Marcus C Ng1. 1. Section of Neurology, Department of Internal Medicine, Health Sciences Centre, University of Manitoba, Winnipeg, MB, Canada.
Abstract
PURPOSE: Rapid eye movement (REM) sleep can help localize the epileptogenic zone in multifocal epilepsy when interictal discharges appear diffuse. However, REM sleep is reputedly rare and easily overlooked in the Epilepsy Monitoring Unit (EMU). The aims of this study are to determine the characteristics of REM sleep in a typical EMU and whether using automated artifact recognition can meaningfully enhance REM sleep detection. METHODS: Artifact-based REM sleep detection was applied to 581 nights of EMU recording from 100 patients over 12 months. REM sleep had been manually detected at the time of recording. The index of suspicion for manual detection was raised after 6 months. Artifact-based detection was compared with manual detection, and the impact on localization was assessed. RESULTS: REM sleep occurred in 77% of EMU nights. Thirty-six patients achieved REM sleep nightly and 62 patients on at least one night. Mean admission was 5.83 days. Mean REM sleep duration was 5.92 minutes over 1.88 mean nightly bouts. Raising the level of suspicion increased manual detection rates from 22.6% to 40.5%. The artifact-based detection rate was 96% and provided additional localizing information in 10% of epilepsy patients. CONCLUSIONS: REM sleep is common in the EMU, but bouts are few and brief. Capturing these bouts to maximize the yield of REM sleep in the EMU is made possible by automated artifact recognition whose results could enhance localization of the epileptogenic zone.
PURPOSE: Rapid eye movement (REM) sleep can help localize the epileptogenic zone in multifocal epilepsy when interictal discharges appear diffuse. However, REM sleep is reputedly rare and easily overlooked in the Epilepsy Monitoring Unit (EMU). The aims of this study are to determine the characteristics of REM sleep in a typical EMU and whether using automated artifact recognition can meaningfully enhance REM sleep detection. METHODS: Artifact-based REM sleep detection was applied to 581 nights of EMU recording from 100 patients over 12 months. REM sleep had been manually detected at the time of recording. The index of suspicion for manual detection was raised after 6 months. Artifact-based detection was compared with manual detection, and the impact on localization was assessed. RESULTS: REM sleep occurred in 77% of EMU nights. Thirty-six patients achieved REM sleep nightly and 62 patients on at least one night. Mean admission was 5.83 days. Mean REM sleep duration was 5.92 minutes over 1.88 mean nightly bouts. Raising the level of suspicion increased manual detection rates from 22.6% to 40.5%. The artifact-based detection rate was 96% and provided additional localizing information in 10% of epilepsypatients. CONCLUSIONS: REM sleep is common in the EMU, but bouts are few and brief. Capturing these bouts to maximize the yield of REM sleep in the EMU is made possible by automated artifact recognition whose results could enhance localization of the epileptogenic zone.