作者: Alan L. Rector , Eugene Ackerman
DOI: 10.1016/0010-4809(75)90035-X
关键词:
摘要: Abstract Two different types of programs have been developed for computer-aided diagnosis. Either all the data can be input and diagnostic probabilities determined algorithmically or are entered sequentially, a set decision rules being used to rank tests remaining, and, if appropriate, stop sequence. This latter type may learning about A special has compiled generate, update modify files which then in other written BRANCHIT author language test strategies sequential From these, new heuristic function satisfies following requirements. It must possible evaluate rule using relatively little computer memory time. The should yield not only best but an ordering scoring available. There threshold is useful both as discriminator against poor stopping rule. And, finally, method reasonably robust minor errors base. final functions were selected work well with either costs. They use program dealing abdominal pain available on CDC 3300 Health Computer Sciences distributed capacity network.