|| Checking for direct PDF access through Ovid
Currently there is no consensus on the derivations that should be used for EEG monitoring during carotid endarterectomy (CEA). The aim of this study was to determine which derivations distinguish the best between patients requiring a shunt and patients who do not need a shunt. Four predefined frequency bands and two regimens for general anesthesia (isoflurane versus propofol) were used. EEG data (16 channels) were obtained from 152 EEGs recorded during carotid endarterectomy. Analog EEG signals of preclamp and clamp periods of 100 seconds were digitized to compute power spectra. Changes in power during clamping were calculated for all possible derivations in four predefined frequency bands and were expressed as Z-scores. For each derivation, the area under the receiver operating characteristics curve was calculated. Derivations with the greatest area under the receiver operating characteristics curve were considered to distinguish the best between the shunt and the nonshunt groups formed in retrospect on the basis of consensus between three independent and experienced board-certified electroencephalographers. The two different anesthetic regimens resulted in different patterns of EEG changes because of clamping. The optimal derivations to differentiate between the shunt and the nonshunt groups also differed for the two anesthetic regimens, although for both conditions, anterior head regions were especially preferred. The optimal derivations are given for each anesthetic regimen.