This paper discusses granular models of medical diagnostic rules which is an extension of rough set rule model. Medical diagnostic reasoning is characterized by three processes: focusing mechanism, differential diagnosis and detection of complications. First, focusing mechanism uses a set of symptoms which are always observed by almost all the cases of a candidate and if a case does not include any one of them, the candidate will be rejected. Second, from selected candidates, a set of symptoms which are highly observed in the cases are used for confirming the differential diagnosis. Finally, detection of complications is a set of symptoms whose occurrence of a candidate is very low but are very important for diagnosis of other diseases. These rule models can be easily described by an extension of rough set model: supporting sets of the first two sets of symptoms correspond to upper and lower approximations of a target concept. The final one is described by interrelations between a target concept and other concepts, which will be a new type of information granules.
展开▼
