A comparative study of the output correlations between wavelet transform, neural and neuro fuzzy networks and BIS index for depth of anesthesia

Abstract

Anesthesia is a vital and important part of any surgical practice and allows doctors to operate safely and painlessly on a patients. The wide variety of available anesthetics allows anesthesiologists to select the most suitable type of anesthesia and anesthetic drug for a patient. Providing balanced anesthesia by testing the depth of anesthesia (DOA) is a way to know sudden awareness or increasing level of anesthesia during a surgery. In this paper, we present several methods to analyze the results of our experiments performed on 33 patients under coronary vessel surgery. In the first method, by applying the wavelet transform on EEG signal, a new index (namely WAI) is obtained, which shows the conscious level of the patient. In the second method, 10 features related to EEG signal during 10 second windows, such as, edge frequency, and beta ratio, are extracted and used by neural and neuro-fuzzy networks as inputs. Then, the value of DOA is calculated for each of the used algorithms. The correlation value of these methods, which is a criterion of the accuracy, is shown by the BIS monitor output. Simulation results show that the highest amount of correlation is achieved using neural networks with respect to BIS index.