Expert systems in gynaecology: a model system to demonstrate how Bayes' theorem might work in a case of vaginal discharge

Publisher Website
Google Scholar
Abstract
Summary A computer model of the presenting symptoms and signs of a case of vaginal discharge is described. This model was used to set up 1000 hypothetical ‘cases’ each of which was analysed by Bayes' theorem to yield the probabilities of one of a series of diagnoses. The following conclusions were reached: (a) use of Bayes' theorem is considerably more efficient than a random choice of diagnosis; (b) Bayes' theorem is of no value in making rare diagnoses (for example cancer); (c) use of Bayes' theorem can lead to overdiagnosis of some common conditions and under-diagnosis of some common conditions with a low incidence of positive clinical features. The model system and its analysis illustrate some of the potential strengths and weaknesses of the application of ‘artificial intelligence’ in medicine.