Minutes DELOS T1.6 Meeting Vienna

Document Sample
Minutes DELOS T1.6 Meeting Vienna Powered By Docstoc
					Minutes DELOS T1.6 Meeting – Vienna                                         21.09.2005


University of Athens:    Harry Dimitropoulos
ETH Zürich:              Peter Fischer
OFFIS:                   Ulrike Steffens
HITT:                    Raimund Vogl
UMIT:                    Gert Brettlecker, Heiko Schuldt




T1.6: Management of and Access to Virtual Electronic Health Records

Agenda
• General Overview
• Presentation of Partner Contributions
• Discussion: Joint Activities
• Next steps


General Overview (Heiko) – Summary of and Excerpt from JPA2, Technical Annex

•   JPA2 work has officially begun January 1st, 2005
    − Duration: 18 months (until June 2006)
    − With 6 months overlap to JPA1

•   This Task is the logical continuation of T1.1 of JPA1. It evaluates the architecture
    and adds in it aspects of electronic health records, which represent an important
    application field for digital libraries. The realization of these goals requires an
    infrastructure that is highly dependable and reliable. Moreover, the infrastructure
    has to allow for the transparent access to distributed data, and to efficiently
    schedule the access to computationally intensive services by applying
    sophisticated load balancing strategies using GRID technology.

     Task activities will address the following issues:
    − Identification of the basic building blocks to access distributed artifacts and to
       intelligently search within a set of these artifacts.
    − Provision of a dependable platform that supports the integration of these
       building blocks into processes (e.g., based on the ETH/UMIT hyper database
       prototype system OSIRIS – Open Service Infrastructure for Reliable and
       Integrated process Support), thereby realizing a virtual electronic patient
       record.




DELOS Meeting Vienna                                                              1/7
Expected Results

•   Management of and Access to Virtual Electronic Health Records Implementation of
    sample building blocks and processes in combination with the specialized
    applications of HITT/TILAK

•   The electronic patient record is a virtual entity that consists of a set of distributed
    artifacts, which cannot be materialized for organizational reasons. The main goal
    of this Task is to identify, design and build demonstrators for the basic building
    blocks needed to access virtual electronic health records, i.e. locate the different
    artifacts, make data from the different healthcare providers available, perform
    the format conversations needed, and present the result to a user.


Del. Deliverable name               WP Lead Nature        Dissemination     Delivery Date
No
1.6.1 Implementation of             1    19     D        PU                 24
      selected building blocks           (UMIT) (Demon- (Public)            (Dec. 05)
      and processes in a rea-                   strator)
      listic healthcare setting



Partner Contributions

ETH Contribution (Peter)

    XL: Development and Operation of Web Services
    • Easier Development of Web Services
    • Easier Administration of Web Services
    • Remove impedance mismatch (SOAP/Java/SQL) by using a uniform XML data
        model
    • Scalable (PDA to Multiprocessor/Cluster) web service runtime environment that
        abstract away most low-level aspects

    iMeMex „Where have all my files gone?“
    • Personal Information management tool
    • Accessible by any program due to file system integration (virtual folders)
    • Enable query processing & data integration on all data files on your system
       (remote sources available soon
    • Extensible plugin architecture for additional features and file formats




DELOS Meeting Vienna                                                                 2/7
OFFIS Contribution (Ulrike)
  • Replication strategy based on peer-to-peer (P2P) techniques combining intra-
      institutional and inter-institutional replication strategies
      o DL systems whose information providers are organisationally closely related
          should be tightly coupled, in order to allow for a high degree of consistency
          to be maintained (intra-institutional)
      o DLs, such as those of separate publishers, should be coupled less tightly, to
          allow for retaining their autonomy (inter-institutional)
  • Goal:
      o achieve an optimal trade-of between tight and loose coupling of
          information systems
      o realized by dynamically switching participating systems between being
          synchronously and asynchronously updated.

