LEXMED (Learning Expert System for Medical Diagnosis) supports
computer aided diagnosis of acute abdominal pain, particularly suspected appendicitis.
LEXMED combines probabilistic
reasoning based on the concept of Maximum Entropy
with the induction of probabilistic knowledge from data.
The basic knowledge source of LEXMED was a database of 15,000 patient
records which has been used to compute probabilistic rules. These rules were
combined with further explicit probabilistic rules elicited from medical experts
to cover cases not represented in the database.
Decision support in LEXMED is provided by the PIT (Probability Induction Tool), a probabilistic reasoning engine
designed to deal with
incomplete and uncertain knowledge. The PIT, which is based on the principle of Maximum Entropy, computes a
unique probability model while maintaining all necessary
additional assumptions as minimal and as clear as possible.
Based on the probability
distribution computed by PIT, the expert system proposes treatments
of relatively low average cost. LEXMED has been undergoing evaluation in the municipal hospital in Weingarten (Baden-Württemberg, Germany)
since June 1999 where an accuracy rate of 88% has been achieved.
Part of LEXMED data entry screen
LEXMED has been in use at the Weingarten Municipal Hospital
since June 1999. About 200 physicians, mainly from Germany, Austria
and Switzerland have registered with Lexmed.
LEXMED is to be extended to diagnose other causes of acute abdominal
pain. The developers of LEXMED are looking for partners in this