While it is generally accepted that quantitative EEG analysis is less subjective and can extract information that would be difficult or impossible to obtain by visual means, it remains uncertain whether it can enable the clinician to make more informed decisions about individual patients. This project will address this question in the contex of selecting patients for epilepsy surgery based on non-invasive EEG findings. The hypothesis is that optimal computerized scoring of scalp- recorded ictal EEGs with allow as accurate or more accurate lateralization and localization of seizure activity than expert visual reading. Two existing computerized seizure analysis techniques with be rigorously evaluated for their lateralization and localization capabilities and compared with visual scoring by expert raters. To do so, we will retrospectively analyze the scalp recorded icta EEGs of a large group of well-documented patients that have been evaluated for surgery will be analyzed retrospectively. Blinded visual and computerized scoring will be applied to the seizures of those patients whose seizure lateralization and localization was verified by longterm surgical follow up and/or the results of invasive EEG recording. Each technique will be applied independently to the same EEG data and the analysis will be converted to lateralizations and localizations, and two-by-two contingency tables of true- and false- positive and negative lateralization/localization with be tabulated. Plots of true-positive proportion vs. false-positive proportion constructed from this analysis will assess the potential accuracy of each technique in the absence of fixed decision criteria and then be used to identify optimal selection criteria for specific applications.
| Battiston, John J; Darcey, Terrance M; Siegel, Adrian M et al. (2003) Statistical mapping of scalp-recorded ictal EEG records using wavelet analysis. Epilepsia 44:664-72 |
