Fuzzy Logic in Clinical Decision Support Systems

Abstract

The diagnostic process is an important activity that is aimed at explaining and treating a patient's health problem. However, this process is subject to diagnostic errors that jeopardize patient safety and healthcare quality. In response to this, diagnostic clinical decision support systems (CDSS) have been developed to aid physicians in clinical reasoning and decision making. Diagnostic systems that have been successfully implemented employ fuzzy logic techniques.

The aim of this thesis is to examine how fuzzy reasoning can be used in these systems. This has led to the following research question: How can fuzzy logic be incorporated in diagnostic CDSS? In order to answer this question, literature on the operation of CDSS and the fundamentals of fuzzy logic has been reviewed. This literature has been used to create an example of a CDSS that uses fuzzy reasoning for the diagnosis of pneumonia (lung inflammation). Finally, the utility of using fuzzy logic in healthcare has been evaluated by discussing the difference between fuzzy and classical reasoning in CDSS.

The CDSS for pneumonia diagnosis has demonstrated that fuzzy logic can be incorporated in CDSS by employing a fuzzy expert system that executes the reasoning process. Furthermore, it has been found that fuzzy-based systems are characterised by robust outputs -- which means that a change in input values will not cause a drastic change in the output. This feature is desirable in healthcare, as it should be prevented that minor alterations in symptoms, patient data or test results cause a major alteration in the diagnosis proposed by the CDSS. In addition, fuzzy CDSS are capable of handling contradictions. This property enables systems to alter their decisions as they obtain new information.

Further research could examine different techniques to implement a fuzzy CDSS. Additionally, the CDSS for pneumonia diagnosis could be expanded by entering more symptoms and patient data into the system and by extending the rules in the knowledge base. The accuracy of the CDSS could then be evaluated by comparing the results with existing medical data about pneumonia diagnosis in patients.