Abstracts

Rationale: High-frequency oscillations (HFOs) are a promising biomarker of epileptic tissue and networks. Results based on the human marking of limited data segments suggest that HFOs during sleep may be more specific. This hypothesis has not been tested in full recordings, nor is it known whether the potential change in specificity is due to the HFOs themselves or due to a change in the number of artifacts. The objective of this study was to assess how the sleep stage influences HFOs and artifacts.

Methods: We detected HFOs and multiple artifacts in intracranial EEG recordings for 31 subjects using previously validated automated detectors. Data were scored for the sleep stage by a sleep technician using simultaneously recorded scalp EEG. We calculated the rate of true positive HFOs (putative HFO detections not coincident with any artifact) and the rate of each artifact type within each sleep stage. We then compared the rates with the seizure onset zone and resected volume using an asymmetry measure. Additionally, we computed 25 quantitative features of the HFO morphology and assessed if they were predictive of given sleep stages utilizing the area under the receiver operator curve (AUROC).

Results: Within NREM sleep, true positive HFO rates were higher in deeper stages (p < 0.01, Wilcoxon Rank Sum). However, HFO rates were still higher when patients were awake than asleep (p < 0.001, Wilcoxon Rank Sum). Muscle and fast transients that were not removed by the HFOs detection algorithm were highest during REM., whereas false-positive HFOs were highest during awake. The association of true positive HFOs with the seizure onset zone and resected volume was consistent across all sleep stages (median change in asymmetry < 0.1). HFO features were also highly consistent across sleep stages, with all AUROC values being less than 0.65.