| Abstract | Faced with overwhelming data, the clinician may focus on a subset of the signals, ignoring other significant information and overlooking problems. A monitor that integrates data from multiple sources could present a high-level synthesis, minimizing information overload and preventing fixation. To build a software system that supports such a monitor, we must achieve predictable, real-time performance, accomodate heterogenous appproaches to the many separable subproblems, and design a useful interface. The need for real-time performance is obvious: A monitor that does not run in real time cannot give early warning of life-threatening situations. Substantial computing power is needed to process many continuous waveforms, convert them into a qualitative representation, correlate information to identify the intermediate-level physiological state, and finally produce high-level diagnoses and suggestions. In this article, we describe the process trellis, a domain- and hardware-independent software architecture, and show its usefulness in building the Intelligent Cardiovascular Monitor, a real-time clinical decision-support system. |