Sleep studies face new challenges in terms of data, objectives and metrics. This requires reappraising the adequacy of existing analysis methods, including scoring methods. Visual and automatic sleep scoring of healthy individuals were compared in terms of reliability (i.e., accuracy and stability) to find a scoring method capable of giving access to the actual data variability without adding exogenous variability. A first dataset (DS1, four recordings) scored by six experts plus an autoscoring algorithm was used to characterize inter-scoring variability. A second dataset (DS2, 88 recordings) scored a few weeks later was used to explore intra-expert variability. Percentage agreements and Conger's kappa were derived from epoch-by-epoch comparisons on pairwise and consensus scorings. On DS1 the number of epochs of agreement decreased when the number of experts increased, ranging from 86% (pairwise) to 69% (all experts). Adding autoscoring to visual scorings changed the kappa value from 0.81 to 0.79. Agreement between expert consensus and autoscoring was 93%. On DS2 the hypothesis of intra-expert variability was supported by a systematic decrease in kappa scores between autoscoring used as reference and each single expert between datasets (.75-.70). Although visual scoring induces inter- and intra-expert variability, autoscoring methods can cope with intra-scorer variability, making them a sensible option to reduce exogenous variability and give access to the endogenous variability in the data.

Study Objectives New challenges in sleep science require to describe fine grain phenomena or to deal with large datasets. Beside the human resource challenge of scoring huge datasets, the inter- and intra-expert variability may also reduce the sensitivity of such studies. Searching for a way to disentangle the variability induced by the scoring method from the actual variability in the data, visual and automatic sleep scorings of healthy individuals were examined. Methods A first dataset (DS1, 4 recordings) scored by 6 experts plus an autoscoring algorithm was used to characterize inter-scoring variability. A second dataset (DS2, 88 recordings) scored a few weeks later was used to investigate intra-expert variability. Percentage agreements and Conger’s kappa were derived from epoch-by-epoch comparisons on pairwise, consensus and majority scorings. Results On DS1 the number of epochs of agreement decreased when the number of expert increased, in both majority and consensus scoring, where agreement ranged from 86% (pairwise) to 69% (all experts). Adding autoscoring to visual scorings changed the kappa value from 0.81 to 0.79. Agreement between expert consensus and autoscoring was 93%. On DS2 intra-expert variability was evidenced by the kappa systematic decrease between autoscoring and each single expert between datasets (0.75 to 0.70). Conclusions Visual scoring induces inter- and intra-expert variability, which is difficult to address especially in big data studies. When proven to be reliable and if perfectly reproducible, autoscoring methods can cope with intra-scorer variability making them a sensible option when dealing with large datasets. Statement of Significance We confirmed and extended previous findings highlighting the intra- and inter-expert variability in visual sleep scoring. On large datasets those variability issues cannot be completely addressed by neither practical nor statistical solutions such as group training, majority or consensus scoring. When an automated scoring method can be proven to be as reasonably imperfect as visual scoring but perfectly reproducible, it can serve as a reliable scoring reference for sleep studies. * EEG : – Electroencephalogram EOG : – Electro-oculogram EMG : – Electromyogram PSG : – Polysomnography BAS : – Baseline night EXT : – Extended sleep opportunity BEF : – Night before sleep deprivation REC : – Recovery night after sleep deprivation V : – visual scorer A : – automated analysis Aseega DS1 : – Dataset 1 DS2 : – Dataset 2

PURPOSE: ASEEGA is a sleep automatic scoring algorithm, based on single-channel (EEG). Compared to visual scoring in patients, ASEEGA sensitivity/specificity for wake detection proved 84%/98% respectively (JSR 2006; 15(Suppl 1):P295). The aim of the present study is to evaluate the ability of the algorithm to measure, in real-time, the amount of recorded sleep time (RST) in patients with Sleep Apnea Syndrome (SAS), and to automatically detect when RST exceeds 2 hours.

STUDY OBJECTIVE To assess the performance of automatic sleep scoring software (ASEEGA) based on a single EEG channel comparatively with manual scoring (2 experts) of conventional full polysomnograms. DESIGN Polysomnograms from 15 healthy individuals were scored by 2 independent experts using conventional R&K rules. The results were compared to those of ASEEGA scoring on an epoch-by-epoch basis. SETTING Sleep laboratory in the physiology department of a teaching hospital. PARTICIPANTS Fifteen healthy volunteers. MEASUREMENTS AND RESULTS The epoch-by-epoch comparison was based on classifying into 2 states (wake/sleep), 3 states (wake/REM/ NREM), 4 states (wake/REM/stages 1-2/SWS), or 5 states (wake/REM/ stage 1/stage 2/SWS). The obtained overall agreements, as quantified by the kappa coefficient, were 0.82, 0.81, 0.75, and 0.72, respectively. Furthermore, obtained agreements between ASEEGA and the expert consensual scoring were 96.0%, 92.1%, 84.9%, and 82.9%, respectively. Finally, when classifying into 5 states, the sensitivity and positive predictive value of ASEEGA regarding wakefulness were 82.5% and 89.7%, respectively. Similarly, sensitivity and positive predictive value regarding REM state were 83.0% and 89.1%. CONCLUSIONS Our results establish the face validity and convergent validity of ASEEGA for single-channel sleep analysis in healthy individuals. ASEEGA appears as a good candidate for diagnostic aid and automatic ambulant scoring.

