Two auto‐detection methods for eye movements during eyes closed

Abstract

Eye movements during closed eyes closely reflect changes of the arousal level during transition from wakefulness to sleep. Because they contain both rapid and slow eye movements (REM and SEM), it has been difficult to detect them automatically. Hiroshige recently developed the method of linear regression analysis for automatic detection of the two types of eye movements, and we have developed a template matching method for autodetection. The aim of the present study was to compare both auto-detection methods and visual scoring for REM and SEM. The results revealed high agreement between the two quantitative methods and the visual scoring, indicating that auto-detection of eye movements is useful for quantitative evaluation of arousal level.

Slow eye movement detection can prevent sleep-related accidents effectively in a simulated driving task

“…Several studies have specifically investigated SEM detection to identify sleep onset automatically (Atienza et al, 2004;Hiroshige, 1999;Magosso et al, 2006;Suzuki et al, 2001;Va¨rri et al, 1996;Virkkala et al, 2007). Va¨rri et al (1996) proposed a SEM detection method using non-linear filtering techniques and applied it to EOG data measured during evening activities before going to bed.…”

Automatic Detection of Single Slow Eye Movements and Analysis of their Changes at Sleep Onset

“…The growing interest in SEMs has led to development of various algorithms for automated SEMs detection in EOG, based on different techniques and with different aims. However, some of these algorithms do not identify single SEMs (Värri et al, 1995); (Virkkala et al, 2007); others identify single SEMs, but the validation procedure either exhibits moderate performance (48% sensitivity) (Värri et al, 1996), or is not examined in depth (consisting only in the autodetection/visual scoring ratio) (Hiroshige, 1999); (Suzuki et al, 2001), or is absent (Shin et al, 2011). Moreover, so far none of these studies has characterized SEMs in terms of their parameters (such as amplitude or duration) nor has investigated how SEMs parameters evolve across the sleep onset period.…”