Representing and using template-knowledge for a medical ontology in protégé

We present a medical ontology that is used for registering health problems at intensive care units in hospitals. Because of its flexible architecture it is necessary to be able to determine equivalence and subsumption automatically. These two tasks require two types of knowledge to be encoded: definitional knowledge and template knowledge. We discuss how both kinds of knowledge can be made explicit and used in Protege, using the OWL language. In our approach, we have to mix description logic (DL) style of knowledge representation with a frame-based style of representation. The consistency between both is an important issue. Our aim with this paper is to demonstrate and discuss approaches to augment DL-based representations with frame-based meta-knowledge. 1 Real-World Usage of Ontologies DICE (Diagnoses for Intensive Care Evaluation) is an ontology regarding reasons for admission in intensive care. Its primary applications are providing a means for registration and aggregation of patients’ reasons for admission. In order to stimulate clinicians and nurses to provide as much detail as is known about a patient, a mechanism for supporting post-coordination has been implemented, based on a (frame-based) representation that allows specification of refinable characteristics. Aggregation is supported by means of definitional characteristics, that enable grouping together reasons for admission that share common properties. As the post-coordination mechanism allows for a multitude of ways to specify a reason for admission, it is necessary to be able to determine equivalence and subsumption automatically. In order to realize this, a DL-based representation is being investigated. Ideally, this will contribute to the development of an ontology that is consistent, supporting highly detailed registration, allowing automated classification and instance querying. As an example of the post-coordination mechanism, consider a patient being admitted to the intensive care department with acute type B viral hepatitis. The mechanism provides various ways of constructing this expression, e.g. as Viral Hepatitis, caused by Hepatitis B virus, having course Acute, as Acute Viral Hepatitis, caused by Hepatitis B virus, or even as Infection, located in Liver, caused by Hepatitis B virus, having course Acute. 2 Description Logic versus Frame-based Knowledge The two tasks of automated classification and registering instance information mentioned in Section 1 require two types of knowledge about a concept to be described: knowledge about its semantics, i.e. its definitional knowledge, and knowledge about the construction of instance data, which we call template knowledge. These two issues are related, but certainly not interchangeable. For example, consider the concept ‘wine’. ‘Wine’ can be defined as “the fermented juice of fresh grapes used as a beverage” (Merriam-Webster Online Dictionary). For knowledge acquisition purposes however, one wants to specify that the name, color, sugar, flavor, grape, maker, and body of the wine are relevant properties of wine [2]. Red wine can be defined as “wine which has a red color”, and for red wine the tannin level can be specified. In the ontology introduced above, we use the description logic (DL) based characteristics of OWL to describe the semantics of concepts. The benefit DL is the possibility of reasoning with defined necessary (and sufficient) properties of concepts. At the same time, we used frame-based representations to represent template knowledge. Frames offer