Automatic Analysis and Monitoring of Burst Suppression in Anesthesia

Objective.We studied the spectral characteristics of the EEGburst suppression patterns (BSP) of two intravenous anesthetics,propofol and thiopental. Based on the obtained results, we developed amethod for automatic segmentation, classification and compactpresentation of burst suppression patterns. Methods.The spectralanalysis was performed with the short time Fourier transform and withautoregressive modeling to provide information of frequency contents ofbursts. This information was used when designing appropriate filters forsegmentation algorithms. The adaptive segmentation was carried out usingtwo different nonparametric methods. The first one was based on theabsolute values of amplitudes and is referred to as the ADIF method. Thesecond method used the absolute values of the Nonlinear Energy Operator(NLEO) and is referred to as the NLEO method. Both methods have beendescribed earlier but they were modified for the purposes of BSPdetection. The signal was classified to bursts, suppressions andartifacts. Automatic classification was compared with manualclassification. Results.The NLEO method was more accurate,especially in the detection of artifacts. NLEO method classifiedcorrectly 94.0% of the propofol data and 92.8% of thethiopental data. With the ADIF method, the results were 90.5% and88.1% respectively. Conclusions.Our results show thatburst suppression caused by the different anesthetics can be reliablydetected with our segmentation and classification methods. The analysisof normal and pathological EEG, however, should include information ofthe anesthetic used. Knowledge of the normal variation of the EEG isnecessary in order to detect the abnormal BSP of, for instance, seizurepatients.