DESIGNING A MODEL TO ESTIMATE THE SEVOFLURANE DOSE FOR A PATIENT UNDER THE GENERAL ANAESTHESIA BY USING ADAPTIVE-NETWORK-BASED FUZZY INFERENCE SYSTEM

Abstract

The field of Depth of Anaesthesia (DOA) is a very challenging area for neuro-fuzzy control since direct measurements are unavailable. During anaesthesia, the blood pressures (BP), the mean arterial blood pressure (MAP) and the heart rate (HR) are monitored to maintain hemodynamic stability and to assess the level of consciousness. The purpose of this study is to find the best input-output definitions in the Adaptive-Network-based Fuzzy Inference System (ANFIS) to control the Sevoflurane dose to patient under the general anaesthesia with the classical MAP and HR parameters. The best models have been found among many possible input combinations. This study helps to provide an alternate control for the dose of Sevoflurane which is widely used as an anaesthetic agent. The models have been trained and validated by clinical data. The results show that the patients can be modelled by ANFIS if sufficient HR and MAP data are provided. Furthermore, the model performance could be increased if the patients are grouped as adults and children. The performance (up to 0.99) in this study is comparable to recent works in similar subject which detect DOA by Electroencephalograms (EEG).