What is good diagnostic practice? The answer is elusive for many medical students and
equally puzzling for those trying to build effective medical decision support systems.
Much of the problem lies in the difficult of 'getting at' diagnosis. Expert diagnosticians
find it difficult to introspect on their own strategies, thus making it difficult to pass on
Traditional knowledge acquisition methods are designed for gathering static domain
knowledge and are inappropriate for the acquisition of knowledge about the diagnos¬
tic 'task'. More advanced knowledge acquisition methodologies, particularly those which
focus on the modelling of problem-solving knowledge seem to hold more promise, but are
not sufficiently practicable to allow anyone other than a knowledge engineer to operate
directly. Given the difficulty experts have in accessing their own diagnostic strategies
what is needed is a tool which would enable diagnosticians themselves to directly formu¬
late and experiment with their own methods of diagnosis.
This research describes the development of a knowledge acquisition methodology geared
specifically towards the exposition of medical diagnosis. The methodology is implemented as a toolkit enabling exploration and construction of medical diagnostic models
and production of model-based medical diagnostic support systems. The toolkit allows
someone skilled in diagnosis to articulate their diagnostic strategy so that it can be used
by those with less experience.