Journal of clinical monitoring
-
Review
Integration of monitoring for intelligent alarms in anesthesia: neural networks--can they help?
Although there has been a decrease in the number of anesthesia-related critical incidents, there are still opportunities for further improvement. We discuss the potential of integrated monitoring and artificial neural networks as a means of vigilantly watching for patterns in multiple variables to detect incidents and reduce false alarms. ⋯ We present artificial neural networks as an approach that is more suited to the type of multivariable monitoring and pattern recognition required. Along with rule-based artificial intelligence, these now have the potential to help develop innovative monitoring in the operating room.
-
We developed a two-compartment model to simulate neuromuscular function and heart rate following the administration of four nondepolarizing neuromuscular blocking agents (atracurium, vecuronium, pancuronium, and d-tubocurarine), three neuromuscular block reversal agents (edrophonium, neostigmine, and pyridostigmine), and two anticholinergic agents (atropine and glycopyrrolate). Twitch depression, train-of-four ratio, and heart rate were modeled during fentanyl, halothane, enflurane, or isoflurane anesthesia, optionally supplemented with nitrous oxide. Simulation results, compared with published values for each drug, fell within the clinical accuracy range (onset time 6.1 +/- 3.9% [mean +/- SEM]; duration, 1.7 +/- 3.5%, 50% effective dose, 0.5 +/- 5.7%; and 95% effective dose, 2.1 +/- 1.1%). ⋯ When inhalational agents are given concomitantly, the task becomes even more difficult, since potentiation changes with anesthetic uptake. Recurarization, tachycardia, or bradycardia may be seen with the simulation if an improper drug regimen is followed. Concurrent simulation of two identical patients allows comparison of different modes of administration, choice of anesthetic agents, and drug doses.
-
The need to incorporate alarms in monitoring systems is related to the growing complexity of monitoring and the large number of variables. For sophisticated alarms, information about the inputs to the patient is of importance; for example, clinical interventions such as drug administration and ventilation readjustment need to be known to the monitoring system. Alarms are triggered by signals or signal features that exceed thresholds. ⋯ Approaches to determine such levels automatically are discussed in this article. Most promising seems the multiple signal approach using an expert system. It seems reasonable to expect that information concerning alarm limits, needed for the operation of knowledge-based alarm systems, may come from integrated departmental data bases.
-
Patients receiving intraspinal opiates should be monitored in the intensive care unit for at least 24 hours to prevent potentially lethal outcomes. These include respiratory depression caused by sequestration of the morphine in the cerebrospinal fluid and migration of epidural catheters in the subarachnoid or intravascular space. At this time, most hospitals are not equipped or staffed adequately to guarantee the safety of these patients outside the intensive care unit.
-
The pulse oximeter, a widely used noninvasive monitor of arterial oxygen saturation, has numerous applications in anesthesiology and critical care. Although pulse oximetry is considered sufficiently accurate for many clinical purposes, there are significant limitations on the accuracy and availability of pulse oximetry data. This article reviews both the clinical uses of the pulse oximeter and the limitations on its performance. The pulse oximeter is generally acknowledged to be one of the most important advances in the history of clinical monitoring.