Abstracts

Rationale: Scalp EEG with quantitative EEG (QEEG) trends save time but come at the expense of reduction in sensitivity for seizure detection. QEEG methods had not been used for seizure screening in intracranial EEG recording.  Intracranial EEG generate larger files compared to standard scalp EEG. The larger files usually mean a need for a slower page-by-page review. We explored the utility of using beta software for the semi-automatic creation of QEEG panels for intracranial EEG for seizure screening and tested it against page-by-page EEG review in terms of time saving and sensitivity. _x000D_
Methods: We identified five consecutive intracranial EEG recordings for analysis. All recordings were processed quantitatively using Persyst EEG software (Persyst 14, Rev. C). The EEG raw data was processed using Persyst’s Proprietary Rhythmicity engine and was displayed using individual custom templates. Custom templates were created using a beta persyst software for creating channel groups.  The created QEEG panels were viewed in 15-minute windows.  _x000D_ _x000D_ Two independent reviewers analyzed the studies and screened them for seizures. One reviewer screened the records for seizures using a “hybrid method.” The hybrid method consisted of reviewing the studies using two monitors (raw EEG viewed in 10-second window and QEEG viewed in 15-minute window). The reviewer screened the QEEG panels for a seizure pattern (using rhythmicity spectrogram patterns from scalp EEG seizures). Patters suggestive of seizure activity were confirmed by a review of the relevant portion of the raw EEG. Events identified as seizures on QEEG and confirmed by the review of the raw EEG were logged as seizures. The second reviewer screened the records for seizures using the default review methods of reviewing the entire EEG page-by-page. Raw EEG was reviewed in a 10-second window. The setting of EEG review mimicked the daily adult EMU review station set up (two 24” monitors).   _x000D_ _x000D_ Recorded parameters: The total duration of the review time, number of seizures per file, duration of seizures, and presence or absence of post-ictal EEG suppression on raw EEG were recorded. _x000D_ _x000D_ Descriptive statistics: The total duration of the review time, number of seizures per file, and the average duration of seizures were reported. _x000D_
Results: Twenty-five files from 5 patients were reviewed by the two reviewers independently using the methods described above.  _x000D_ _x000D_ Both reviewers detected 18 seizures independently. The detection rate (sensitivity) of the hybrid method was 100% compared to the classic method for the detection of seizures. The nature of the hybrid review guaranteed that possible seizure events detected on QEEG were not erroneously classified as seizures resulting in a specificity of 100%. The average review time per hour of EEG was 10 seconds for the hybrid review, while the average time was 38.2 seconds per hour for the classic method. _x000D_
Conclusions: This study supports the utility of using quantitative EEG for identifying seizures using commercially available software with substantial time savings and excellent sensitivity. The obstacle of creating custom quantitative EEG panels can be overcome by using semi-automated software (currently in beta) reliably. 

Funding: None