We have developed a consultation system on a personal computer to help radiologists in clinical decision makings. The system was applied to the CT diagnosis of pulmonary nodules (theme 1) and to the prediction of the outcome of radiation therapy to lung cancers (theme 2). It is our policy to construct a knowledge base with the cases which have been examined or treated in our department and for which detailed information has been derived. Knowledge base for theme 1 contains 127 pulmonary nodules, and that for theme 2 only 40 cases, which we think as a preliminary one. A knowledge base is a collection of the name of descriptive criteria, their significance, names and distributions of diseases to be differenciated. Significance of findings was appreciated by multivariate analysis (quantification analysis type 2).
In system design, it was considered that the system is expected not o n ly to tell a doctor what the diagnosis is but to communicate with him about the daignostic processing, it should be friendly to clinicians, and it can be easily accessed at the site of film reading.
