We evaluated three techniques of analyzing 18FDG-positron emission tomography-(PET) scans in 23 cases of presumed frontal lobe epilepsy (FLE): routine visual or "qualitative linear," "qualitative normalized," and quantitative normalized approaches. Patients were then classified as having pure frontal, probable frontal, frontoparietal and frontotemporal epilepsy based on prolonged surface EEG monitoring with video, magnetic resonance imaging (MRI), chronic intracranial recording (CIR), and results of surgical excision. Overall sensitivity and accuracy of the scans was 52 and 48%for qualitative linear analysis, which was equivalent to that of MRI, and 69 and 43% for qualitative normalized analysis. Quantitative normalized analysis had 96% sensitivity and 74 - 78% accuracy and also detected 9 of 11 (81%) abnormalities in nonlesional cases, improving routine sensitivity from 1 of 11 (9%). We conclude that qualitative linear (routine) analysis is inadequate for diagnosis of FLE lobe or "frontal-plus" epilepsies and does not add to the MRI scan. Because qualitative normalized images improve on routine analysis only slightly, quantitative techniques should be applied for preoperaive evaluations.