Clinical communication and telemedicine

Abstract

Human operators of monitoring and control systems typically perform knowledge intensive tasks requiring human intervention in diverse and sometimes critical situations. The cognitive overload to which they are submitted may motivate human errors which compromise safety and performance. Expert Systems offer the means to codify and automate the use of the knowledge of the best experts in a given domain, and may provide support to monitoring, diagnosis and control tasks. In this text we will make an overview of what expert systems are, what they may offer and what they don't. We will present their typical architecture and the most common knowledge representation formalisms and reasoning mechanisms, and discuss how they may deal with uncertainty. The lifecycle of an Expert System will also be presented, special attention being devoted to knowledge acquisition activities. An overview of some applications of expert systems to monitoring and control will be given, as well as some alternative approaches. INTRODUCTIONThe control needs of complex and expensive processes have motivated the development of sophisticated computational systems which allow the monitoring and control of complex processes by small crews of specialised human operators. Typically, these operators follow the process evolution through a set of instruments connected to a computer system which interacts with sensors, alarms and actuators. The control system is intended to minimize human intervention and to provide the operators with the maximum information on the controlled process. The operators' task essentially consists in trying to minimize failure risks while maximizing performance. The required degree of expertise depends on the complexity of the process and of the control system, on the importance, to people and equipment safety, of keeping the process in normal operation, and on the losses that an incorrect process control or its interruption may cause.In normal steady-state conditions, the control system takes charge of keeping the system in a given set point, and the operators confine themselves to monitoring the most important variables. Human intervention is required during system start-up and shut-down, and also to regulate the process to given resources and needs.Another, more critical, kind of situation occurs when equipment or process failures, or human errors, originate anomalous conditions. In such circumstances the control system presents a great quantity of information to the operators in a short time, originating in the most diverse sources, with the process conditions quickly changing and alarms escalating. In such a stressed situation, operators must discriminate actual faults from false alarm states, perform malfunction diagnosis, and plan and accomplish corrective actions 1 . This requires them to intensively resort to their knowledge on the process and on the control system. In particularly complex, stressed or time-critical situations, they 1 According to [1], "monitoring activities track process variables intelligent...