UNLABELLED: There are many closed-loop control systems for muscle relaxants reported, but only a few could cope with the introduction of the latest shorter acting neuromuscular blocking drugs. These new muscle relaxants such as mivacurium require a fast adapting closed-loop system for controlling an adequate infusion. METHODS: After approval of the local ethics committee and having the patients' informed consent a total number of 75 patients [ASA I and II] were included in the study and assigned either to a training-, prediction-, prediction-/feedback- or a validation phase, as needed. Anaesthesia was induced and maintained with propofol in a TCI-mode with a plasma level of 3 to 5 micrograms/ml and 0.1 mg fentanyl boli as needed in all patients. In the last validation phase, having 20 patients, the prediction error and the error of the whole system was taken and analysed. RESULTS: A closed-loop system using a neural network as a predictor could be established. In the final validation phase consisting of 20 patients the mean square prediction error was found to be 0.1% +/- 0.2% [mean +/- SD]. The mean square error of the whole system was 0.55% +/- 0.59% [mean +/- SD]. CONCLUSIONS: A closed-loop system for control of a mivacurium infusion could be established. The system proofed to be reliable for a closed-loop infusion of mivacurium in order to maintain a predefined degree of neuromuscular blockade of 95% during routine surgery. The performance of the described controller is comparable to all recent attempts and could therefore be useful for scientific studies. It should be further validated and established for other muscle relaxants, as well.
UNLABELLED: There are many closed-loop control systems for muscle relaxants reported, but only a few could cope with the introduction of the latest shorter acting neuromuscular blocking drugs. These new muscle relaxants such as mivacurium require a fast adapting closed-loop system for controlling an adequate infusion. METHODS: After approval of the local ethics committee and having the patients' informed consent a total number of 75 patients [ASA I and II] were included in the study and assigned either to a training-, prediction-, prediction-/feedback- or a validation phase, as needed. Anaesthesia was induced and maintained with propofol in a TCI-mode with a plasma level of 3 to 5 micrograms/ml and 0.1 mg fentanyl boli as needed in all patients. In the last validation phase, having 20 patients, the prediction error and the error of the whole system was taken and analysed. RESULTS: A closed-loop system using a neural network as a predictor could be established. In the final validation phase consisting of 20 patients the mean square prediction error was found to be 0.1% +/- 0.2% [mean +/- SD]. The mean square error of the whole system was 0.55% +/- 0.59% [mean +/- SD]. CONCLUSIONS: A closed-loop system for control of a mivacurium infusion could be established. The system proofed to be reliable for a closed-loop infusion of mivacurium in order to maintain a predefined degree of neuromuscular blockade of 95% during routine surgery. The performance of the described controller is comparable to all recent attempts and could therefore be useful for scientific studies. It should be further validated and established for other muscle relaxants, as well.