Design, Development, and Adoption of Ontology-Driven Clinical Software

Abstract: This thesis addresses how ontology-driven applications can be designed and developed to support distributed clinical knowledge management in oral medicine, where geographically dispersed practitioners need to share practical clinical knowledge. A step in developing tools for knowledge management is representing knowledge in a machine-processable and sharable manner. We investigate the use of the World Wide Web Consortium's recommendations of the Web Ontology Language (OWL) and Resource Description Framework (RDF) to describe clinical knowledge in oral medicine. The MedView project, a cooperation between clinicians in oral medicine and computer scientists, provides a basis for this work. Limitations of MedView's original knowledge model are identified, together with a list of requirements for a new model.

In this new model, OWL and RDF are used for representing examination templates, value lists, aggregates of values, and individual examination records. Further, we give a description of how the ISO/IEC 15288 system life cycle processes can be used to structure the ontology development process. To support distributed knowledge management in oral medicine, an online community for sharing and discussing difficult and interesting cases was built, which uses OWL and RDF for representing examinations, users, meetings, news, and case meta-data. OWL and RDF were found to support the requirements of making ontology reuse possible, and different language versions and meta-data are more easily represented than in the original model. One of the requirements not readily provided for is capturing interactions between different parts of templates, for which the Semantic Web Rule Language could be used. In practice, we found it hard to find ontologies to reuse. Further, there are differing opinions on the appropriate use of constructs, and OWL doesn't directly provide the expected validation facilities. Also, we found a lack of guidance for developing OWL ontologies at different levels of sophistication.