DIAGAID: a connectionist approach to determine the diagnostic value of clinical data

摘要

In clinical medicine the diagnosis is usually based on several signs and symptoms and some laboratory test results. It would be of great benefit if we could learn the diagnostic criteria from example cases and represent systematically the most important findings supporting or rejecting the diagnosis. In rare diseases where the physician may not be familiar with the ailment the scoring of symptoms would be very useful. In this paper we describe a connectionist approach based on Minsky-Papert's perceptrons for evaluation of the information value of clinical data in diagnosing the Nephropathia epidemica. The scores are compared with those from the Bayesian approach and from the evaluation by an experienced clinician.