Recognition of the deep anesthesia stage from parameters of the approximated entropy of EEG signal

The paper considers theoretical problems and results of experiments on recognition of anesthesia stages from electroencephalograms (EEGs) by methods based on analyzing the parameters of approximated entropy. It is shown that a discrete sequence of point entropy estimates can be efficiently used for describing regular and chaotic components of the EEG signal and for identifying a patient's functional conditions during surgical operations. The minimum length of the time interval is determined for which the obtained entropy characteristics of the signal are reliable and stable. The computational resources required for obtaining entropy parameters of the signal are estimated in online EEG analysis. On this basis, restrictions on the signal sampling frequency are introduced and the number of point estimates of the approximated entropy is selected for which the proposed analyzing algorithms can be implemented in medical systems with midlevel computers.