NESTOR: A Computer-Based Medical Diagnostic Aid That Integrates Causal and Probabilistic Knowledge

摘要

In order to address some existing problems in computer-aided medical decision making, a computer program called NESTOR has been developed to aid physicians in determining the most likely diagnostic hypothesis to account for a set of patient findings. The domain of hypercalcemic disorders is used to test solution methods that should be applicable to other medical areas. A key design philosophy underlying NESTOR is that the physicians should have control of the computer interaction to determine what is done and when. In order to provide such controllable, interactive aid, specific technical tasks to be addressed. The unifying philosophy in addressing them is the use of knowledge-based methods within a formal probability theory framework. A user interface module gives the physician control over when and how these tasks are used to aid in diagnosing the cause of a patient's condition. This dissertation presents the problems that are addressed by each of the three tasks, and the details of the methods used to address them. In addition, the results of an evaluation of the hypothesis scoring and search techniques are presented and discussed.