Development of an expert for medical diagnosis most common lung and ear nose throat(ENT) diseases

Abstract

An expert system for medical diagnosis of the most common lung and ENT (earnose-throat) diseases was developed. This system diagnoses tuberculosis, pneumonia, lungcancer, chronic obstructive pulmonary system diseases (COPD), acute sinusitis, chronicotitis media, chronic tonsillitis and chronic bronchitis.The system runs under PC and PC-Compatible machines and MS-DOS operatingsystem or Windows 3.x. It has been written in both C and Pascal programming languages.The knowledge base of the system consists of medical knowledge involving illnesssymptomrelationships which were established from questionnaires and archive work. Theinference engine of the system uses a rule-based approach. Uncertainty management isachieved by two different methods llsing either Subjective Probability Theory (Bayes'Theorem) or the certainty factor meth09. The expert system produces two different reports.The program utilizing the Bayes' theorem reports the illness probabilities as percentagevalues, while the one using the certainty factors (CF) method assigns values in the range of-1 to + 1 to the illnesses with -1 implying absolute health and + 1 implying absolute illness.Comparison of the expert system results with those obtained from the doctors forlung diseases such as iung cancer, COPD, pneumonia, and tuberculosis, show that thecomputer diagnosis can be as good as 60 per cent or better in a population of 62 patients.The expert system diagnosis ofENT diseases, however, can be as reliable as 88 per cent to100 per cent since these diseases have no symptoms in common.