ABSTRACT
Knowledge engineering has shown that besides the general methodologies from software engineering it is useful to develop special purpose methodologies for knowledge based systems (KBS). PROforma is a newly developed methodology for a specific type of knowledge based systems. PROforma is intended for decision support systems and in particular for clinical procedures in the medical domain. This paper reports on an evaluation study of PROforma, and on the trade-off that is involved between general purpose and special purpose development methods in Knowledge Engineering and Medical AI. Our method for evaluating PROforma is based on re-engineering a realistic system in two methodologies: the new and special purpose KBS methodology PROforma and the widely accepted, and more general KBS methodology CommonKADS. The four most important results from our study are as follows. Firstly, PROforma has some strong points which are also strong related to requirements of medical reasoning. Secondly, PROforma has some weak points, but none of them are in any way related to the special purpose nature of PROforma. Thirdly, a more general method like CommonKADS works better in the analysis phase than the more special purpose method PROforma. Finally, to support a complementary use of the methodologies, we propose a mapping between their respective languages.
Subject(s)
Artificial Intelligence , Databases, Factual , Family Practice , Humans , Netherlands , Programming Languages , Task Performance and AnalysisABSTRACT
This paper reports on research for decision support for anaesthesiologists at the University Hospital in Groningen, the Netherlands. Based on CAROLA, an existing automated operation documentation system, we designed a support environment that will assist in real-time diagnosis. The core of the work presented here consists of a knowledge base (containing anaesthesiological knowledge) and a diagnosis system. The knowledge base is specified in the logic-based formal specification language AFSL. This leads to a powerful and precise treatment of knowledge structuring and data abstraction.
