EXPERIENCE WITH BAYE'S THEOREM FOR COMPUTER DIAGNOSIS OF CONGENITAL HEART DISEASE *

The problem of computer diagnosis is an example of the general problem of pattern recognition. It is the purpose of this paper to describe a mathematical model of medical diagnosis fundamentally based on Baye's Theorem but incorporating certain expansions and modifications which were added as the result of experience in applying the model to the diagnosis of congenital heart disease.' :l A computer program has been written which permits translation of statistical data concerning the incidence of symptoms in diseases into a prediction of the probability of particular patients having particular diseases. Refinements in the mathematical model of diagnosis have been introduced as the result of experience in the application of this program to a patient population. On the other hand, experience with the program has yielded insight into the intricacies of diagnosis in a particular field of medicine and has resulted in improved accuracy of diagnosis, both by the computer program and the participating physicians who feed patient data to the computer and who review the computer's diagnoses. The expanded form of Baye's Theorem, which forms the basis of this mathematical model and has been used for the diagnosis of congenital heart disease in this laboratory over the past two and one-half years, is shown in: