(Projektbeschreibung nur in Englisch)
K. Boegl, K.-P. Adlassnig
Department of Medical Computer Sciences, Section on Medical Expert and Knowledge-Based
Systems, University of Vienna Medical School, Spitalgasse 23, A-1090 Vienna, Austria
e-mail: karl.boegl@univie.ac.at
Background
Knowledge acquisition is a central concern in the development of medical expert systems.
Beside the fact that the elicitation and acquisition of expert knowledge is a difficult and
time-consuming task (even more so when unfamiliar representations like fuzzy membership
functions are to be acquired), one of the major difficulties in building medical knowledge
bases are the lack of time and the limited availability of medical experts.
Objective
The goal of this project was to develop a knowledge acquisition tool, the MedFrame/KBuilder Toolkit,
that allows a collaborative construction of medical knowledge bases by
multiple experts via the Internet [1,2]. The experts should be enabled to perform this task in
a collaborative manner and without support by a mediating knowledge engineer.
Material and Methods
The MedFrame/KBuilder Toolkit is a Java application that is integrated in the MedFrame
system, which is a common platform for a variety of medical consultation systems that are
designed to support diagnostic and therapeutic decision making in various medical
domains.
Prominent features of MedFrame are its client-server architecture, its fuzzy- and rulebased
knowledge representation and inference framework, and its object-oriented data-base management
system and data model.
The MedFrame/KBuilder Toolkit supports all subtasks of knowledge acquisition: The definition of medical concepts (findings, diseases, therapies), of (fuzzy) logical rules, of ‘disease
profiles, of data-to-entity conversion rules (translation of “raw” observational and
measured quantitative data into symbolic — but fuzzy — concepts), the encoding of medical concepts
(SNOMED, ICD-10), and many others (Fig. 5).
Special emphasis was laid on the design of a conceptual system model that reflects the
analysis of the tasks and concepts that are involved in the knowledge acquisition process.
Among others, this conceptual model comprises an intuitive step-by-step procedure for the
definition and refinement of medical relationships, a model of how to employ fuzzy set theory
for the interpretation of quantitative medical data, and several other aspects of fuzzy
knowledge modelling. Based on the conceptual system model, the actual implementation
offers a suite of tools that enable the experts not only to build knowledge bases but also to
update and refine the knowledge bases continuously during knowledge acquisition.
Results
The MedFrame/KBuilder Toolkit was tested at the Department of Radiodiagnostics of the
Vienna General Hospital [3]. Knowledge bases in the fields of diagnostic radiology were
established. It was shown that the experts were able to transfer their knowledge into a
“computerized” form with little or no support by a knowledge engineer.
Technical Specification
MedFrame/KBuilder was programmed in Java (Java Development Kit (JDK) 1.1.7) and
was tested on Windows95/NT and UNIX platforms.
Conclusion
Our preliminary results indicate that we are able to reach a significantly higher acceptance
rate by the involved experts as compared to previous projects where most of the knowledge
acquisition task has been performed in a paper-based manner. The availability of the program
via Internet and the intuitive and easy-to-handle graphical user interface are highly
appreciated by the medical experts. As a result, building medical knowledge bases is accelerated
and the maintenance cycles of existing knowledge bases can be shortened.
References