Skip to main content

A General Architecture for Medical Knowledge-Based System


- to contribute to the delivery of health care by enabling medical expertise to be made generally available via knowledge-based computer systems (KBS)
- to study the possibility of cross-fertilisation between this project and results of ESPRIT and other European initiatives
- to apply innovative methodologies and software tools to a representative range of medical problems and to evaluate their effectiveness
- to develop a general architecture and a corresponding set of software tools for the design and construction of medical KBS
- to assess the potential of GAMES II methodology within the health telematics environment
- to monitor relevant telematics developments for telecommunications, cost-effective delivery platforms, integration with hospital information systems or networks and effective human-computer interface design
- to develop and refine epistemological and computational models for medical KBS
- to develop demonstrator medical prototype systems
- to categorise medical activities and decision-making problems for easy selection of appropriate SEMAG formalisms for building KBS
- to develop formalisms and software tools for temporal reasoning, state models (belief networks) and systems models (qualitative models)
- to develop a medical knowledge acquisition framework based on a sound epistemological model of medical reasoning.

Technical Approach

This project aims at developing a comprehensive, commercially viable methodology for medical knowledge-based systems (KBS) construction. Problems that frequently occur in conventional information systems development projects are amplified in the case of KBS development. To avoid them, the project aims at designing a KBS starting from an epistemological model of medical reasoning. It will specify the desired problem-solving behaviour for a KBS through a sound epistemological analysis of the different types of knowledge required to generate this behaviour.

Then, a computational model of the KBS will be built based on the epistemological previously defined. There should be a correspondence between epistemological model elements and computational constructs.

The design of a KBS is thus viewed 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 careful analysis of formalisms and methods for representing causal medical knowledge will be carried out as well as a set of advanced formalisms requiring different methods (qualitative and quantitative models, probabilistic causal networks, temporal networks). Knowledge so elicited will be used by the target KBS according to the desired behaviour conceptually described by the epistemological model at the knowledge level and by the computational model at the symbol level.

Towards such a goal a comprehensive and advanced tool for medical knowledge acquisition will be designed and constructed. It will help the knowledge engineer in the definition of the domain ontology, the inferences and their inter-dependencies, and the task to be executed. Moreover, it will advise the knowledge engineer in the selection of the appropriate computational techniques to implement the epistemological model. Then, the domain expert will be able to use the knowledge acquisition tool to enter the full domain knowledge necessity for task execution.

The medical value of the tools developed for executing generic medical tasks (diagnosis, therapy planning and monitoring) will be addressed and evaluated by a panel of physicians with great expertise in different domains.

The role is twofold:

(a) to select a limited number of medical problems whose solution appears challenging for developed tools;
(b) to evaluate the need of further extensions of the methodologies developed to solve new and more ambitious problems.

Key Issues

Models of medical reasoning, representation of medical knowledge, explanation, medical decision making, prototyping of KBS in exemplar domains.

Relationship to Previous Work

GAMES II continues a previous AIM project by adding new methodological issues whose solutions are expected to increase medical KBS diffusion.

Expected Impact


One of the main impacts that the project could have is that of contributing to the delivery of health care by enabling medical expertise to be made generally available via Knowledge-Based computer Systems (KBS).

- Participation in consensus. Knowledge-Based Systems are a valid and basic contribution to the development of advanced informatics in a medical reference background.
- Demonstration of progress and utility value-added of project. GAMES II proposes a new approach to knowledge acquisition, based on the epistemological model of medical reasoning. Knowledge is acquired separately for each inference task, and the definition of ontological entities of the domain is highly facilitated. This allows physicians themselves to put their knowledge into the expert system, so facilitating knowledge exchange.
- Contribution to standardisation. It is believed that as KBS development is still an advanced topic, it could play an important role in pre-normative work. This work is very important for defining reference terms (standard terminologies of diseases, multilingual terminology, etc.). A further line of work will be the contribution in the field of standards relevant to the problem of communication between GAMES II KBS and the Patient Information System Standard syntax for exchange of laboratory information, patient record exchange format, etc.).
- Participation in consensus formation. GAMES II has provided arrangements and a budget for organising workshops which aim to present the results at different stages of the project to the industrial, scientific and public sectors. This should help in building consensus by taking into account the suggestions emerging from these workshops and by diffusing the outcomes of the European support to research.

Relationships to other Projects and Actions

The main relationships and synergies foreseeable are with the following projects:

GALEN (A2012) Generalised Architecture for Language Encyclopedias and Nomenclatures in Medicine
OPADE (A2027) Optimisation of Drug Prescriptions using Advanced Informatics
HELIOS (A2015) Hospital Object Software Tool
SHINE (A2035) Strategic Health Informatics Networks for Europe
EMDIS (A2006) European Mar-row Donor Information System. A Model for International Communication in Transplantation
KAVAS (A2019) Knowledge Acquisition, Visualisation, Refinement and Assessment System
DILEMMA (A2005) Logic Engineering in General Practice. Oncology and Shared Care
MARGOT (A2021) Medical Archive Generation with Object-Oriented Techniques
ESTEEM (A2010) European Standardised Telematic Tool to Evaluate EMG knowledger-Based Systems and Methods

Testbed and Verification

Two prototype systems will be implemented for aspects of patient management and treatment evaluation. The implementations will use and test the general software tools of GAMES II. The prototypes will first be evaluated in laboratory settings, comparing the advice from prototypes with that of panels of clinical experts and with relevant data.

After necessary refinement, the prototypes will undergo field testing in at least two clinical independent sites defined through suggestions of the clinical expert panel. The main focus will be usability and clinical relevance by clinical staff.

The goal of this testing is to evaluate the GAMES II methodology from a medical perspective and to recommend enhancements to the methodology. The emphasis is on evaluating GAMES II and not the specific prototypes developed.

A medical evaluation methodology will be drawn up to integrate the following aspects:

i) The results of classifying medical decision making activities
ii) The GAMES II methodology in operation (i.e. applying it to exemplar areas)
iii) Feedback from prototype systems.

By comparison of the results obtained in the exemplars, gained by implementing the different prototypes, the GAMES II methodology will be evaluated within the defined framework.


Sago SpA
Viale A. Gramsci 22
50132 Firenze

Participants (6)

Daidalou 36 P.o. Box 1385
71110 Heraklion
Rue Micheli-du-crest 24
1211 Geneve 4
Roeterstraat 15
1018 WB Amsterdam
University College London
United Kingdom
Gower Street
WC1E 6BT London
Università degli Studi di Pavia
Via Abbiategrasso 209
27100 Pavia
Universität Ulm
Oberer Eselsberg
89010 Ulm