Multi-Agents for Biomedical Monitoring and Analysis

摘要

This paper describes how a multi-agent system displays flexibility, system integration and interactivencess desirable for biomedical modeling and analysis. This approach combines physiological modeling, distributed agent-based processing and artificial intelligence methods. A cardiovascular model has been implemented as a rigorous testbed application within the SimAgents multi-agent system (Mabry). Three classes of agents are utilized, flow agents, signal agents and monitoring agents. These agents have a local perception of their world, are location sensitive, perform coordination and goal-oriencted actions within their synthetic cardiovascular world. Flexibility and interactiveness found in the system support study of individual pathophysiologies, interactive introduction of events or state changes, dynamic temporal states and return of analysis based on composite calculations and dynamic causal networks. A multi agent-based approach also facilitates a "live" environment, that is all of these feature during runtime with-out restarting the system to accommodate change of state. A high degree of expressivencess results from the close alignment of computing processes to actual physiological behaviors. There are several aspects of this system, making it attractive for "real-world" usage. Applications developed in this paradign have potential to be adapted for laboratory or clinical usage such as adaptive monitoring, interactive testing, device prototyping, device validation and integrated analysis.