The purpose of this prospective study was twofold: 1) to determine the sensitivity and specificity of computer-derived density spectral array in detecting analog electroencephalographic (EEG) ischemic pattern changes during carotid artery cross-clamping in patients undergoing carotid endarterectomy; and 2) to assess the ability of density spectral array to identify such changes in comparison with the degree and type of change seen in the analog EEG ischemic pattern. Sixteen channels of anteroposterior bipolar and two to four channels of referential electroencephalography with four channels of density spectral array were monitored simultaneously during carotid endarterectomy in 103 patients under general anesthesia. Two "observers" interpreted the density spectral array and the analog electroencephalograms, one during and immediately after the operations and the other 6 months after completion of all surgery. Analyses were conducted to establish both the number of patients with analog EEG ischemic changes and the number of ischemia events during carotid artery cross-clamping. Observer A indicated that the density spectral array identified analog EEG ischemic changes in 21 of 29 patients, for a sensitivity of 72% (specificity 99%), whereas Observer B's results showed that the density spectral array identified analog EEG ischemic changes in 16 of 27 patients, for a sensitivity of 59% (specificity 96%). Density spectral array detection of analog EEG ischemic changes based on severity classifications were 61% and 18% in the mild group, 70% and 71% in the moderate group, and 95% in the severe group, indicating a relationship between density spectral array sensitivity and severity of analog EEG ischemic change, with p = 0.02 and p = 0.004 for the two observers. The kappa statistics for observer reproducibility were highly significant, with k = 0.95 for analog EEG ischemic changes and 0.85 for density spectral array changes. It is concluded that density spectral array does not reliably detect mild analog EEG pattern changes of cerebral ischemia and is not a reliable substitute for 16-channel analog EEG monitoring of cerebral ischemia during carotid endarterectomy.