Network Database Representation of Medical Knowledge Production Rules: An Application to Pediatric Diagnosis

This thesis is a representation of a specific set of artificial intelligence production rules with the CODASYL database management system "SEED" available at the University of Pennsylvania. Among the advantages of this database representation over embedding the rule in a higher level language program are: 1. The database management system (DBMS) additionally performs commercial and information retrieval functions relating to the environment in which the artificial intelligence representation is made. 2. The end user interface is potentially friendly, permitting easy updates and queries using various standard data manipulation routines and higher level processors of the DBMS. The application area chosen for this demonstration in a simple hospital information system covering medical diagnosis and the patient-doctor interface. Indicative preliminary decision rules for diagnosis of a certain class of pediatric ailments were obtained from Dr. B. Athreya of the Childrens' Hospital of Philadelphia (affiliated to the University of Pennsylvania). The assistance of Dr. Athreya is gratefully acknowledged. It is important to note that the clinical data used here is merely indicative and has been used purely as a model. The same is not intended to represent correct or incorrect handling of diagnostic situations.