An uncertainty aware medical diagnosis support system Krzysztof Dyczkowski, Anna Stachowiak, Andrzej Wójtowicz, Patryk Żywica Department of Imprecise Information Processing Methods Faculty of Mathematics and Computer Science, Adam Mickiewicz University Umultowska 87, 61-614 Poznań, Poland chris@amu.edu.pl Abstract. In the paper we describe a system that store and process uncertain data in such a way as to be able to obtain information essen- tial to make an effective diagnosis while also indicating the uncertainty level of that diagnosis. We consider the problem of incompleteness and imprecision of medical data and discuss some issues connected with such kind of information - like modeling, making decision that is aware of the imperfection of data, evaluating results in the context of uncertain med- ical data. As an example we describe a method of supporting medical decision that is based on interval-valued fuzzy cardinality and that was implemented in the OvaExpert system. Keywords: diagnosis support, medical data, uncertainty, imperfect in- formation 1 Introduction Computer decision-making systems are highly effective in terms of prognosis when solving many diagnostic problems. This is true especially for common dis- eases for which there is access to large number of cases. The situation is less satisfactory for diseases which are less common and thus the access to large number of well-depicted cases is limited. Lack of centralized system for gath- ering uniform data from many medical institutions is also a problem. If such databases exist they are gathered in a specific medical center and are not ac- cessible to others. Another problem is lack of access to full required diagnostics (e.g. due to unavailability of proper diagnostic equipment or high cost of diagnos- tic examinations), which contributes to ambiguities and omissions in patient’s record. In addition, by their very nature, medical descriptions are often impre- cise and ambiguous. In most cases, they are descriptive and terminology used in them is not standardized. Their quality often depends on the education of the doctor (including the center where he or she was educated) as well as the doctor’s experience. The existing situation calls for the use of unconventional data modeling and reasoning methods. It requires methods factoring in both the imprecision and incompleteness of the data. Those methods must ensure high