MENINGE: A medical consulting system for child's meningitis. Study on a series of consecutive cases

摘要

meninge is an expert system developed using Prolog and applied to the diagnosis and treatment of acute child's meningitis. The knowledge base is a set of unweighted logical rules and is modularly structured. In the discrimination phase between bacterial and viral meningitis, the system uses scores based on the statistical analysis of clinical cases. Therefore the system includes a linear model as a subsystem of an expert system. meninge was tested using a series of 212 consecutive clinical cases. The discrimination between the two types of infection was found in 98,2% of the cases. The correct germ had top priority in 92% of the cases. The treatment proposed by the system is sensible and efficient in 94,8% of the cases. 4,2% of the treatments are useless and 0,9% are insufficient. The medical team committed significantly more therapeutic errors (11,3% of the cases) than the system. These encouraging results allowed us to install meninge in a medical care unit to evaluate its clinical relevance.