抄録
Bayes' theorem and multivariate analyses have been used to simulate the process of medical diagnoses. However, in these studies some mathematical models had to be assumed, which are not necessarily related to the real process of physicians' diagnoses. Because of this fact, the results of these studies Were sometimes difficult to evaluate.
Physicians usually make decisions according to some rules which were obtained in their, long experiences. If there are some methods which can simulate the same process as physicians, the methods will be more acceptable in the field of medical decision making by computers. “MYCIN” developed by Shortliffe and others demonstrated this possibility by using the methods of artificial intelligence techniques. On the basis of similar considerations, we have developed a program named “MECS-AI” (Medical Consultation and Education System by Artificial Intelligence) which makes diagnoses and suggests treatments in cardiovascular diseases.
In this study rules which physicians use for the diagnosis and treatments were collected. The rules were organized in the form of premises and actions. When some data of a patient are given to the system, the system searches the premises of the rules and find the ones to apply to take proper actions. Then with this tentative conclusion, it continues to search for the next rules until it reaches some conclusions. The program is written in LISP language.
This system is now being evaluated by cardiologists of the University of Tokyo Hospital. It seems that this system has wide area of applications in physicians' consultation and medical education.
