Reasoning about Therapy from a Physiological Model

In many medical domains a causal physiological model provides a knowledge base of relationships useful for reasoning about diagnosis, management, prognosis, and basic understanding of the processes. This paper reports on our e orts to develop algorithms for reasoning about the potential e ects of therapy in the context of a system designed to assist the physician in diagnosis and management of patients with heart failure. The primary result has been the development of an algorithm based on signal ow analysis for predicting e ects and the implementation of this algorithm as a mechanism for handling multiple e ects, changes over time, non-linear relationships, and in providing explanations. When applied to a model of the cardiovascular system we are developing this methodology predicts drug e ects consistent with the medical literature.