Gregory S Weinger, Presenter: Nothing to Disclose

Susan Barretta BS, Abstract Co-Author: Nothing to Disclose

Alex Anh-Tuan Bui, Abstract Co-Author: Nothing to Disclose

Ricky Kiyotaka Taira PhD, Abstract Co-Author: Nothing to Disclose

Hooshang Kangarloo MD, Abstract Co-Author: Nothing to Disclose

To establish a peer-to-peer (P2P) network of XML (eXtensible Markup Language) gateways, enabling integrated access to patient records from multiple institutions for users in clinical and research environments.

A modular infrastructure, DataServer, was designed to provide XML-based query and retrieval to databases at UCLA, including RIS, HIS, PACS, and other research databases. DataServer facilitates access to different data sources (DICOM, relational databases, HL7 streams, SOAP (simple object access protocol)), abstracting the different underlying query mechanisms into a uniform representation. Translation of data between sites was accomplished using newer semantic web constructs. DataServer incorporated methods for automatic audit and de-identification of patient information, using natural language processing (NLP) to scan and replace free-text reports for HIPAA identifiers (e.g. names, age, address). Distributed access across multiple DataServers was pursued: 1) using UDDI (universal description, discovery, and integration) to register systems; and 2) using a P2P network topology to provide access to site-specific patient records. Distributed caching was employed to enable prefectching for clinical applications and larger data sets. Additionally, access to grid computing resources (e.g. for image processing) was facilitated through a Globus framework.

DataServer has provided integrated access to patient demographics, encounter information, clinical documents, laboratory and imaging. Access to live patient data for research purposes is handled through its automated de-identification methods. As an application framework, it has successfully been used as a data engine for two applications: TimeLine, a problem-centric view of the patient medical record; and a new Java-based DICOM image viewer for telemedicine. DataServer has also been successfully demonstrated outside of UCLA, leading to its release as an open source project this past year (www.mii.ucla.edu/dataserver/).

DataServer's extensible architecture and support for web and XML technologies provides a foundation for integrating data across medical environments for both clinical and research purposes.

Weinger, G, Barretta, S, Bui, A, Taira, R, Kangarloo, H, DataServer: An Open Source, Peer-to-Peer XML Patient Record Network.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4404726.html