Presented at the 2000 APIII Conference Return to 2000 Abstract Index
AN INITIAL METADATA MODEL AND ITS DTD / XML FOR SURGICAL PATHOLOGY REPORT
University
of Pittsburgh
Pittsburgh, Pennsylvania
Fanglin Zheng
Fanglin Zheng, MSCS, John Gilbertson, MD, Michael J. Becich,
MD, PhD
Department of Pathology, Division of Pathology Informatics,
University of Pittsburgh, Pittsburgh, PA
Metadata is Data about data. It provides a means by which the multimedia semantics can be described in a structured fashion for use by many varied applications or services. In a pathology relevant context, as well as many others, this facilitates searchability, extensibility, re-usability and scalability but only where the metadata can be exchanged between systems.
In the development of a Tissue Bank Information System, the authors faced the problems of unstructured, un-enumerated and incomplete data when populating the Tissue Bank System from the Laboratory Information System (LIS). Though a Metadata implemented as a standard DTD/XML could provide structure, enumeration of data and check for completeness, a single DTD for all pathology reports would rapidly become unworkable as data elements like "GLEASON GRADE" necessary for prostate cancer would be unnecessary and wasteful in all other diseases.
Given the above, we proposed an initial metadata model for pathology report and implemented in DTD/XML, known PARMeta (Pathology Report Metadata) internally. PARMeta includes not a single, monolithic DTD, but rather a sparse 'stem' DTD and a library of sub or 'branch' DTDs (which may have their own branches). The basic data elements for the Stem DTD derive from the UPMC Copath C/S to MARS HL7 v 2.2 Final Results Interface because it was available and provided a relatively compete, understandable structure from which to build. The stem includes fairly consistent elements such as accession number, pro-operative diagnosis, etc. The branch DTD's are defined using three parameters - disease (ie malignant v benign), organ (ie prostate v others) and procedure (ie. resection v biopsy v TURP), and include data elements specific to a given disease, organ and procedure. A human, or machine, could construct a complete and valid DTD of any desired granularity by the appropriate selection of sub DTDs.
This work is part of the UPMC Tissue Bank Information System project. It is expected that the document will be reviewed, added to, improved, and shared by wide range of contributors before being implemented of the basis for structured reports in the UPMC LIS.
Keywords: metadata, tissue bank, database, Pathology, XML, DTD