DICOM Structured Reporting: An object model as an implementation boundary David A. Clunie ComView Corporation mailto:dclunie@dclunie.com ABSTRACT DICOM Structured Reporting (SR) provides for encoding and interchanging structured information that may reference images, waveforms or other composite objects, in traditional reporting applications as well as for logs, measurements and CAD results. DICOM SR differs from generic content encoding approaches like XML, in that it supports coded entries, values that are strongly typed, and explicit relationships. DICOM structured reports (like images and waveforms) are composite objects that can be stored, transmitted and queried. The traditional DICOM binary encoding is used to encode structured reports. The structure and content of the SR tree should be accessible regardless of the internal or external representation. XML parsers and XSL-T tree transformation engines that can be used for data entry, presentation (display and printing) and trans-coding (to HL7 2.x and HL7 Clinical Document Architecture (CDA)) need to be interfaced with DICOM tools that support encoding, transmission, storage and retrieval. Issues associated with establishing the appropriate boundaries between tools are discussed, as are how and when to internalize a DICOM SR in an actual or virtual XML representation, the characteristics of such a representation, and the use of SAX events or the Document Object Model (DOM) to drive style-sheet driven tree transformation engines. Keywords: DICOM, Structured Reporting, XML, XSL, Document Object Model 1. INTRODUCTION Structured Reporting (SR) is a recent addition to the DICOM standard. 1 It provides a mechanism for encoding and interchanging structured information that may reference images, waveforms or other composite objects. Though applicable to traditional reporting applications, SR can be used for any type of content. Examples include procedure logs, measurements generated by devices and applications, and computer assisted detection (CAD) results. Unlike XML, DICOM SR has domain specific mechanisms for describing coded entries. Nodes are name- value pairs and values are strongly typed. Explicit relationships other than containment are defined, such as “has properties” and “inferred from”. The content can be organized as a directed acyclic graph rather than as a strict tree, by using by-reference relationships. In common with other DICOM objects such as images and waveforms, DICOM structured reports are composite objects. They can be stored, transmitted and queried just like any other object. The use of traditional DICOM binary encoding to encode structured reports allows for re-use of the existing installed base of DICOM applications and toolkits. It does create a barrier to entry by implementers with no previous DICOM experience. Furthermore, the skills required to build high quality report generation, rendering and analysis applications differ from those required to build DICOM encoders. In the past, DICOM “tool kits” have been created to ease the burden on application designers. These have been made available either as free libraries or as commercial products. However, each has a different architecture and API. DICOM structured reports are “layered” on top of existing DICOM encoding mechanisms and services. There exists a need for “structured reporting toolkits” which allow access to the structure and content of the SR tree independent of the internal or external representation. Using the W3C recommendation for an XML Document Object Model (DOM) 2 as a basis, a “Structured Reporting Object Model” is proposed. The DOM cannot be used directly. It contains XML specific features that are not applicable. It also lacks support for Medical Imaging 2001: PACS and Integrated Medical Information Systems: Design and Evaluation, Eliot L. Siegel, H. K. Huang, Editors, Proceedings of SPIE Vol. 4323 (2001) © 2001 SPIE · 1605-7422/01/$15.00 207