University of Athens Contribution (Harry)

  Search of Composite Health Records
  • Query Optimization, in particular for queries accessing parts of health records.
      Important issues: access to parts of a composite record must be
      o highly dependable, reliable & secure
      o allow for transparent access to distributed artifacts
      o allow to intelligently search within artifacts for specific medical information
      o deliver relevant medical information on-demand, in real-time, to the point
         of care; need to be immediate, with no delays
  • query methods must be optimized for the many different types of searches
      expected, e.g.:
      o complex searches over large sets of distributed, heterogeneous data, will
         need to be supported, including content-based similarity searches over
         many virtual shared electronic health records, and multi-object multi-
         feature queries. Example: “Get all chest x-rays of patients between the
         ages 20-30, that have blood type ‘A’ and are allergic to pine trees
  • Plan for query processing and optimization strategies:
      o heuristics
      o cost functions
      o semantic query optimization
      o distributed query processing, optimization and decomposition methods (e.g.
         aiming to reduce the amount of data transfer)
      o temporal queries
      o multimedia querying strategies
      o high-level, complex, queries supported
      o predefined query templates provided
      o In summary: for medical images, medical video data, and audio data, this
         opens up many issues, among them efficient query formulation, query
         execution, and optimization. Therefore, standard optimization techniques
         need to be further modified to work for multimedia data types


DELOS Meeting Vienna                                                             3/7
HITT Contribution (Raimund)
  • Implementation of a regional, shared electronic patient record in the Austrian
      state Tyrol (Health@Net). Partners are:
      o Innsbruck University Hospital (IUH)
      o University for Health Sciences, Medical Informatics and Technology (UMIT)
      o Health Information Technologies Tyrol (HITT)
      o icoserve
      o Information Technologies for Healthcare (ITH)
  • Status:
      o Connection to existing health care networks and creation of web portal
          system (step 1) completed, access by patients and realization of document-
          inbound currently in test period.
      o Common unique patient identifier (master patient index) necessary for
          cooperative care (planned in step3)

UMIT Contribution (Heiko)

  OSIRIS (Open Service Infrastructure for Reliable and Integrated Process Support)
  • Distributed Infrastructure for process-based applications
      o Peer-to-peer process execution
  • Dynamic adaptation to changing environments
      o Load balancing, service deployment on demand
  • (Transactional) Execution guaranteed
      o failure handling, alternative executions, concurrency control for processes

  Efficient and Effective Matching of Compound Health Records (M. Springmann)
  • Patient Records comprise many multimedia objects of various media types.
      Goal: Contend-based / Similarity search for
      o Teaching, Research, and Diagnosis
  • Needed: Computation of document similarity based on similarity of
      component objects
      o Assignment Problem: “Find complete match with best overall domain
          similarity”
  • Two-stage filter-and-refinement algorithm for efficient and effective retrieval
      has been implemented



Joint Activities

1. An Infrastructure for virtual electronic health records (all)

  Contribution of Partners
  • HITT: Health@Net Prototype
  • OFFIS: P2P metadata management in eHealth Infrastructure; replication of
     healthcare provider data

DELOS Meeting Vienna                                                           4/7
    •   ETH: XL Services for health-related services; optimization of queries on top of
        P2P metadata
    •   UoA: optimization of queries on top of P2P metadata
    •   UMIT: service composition and process management (application of OSIRIS)




•   Planned cooperations / exchanges:
    − OFFIS – UMIT: replication management
    − HITT – UMIT: health@net project



2. Multi-object Multi-feature Queries (UoA, HITT, UMIT)

    Support similarity search in compound health records. One functional building
    block in the overall eHealth infrastructure (1.)

    Contribution of Partners
    • HITT: provide (anonymized) real data of compound health records (pdf
       documents)
    • UoA: optimization of multi-object multi-feature queries
    • UMIT: similarity search in multi-media health records



DELOS Meeting Vienna                                                             5/7
•   Planned cooperations / exchanges:
    − UoA T UMIT/HITT



3. Imemex of healthcare data (ETH / UMIT ?)

    Management of local files stored at the general practitioner’s sites. One building
    block in the overall eHealth infrastructure (1.) – makes these files available within
    the infrastructure

    Contribution of Partners
    • ETH: application of imemex to healthcare data
    • UMIT: provide data from general practitioner system (?)

    Dependent on the availability of these data




DELOS Meeting Vienna                                                                6/7
•   Cooperations:
    − ETH ¨ UMIT?



Next steps:

•   HITT keeps partners updated on implementations in health@net

•   HITT to provide data for Compound Health Records

•   UMIT attempts to get access to (sample) GP software / GP files for imemex data
    management

• AudiConferece: date to be scheduled




DELOS Meeting Vienna                                                           7/7