An Adaptive Fuzzy Logic Based Diagnostic Decision Support System for Critical Care Medicine

S.R.Liyanage, K.S.Walgama and C.D.A.Goonasekara


This paper describes the fuzzy logic based framework of the decision supporting software system “Critical Care Assistant”. As in the predecessor prototype system published in [5], fuzzy medical knowledge bases are used to model the uncertainty and the vagueness of medical concepts. Fuzzy logical reasoning mechanisms provide the inference process similar to [6]. This system is capable of handling four hundred and seventy clinical features and can diagnose seventy diseases commonly encountered in critical care medicine. The medical knowledge in the system is stored in the form of fuzzy logical relationships between clinical features and diseases. The fuzzy inference is performed using the min-max compositional rule to calculate indication relations expressing occurrence, confirmability and negation. These lead to confirmed and excluded diagnoses as well as diagnostic hypotheses. The diagnostic hypotheses are ranked according to their respective support values. The self learning mechanism incorporated to the system allows it to adapt to the prevalence of diseases in the environment in which it is being used. A novel neuro-fuzzy learning mechanism is attempted by means of the least mean square learning rule. The developed framework performs reasonably well in diagnosing the considered diseases. Further, it is expected that the performance will improve with the utilization of the self adaptive mechanism effectively.