Autonomously Learning Neural Networks for Clinical Decision Support.

The purpose of this contribution is to motivate the use of artificial neural networks in "intelligent" clinical decision support; to examine the advantages and limitations of two important classes of artificial neural network; to highlight the potential of intelligent decision support in the early diagnosis of heart attack; and to outline results which indicate, in particular, the potential of fuzzy ARTMAP network in this acute setting. The work to be described, demonstrates that this neural network can overcome problems in knowledge acquisition and portability, which may open the way to neural-network-based "apprentices" which learn autonomously whilst providing useful decision support.