A Framework for Verification of Fuzzy Rule Bases Representing Clinical Guidelines

The increase of expert knowledge is characterizing medical domain and determining a constantly growing and interacting number of relevant standardized specifications for care known as clinical guidelines. However, most clinical guidelines, especially when expressed in the form of condition-action recommendations, embody different kinds of structural errors that compromise their effectiveness. With this respect, this paper presents a framework to represent condition-action clinical recommendations as "IF-THEN" fuzzy rules and to verify the presence of some structural anomalies. In particular, we propose a method to detect redundancy, inconsistency and contradictoriness--a structural anomaly introduced in this paper for the first time--in a very simple and understandable way by using the concept of similarity between antecedents and consequents. Formalization in fuzzy degrees for these anomalies can be straightly interpretable as measurements suggesting how to suitably modify the clinical rules to eliminate or mitigate undesired effects. The framework has been assessed on a relevant sample set identified from the clinical literature with profitable results.