An expert system for teaching cytopathologic diagnosis of lung cancers.

ABSTRACT

An expert system for teaching cytopathologic skills is described. It uses Tao's cytologic criteria as graded feature parameters for comparison against a set of expert descriptors of known carcinomas. The system scores each diagnosis according to the degree of agreement along each feature axis. The system also allows the user to compare his/her description with that of the expert for any of the tumor entities. The expert knowledge base is abstracted into a numerical matrix within the system. This matrix is used for parameter comparison and score interpretation. This design is simple and compact, allowing portability and broad compatibility for small personal computers. Although testing is only at the initial stages, preliminary studies have shown that the system provides reproducible, consistent results. It is useful as a teaching tool for students, residents and staff pathologists.