Model of patient related data and a structure of medical terms for knowledge-based systems in medicine.

Modelling medical knowledge is still a significant problem in Artificial Intelligence in Medicine. The topic of our project is to develop a general reference model for medical knowledge and decision making. This development requires a thorough analysis of the structure of patient related data. To support modelling patient related data, we also developed a repository for all relevant medical terms. For both we defined a model, derived a frame-like representation and coded a prototype implementation. In the repository, terms are described by specialisation, components, attributes, and possible values. The implementation of the repository, also consists of an acquisition tool and a definition browser. The model of patient related data is represented in a frame-like form, the Case-Data-Sheet. Case-based-data can be formalized in the Case-Data-Sheet by simple object-attribute-value-triples as well as by more complex structures with iterated attributes and multiple attributes. The above complex structures can be augmented by context-dependent information about patient-object and/or object-object relationships. We developed our prototype implementation to evaluate whether our model of patient related data is adequate and complete. This implementation consists of a comfortable user interface, that lets the user enter information about a patient with familiar terms and generally accepted descriptions. Our system then creates Case-Data-Sheets as necessary and automatically transforms the user's input into the corresponding data structures. The models for both, the patient-related data and the repository of terms, are implemented fairly straightforward in Smalltalk/V on PC. Future research should investigate, whether these models and the derived structures can be an adequate base for the formalization of patient related data in medical knowledge based systems