Towards a general architecture for medical expert systems

by M. Stefanelli, R. Bellazzi, C. Berzuini, L. Ironi, S. Quaglini



in Proc. AAAI Spring Symposium on Artificial Intelligence in Medicine, Stanford, (1992), 105- 112.

ABSTRACT

GAMES II is a project supported by the European Commission within AIM (Advanced Informatics Technologies in Medicine) aiming at developing a comprehensive methodology for medical knowledge-based systems (MKBS) design and construction. The main idea behind GAMES II is to design a MKBS starting from an epistemological model of medical reasoning. It specifies the desired problem solving behavior for a MKBS through a sound epistemological analysis of the different types of knowledge required to generate such behavior. Then, a computational model for the MKBS will be built based on the epistemological model previously defined. There should be a correspondence between epistemological model elements and computational constructs. The design of a KBS is thus view as a process of adding symbol-level information to an epistemological model of medical reasoning. For each inference type corresponding computational techniques need to be selected. Moreover, a set of suitable knowledge representation techniques (frames, production rules, qualitative and quantitative models, bayesian belief networks, temporal networks, etc.) may need to be used. this paper illustrates how GAMES II could solve problems which arise in designing and building a MKBS potentially able to menage patients who developed a graft-versus-host disease (GVHD). A recent article By Ferrara and Deeg (1991) reviewes what we presently know about GVHD. Different types of knowledge are available: they range from definition and causes of GVHD to observed clinicopathological spectrum, from proposed immunopathophysiological theories to experimented prophylactic and therapeutic interventions. The diversity of these types of knowledge is particularly challenging for the methodology under development within GAMES II. Thus, after a short illustration of the main epistemological features of GAMES II, we will describe how portions of knowledge on GVHD could be represented both avoiding to distort its nature and pursuing a computationally efficient use.



  • Back to publication list

    Liliana Ironi June 25th, 1996