MUM - Managing Uncertainty in Medicine

1. Objectives

This Project will extend over two years terminating in April 1996. The objectives are to contribute to the emerging unifying theory of uncertainty management, including probabilistic, fuzzy, and belief functional approaches; to incorporate these findings into diagnostic knowledge-based systems and intelligent systems of data analysis; to gain practical experience in medical data/knowledge processing by introducing software tools into selected medical institutions; and to develop Decision Support Systems for use in various clinical departments with the aim of providing better patient care, especially in those areas where there is a shortage of clinical expertise.

The reason for the existence of this Project is the need for the benefits of the advanced technologies of Western Europe to be made available to the participating countries within Central Europe with the firm aim of the eventual benefit to the patients.

2. Description of the Project

Inference under uncertainty is a fascinating topic on the borderline of Artificial Intelligence and Data Analysis. Knowledge, data, information - everything can be uncertain. There are several competing approaches to approximate reasoning (probabilistic, fuzzy, belief functions, many valued logics and so on) but, especially in Europe, contacts between the various schools are being developed and there is mutual enrichment. Significant work has been done in the Analysis of Data which is incomplete and/or carries a degree of uncertainty. In England, statistically reliable tree classifiers have been developed and tested, and applied in various fields and specialities. In the Czech Republic, a method of automated generation of hypotheses from incomplete data (GUHA) has been developed. In Spain, thorough theoretical work in fuzzy and many value-logic has lead to the construction of an inference engine (MILORD) with approximate reasoning. Theoretical work in the Czech Republic was done in uncertainty management using probability theory and Dempster-Shafer theory of evidence. Some rather important results have already been obtained and some experimental system shells constructed. It is proposed to continue and co-ordinate theoretical research in management of uncertainty and implementation activity in constructing software tools for building knowledge-processing systems working with uncertainty.

Much of the data occuring in the biomedicine by its very nature contains a degree of uncertainty There is considerable experience available in a few research centers to permit the handling of such data in a logical and scientific manner. Likewise with both hard and soft data there are newly developed techniques for data analysis. Expert systems, often referred to as Decision Support Systems, are now at the forefront in many areas, including medicine. It is proposed to use the experience of the project members and to use the above mentioned techniques in order to investigate the applicability in medicine and health care. typical areas where benefits can result are : intensive care, cardiothoracic surgery, diabetes control, hypertension treatment, health care management, and studies of association between risk factors and diseases.

Considerable work has already been done in Mechanizing of Hypothesis Formation and Testing in the presence of uncertainty; expert systems in Open Heart Surgery and Intensive Care have been developed and brought into routine use (England). MILORD was used to construct an expert system PNEUMON-IA for diagnosis of pneumonia (Spain). However, much work in these fields is needed and the metodology needs to be tested across national boundaries. This can be accomplished with significant benefits for patient care especially in those areas of medicine, and in those countries where there is a shortage of consultant expertise.

The proposers complement each other in a sound way; they cover various approaches to approximate reasoning and various attempts to create mature expert systems and data-analytic systems for medicine.