
University of California, San Francisco (UCSF)
Mount Sinai Medical Center (NYC)
Mayo Clinic (Rochester, MN)
University of Chicago Medical Center
University of Maryland Medical Center
Phase Two Sites:
Gillette Children’s (St. Paul, MN)
St. Barnabas Health (NJ)
Stanford
Advanced Radiology (CT)

9. Design Principles Based on IHE XDS model
Data to be shared images (CT, MR, X-Ray) and diagnostic reports
Phase One: All data access will be through PHR accounts
Phase Two: Support research and site-to-site clinical transfers

10. Steps in Sharing via PHR
At UCSF:
Coordinator obtains patient consent
Coordinator sends images & reports into network
Coordinator s patients instructions and credentials for retrieving their images & reports
At home:
Patient logins into PHR and retrieves their images and reports
Using PHR patient shares images with their consulting physician(s)

11. System Components
Edge Device at each site to bridge local imaging and information systems with image sharing network
A Clearinghouse to temporarily store images and reports
One or more Web-based PHR providers capable of securely retrieving images and reports from clearinghouse into patient account

12. Data Flow PHR‏ Clearinghouse PHR ‏ Imaging Center/ Hospital PACS DICOM
Retrieve
Edge Server
Upload
HL7
Retrieve
PHR ‏
RIS

13. Edge Device Created by phase one consortium
Open source, available on GitHub
UI for site personnel to select exams to share
Generates security token for retrieving images
Assembles and sends payload (reports & images) to clearinghouse
HL7 interface from RIS for reports
DICOM query/retrieve from PACS
SOAP calls to clearinghouse for submission

14. Clearinghouse Created by LifeImage
Pass-through for all images & reports
Receives submissions from edge servers
30 days of temporary storage
Responds to queries from designated PHR systems for images & reports
Has no customer facing UI
Add life image

15. Personal Health Record (PHR)
Internet accessible
Web based UI for patients to retrieve, view & share their images and reports
Provides persistent storage for images and reports
Owned and controlled by patients
Independent of sending institution. This is NOT a patient portal
Four vendors: LifeImage, Dell, itMD and DICOM Grid

16. IHE®: Integrating the Healthcare Enterprise
Initiative between industry and healthcare professionals
100+ member organizations & individuals
Is a framework not a standard.
Relies on existing standards:
DICOM for imaging
HL7 for clinical messaging
Web services (SOAP = XML + HTTP) for enterprising messaging
Create consensus on how to use standards
Profile: describe the solution to a specific integration problem, and document the system roles (“actors”) and design details for implementers (“transactions”)

17. Key IHE Concepts Profiles: workflow models of business processes
Cross Enterprise Document Sharing Profile : For exchanging documents between organizations
Import Reconciliation Profile: For importing outside images into an institution's archive
Actors: systems that create or process data.
Examples:
Image Manager & Image Archive = PACS (moves & stores images)
Order Filler = RIS (schedules ordered exams)
Transactions: interactions between actors/data flow
RAD-4 [Procedure Scheduled]: RIS schedules an previously ordered exam
RAD-8 [Image Stored]: When a scanner stores a new images to PACS

18. XDS Profile XDS = Cross Enterprise Document Sharing
Defines actors & transactions for exchanging clinical documents between healthcare organizations (affinity domain)
Actors:
Document sources (edge servers) & consumers (PHRs)
Document registry & repository (clearinghouse)
Transactions are all SOAP based web service calls:
ITI-41: Provide & register document set (upload to CH)
ITI-18: Registry stored query (PHR query of CH)
ITI-43: Retrieve document set (PHR retrieve from CH)

19. Alterations of XDS Model
Normally XDS relies on patient identifiers to locate documents in registry
ITI-18 supports queries on demographics (name, MRN, etc)
OK in a regular XDS deployment since all nodes within an affinity domain can access PHI
RSNA Network utilizes a one time use transaction id to retrieve content
64 character string: ce401a161c6061ae74baf7cb96593d1a2fdd7c58d951b7439e67f96099b1e07a
Used in ITI-41: Provide & Register Document Set transaction
Eliminates the need to store PHI in the registry

20. Tokens & Transaction IDs
Transaction id is generated from SHA-256 hash of:
One time token: Created when images are selected for transmission: ce401a16
Password: Selected by patient or site staff when images are queued for transmission:
Patient’s DOB: 11/7/1972
To retrieve images and reports from PHR:
Patient enters token & password (PHR already has DOB)
PHR calculates transaction id: SHA256(token + pw + dob)
PHR uses transaction id to locate documents (ITI-18) and retrieve content (ITI-43)

21. Queuing Exams at Edge Server

22. Retrieving Exams via PHR

23. Usage of PHR Sharing 8 sites are currently enrolling patients
6015 patients (UCSF: 1404)
23989 exams (UCSF: 9386)
At UCSF majority of exams were cross sectional:

