We constructed a syntactic parsing system as an intelligent reporting
interface for radiological diagnosis of the alimentary tract, into whi
ch radiological diagnoses could be put using ordinary language after w
hich they were recognized and categorized automatically. The input dia
gnoses were limited to disease names, including both pathological and
anatomical entities, and purely radiological description was excluded.
The system was equipped with a table of possible diagnostic expressio
ns composed of hundred of terms used in routine radiological work. A t
otal of 95% of the CT and MRI diagnoses actually input by three radiol
ogists were correctly recognized and corresponded to the code of the m
odified ''Index for Radiological Diagnoses of the American College of
Radiological Diagnoses of the American College of Radiology.'' The rem
aining 5% were not categorized because of either over- or underdescrip
tion. This method was quicker and more user-friendly than hierarchical
menu selection by mouse and/or Windows.