
Abstract
Hemodynamic abnormalities such as hypovolemia typically progress through a sequence of discrete clinical phases or “scenes” (e. g., intravascular volume depletion, vasoconstriction, hypotension). Each scene can be defined by a cluster of hemodynamic trends. A natural approach to modeling the process of hemodynamic monitoring involves identifying these scenes and the temporal relationships among them. This approach has been utilized in the development of DYNASCENE, a parallel programming implementation of a computer-based intelligent hemodynamic monitor. This paper discusses: (1) The rationale for utilizing sequential clinical scenes to represent knowledge of hemodynamic behavior, (2) the design of the DYNASCENE system, and (3) preliminary tests of the DYNASCENE system.
Key-Words
Expert Systems - Computer-Assisted Diagnosis - Knowledge Representation - Real-Time Systems