historical perspective - dicom native models from wg-23
TRANSCRIPT
Historical Perspective - DICOM Native Models from WG-23
Model Discussions in WG-23 Occurred in over a dozen meetings and t-cons
Began in early 2007, continuing through summer 2008, minor tweaks since then
Leveraged earlier work Suggestions from Guenther Zeilinger
(father of dcm4che, a widely used DICOM toolkit for Java)
David Clunie in his enhanced MR validation suite (also used in PixelMed’s DICOM toolkit)
Dongbau Guo and Oracle’s schemas for DICOM in XML Lessons learned from other image formats (e.g. NFTI)
Participants from most major vendors, several smaller vendors, and from academia
Ideas presented and feedback solicited at multiple major conferences.
Why DSDL?
ISO/IEC Standard Politically correct, as DICOM is an ISO Standard ISO rules say ‘use ISO Standards when possible’
The clarity of the Relax NG Compact form Part of target audience not well versed in XML Separating out complex validation rules aids
clarity Rich validation capabilities of Schematron Simple translation to other schema
languages Several tools available to translate Relax NG into
XSDL, DTD, and other languages Can use Schematron rules independent of
schema
Alternatives Considered
Use XML Element names derived from DICOM Data Dictionary names Similar to suggested schema from Emanuel Problem with unknown DICOM Data Elements
Use XML Element names derived from numeric tag Not as easy to work with Strong validators could fail with unknown DICOM
Data Elements – schema skew highly likely Use VR as XML Element, with tag and name
as XML Attributes Easy to support strong type checking Not natural to most people
Consensus Reached
Simple grammar matching DICOM encoding Mechanical, bi-directional translation between
binary DICOM and the XML Infoset model Allows searching by either numeric tag or
keyword (i.e., DICOM Attribute Name) Stable Schema – need not change Dictionary driven Allows for private DICOM Data Elements Leverages VR for potential validation
Separately defined enhanced validation using Schematron rules and assertions
Open to Suggestions, but
Any suggested changes must take into account previous decisions: Must be bi-directional Must take into account Private Data Elements
(important for research use) without breaking Must allow transparent pass-through (e.g.
through Hosting Systems) of unknown DICOM objects
Must not break if Hosting System and Hosted Application are working off different versions of the DICOM Data Dictionary
Must not be onerous for the uninitiated to use
Possible Suggestion
Instead of a generic “Value” XML Element inside the DICOM Data Element, use a VR-specific XML Element (e.g. PNValue, LOValue, SQValue, etc.) Still a mechanical, bidirectional translation
from binary DICOM, given the UN VR Allows for VR-specific constructs (e.g. names) May be better for strong type checking
(This was considered by WG-23, but was not incorporated. It could be presented again, if that brings a convergence.)
Should WADO use the WG-23 Model?
Having a consistent methodology for representing DICOM in XML is desirable
But
Goals may be different
The two WGs should converge, but only if their differing goals can be met with a single methodology.