24. Summary Electronic image exchange is viable and growing industry
Facilitated by pervasiveness of cloud infrastructure
Standards needed to prevent vendor lock in and information silos
Tower of Babel problem
PHRs demonstrate feasibility of granting patient access to and control of their images
Number of signups show patient interest
Meaningful Use will require greater patient access
22 patients at UCSF

25. Next Steps Expand functionality of the network:
Use for exchange of research images such as in a clinical trial
Direct site-to-site exchange for trauma or urgent transfers
Data flow changes, but transactions remain the same
Final destination of images and reports are no longer a PHR, but a research or clinical archive
Research data must be anonyimzied due to IRB requirements
Utilize RSNA’s Clinical Trials Processor (CTP)

26. Sharing Medical Images Researchers’ WAY
Maxwell Cheong, M.S.
Medical Imaging Informaticist
Quantitative Image Processing Center
Department of Radiology
& Biomedical Imaging
UCSF

27. Once Upon A Time… The story of Dr. Felonius Neurosurgery researcher
Mild traumatic brain injury (mTBI) study
“I need more data to support my hypothesis”
Multi-site study of patient’s CT and MRI scans
Graphics:

28. Challenges HIPAA Privacy Rule Compliance – Research
Protected Health Information (PHI) Identifiers
patient name, birthday, medical record number
Institutional Review Board (IRB) approval at each participating site
Consistent MRI scan quality
Efficient and cost effective
Scalable for big data
CDs from film library
Manual de-identifcation of imaging data
An easy but yet secure way to share data
Graphics:

29. Solution: Research PACS (rPACS) at UCSF
Started in 2008 to address the challenges of enabling HIPAA compliant mult-site studies
Dcm4chee Image Management Server
DICOM compliant
Open source with Jboss, PostgreSQL backend
Scalable with J2EE multi-server deployment
Redundant Array of Independent Disks (RAID)
Daily backups
But how do you get the data?
Graphics:

30. rPACS + QUIPC DICOM File Uploader
Powered by CTP
Medical imaging data anonymization and upload all in one
Open source
Easily configurable – XML, Java
On-demand anonymization
Pre-defined mapping table
Automatic anonymized ID assignment
Many different workflows configured to suit every need

31. Secured VPN Connection QUIPC DICOM File Uploader (powered by CTP)
rPACS Data Flow
PACS
Researcher
DICOM (PHI)
Internet
DICOM
(De-Identified)
Secured VPN Connection
rPACS
QUIPC DICOM File Uploader (powered by CTP)
(De-Identified)

32. On Demand Anonymization

33. Pre-Defined Mapping Table

34. Automatic Anonymized ID Assignment

35. Phase Two: rPACS + CTP + RSNA ISN
Hospital 1
Researcher
CTP Enabled Edge Server
DICOM
(De-Identified)
Clearing House
Hospital 2
CTP Enabled Edge Server
rPACS
Hospital 3
(De-Identified)
CTP Enabled Edge Server
Hospital 4
CTP Enabled Edge Server

36. Where do data come from?

37. Where do data come from?

38. Where do data come from?
24 medical centers providing imaging data

39. How do I download the data?
DICOM Viewer: Osirix
Network File Servers
Direct Download via FTP
Laptop
Desktop
iPad and other mobile devices

40. DICOM Viewer - Osirix
DICOM compliant on-demand data query and retrieve
Clinical viewing / FDA cleared

42. Network File Servers
Automated image processing pipeline friendly

43. Direct Download
s sent to collaborators with password protected links to the anonymized imaging data

44. Direct Download (con’t)
ftp://ftp.radiology.ucsf.edu/pub/quipc/sharecase/brain.zip
pw: sharecase

45. rPACS - Metrics 2.3 TB 9,936,374 images 15 concurrent servers
Multiple modalities
CT, MRI, MicroCT, X-Ray, Angiogram…
4 multi-site studies
Mild Traumatic Brain Injury (mTBI)
Austism
Osteoarthritis
Epilepsy

46. Summary Dr Felonius’ brain study is a success rPACS + CTP + RSNA ISN
Exceeded patient enrollment goals by recruiting patients through multiple medical centers across the world
Large dataset of anonymized imaging data collected
HIPAA compliant
rPACS + CTP + RSNA ISN
Collaborative research platform
Scalable
Secure
Graphics:

47. Future Plans Social sharing via web portal integrations
Box?
Facebook?
Research system integrations
The Federal Interagency Traumatic Brain Injury Research (FITBIR) informatics system

48. Questions? Wyatt M. Tellis, Ph.D.
Laboratory for Radiological Informatics (LRI)
University of California San Francisco
530 Parnassus Ave, CL-158, San Francisco CA 94143
Tel: (415)
Maxwell Cheong, M.S.
Quantitative Image Processing Center (QUIPC)
185 Berry St, Suite 350 Lobby 6, San Francisco CA 94107
Tel: (415)