Une méthode d'analyse automatique de l'électroencéphalogramme de sommeil est présentée. Cette méthode est constituée de deux parties distinctes: une partie analyse de l'électroencéphalogramme, suivie d'une partie classification de la nuit en stades de sommeil (hypnogramme). L'analyse est réalisée par modélisation autorégressive (AR) à deux échelles de temps. La classification est itérative. L'initialisation de la procédure s'effectue à l'aide d'une base de données fixée a priori. Cette base de données est modifée au cours des itérations afin de s'adapter à l'enregistrement analysé. Les paramètres utilisés sont les quantités de puissance normalisée contenues dans des bandes de fréquence prédéfinies. L'originalité de notre approche est de n'utiliser qu'une seule dérivation électroencéphalogramme, sans tenir compte des informations pouvant provenir de l'électro-oculogramme ou de l'électromyogramme. La partie analyse, lourde en calculs, est réalisée en ligne. La partie classification ne prenant que quelques minutes, on peut envisager une implémentation temps réel de la méthode.

Filter banks analysis is an easy and quick computing way to implement time-scale methods, these being well suited for short events as well as for large waveforms detection tasks. We propose a non-uniform oversampled filter banks to analyze sleep EEG single channel signal, the different subbands matching the classical EEG rhythms. In order to preserve the temporal shapes information, filter banks are oversampled. Coupling the informations conveyed by the different outputs allow to perform an automatic hypnogram.

An automatic procedure for the spectral analysis of an all-night sleep electroencephalogram (EEG) is presented. This method relies on a fixed database initializing a procedure which adapts parameters so that they match to the signal to be analyzed. Parameters coming from database are normalized power spectra in predefined frequency bands. The novelty of our approach is to use flexible sleep stage patterns rather than fixed ones, these patterns being iteratively updated (thanks to a short/long term analysis) in order to cope with the EEG variability. The main part of the procedure, performed on-line, is followed by a very short off-line processing yielding a real time implementation. The procedure adaptation ability is shown on detection of Rapid Eyes Movement (REM) events, the latter being well known for their inter as well as for their intra individual extreme variability.

Time-frequency or time-scale methods are well suited for short events as well as for large waveforms detection tasks. Filter banks analysis is an easy and quick computing way to implement these methods. We propose a non-uniform oversampled filter banks to analyze EEG single channel signal, the different subbands matching the classical EEG rhythms. We use oversampled filter banks to preserve the temporal shapes information. In order to illustrate this method, we show different spindle temporal structures as also their meaning in terms of sleep stage detection.

The aim of this study was to assess the effects of exposure for one night to moderate bright light (BL) on subjective and objective measures of alertness, performance, and mood. Eight subjects were exposed to either BL or dim light (DL) during one night. During the previous day, they were exposed to bright light in both experimental conditions. Nocturnal BL suppressed melatonin secretion and increased body temperature. The ability to stay awake during the night, as measured by the power density in the alpha band, was significantly improved by BL exposure. BL also improved subjective alertness and performance during the first part of the night. However, improvements in the last two variables were followed by marked disruption in early morning patterns, with deterioration of mood and motivation.

The circadian system is synchronized on 24-hour by the light-dark synchronizer and by the social time cue. The circadian rhythm sleep disorders share a common underlying chronobiological basis. These disorders may be due to either jet-lag, shift-work or to lesions of the peripheral visual pathway. The delayed sleep phase syndrome, the advanced sleep phase syndrome, the non 24-hour syndrome as well as the irregular sleep-wake patterns are described. Therapeutic approaches, in particular with melatonin and (or) bright light, are presented.

Eight healthy subjects were studied during 39-h spans (from 07:00 on one day until 22:00 the second) in which they remained awake. During one experiment, subjects were exposed to 100 lux of light between 18:00 and 8:00, and during a second experiment, they were exposed to 1000 lux during the same time span. Throughout the daytime period, they were exposed to normal daylight (1500 lux or more). The nighttime 1000-lux light treatment suppressed the melatonin metabolite aMT6s, while the 100 lux treatment did not. On the treatment day, the 1000 lux, in comparison to the 100 lux, light treatment resulted in both an elevated temperature minimum and a delay in its clock-time occurrence overnight. No real circadian phase shift in the temperature, urinary melatonin, or cortisol rhythms was detected after light treatment. This study confirmed that nocturnal exposure to lower light intensities is capable of modifying circadian variables more than previously estimated. The immediate effects of all-night light treatment are essentially not different from those of evening light. This may be important if bright light is used to improve alertness of night workers. Whether subsequent daytime alertness and sleep recovery are affected by the protocol used in our study remains to be determined.