A comparative study of three expert systems for blood pressure control

摘要

Regulation of mean arterial pressure (MAP) using sodium nitroprusside (SNP) infusion is common in many hospitals. We comparatively evaluated the performance of three types of expert systems controllers to automate this task: rule-based, fuzzy, and artificial neural network. For meaningful comparisons the three systems were based on the same set of rules. Their performance was tested on a nonlinear blood pressure model derived and scaled from canine data that simulated typical patient responses to the drug. The controllers were tested for differing patient sensitivities, baseline drift, and noisy blood pressure readings. The controllers were able to regulate the MAP in the target region about the set point for more then 90% of the time. The rule-based controller reduced MAP the fastest, while the fuzzy and neural controllers regulated MAP better over longer periods. Overall, the performance of the expert system controllers, while being intuitive and easier to design and implement, was comparable to more traditional controllers.