A generic expert system framework with fuzzy Bayesian inference for medical classification and diagnosis in cardiology

Medical knowledge is expanding fast and it is difficult for general practitioners to remain abreast of all medical domains. Also, access to domain specialist is limited due to availability and geographical constraints. In many situations the diagnosis in [sic] upon the decision of the general practitioner and in cases this has resulted in the problem of patient's misdiagnosis. The purpose of this research is to create an expert system as a decision support model which is capable of risk analysis for diagnosis based on the patient's demography and laboratory tests. The expert system is designed in compliancy with medical communications protocol such as HL7 and can be integrated to any HL7 compliant Electronic Medical records system to provide more intelligence in diagnosis. Using linear scoring models and Fuzzy logic, the patient's demography and laboratory results will be used as rule bases. Such knowledge will be used as priors for a Bayesian engine to create the diagnostic spaces. Patient's information is compared in the space and the general practitioner can select between the possible hypotheses. Each diagnostic decision will be associated with a risk value. Using such scoring model provides a new semantic in diagnosis by providing risk values for every diagnosis made and by suggesting the most suitable treatment. Unlike many other existing expert systems, the architecture is designed in a generic standard which provides the capability to use the system for all medical domains. Achieving this generality has been a major goal achieved and its details are discussed in this document.