Coping with uncertainty [in medical diagnostic computing]

A number of different kinds of uncertainty need to be faced in computer-aided diagnosis. For example, there may be both qualitative and quantitative doubts concerning the background knowledge, and data on a patient may be fragmentary or of dubious quality. The author describes a methodology intended to handle both the propagation of evidence through the network as data on a new case is obtained, and to allow monitoring, criticism and adaptation of both quantitative and qualitative aspects of the model as the database accumulates