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 estimate that half the anesthesia-related events could be detected with integrated monitoring using only 5 variables. A review of research using artificial intelligence/expert systems indicates limited potential for success using these tools alone for integrated monitoring in the operating room. 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.