Eeg Based Automated Detection of Anesthetic Levels Using A

Continuous monitoring of the anesthetic drug dosage administered during a surgery is very important to avoid the patient's interoperative awareness due to inadequate levels of anesthesia. The traditional methods of assessing the anesthetic depth levels which are based on the qualitative physical signs such as heart rate, blood pressure, pupil size, sweating, etc are not very accurate as these autonomic responses may differ from patient to patient depending on the type of surgery and the anesthetic drug administered. Further it is also possible that these autonomic activities may get attenuated due to premedication. For these reasons, Electroencephalogram (EEG) based methods of anesthetic level detection have been gaining prominence in recent years. This paper discusses an automated detection of anesthetic levels based on EEG signals by using a special type of recurrent neural network known as Elman network. A frequency domain feature, namely, normalized spectral entropy is used to characterize the anesthetic levels. Experimental results show that Elman network is capable of detecting the three different anesthetic levels (low, medium, and high) with an overall accuracy rate of 99.6% which is better than the results reported earlier.