11 DICOM Image Communication in Globus-Based Medical Grids Michal Vossberg, Thomas Tolxdorff, Associate Member, IEEE, and Dagmar Krefting Ting-Wei, Chen

22 Outline  Introduction  Related Work  Methods  Results and Discussion  Conclusion and Future Work

33 Introduction (cont.)  Grid environment (Medical grid)  Secure  Reliable  Highly efficient data transport  Grid Middleware  Globus toolkit  Lack the integration the world-wide medical image communication standard Digital Imaging and Communication in Medicine (DICOM)

44 Introduction (cont.)  DICOM’s Advantage:  Interoperability  Asynchronous communication  Integrity  From the DICOM protocol to the FTP protocol’s Disadvantage:  Reduce most of the advantages and security an integrated network of DICOM devices offers

55 Introduction (cont.)  Problem  Incompatible between the different imaging devices  Solution  Adapts the DICOM protocol to the Globus grid security infrastructure

6 Introduction (cont.)  Standardization  Ensure compatible  Correct representation  Imaging equipment of the different vendors  Expect  Healthcare business  The way the various healthcare actors interact with one another

77 Introduction (cont.)  Medical grid projects  European Enabling Grids for E-Science in Europe (EGEE)  U.S. cancer network caBIG  MediGRID

88 Related Work (cont.)  Toolkit’s common security infrastructure  Encryption and integrity verification of the data  Authentication user or host  Authorization based on the host

99 Related Work (cont.)  Globus components  Grid Security Interface (GSI)  Grid File Transfer Protocol (Grid-FTP)  Grid Services and HTTP  DICOM Grid Interface Service (DGIS)  Medical Data Manager (MDM)  Others: Storage Resource Broker (SRB)

10 Methods (cont.)  Grid-DICOM  Upper layer messaging protocol for message and data exchange  Allow secure communication through an encrypted transport protocol TLS/SSL  Use a Java implementation of the DICOM standard  Dcm4che2 toolkit

11 Methods (cont.)

12 Methods (cont.)  Grid-DICOM Router  Act a proxy and translates between the plain and the grid protocol  Service class  Verification: Forward a C-ECHO message  Storage: Forward C-STORE  Query: Forward C-FIND  Retrieve: Forward C-GET and C-MOVE

13 Methods (cont.)

14 Methods (cont.)  Keep router mostly independent of the architecture of the hosting system  Design the application according to the Java Management Extensions specification  JBoss JMX  Implicit clustering capabilities improve the scalability and fault tolerance of the router application

15 Methods (cont.)  A number of design optimization improve the performance and stability  Optimal thread reuse and performance scalability  Minimize the initial handshaking  All incoming DICOM messages are processed in buffered memory blocks

16 Methods (cont.)

17 Methods (cont.)  Test Scenarios  Have been tested in a partial environment of the MediGRID test bed  The security level  Full transport level encryption  Mutual user/host certification  Authorization against the gridmap file  Full delegation support of credentials

18 Methods (cont.)  Three typical scenarios based on the grid image processing applications  Scenario 1: Distribution  Scenario 2: Storage  Scenario 3: Moving

19 Methods (cont.)  Scenario 1: Distribution. A user distributes images from a modality.  a) Conventional DICOM transfer  b) Encrypted DICOM Transfer  c) GSI-based transfer  d) GSI-based transfer through a router  e) The DGIS imaging solution of the Globus incubator project MEDICUS

20 Methods (cont.)

21 Methods (cont.)  Scenario 2: Storage. A user sends images from an imaging device to an off-site image archive (C-STORE)

22 Methods (cont.)  Scenario 3: Moving. A user requests the off-site image archive to move images to a different archive

23 Methods (cont.)  Three different set:  One Magnetic resonance (MR)  5 series of 100 images each (512*512, 16 bit, total 250MB)  One Computed tomography (CT)  50 series of 10 images each (512*512, 16 bit, total 250MB)  Ten Computed radiology (CR) chest image  10 series of 1 image each (2140*1760, 16 bit, total approx. 800MB)

24 Results and Discussion (cont.) 24 Transfer Rates of Scenario 1-3 In MB/s

25 Results and Discussion (cont.)  DICOM throughput increases with a lower number of single images (CR > CT = MR)  The transfer rate decreases when engaging the TLS 3des encryption  Engaging the Grid-DICOM transfer results in an almost equal, if not slightly lower transfer rate than plain encryption

26 Results and Discussion (cont.)  Connecting devices through a router further reduces the transfer rate through the additional message processing costs, depending on the number of images transferred  The router solution performs in the same range as the DGIS

27 Conclusion and Future Work (cont.)  Proposed a solution to integrate legacy DICOM- capable system  Developed an adaptation of the DICOM protocol stack to the GSI

28 Conclusion and Future Work (cont.)  Employed a system of routers that transparently convert any traffic from pure DICOM protocol  Show the setup is a promising solution for grids based on the Globus middleware

29 Conclusion and Future Work (cont.)  Future work  Replace the command line clients by a user interface  Improve the router software in terms of stability and transaction ratio

30 Conclusion and Future Work (cont.)  Add modification chains for the DICOM data when passing the routers  Enhance the system by a Web service for a reliable DICOM transfer

31 Thank you for your attention